off-topic-tech

[oblique hamlet]

my ryzen 7 7700x has one

[night girder]

I mean, a CPU with iGPU for sure. They exist.

but reversed?

[oblique hamlet]

oh

[night girder]

Zeus is a GPU with a CPU.

[oblique hamlet]

that is an interesting idea

[night girder]

But also, extra heat? Curious how they want to pull it off.

[oblique hamlet]

but with something GPU size, it would be easier to immersion cool than a whole case

[mental oriole]

They making it so you can upgrade the memory on it... that's neat

and 2x pcie lanes so they're doing something funky for sure.

Let's hope it's not a fluke.

[mental oriole]

I wonder if it is made to offload certain computations that's not worth doing on the gpu side.

e.g model view matrix updates

[oblique hamlet]

it looks like one of their main focuses is 3d rendering

and ray tracing

[mental oriole]

Ray tracing is a part of 3d rendering so makes sense. Less for gaming (at least for the moment)

But I'll believe it when I see it.

[visual tree]

@night girder Blame this video for high RAM prices

The only person I am clueless about is that blonde woman who is apparently Kirk's wife

[oblique hamlet]

ram prices are insane

32 gigs of ddr5 cost a whopping $200 2 weeks ago

[visual tree]

Don't want to know current price 🙈

[slow flint]

Oh it's these guys
I remember hearing about them about 1-2 years ago
I didn't think they where targeting gamers but looks like they are now

Such a strange card, will be interesting to see it launch

[tribal kraken]

Litte dissappointed for Bayerndynamic MX230 wireless headset that arrived today. Does not sound as good as my 7 years old MMX300 gen2 wired one. Feels like the treble end is choked even if try to fix it with EQ

[slow flint]

I had a look at what RAM I have
I bought a set of 32GB (2x16GB) 6000MHz CL30 Klevv RAM for $162

Now its $759

[tribal kraken]

My 96GB kit seems to hover at 1700-1800€

Better to put it at use

Processing some 3D scan data again

[oblique hamlet]

glad to see g skill putting the blame where its at

they said:
"DRAM prices are experiencing significant industry-wide volatility, ... driven by unprecedented high demand from the AI industry."

[slow flint]

Wish I got 64gb
If I had a spare set I would so sell rn

[soft bloom]

Who was whining about yt today?

I am not usually wathching it on the phone, but power struggles forced me...

Noticed that when you try to get back from fullscreen in landscape orientation, first press of the button does else other than simple displaying an add for half the screen!

I started reading Godot docs and for some reason I can't quite grasp it's architecture
As software dev, I don't necessarily want to dive into hw details of rendering, but just to get a sense of what are building blocks in the engine and how to efficiently work with them.

[bronze jasper]

the zeus gpu's cpu is a riscv system. low power. very little extra heat or power draw over head. it's a neat design choice. wonder what they're going to do with that

[slow flint]

I'm curious as to how it will actually work
Can it be plugged into a Linux or Windows x86 system and have the RISCV processor run independently? Or does it have to use a RISCV system to be compatible

Of it is only comparable with RISCV systems its gonna be a long time till its something we can use and it'll be more developers

[twin dew]

Its server as a card thing.
That specific one they showcased has nothing to do with gaming, it is supercomputer slice with that onboard 400Gb/s NIC and separate BMC NIC.
Which makes those display connectors on it just weird, same for the PCIe power connector instead of EPS power.

Making it seem even more like complete vaporware they are trying to use to get investor money.

[bronze jasper]

their gaming card whatever it will be will have the same riscv setup

[twin dew]

What they show is weird mashup of consumer GPU, server GPU, and supercomputer slice.

Display-connectors and power connector are consumer GPU.
LANs are supercomputer slice.
Rest is server GPU.

[sharp matrix]

in other words who the hell knows what this is supposed to even be, assuming anything even ever makes it to market

[twin dew]

Like I said, seems like investor bait, not actual product.

[sharp matrix]

personally im just considering this vaporware for now, because what the showcased doesnt make alot of sense

[bronze jasper]

i'm cheering them on because the space needs more movement

does seem kinda vapory though

[twin dew]

Would be easier to cheer on them if the specs were reasonable, or the design made any sense.

[sharp matrix]

agreed but i cant help but think this very weird mashup of both server and consumer together in hardware, that is suppose to some gpu that maybe in theory consumers could somehow use in their pc, idk atm this just doesnt seem like a good idea

[twin dew]

Because the specs don't make sense compared to the die area shown.

For the record, new 400Gb/s NICs cost in range of $2000 for themselves.

And anything on server side would use that 8-pin EPS power connector, not 8-pin PCIe.

[sharp matrix]

yeah and somehow consumers are going to be spending thousands of dollars on a gpu, sorry check this just doesnt look like its going to end well, even from a cost perspective

unless this company is aiming for the absolute high end aka trying to compete with the 5090, but still the cost of this thing is going to be so high that the average consumer wont be able to afford it

[bronze jasper]

they must have some interesting way to mount them all that they're not showing yet. whats up with 2 pcie interfaces

[sharp matrix]

idk im going to try not to give this vaporware much thought....

call me when this company actually comes with a reasonable idea, this aint it

[twin dew]

The left side is normal with hook.
Right side is either vaporware, or for some kind of riser. For either something weird or for the on-board CPU.

But yeah, current marketing is just investor bait.

Will see if they are scam and run, die, or actually make something.

But I don't have high hopes for that last one.

[sharp matrix]

me either, so far what they showcased nothing is reasonable, this surely has to be investor bait, like showcase this cool weird thing that no other company is doing, ignoring the obvious facts of why no other company is doing it to begin with

[twin dew]

So 3x 5090 FP64 performance, with about 200W power budget (150W from single PCIe power connector + slot power)
The power efficiency would need to be even more insane than the performance per die-area XD

[sharp matrix]

yeah im calling bullshit on that to XD

[twin dew]

Ah, ok, pretty easy to get 3x 5090 FP64 performance.
As 5090 has the gimped consumer card 1/64 FP64 performance compared to FP32.
When professional cards have 1/2.

[sharp matrix]

it all sounds like complete marketing bs so far, now maybe if they had a actual product to showcase this insane performance to power efficiency....but i dont think thats going to happen as it sounds all very unreasonable

[twin dew]

So they are basically lying for that part.
And the left-most graph is just weird.

Also note that that 10x pathtrace is with 4 cards basically.
2.5x with one, and you will never get that kind of linear progression with multiple dies.

[sharp matrix]

well thats why this has to be investor bait, investor sees insane claims, throws money at it, because investor think magic is possible XD

[twin dew]

But from that middle one, I would expect even more this is pure bullshit.
And the people making the slice didn't even realise that 5090 has crap FP64 performance, just used as it is very known name.

When comparing to that left.
Where somehow they have 300x FP64 performance on 4-core vs. Nvidias B200 supercomputer parts.

Basically there is complete disconnect between left and middle for what kind FP64 performance there is.
Middle says terrible.
Left says insane.

Because card that has 5090 FP32 performance, with 1:2 FP64 performance, would have 32x better FP64 performance than 5090.
But B200 has more FP32 performance than 5090, and has 1:2 FP64.

[sharp matrix]

32x better performance compared to the 5090, sure this sounds reasonable, in some insane alternate reality

[twin dew]

No?
That is pretty common that some cards have 1:64 FP64 performance compared to FP32 (basically no proper FP64 support at hardware level).
And some cards have 1:2 FP64 performance compared to FP32 (proper hardware level FP64 support in all FPUs).

And with same FP32 performance, second will have 32x FP64 performance of the first.

Been Nvidia silicon differentiator for ages.
Where only the best die has 1:2 FP64, rest have 1:64 FP64.

So that you cannot use the lesser cards for scientific etc. work, as the performance at necessary accuracy is crap.

[twin dew]

Basically 5090 is consumer card with consumer die, and so has 32x worse FP64 performance than it could have had.

With some additional die area required.

Intentionally gimped in FP64 by Nvidia.

9070 XT has almost same FP64 performance (1.521 TFLOPS) as 5090 (1.637 TFLOPS), as NAVI48 has 1:32 FP64.

7900 XTX is faster than 5090 in FP64: 1.918 TFLOPS (1:32)

Basically those mean that single FP64 operation takes 64 times more time, 32 times more time, or 2 times more time than single FP32 operation.

Where that 1:2 is "optimum", where you can do either one FP32 per clock and one FP64 per two clocks (32-bit FPUs), or two FP32s per clock and one FP64 per clock (64-bit FPUs).

And anything below is combination of minor die size savings and segmentation where you gimp that FP64 performance.

Also makes so that Nvidia 1000-series NVIDIA TITAN V has 7.450 TFLOPS (1:2) FP64 performance, so about 4.5x FP64 performance compared to 5090.
That was the last non-gimped "consumer" card from Nvidia.
Then they switched to 1:32 for the highest non-server dies, and then later to 1:64.

For AMD, 9070XT and 7900XTX have 1:32, 6950XT and before have 1:16 FP64.

[tribal kraken]

3D scanning with 3DMakerpro Toucan, shiny black surface no problems with the 3R class blue lasers, no spray needed on this. Resulted about 20GB data and used about 80% battery. No laptop used on scan site.

[bold tapir]

Dont all the gpus count in 16 but then they divide it by 2 and call it 32?

[bold tapir]

Or is it about 32 and 64

Like core difference between gpu and cpu is thats gpu has many stupid cores and cpu has like 8 good cores those can do all the variety of operations

[twin dew]

GPUs traditionally work in FP32.
And that is what gaming etc. uses.

Then scientific stuff etc. uses FP64.
And then for AI and limited precision you use FP16, FP8, INT8 etc.

[bold tapir]

Oh ok i misremembered 32 as 16

[twin dew]

So 5090 has 21760 FP32 FPUs (shader units) for example.

[bold tapir]

Oh and last nvidia gen is for ai so its even more stupid

[twin dew]

But those FPUs intentionally suck at FP64, only being able to do FP64 operations at 1/64th speed.

Of FP32 operations.

[bold tapir]

And you say that if they dont try to save on die they can reach that theoretical 64 being twice harder than 32

[twin dew]

You can do two-cycle FP64 on FP32 unit with limited amount of extra circuitry.

But that would be competing with the extra high price enterprise server GPUs.

[bold tapir]

But they dont make that for consumer 5090

[twin dew]

Been pushing it up and up and up in the stack for segmentation.

From that 1:2 in Titan V, to 1:32 in highest model, to 1:64 even in highest model outside of the specialist server GPUs.
Before it was limited to Titans and to some of the Pro-line cards.

[bold tapir]

So it takes extra clock cycles

[twin dew]

Basically to nerf the performance to point where you can use the non-server stuff for code development, but actually using the code needs the high-price GPU servers.

As shown in that that Titan V from 2017 has faster FP64 than 5090.

[bold tapir]

Oh they intentionally limit again like with vram

[twin dew]

And it had "Titan V CEO Edition" from 2018 that has 32GB of VRAM.

[twin dew]

Yes.

Segmentation.

[bold tapir]

Why do they make 8gb with pcie x8 tho

Its like death in 2025

[twin dew]

But ok, that CEO edition was non-sold 20 unit special run.

The base Titan V had 12GB of VRAM.

[bold tapir]

I thought its easy to manually add vram

[twin dew]

No.

[bold tapir]

Or is it bios bound

[twin dew]

Even physically doing it needs lot of specialist equipment.
And then there are those VBIOS limits, need VBIOS that supports that VRAM amount too.

[bold tapir]

Yes replacing chips manually

[twin dew]

If there isn't official similar GPU you can steal VBIOS from, not going to work as Nvidia has locked those up hard.

[bold tapir]

To double the amount

[bold tapir]

Can amd?

[twin dew]

No idea about current state.

[bold tapir]

Btw from vendor perspective am i supposed to buy new gpu each 2 years or smth because thats how long heat interfaces may live

[bold tapir]

Now i know why they state 16 32 and 64 flops on tech power up

[night girder]

That's the worry of journalist too, they mention it in the article. The young age of the company. No actual hardware. No prior experience.

But we can hope a little bit (copium).

[twin dew]

Yeah, because the ratios between them aren't static.
You could have FP16 being twice as fast as FP32 and FP64 being half as fast as FP32 as "gold standard".
But for that 5090 for example FP16 is same as FP32 (throwing half the performance away for simplification).
And FP64 is just 1/64:th of FP32, for segmentation reasons.

As it is easier to just run each FP32 unit as FP16 unit, than try to make it work in way that would allow you to do two different calculations at same time.

So that that FP32 unit would work as two FP16 units.

[languid gulch]

new challenge: find something that the 1660 does better than a 5090

[twin dew]

32bit PhysX

[languid gulch]

true but that also wasn't a challenge 🤣

[pure karma]

price

[bronze jasper]

the price is inconsequential after you own the gpu. the gpu doesn't "do" the price.

it's not part of the spec sheet at all

[dire igloo]

Then power consumption

[bronze jasper]

depends if you're going for more power or less power

some people think more is better

gruntgruntgruntgrunt

[night girder]

which can affect your wallet

[tough owl]

like my server drawing 90w at idle

[dire igloo]

then let's call it cost of operation

[oblique hamlet]

less ai bs

[soft bloom]

How much of a good idea is to try to individually charge battery packs inside bluetti to 'revive' it?

[jagged snow]

You mean individually charge cells?

[twin dew]

Manually: very bad idea.
With suitable intelligent charger for the battery cell type, that supports charging single cells, completely normal.

[soft bloom]

One of the bluetti units threw "battery pack communication failure" and when it occasionally decided to rshos anything on display - it shows 100%, immediately shuts down at any load when not powered from the grod.
On bluettinforums some ask for firneare update in this case, but I couldn't manage to apply it. Some just add new battery controller to manage charge of cells.
Question is if there's some way to temporarily restore function by stabilizing charge of cells

[soft bloom]

They are in series of 2 and 3, but essentially yes

[twin dew]

Which means it cannot talk with that battery controller.
Not that the battery cells are in any way faulty.

And when it cannot talk with that battery controller, it doesn't know the state of charge for the battery, and it will not use it, or might not even be able to use it, for safety reasons.

[languid gulch]

your next car is gonna be for sale near me this week if you're interested

[twin dew]

What?

[languid gulch]

1955 300SL

www.barrett-jackson.com/2026-scottsdale/docket/vehicle/1955-mercedes-benz-300sl-gullwing-coupe-295949?origin=docket

[twin dew]

And makes you think I would be either interested, or be able to afford, that?

[languid gulch]

🤣

just a gorgeous german coupe in pristine condition halfway around the world

this auction gets so many Mustangs & Corvettes, it's just nice to see some variation

[twin dew]

Car collectors dream.
But I'm not car collector, cars are just utility things for me.

[languid gulch]

hey now, there's some german utility that'll be available

[soft bloom]

Well, yeah, that's what I would expect to - simple connection failure.
But, it had this behaviour before, though a bit less persistent. And I wouldn't suspect mechanical failure. Some fix it by using another controller. In mine battery packs do have different voltages in the vicinity of 0.1V atm

[twin dew]

Doesn't mean that the cause would be just connector.
Can be the communication IC at either end, can be something else at either end PCB etc.

Almost all Li-Ion batteries have battery management PCB inside the battery, with the cells.
Taking care that the individual cells don't get overcharged or into too low charge state.

0.1V differences don't matter much.

[languid gulch]

ok dumb question time: when you undervolt a cpu/gpu, does it try to counter it by raising amps?

[soft bloom]

no, that's kinda the point

[dire igloo]

Speaking of cool German cars

[languid gulch]

the shadow on the ground makes that look fake 🤣

[soft bloom]

just a couple additional lights, nothing strange

[night girder]

Today I learned.

That YT put playback speed 4x behind paywall.

Why the fuck 🤷‍♂️

[sharp matrix]

well

[twin dew]

Jensen Huang says relentless negativity around AI is hurting society and has "done a lot of damage"
Won't somebody think of the CEOs?
https://www.techspot.com/news/110879-jensen-huang-relentless-ai-negativity-hurting-society-has.html

[languid gulch]

"hey jensen, people hate this thing we're trying to force on them, do you think we should listen to them and stop doing that?"

"NO. FORCE IT HARDER."

[night girder]

whats up with all these crazy ceos living on another planet than us?

[languid gulch]

there's a point at which you have so much money that you don't know what to do with it, so the only thing left is to spend money to stop others from making as much money as you

it's like that with every resource now

[soft bloom]

Damn,
Either voltgae is too low or somehing else but the working Bluetti keeps switching input relay on and off like 10s ... Not really charging

[soft bloom]

Heh, so what could be the damage from manual charging? Turns out I asked too late...

Need to figure out some way to bring it back to life or pinpoint the cause. Batteries hang around 3.2V at the moment

[twin dew]

Overcharging Li-Ion cells damages them permanently, and can cause them to go into thermal runaway.

Aka fire or explosion.

[soft bloom]

F. where's he danger zone?

wait, even if there' builtin controller?

[twin dew]

Way higher than that 3.2V.
But also charging speed matters.

Manual charging of Li-ion is not to be done, only by actual Li-ion charger with logic.

As they can take charge way faster than is safe, if the charging voltage isn't controlled to limit the charging current.

[soft bloom]

they now been manually charged to 3.6v, controlling both voltage and amps... it's just been done sadly 🙂
and that didn't help,
so need to figure out next step that isn't just sending to repair shop that... who knows what will do given recomendations on the internet

[twin dew]

Like I said, not the cells, but that charge monitor.

Or whatever reads the charge monitor at the main PCB end.

[soft bloom]

full on replacement is best course of action?
data bus seems to be doing fine, not hanging around or anything

[twin dew]

How would I know?

[soft bloom]

I don't dare try continuity test in a system that has large capacitors 🙂
Even though, data stuff is most likely decoupled from them

[rustic panther]

Peer pressure

I mean with nvidia it's different cuz they bet the farm on AI but most people high up in org charts are peer pressured into pushing AI

[night girder]

Tesla will stop selling FSD after Feb 14.

FSD will only be available as a monthly subscription thereafter.

• Elon The Idiot

FSD: Full Self Drive.

What will be next they make a service off. Breathing air as a service?

You have to pay us monthly else you aren't allowed to breath in air.

[soft bloom]

Ability to recharge the car

And for cybertrucks, if you don't pay every weak it will coat itself in water and get extra rusty

At this point, there's no reason for them not to
It's just a big mob doing mob stuff

Create demand and scam buyers

[soft bloom]

Should I expect it work it's thing when it just receives power?
There's power of well insulated power cables, tiny 4 wires data bus connected to the main pcb, and a free port with more contacts.
There's an idea that maybe... If I supply correct voltage on those power cables, it will at least be good? Though, charge on power cells doesn't seem to be the problem atm

[twin dew]

If it was ok, it would work.
Something is dead in the monitoring.
You just know that the main CPU isn't getting data.
But that can be because the receiver in main-CPU is dead.
Something is dead in the battery monitor PCB.
Or something is dead between the two.

[soft bloom]

Oh nvm no way i could connect power without removing insolation...

[twin dew]

And very certainly two of those four small wires are what is powering that monitoring PCB.
And two are data.

And the main power is only connected over the large wires when everything is ok.

Basically: electronics are broken.
Poking around with multimeter doesn't help, and neither will randomly connecting stuff.

[soft bloom]

Well, that pcb on battery pack is proprietary, judging by name of the whole model on it...
At least it has 'bms' on it

[twin dew]

BMS = Battery Management System.

Generic name for that lithium-ion protection and monitoring stuff.

But yes, with such integrated unit that is usually part of the design.

[soft bloom]

Monitoring pcb is bms?
There technically 3 different ones
Largest is one with all the ac inout stuff, medium is the bms, and smallest is orthogonal to the largest, has 2 wide busses connected to the largest and that bus from Bms

Given that whole system doesn't power on when not connected to frid, guess it's right to assume that bms is not giving power to main board... Unless there's something more complex in how they share power

[twin dew]

Yes? There is something more complex.
That is the whole point of that system.

And remember, the actual main stuff has power, when it complains it doesn't get status from BMS.
And without data from BMS, the battery output will not be turned on.

[soft bloom]

I mean without external power connected, how would front panel react to anything? That's why I assume that it would draw power from battery pack at all times...
Except, if bms thinks "no, we can't discharge further" then it would cut the power, and then main board would need external power to do anything, including reading data from bms...
That's why I assumed also that bms's mind is powered from batteries directly, and not through main board, cause then there could be condition like this exact one, I guess ... When bms cuts the power and main board cannot do anything now without external power.
But even if thAt was the case, it would always react when external power present, which it doe not...

[soft bloom]

Yeah...
Though not sure about "while point of that system" tbh

[twin dew]

Other way around.
Battery isn't used if it isn't known to be good.
It will never be used to directly power the BMS.
Instead when BMS says the battery is good, the battery output is enabled, to feed the main electronics, which feed the BMS.
And as long as BMS says that the battery is good, the main electronics will get power from the battery, and is able to feed the BMS.

But once BMS says that the battery is not good, the battery output is disabled.
And if there wasn't mains power, the device dies until mains power is connected.
And then sees if it can start charging the battery, in which case it would be reconnected and charged.

[soft bloom]

Wouldn't it be a lil better if bms could shut itself off if there's need to cut battery consumption, but be powered from battery directly during normal operation?
Though I guess that would require some voltage conversion, which is on main board anyway

[twin dew]

Using the battery needs power from the "outside the battery".
To keep the output connected.
Just once bootstarted, it can then use the battery power to keep that output connected.
Point is, you need to be able to completely cut all consumers from the battery if needed, to keep it from discharging completely and bursting into fire.

[soft bloom]

Okay, so basically no point in even trying to turn it on without external power ...
And no point trying anything with batteries themselves...
Perhaps that unused bus on bms is for debugging (monitoring etc?), but good luck making that work, hehe.
So, huh, I guess I have no idea what to do other then replug that bus from bms

[twin dew]

And you need to monitor that on per-cell level, so a BMS that monitors all individual cells.

But it cannot be powered by the battery directly, as that would allow it to drain the battery and make it burst into fire.

[soft bloom]

I would imagine having mechanical switch that just pops when bms decides it's time to stop any consumption and it would be possible to activate it back by bms when external power is passed to it from main board

[twin dew]

And you don't have schematic for that thing, so you don't know how the various bits are connected.

So you have been poking in the battery side blind.
And disconnected important part of the chain then if the battery to main electronics smaller wires aren't connected.

[soft bloom]

Replug as in unplug and plug back 🙂

Some guy on internet says that bms can be "jumpstarted" if you supply 24v on 2 ends of the pack.
I feel like it's just a cheap way to blow up whole thing...

Especially since there's no way there should be 24v with series of 5 groups of 3 3 2 3 3 cells

[twin dew]

Depends on device.
Some did use battery side powered ones, which go braindead if the voltage drops too low.
And need such stuff into RIGHT place at right voltage.

Usually in that case there is double shutdown.
And the BMS takes the power between the shutdowns.

[soft bloom]

Double as in 2 boards are shutdown or 2 consecutive shutdowns?

[twin dew]

Basic idea, but usually not using relays anymore:
Battery - relay - bms power - relay - main electronics

[soft bloom]

That might be case cause there's seemingly some kind of reset when power is supplied through 'solar panel' port.
Though it didn't seem to work exactly as expected - wouldn't really revive, but at least boot with external power

[twin dew]

Where both relays are powered from that bms power island.
So if both are ever shut down at same time, the whole system breaks down and need jumpstarting.

Which is why that design is very frowned upon.

[soft bloom]

Always good when something can work with transistor 🙂

[soft bloom]

Hehe, sounds like the stuff I saw ...

Still can't wrap my head around why supplying voltage, that is higher then normal operational to the system with battery cells, would be the way to go forward...

[jagged snow]

What's the pack arrangement?

[soft bloom]

Isn't that same as trying to overcharge batteries m

[twin dew]

That jumpstarting needs the right voltage, in the range of the normal battery output.

[soft bloom]

3 3 2 3 3

[jagged snow]

So you're saying it's a 3p5s arrangement with one group missing a cell?

[twin dew]

Which is impossible to be working set.

[soft bloom]

5 groups of 3.2 nominal voltage...
Idk how 24v is inside normal

[soft bloom]

Yes 🙂
I mean, exactly that

[twin dew]

Some of the cells are in parallel, some are in series.
Are the series conections inside the parallel ones, or parallel ones inside the series ones?

[jagged snow]

What chemistry is it?
For LiIon 3.7 is nominal, for LiFe(Po) it's 3.2

[soft bloom]

Lifepo4
Wtt 32700

[jagged snow]

Are you wanting to put that voltage directly across the pack or just supply it to the bms without the pack connected?

[soft bloom]

2 diametral contacts, 6 numbered ones

[jagged snow]

Oh, I see what's going on there

[twin dew]

So you cannot see how they are connected.
As that plate you see isn't the actual conductor.

[soft bloom]

Hypothesis is that jumpstarting bms requires giving it that 24v
And i want to avoid doing anything (more) with cells

[twin dew]

The actual contacts are hidden behind that and the foam backing it has.

[jagged snow]

Yep
This is very likely a 7s2p pack

I would not roll with that assumption until I actually validated it though

[twin dew]

Yeah, 7 in series from the pad numbering.

[soft bloom]

Fair, yes. Would make sense with 6 numbered contacts

[twin dew]

B- to B1, ... B6 to B+.

And the BMS is monitoring all the intervals.

[jagged snow]

And nominal voltage would be 22.4v

So 24v would be a reasonable voltage for that pack

[soft bloom]

So, 7 times 3.2 is about 22v, I guess 24v isn't too bad

[jagged snow]

25.2v would be fully charged on that pack.

[soft bloom]

I gog bumboozled by somebody saying it's 33233 and retold that 🙂

[twin dew]

But as it is expected to work down to around 16V or so, feeding 20-22V would be better.

[jagged snow]

I missed the earlier explanation
Why can't you wake it from wall power?

[soft bloom]

I didn't actually poke voltemeter at it myself. Let me check what it says atm

[twin dew]

Either electronics are dead.
Or it is one of the stupid designs that goes permanently dead if the battery ever gets too low, without such jumpstarting.

[jagged snow]

Gotcha
Could very easily be either

[jagged snow]

Let us know what the full pack voltage is as well as the individual cell group voltages if you would

[twin dew]

But such jumpstart would be somewhere where those big power cables go.

Ah, sorry

No, that was wrong.
The eight fets near the center are the shutdowns.

[soft bloom]

It shows 100 charge, dies immediately when power is out, and once showed error specifically about commuting with BMS, and even attempting to update firmware (once it didn't show dsp version at all, the other time it did) made it die even with external power. It doesn't always work with external ac power too - seems to only reliably turn on with dc input

[jagged snow]

Some charging systems will enter a failsafe mode if a cell group falls too far below a threshold voltage and will refuse to charge it without external intervention

[twin dew]

Between B- and the Bat-

[jagged snow]

Gotcha
That does sound like control circuitry failure

[twin dew]

So need to connect - to that between them areas, and the bat+

to jumpstart.

Left is the battery to BMS shutdown.
Right is the BMS to output shutdown.

[soft bloom]

21.67v minus to plus

[twin dew]

At least very likely.

Basically if both ever went into shutdown, needs jumpstart to between.
But would need more measurements to be sure.

So would seem to be one of the stupid designs that can go into permadeath if battery voltage drops too low when there isn't mains connected.

Also, even the PCB says it is 7S:

[soft bloom]

Oh, missed that

[soft bloom]

Hm, why 4?

[twin dew]

Version 4.0.

fourth major design revision.

[soft bloom]

Measure between - or + and those chips? Would need to remove isolation coat thingy

[soft bloom]

my brain was thinking 4volts phaha

[twin dew]

But basically the battery - is on the bottom left, then it passes through those four very low resistance resistors to the left-middle area, then through the first set of fets to the center area, which is also tapped for BMS electronics ground, then over the other set of fets to center-right and to bat- wire.

So you need to get patch cable from B- to that center area momentarely, when everything is otherwise connected.

Or from the Bat- side to center, with power connected as other option.

[soft bloom]

B- to b1 3.159v
B1 to b2. 3.013v
B2 to b3 3.155v
B3 to b4 2.951v
B4 to b5 3.172v
B5 to b6 3.050v
B6 to b+ 3.175v
B- to b+ 21.67v

[twin dew]

Shot that shows the area above the visible?

[twin dew]

So the pair of cells between B3 and B4 is in worst state of charge.

[jagged snow]

Significantly out of balance then but that can happen from sitting long enough

[soft bloom]

Other side of batter pack ot other boards?

[jagged snow]

Typical target for a lot of systems is max deviation of .1v between cell groups

[twin dew]

What is upwards?

Ah, sorry, the PCB ends there.

Right off frame.

[soft bloom]

It was on netwroking stuff duty for the past couple months
Aka either powering 2w or being plugged into external ac, recharging, and giving same 2w dc

[soft bloom]

Yes. Did cut a bit low but there's just data bus right where arrow is and that's it

[twin dew]

And that left-side connector is just tapped to those direct battery connections.
For external measurement and/or battery balancer.

The B-, B1 etc.

[soft bloom]

Huh, 8 pins indeed

[jagged snow]

Usually used for in-factory pack conditioning

[soft bloom]

So jumpstarting is not where b- and b+ but somehow need to pass it to those shutdowns?

[twin dew]

Need to power that BMS, which should be gettings its ground from either that BAT-V connector, in which case the electronics are dead.
Or from that center area, between the two sets of cutoff FETs.

So that it reopens those FET sets.

You already have good voltage between B- and B+/Bat+
But B- isn't connected to Bat-, or that center area.

As the eight fets are not conducting.

Really no idea why that second set is there, when without it, the main side could power that BMS up and you couldn't go into this state.

If that BMS power isn't via that BAT-V connector.

[soft bloom]

I was measuring b- and b+, but nothing with bat- bat+ , there are no open contacts for these 2

[twin dew]

Like I said, B- to Bat+ should show same as B- to B+.
But you don't have connection between B- and Bat-, as those cutoff fets are off.

[soft bloom]

Bat-v does have 3 black and 1 blue so 2 data 2 power seams reasonable

[twin dew]

As my expectation on how that B- to Bat- is connected via the two FET banks when the BMS is working

[soft bloom]

So the would be jumpstart is between b- and b+ OR bat-v power pins?

[twin dew]

Would need to connect B- to center when mains aren't connected.
Or Bat- to center when mains are connected.

You might be able to connect to some of the legs of those FETs on the center area, and try to measure if they are connected to anything else.

Top right ones might be touchable for example.

From the pic.

Or scratch enough to get multimeter probe connection to the area.

[soft bloom]

They all have that yellowish coating

It looks like the 2 right pins are going to the right, while other 2 are going left to the chip

[twin dew]

But the two right-most also go into vias already below the connector.

[soft bloom]

The pins of soltered bat-v connector seem naked

[soft bloom]

Oh, will have to explore that. Supects are any of the B-+17 contacts?

[twin dew]

No.
If that center area is connected to any of the BAT-V pins for example.
Or any of the left-side main chips pins.

Probably one of the corner pins.

Those small fets on the left of that chip, around B1 with that white foam, are connected to the B#:s

Point is to find where this chip gets its + and -:

[soft bloom]

It should be safe to poke power pins on bat-v when external ac is on, right?
Just to check if power even gets there

[twin dew]

There should be dot on one corners.
Usually the pin on opposite side, on same end, is +
And the pin on same side, at opposite end is -

But depends on the exact chip.
Are there any markings on that chip?

Ah, sorry, usually the pin closest to marked is +, and same end opposite side is -

But basically newer the chip, less likely either of those are to be true.

[soft bloom]

Like 1?
Ignore 2 3
||2 was me looking for dot before i read further||
||3 i guess is some passthrough pokeable contact?..||

[twin dew]

Ah, yeah, that white dot should be mirrored by dimple in the actual chip.

[soft bloom]

Newer as in decades? This unit is from 2022

[soft bloom]

Yes, can be seen when looking around

[twin dew]

That marks the pin 1 on the chip.
Which was traditionally power pin.
And then goes down the same side, jumps to other side at the bottom and goes up to last pin that was traditionally ground.

[twin dew]

But in this case, from the PCB images, the two pins near the pin 1 are probably + or -, and the opposite corner two are probably the other.

But ok, that new pic made that less likely.

[soft bloom]

That beafy brown thing stands out 🙂

[twin dew]

But seems that the BMS side ground is directly from B-:

[soft bloom]

Took me a while to notice there are 3 rows of markings on the chip, but really hard to read...

[twin dew]

Or might just be the reading of that B- instead.

[soft bloom]

The brown thing (resistor?) seems to be soldered onto same plate

[twin dew]

Check if these are connected together, and where else they connect to.
Seem to be filtering caps and those should connect to the main ground/- plane.

[soft bloom]

Bottom right is same plate as pad0
And that plate goes under the chip to the top left pin

[soft bloom]

They all seem to go to that foam on the left, with 3-way stuff
Would that even go to cells?

[twin dew]

No, the upper pads go both to the chip, and the resistors on right.

Those are the lines that are used to discharge individual cell pairs to keep the cells balanced.

Via those fets on left, there is one more hidden off left of the pic with C10 and C9 leads going to it.

[soft bloom]

Oh wait, paths only touch them,after supposed filtering caps they likely go onto other side of board

[twin dew]

And the bottom cap pads are connected with vias to the ground plane.

So are these all connected?
Or is that top-right B- fill not connected to the cap bottom legs?

And is this connected to either B+, or to one of the BAT-V pins?

[pure karma]

Worlds longest troubleshooting again?🤣

[twin dew]

And if the top-right mark isn't connected to the bottom marks, find if those bottom marks connect to the bottom left chip pin, or to any of the BAT-V connector pins.

[soft bloom]

I am going to probe them now

To be clear, I should test for voltage across and not connectivity?
No wait your phrasing suggest the latter

[twin dew]

Resistance with no outside power connected.

But my current expectation would be:

With stuff between that + pin and the actual source that messes up measurements, so that was reason for the test dot more to right on the right-side leg of the resistor

That that D1 is something to drop the voltage, and measuring from R1:s right-side leg will find where the + feed is coming from

And measuring from the big caps lower leg will find where the - feed is coming from.

So those two top dots would probably go into the two leftmost BAT-V pins.
The next one down might go to the right-most, and the last to second from right.

Or not, or in different order etc.

[night girder]

Bandcamp forbids* music generated with AI.

blog.bandcamp.com/2026/01/13/keeping-bandcamp-human/ 🦾

[sharp matrix]

I wonder if thats purely just for show. How easy would it be to actually determined what music is AI generated, doesnt sound exactly straight forward

[night girder]

No. They taking a stance.,

They might backpaddle though or go under

[mental oriole]

Good

[soft bloom]

Multimeter beeped on me on this

[night girder]

That's why I hope people support bandcamp. I buy my music a lot through that website. A bit mor fair than apple, spotify etc imo.

[sharp matrix]

I guess, it just doesnt seem like a stance that can realistically be enforced

[night girder]

There is AI hunting down AI.

[sharp matrix]

aka why i think this is purely for show

[night girder]

Bandcamp exists for a while now.

[twin dew]

So that and B+ are low resistance to each other?
But not to any of the BAT-V pins?
Or other way?
Or both?

[sharp matrix]

probably already AI music on it already, because its going to be impossible to realistically detect

[night girder]

If every platform goes AI music.

They have a niche position that will appeal to die hard music fans that want music not produced by 🤖

[sharp matrix]

It's already over

[night girder]

Ok, so whats the other option?

The whole internet has AI garbage in it now, so we just allow the flood then?

[sharp matrix]

there realistically isnt one, even universities cant do anything realistically about students using AI either

[night girder]

You know what happens if there is no more human creative input right? AI will generate music based on AI generated songs that are based on AI generated songs.
You get some sort of AI circle jerking.

Creativity and cultures will stagnate.

[sharp matrix]

sure they try to take a stance and do something but from what i can tell 99% of the time its just for show

[night girder]

Oh, I get your sentiment. Most is PR, for sure.

[sharp matrix]

and yes universities arent happy about it, but little can be done

[night girder]

Want to hear a crazy story from my country friend?

[sharp matrix]

not that they dont try, its just a losing battle

[night girder]

We have a politician, who got promoted to become head of one of the most presitge universities of my country.

So, out she goes to give a speech.

And in that speech are multiple quotes. From all sort of famous people.
And the quotes were all fucking made up. Turns out... she used fucking AI to write her speech!

[sharp matrix]

well only because what we have isnt true AI, it just atm only mimics at best

[night girder]

How can you tell students, NOT to use AI ... and then use AI to write your own speech. For a university. And it contains all sort of fake quotes.

[mental oriole]

Hope she got fired.

[night girder]

Nah, ofc not.

She is a politician.

www.vrt.be/vrtnws/en/2026/01/08/ghent-university-rector-petra-de-sutter-uses-fabricated-quotes-i/

[soft bloom]

I checked manual to multimeter nad didn't find any specific mention if beeping during resistance measurement... The 'electricity' icon that appeared also scared me tbh 🙂
So i am not sure what exactly it should have meant.
I was testing that red dot that is slightly aside from the chip with b+

[night girder]

Dogma is the enemy of progress - Albert Einstein

en.vijesti.me/news-b/science/790935/artificial-intelligence-hallucination-einstein-never-said-dogma-is-the-enemy-of-progress

[twin dew]

Then measure in voltage mode first to check.

[sharp matrix]

well

[jagged snow]

Almost always, a beep in resistance mode means closed circuit/zero ohm

[twin dew]

Usually that beeping in resistance mode is the continuity testing mode sound for very low resistance.

[night girder]

That sums up our society in 2026 🤣

Or maybe just the whole human history.

[twin dew]

Depends on multimeter if it is always enabled or is subset etc.

[night girder]

Humans telling other humans what not to do. While they do it themselves.

But I shut up now for the electricity talks.

[soft bloom]

Directly touching probes does not do that

[jagged snow]

Interesting, unsure what it would be on your specific unit then

[soft bloom]

That's resistance check though.
On continuity beeping would mean low res

[sharp matrix]

just wait until AGI

its super over at that point

[soft bloom]

Also good way to get scared during this is to smell baking without realizing someone is doing it...

[sharp matrix]

im going to guess AGI is when society itself collapses right

because when machines can do all of the same thinking as a human, arent we just very disposable, replace humanity with computers, we arent needed anymore

[night girder]

We aren't replaceable immediately. They need us for repairs.

Unless some idiot makes them 100% self produceable and repairable.

Servers need fixing. Parts need to be replaced.

[sharp matrix]

even if we dont, AGI will do it itself, so it can replace us sooner....

thats the truly terrifying thing about AGI, it will basically just do its own thing at that point

[soft bloom]

I probed between B+ and 2 of the bat-v right contacts
Same reaction
It beeps, displays lightning, quickly tosses some numbers, and goes to zero
I didn't dare to keep contact for longer

[twin dew]

So use voltage mode instead.
It sees external voltage difference and protects itself.

But remember that when there isn't voltage difference, that will show nothing, as in when there is actual connection it shows nothing.

Voltage mode use to map circuits is much harder.

[soft bloom]

For all I could find the lightning is high voltage indicator

[twin dew]

Because that resistance mode is expecting max few volts or less from the devices own battery being fed as test voltage.
Not the around 20V from the battery pack itself.

[night girder]

I don't know. If you have a AGI which is limited to a room or anything else, with 0 connections to outside.
What can it do? It's only if we are stupid enough to give it access to things that it shouldn't have.

We can still go inside the room, ask it a question. Get an answer. But never from the outside. And vice versa. Just lock it in. And provide 100 kill switches for it.

[soft bloom]

Rightmost bat-v to b+ is 5.5v

[twin dew]

Those voltages mean almost nothing.
Only full battery voltage and (almost) no voltage mean anything at this stage.

[soft bloom]

Wait this makes no sense to me...
The 3rd bat-v pin is also 5.6v to B+

[twin dew]

Because that just means there is connection over some components.
The voltages themselves will make no sense without schematic at this state.

[soft bloom]

Congratulate me it sparked

[soft bloom]

So, i tried that against bat-v pins
And I think it slipped onto 2nd pin or just right away made spark around bat-v pins

[twin dew]

Yeah, you shorted two different pins together with one probe.

Basically you would want to desolder this fuse:

But easier said than done with those fills.

And even that wouldn't help if the B+ side is the one connected to the rest, and not that BAT+ side.

[soft bloom]

Well, it should have been ground to one of data pins, aka not big deal?.. but it wasn't too big of a spark though

[twin dew]

And with the way those fills are, very likely that would not do anything to help for on-PCB stuff.

But I never asked you to measure between B+/BAT+ and the BAT-V connector.
But to measure from that single point near the chip, to those different places.

One probe stays on the resistor leg, other tests various places.

[soft bloom]

Was doing that.

[visual tree]

Microsoft: "Please add your phone number"

Also Microsoft:"Your country code is not supported"

[soft bloom]

I guess if I hear beep during resistance test that should be treated as continuity confirmation?

[twin dew]

No.
Test first in DC Voltage mode.
If the result is 0.3V or less, then test in resistance mode for the same pair.
If the voltage is more, skip resistance check.

[soft bloom]

Is that confirmation that there is some power going from BMS to the main pcb over bat-v?

[twin dew]

No.

[soft bloom]

To recap a bit
We now one of cell pairs is quite below in voltage (b3 to b4 below 3v)
But, regardless, individual charging should be avoided and first need to understand if bms in a state of double shutdown,
By poking around and figuring out which pin on controller are connected to bat-v and b- and b+
Correct?
Did i miss something?

[twin dew]

Basically trying to find where the power to the BMS is coming from.
And trying to rule out things by for example finding which pins on the BAT-V connector are for data.

[twin dew]

Probably the two topmost dots in this:

[soft bloom]

That might help with jumpstarting by supplying power to controller instead of whole battery pack?

[twin dew]

Yes

And trying to differentiate if you have broken BMS, or just unpowered one.

[soft bloom]

I missed : why remove fuse?

[twin dew]

To cut that B+ from anything else, to unpower the PCB so that voltage mode wouldn't be needed as safety check first.
But it seems that that would only cut the power from the output cable, not from the PCB.

And also, wouldn't help that much with all the other taps too in there.

[soft bloom]

My mind is exploding trying to figure out how to notate measurements in a readable way...
Even here in chat sometimes it's just "between that and this I measure 0v"

[twin dew]

So don't try.

Meant don't try removing that fuse

[twin dew]

And I did say to do that voltage measurement first as safety check.
And then measure in resistance mode for the ones where the DC voltage was 0.3V or below (basically no voltage).

[soft bloom]

Yes.
Method of how to measure is now understood.
But noting down results...

So right now I redid resistor that is likely to be between bat-v red wire and + on the chip
First voltage - nothing
Then reistance
2.4kOm

Was a bit hard to get that reading to be stable

[twin dew]

If it isn't almost immediately near 0, then it doesn't matter.

[soft bloom]

Well, by stable I mean "not 0 M Ohm" aka broken circuit

Haha, so the fun part:
So wtf does that result mean? Some resistance in between but otherwise connected?

[twin dew]

Like I said, not usable without schematic.
Just means, some kind of connection via some components, not direct one.

[soft bloom]

Between the brown bulky thing's bottom and supposed bat-v minus: -6.2V

Before I go near data ports agains...
Where did it draw power to create the spark? Unless there's some residual charge still inside... It must have been from batteries through bms? I feel like I asked this before...

[soft bloom]

Meant to say: Slipped from 3rd to 2nd making spark between them

[twin dew]

Yes, the battery + is connected normally, and the battery - is connected via the electronics in some form, via that trace that goes uunder the chip.

[soft bloom]

Ok ngl I probably did stupid thing
I was going to figure out how to hold it at 2nd pin to measure voltage against one of those 2 resistors near chip.
And before doing so I just used a black probe of turned of multimeter to quickly short 1st and second pins on bat-v
It sparked harder, leaving burning mark...
Good reminder to not touch anything on those electronics with your fingers
But, anyway, why the heck? I would never expect this from supposed data pins.

[twin dew]

Stop testing at this point.

[soft bloom]

So that should rule out possibilty of complete shutdown between batteries and bms?

I might be wreckles

[twin dew]

Seeing about the people who have revived them right now.
To check their work.
But based on the first bad one, you would need two 24V supplies, one kind of controllable.

[jagged snow]

Touching it with your fingers doesn't matter. Your fingers are effective insulators at those voltages. Just don't short stuff with an aluminum probe...

[soft bloom]

If I should reread chat just tell me, I probably am missing something

[soft bloom]

I might need to figure out better way to keep contact with these cause it really is pure luck I didn't create more sparks with how easily it slides and how small gap is compared to probe

[jagged snow]

It just takes practice, you can probe very small points with a standard multimeter

[soft bloom]

Well, I have one where I can change v and amp and it can go to 24v

Why second one though? To emulate main boards power input?

[twin dew]

Seems it should be enough to first charge the cells to normal voltage range from the low charge limit.
Then input power via those normal battery wires so the BMS turns on.
Seems this device doesn't do that from the main PCB end via mains input when BMS isn't talking for some reason or other.

Based on several different people getting them back working via different methods.
Another used small resistor and button to short that B- and BAT- together for a moment while the device was assembled, with mains disconnected, and then reconnected mains for it to start charging normally.

Another was that who just had the battery and main unit not connected together via that battery power cable, and just fed both separately 24V until charged, then reassembled.

So that BMS revival only needs power from the battery cable connector end.
But for some reason or other the device doesn't do that itself automatically.

Probably as protection for the cases when the communication isn't working, but the battery is connected, to not overcharge.

[soft bloom]

Bat- that is on same board? There a second one on main largest board that goes to front board

[twin dew]

Two completely different methods.
The one you read where you input 24V into the battery connector.
And second where he momentarely connected those two points on that BMS PCB together, bypassing all 8 fets.

So the battery powered the device up, set the FETs to conduct, and then he removed the jump.

And plugged in cable for it to finish charging normally.

For that first:
Have everything connected except that XT60 connector from the battery.
Power the main unit somehow, then input 20-24V into that battery connector.
Keep that applied until there is charge in the battery.
Disconnect the charger from that battery XT60, and plug it back into main unit, while it isn't powered.

Which was the one you had hit on.

[soft bloom]

Hit?

To clarify, when it became clear to stop poking around?

[twin dew]

When I started looking on internet about others with exact same device and problem.

And when you kept making dangerous mistakes.

[soft bloom]

Hehe, guess bms isn't too dead

[twin dew]

So as you already charged the cells to near 3.0V or so:
Assemble the device so that all the cables are connected.
Don't plug in mains power or like, but have it ready.
Connect these two together with piece of wire.
Turn the device on.
Remove the wire.
Connect mains power.

Or do that 20V or so input to battery connector one instead.

Once the BMS talks with the main device, disconnect mains, connect battery cable, turn back on.

[rustic panther]

Are we making unintended ieds here?

[soft bloom]

So, one way is to somehow get to bat - and + with controlled 24v
(That cable seems to be under white foam, despite yellow connector on the larger board)
And give power to the front via ac?
And wait untill it charges, presumably getting correctly battery capacity displayed?
Then disconnect all power, reconnect bat - and + and be good

[soft bloom]

Apparently, that one b3 to b4 pack lost a bit of charge again since it's 2.9v, but not tooo far off I guess?

[twin dew]

For that second one.
Have the device connected except that battery cable from battery to main PCB.
Connect 20-24V to battery via the connector.
Connect mains to unit, turn it on.
Once BMS talks with the main unit, disconnect mains.
Disconnect extra 24V.
Connect battery to main unit.
Start the unit
Connect mains.

[twin dew]

Just needs to be above BMS minimum allowed cutoff voltage.
And closeish enough for it to not care about imbalance.

If those aren't met, it will not keep the fets enabled and the battery will not feed.

[soft bloom]

So, jump can be removed as soon as display turns on?

[twin dew]

Yes.

[soft bloom]

If it won't work then will need to balance batteries more?

[twin dew]

Yes, if it turns on while jumped, and turns off when jump is removed.

[soft bloom]

That seems like an all around easier way then disconnecting bat-+ from mains

[twin dew]

Also, you can jump from that B- to the main PCB side, just make sure it is part of the same Battery - plane.

Don't need to remove the white stuff on this end, just need to be functionally same thing.

So that PCB battery connector reverse side pin etc. are completely valid instead.

But make very sure to not connect to the + even for subsecond.

So position the B- pad connecting cable second.

[soft bloom]

Am trying to figure out if that's separate thing from previous sentence, and if it is - then how

[twin dew]

Moment.

[soft bloom]

I should connect B- to Bat -
There's not +, and which one to touch second and why?

[pure karma]

how have you guys been saying the same stuff over and over for 12 hours now 💀

[twin dew]

Point was that if you are connecting to this connectors reverse side, to not have to remove the white gunk from around the BAT- wire on the BMS PCB, the + is very close.
So in that case position the cable on that end first, as that point the - doesn't have anything on it, so no dangerous short.
And then second attach the other end to BMS PCB B- for the start.

But you can just remove some of the green solderresist instead and connect those:

That you get to the copper under the green mask.

Connecting B- and B+ together would be very bad.
So working in a way that has lesser likelyhood of that happening because hand or wire slipped is good.

[soft bloom]

Phehehehehehehe
Not allowed to have fun our way, huh?

[soft bloom]

Wouldn't i have to break white foam anyway? If I understood correctly there's risk to short b- to bat+ instead of poking through white foam to bat-?

I do finally see separation of xt60 connector... But yeah it has it's own white foam connecting 2 sides 🙂

[twin dew]

Three options:
Scrape some of that green solder resist off from that PCB ground fill. Connect to that.
Remove some/all of the white stuff from that BAT-.
Connect to the main PCB backside power connector pin end (danger of touching neighboring + pin).

The solder bump for that main PCB side connector for that battery cable, on the backside of the main PCB.

[pure karma]

No but i feel like you guys are going in circles mostly

[soft bloom]

I had impression that green plate that has bat-, b4 and b5 on it is one of cell groups for some reason
Probably that's why confused.
You sure it is bat-? There's probably some way to test that

[twin dew]

There are cuts in the copper under the white:

So that those B4 and B5 are separate from the large fill that has the BAT- on right side, and one set of fets on the left.

[soft bloom]

Btw, white foam is insulation? or just waranty?
Cause it is over some places, but not others.
Anyway opening Bluetti did break of some of it, as seen on that pick you just sent - those white patches on cells

[twin dew]

Insulation and mechanical support.

[twin dew]

That was there just as a glue.

On those cables it is mainly as vibration dampening and extra support.

If people handled the thing badly.

[soft bloom]

Yeah they should be separated from each other at the very least, evident by my own measurements. Just wasn't sure of that green plate is bat-
I might check first the xt60 cause there might be something I could safely put it in...
If not then likely will do the scratching until copper. It's not too thin I hope

[twin dew]

All that is BAT- fill under that solder resist.

Also the lower left of that B4, but I didn't bother marking.

But I need to go to bed now, and I would also recommend leaving the actual jumping to when you are fully awake, and have several hours after, where can be near that device when it charges normally.

[soft bloom]

Okay, thank you very much.
Due to some ~power destribution stuff will have to be awake anyway

[soft bloom]

So... Just trying to understand, for curiosity sake.
That b- to bat- will bypass BMS' shutdown and make main board think it's fine and give power back to BMS via bat-v?

[twin dew]

It will allow the BMS and main board to get power from the battery and boot up normally.
And the BMS will enable those eight fets to conduct normally from then on.
We are just bypassing those control FETs near the middle for a moment.

And as the BMS now has power, it can talk with the main PCB and controller etc.
So the bypass isn't needed except to allow for that initial boot.

Was needed for few seconds on the other video using this method.

[soft bloom]

I guess question was if bat- and + go strictly one way or switch depending on situation?
Don't let that keep you today, hehe. Just me wondering

[twin dew]

He soldered wires to B- and BAT- and put a momentary push-button switch between.

[soft bloom]

Btw you initially said about using button and resistor, but later it was just wire? Wonder if that's alright

[twin dew]

Wires, 33 ohm resistor because he was afraid of those caps on the main PCB, to limit inrush current, and button.

Just something he happened to have, value selected as smallest he had.

BAT- is always -.
BAT+ is always +.
But they can work either as main PCB feeding the battery, to charge it.
Or battery feeding the main PCB.

But that main pcb charges battery will not happen if the main pcb cannot talk with bms.
And you are in that gotcha right now.

So need to bypass that BMS lockout with fets not conducting with that bypass wire between B- and BAT- equivalents.
Start the unit on battery.
Remove the bypass.
Connect mains.
Allow the device to switch to charging the battery.

But that sleep, now ->

[soft bloom]

I now have doubts about what's "mains disconnected".
I was thinking i can have everything connected (including bat-v and bat cable) except for external power.
But now it seems more like I should disconnect bat cable, keep bat-v in, short bat- and b- for a memont and as display turns on - disconnect it, connect bat cable and then turn on external power.
Maybe it is better done at daytime.. but then it will be harder on the execution side

(ping off so hopefully it will hang there till morning)

[soft bloom]

Haha i used image to try and find the source but only found the one for method I don't want to use
Though I guess idea is anyway to give power to the bms via bat- (either proper external dc or shorting with b-) to unlock fets
Aka bat cable should not connect bms with main board for that moment

So, turns out you can just put probe inside xt60 connector on BMS side and and short with B-
No button no resistor

[soft bloom]

I guess some people just want to die a lil earlier...

[gentle osprey]

i need help fixing a problem with hdr

when ever i play a game its normal but as soon as i press windows to go to the desktop the desktop becomes dark and the game becomes darker than before what should i do

Tag me if you have a solution please

[umbral cliff]

turn off HDR @gentle osprey
Honestly I wish I had a solution, but really unless you're playing on an OLED or a proper HDR monitor you're not getting much benefit out of it anyways.
HDR on windows has been as you describe for a long time. If I really want HDR I just single monitor it and use a laptop for other stuff.

[languid gulch]

my old monitor acted like that, and it was because it had an HDR rating, but it was 250 nit peak brightness, so it was one of those "HDRn't" monitors

and yea, once i turned HDR off it was fine

[umbral cliff]

ya so many hdr "ready" monitors just... HDRn't. I love that term

[gentle osprey]

So should i just turn it off

[languid gulch]

it seems like it could have better color variability with it on, but the other weirdness made it not worth it

[languid gulch]

if it functions better, that way, yea

[gentle osprey]

I started playing rdr2 and it alot better with it on

But i well try to get used to it off

[umbral cliff]

what monitor do you have?, @gentle osprey but most likely ya. windows HDR is funky. Or just remind yourself not to alt-tab out of the game and just immerse yourself

[gentle osprey]

An 280hz acor monitor

[languid gulch]

check to see if it has its own HDR stuff built in that you can use, it might be causing conflicts with windows HDR BS

[umbral cliff]

You're definitely not the only one having issues with HDR.

[gentle osprey]

I have it turned on on the monitor and on in windows

[languid gulch]

try off in windows

[gentle osprey]

Ok

[languid gulch]

it might be doing the "3 way switch" thing, where turning it on in the monitor & on in windows is actually turning it off or messing with it

[gentle osprey]

It worked ty all

I just left the hdr on the monitor on and windows off

[umbral cliff]

nice!

Should be a lot more reliable that way

[languid gulch]

& yes, it is 100% stupid that they choose not to make it consistent

oh yea, doing both on my monitor fucks it up

both off: meh
monitor HDR: better
both on: loses color depth, things blurry

[glossy glacier]

Windows and HDR is more miss than hit

[twin dew]

Mains power was that external power.
Mains power means generic wall AC.

So basically meant: don't connect the external power to the unit, that is used for charging, before the unit is running on battery power, and bypass is disconnected.

[twin dew]

Ok, didn't think that but yeah, that would work too.

Basically you were in catch 22 situation.

For the unit to work, BMS (1) needs to communicate with the main PCB.
For that to work, BMS (1) must have power.
For BMS to have power, there must be 16-24V between BAT- (2) and BAT+ (3).

For BAT- (3) to have anything and not be floating, either:
Main PCB must feed power via cable (4).
FET-set (5) needs to be conducting.

But:
If BMS isn't working, main PCB will not feed power (4).
If BMS isn't working, FETs (5) will not conduct.

So Catch 22: If BMS isn't already powered, BMS will not be powered, and battery will not work, even if everything is ok.

So to get BMS working, connecting B- (6) to BAT- (2) momentarely, makes the BMS get power, and allows it to drive the FET array to conduct and do that connecting naturally.

.

On completely different note, CapCom fucked up DLC unlock check code causing massive performance issues if you didn't own most of the DLC, running it again and again during gameplay for some insane reason:
https://old.reddit.com/r/MonsterHunter/comments/1qcy3hn/mh_wilds_bad_performance_mystery_solved/

[glossy glacier]

On one hand I'm laughing my ass of at the stupidity, on the other I'm crying in a corner because of it

[languid gulch]

my mmo nearly cooked itself on the server side of things by accidentally setting the "check all the mail in the game's system" every second instead of every hour

[soft bloom]

So, what was done should have told bms that main board gives power, unlocking it, and after connecting bat cable to the main board...
Bms should see that there's no longer given to it, so it supplies some over bat cable?
What was observed though, is that display was working, flashlight was working, bluetooth was working BUT:
edc, ac would not turn on despite physical button presses on the front nor bluetooth commands;
Over bluetooth it was reporting same a006 error "battery pack communication "

In hindsight:
I might have forgot to reconnect bat-v (after it was removed to more easily eject bat cable from main board)
Should have tried plugging external ac, which i could not at the time
Need to recheck voltages.

Anyway, unlocking does work at least partially. Battery pack's 21.6v is enough for it to unlock.

[twin dew]

If it kept working and feeding after you disconnected the bypass, you probably forgot to reconnect that BAT-V

That bypass allowed the BMS to get power from the battery, and put the FETs into normal mode, which then means the bypass isn't needed anymore to go around the FETs.

[soft bloom]

I need arguments to epxlain why it's a bad idea to manually charge cell pairs while main board is powered up and underlight load

[jagged snow]

a. From what I've seen you don't know what you're doing well enough to manually charge a cell at all and
b. Doing it while the board is on means it's much easier for you to damage components on the board

I work with battery systems professionally in some capacities and I wouldn't manually charge a battery connected to active power delivery/logic circuitry

[twin dew]

Also that BMS might get mighty confused.

[jagged snow]

Indeed, it will also mess up your state of charge measurement enough as to be useless

[soft bloom]

hehe, I probably need more dumbed down explanation...

or maybe I should give up and just accept that somebody might damage batteries

[twin dew]

Two cooks fighting each other results in bad food.

And possibly the kitchen burning down.

[soft bloom]

🤷‍♂️

can't do shit against "oh, nothing bad will happen, everyone does that" arguments

anyway, BMS is no longer in double shutdown

[twin dew]

And do note that that BMS seems to have ability to do balancing when trickle-charging.
Basically discharging some of the cells slowly to bring the least charged ones up safely.

But I might be mistaken what that top left near B1 stuff does.

No, seems the FETs might be there just to lower power leakage when BMS is unpowered, to keep the cells from discharging.
Each inter-cell tap comes in over 100 ohm resistor, then into drain of a FET.
And both gate and source of that FET are tied together via different value resistors and then go to the IC.
So probably just voltage reading, not draining.

[soft bloom]

trickle?

[bronze jasper]

It's a slow charge that helps battery health in some cases

[wanton orchid]

keep in mind batteries are chemicals
its surface behavior depends on random thing and can be easily un heaven
basically : microscopically each part are fighting with neighbors part for electrical charge potential and molecular content
which means some parts are actually more charged than others, and they are exchanging electrical potentials as the whole cell states change

to better let it heaven out, you only apply very small charging, this powers parts that are still at lower charge state while maintaining already charged one
you can also do the reverse
it's known as repair charging, it tries to improve cell heaveness at chemical level

for ion cells it almost work like electrical plating and can even actually do an internal one resulting in catastrophic cell failure

[twin dew]

Also just charge type that is used to keep the cells topped up, to compensate for natural discharge.
What happens with more intelligent chargers when the cells are full, but charger is still connected.
It switches to trickle/upkeep mode.

[languid gulch]

ok why does AMD have a weird spammy command screen pop up every once in a while, but do nothing

lists itself as an autoupdater

[twin dew]

Probably this thing?

[languid gulch]

maybe

lemme see if updating it makes a difference

i completely forgot that this cpu has an iGPU

[visual tree]

https://www.reddit.com/r/nvidia/comments/1qe7ind/official_asus_public_statement_to_clarify_recent/

"We would like to clarify recent reports regarding the ASUS GeForce RTX™ 5070 Ti and RTX™ 5060 Ti 16 GB. Certain media may have received incomplete information from an ASUS PR representative regarding these products.
The GeForce RTX 5070 Ti and GeForce RTX 5060 Ti 16 GB have not been discontinued or designated as end-of-life (EOL). ASUS has no plans to stop selling these models.
Current fluctuations in supply for both products are primarily due to memory supply constraints, which have temporarily affected production output and restocking cycles. As a result, availability may appear limited in certain markets, but this should not be interpreted as a production halt or product retirement.
ASUS will continue to support the GeForce RTX 5070 Ti and RTX 5060 Ti 16 GB and is working closely with partners to stabilize supply as conditions improve."

Not that I was planning on buying anything from Nvidia but still....

[twin dew]

Yeah, Nvidia used to force card manufacturers to get their VRAM from Nvidia with the GPU chip.
And not allow them to source it themselves.
But shortly after the RAM shortages started, stopped including VRAM and forcing the card manufacturers to get it themselves.
With no previous purchase agreements or any kind of previous ties to the VRAM sellers...

Another giant FU from Nvidia to card partners.

So there is going to be giant amounts of pain on that front.

(Both versions of the policy, forcing card makers to get shafted by Nvidia on the VRAM, and then dropping that arrangement when VRAM got hard to get)

[twin dew]

And such VRAM chip agreements, even when you are able to get one, might have 6-12 month lead time.

[languid gulch]

what's extra dumb about all of this is chip manufacturers refusing to move on plans to build more fabs to make more chips, which is something they'll have to do in the near future anyway since demand for it will never go down

[twin dew]

Not trusting that AI isn't a bubble.

[languid gulch]

that bubble may pop, but memory chips will always have other expanding applications

[twin dew]

But the current issues are just from AI datacenter builds.
Everything else was already covered.

[languid gulch]

yup

but they should announce that they'll be breaking ground on new fabs now, for post-AI demands or whatever good publicity bullshit reason they wanna claim, so that people know it'll do nothing to alleviate current issues

[twin dew]

So starting to build more fabs, to cover the suddenly 40+% larger "market", that might not be there anymore once the fabs would be done in 2-3 years is not good business plan.
Memory has always been market with few year boom-bust cyclic model.

[languid gulch]

and that should be the clearest evidence that it is a bubble

which to me should result in every tech CEO's firing claiming that it isn't

[twin dew]

Same tech CEOs that in AI Hype Circlejerk seem to think that we are just on the cusp of AGI, if we just throw more hardware at the problem....

And "If we build (AI datacenters), they will come (customers)"

[visual tree]

CEO: "How much of our code is AI-written?"

Employee: "It's 2,3% but we tell the media it's 32%"

CEO: " Make that 52% in the next press release. The Investors will love that!"

Employee: "Btw, our automated AI-workflow deleted the inventory database yesterday"

[mental oriole]

[soft bloom]

hmmm
it still displays 100% despite not pulling any real power from AC (even when plugged into another unit that shows 16W drain; an clearly turning on some circuitry as indecated by hidden led near big coil)
and discharging with 2w flashlight does not seem to drain it at all either
I suspect more problems than just that double shutdown...

[wanton orchid]

bmc fried ?

[twin dew]

Might need to discharge significant amount before that charge monitoring gets back on track.

[soft bloom]

also it still reports a006

[twin dew]

Then it will not show charge.

So most likely the cause for the battery drain was that communication frying itself, which then made the unit never charge the battery anymore, which then drained over time to point that BMS shut down.

Both parts work, but the communication between them doesn't, so the battery doesn't get charged.

Or those zap shorts while trying to diagnose fried something in that part.

[soft bloom]

I just got random idea if supplying power over bat-+ cable would be a better way to charge the pack...
but that leaves problem of finding xt60 connector or creating other contact... without solving issue of communication which is crucial for unit

[soft bloom]

could that happen without any visual damage?

[twin dew]

Yes

[soft bloom]

😢

and I guess the only way to even partially check if that data commute is still working is by pocking around again... though would need to disconnect battery pack completely (both bat-v and bat)

[twin dew]

But more likely something was already dead, if the unit had worked for ages at the same load, and then discharged too much.

That that communication failure was the original fault.

Causing the over discharge from battery charging stopping.

[soft bloom]

to be exact, at some point I checked it had 64%, and about 30 minutes later it was unresponsive
which happened about week after similar episode, when it had some~ charge left but went silent, though somehow recovered after being plugged into AC for half a day or so

[twin dew]

And that 64% might have just been the charge when that communication failed?

And it then showed that until dead.

[soft bloom]

yeah. given 30 minutes delay still leaves about 60% of charge

aka if bms died, it should have died at 60% charge

or it shutdown cause it saw too much disbalance between cell pairs?

but that alone doesn't explain why it stopped commuting with main board

[twin dew]

Electronics can just die.

[soft bloom]

in theory, could connect to the pins of bat-v and check with multimeter if any activity happens there

[twin dew]

Not really, way too fast for the averaging digital multimeter does.

[soft bloom]

yeah... not with this one - it won't work

[twin dew]

Not going to read signals that might start at kilohertz rate and be anything from that to megahertz.

[languid gulch]

ok dumb question

what's the highest frequency "tech thingy" that's along the lines of ram/processing, rather than something like a scientific instrument or manufacturing implement

[night girder]

Consumer hardware with high frequency?

[languid gulch]

basically

[wanton orchid]

what kind of frequency are you talking ?

if it's wave your wifi antennas are pretty high

[languid gulch]

transistor

[wanton orchid]

it doesn't mean much
cpus use multiple of them

your network cards may still hold records though about that

[twin dew]

And electronics don't use transistors much anymore.
Outside of some audio amplifiers.

[wanton orchid]

what ?

[twin dew]

If you mean generic switching frequency.

[twin dew]

Various FET flavors all the way.

[languid gulch]

i'm including stuff like processor transistors

[twin dew]

Transistors are current driven, FETs are voltage driven.

[wanton orchid]

fet T still is transistor though

[twin dew]

Ok, I'm corrected.

[languid gulch]

stuff like intel/amd CPUs getting up near 7GHz these days, just kinda wondering if there's a product that looks at that frequency and giggles

[twin dew]

Those WiGig WLAN transceivers.

Using 60GHz band.

[wanton orchid]

I mean from an history standpoint I understand the confusion
but these are still specialized type of transistors
you can have V driven amplifiers

[languid gulch]

does the processor keep up with each Hertz on that?

[wanton orchid]

multiple transistors trigger each "cycle"

[languid gulch]

almost like staggered processing of some kind?

[wanton orchid]

yep

[twin dew]

Lot of the CPU isn't working based on that generic main frequency, but just gets to work when previous thing stops.
So the actual switching is done way above the frequency, just in waterfall type thing where the activity over single clock cycle moves along the fabric.

[wanton orchid]

that's how errors goes in mostly when overclocking
some cycle triggers when the last cascade didn't end

[languid gulch]

can that kind of processing be applied to something like CPU to "increase" its speed?

[twin dew]

That is what CPUs have done for ages.

[wanton orchid]

? we told you it is what the cpu does

[old bridge]

sloop a cpu lmao

[languid gulch]

oh, i didn't know if it was a transceiver-specific processor that did it

[wanton orchid]

some processor don't have "clock" at all

[wanton orchid]

I was talking about cpu with clock standard

and some asics aren't even processors

[twin dew]

And transceivers aren't CPUs but special signaling circuitry with part of it analog.

[wanton orchid]

they are electrical pipelines
not algorithm executing thingy

[twin dew]

But basically most of CPU internals are now in asynchronous areas, and the main clock is just used to synchronize transfers between them.

[soft bloom]

One thing I was surprised about
Some time ago before disassembly
I pluged in ac and it turned on and I tried firmware update
There are 2 separate versions
Arm and DSP
Attempted DSP - it turned off instantly
Was worried it would break since in my mind it was like bricking gpu without second firmware chip.
And now when bms is unlocked and can power main board, updating both didn't change anything sadly

[twin dew]

From very long time ago, when everything took multiple clock cycles to do because of that full synchronization.

ARM = main CPU.
DSP = digital signal processor, specialist accessory CPU for waveform generation etc.

[wanton orchid]

the fastest switchs iirc are currently about 1.5Thz

[soft bloom]

One of the interesting questions is what would happen if we loose ability to design and make x86nand have to start over whole process
Even if arm and risk is in the same state it is atm
I mean, that would be huge impact... And, correct me if not, most of the used tech is proprietary through and through

[wanton orchid]

you probably should have update the arm part first

[wanton orchid]

don't buy the IP hoarders story
some guys out there are able to perform
most of r&d is technically out sourced
it just goes hidden by mountains of NDAs that will break under institution breakdown

nor Lisa nor Huang know how to build all of this
little guys working together do

btw that's when it starts to feel good for EU
lot of actually knowing people are still there

[soft bloom]

Mayhaps....
Well, i started feeling suspicious, and went to double check
Nope, it has old version of arm and dsp is reporting v0
It didn't update...
Well this could be a new issue then, though probably not the reason for a006 judging from what dsp stands for

[twin dew]

Completely possible that that goes through the DSP.

[wanton orchid]

imo new arm version is needed to understand and manage the new dsp version

arm part is the brain and dsp the arm basically

[soft bloom]

Well it's v2063.2 against v2063.3
Idk if that's major and minor or major and patch, but need to test

Don't like that system that you need account to do udpates

[soft bloom]

Would be a bit of cheating to just grab people who did it and ask to do that again/better

I mean, somebody who saw wheel probably can make another wheel.
Somebody who saw memory (set reset) switch and knows how to make transistors can likely make another one.
How much does team need to know to make a cpu for a fully functioning computer? On raspberry level if functionality

[night girder]

Oh wow, this conversation blew up.

[languid gulch]

yea didn't think it'd get this complicated 🤣

[night girder]

So ... who won? What hardware? 😄

The WLAN transceivers at 60GHz?

[languid gulch]

no idea 🤣

[wanton orchid]

you need a fet knowledge, a memory knowledge, a software architecture knowledge, a circuit design/architecture knowledge, lot of means for sourcing meterials and tools, a bigger team if you need to design them, and one or 2 decades

[languid gulch]

well that's easy

[wanton orchid]

and a backup of wikipedia and electronic and physics papers

[twin dew]

Also, something like 8-bit AVRs etc. would be much easier, and fine for most digital control stuff that is actually needed if all went to hell.

Stuff like modern ARM or x86 CPUs are luxuries.

[wanton orchid]

you'll then need another or 2 decades for that to be affordable to public

[twin dew]

Stuff used in classic Arduinos etc.

[soft bloom]

And a way to manufacture
Aka what veritasium made video, which is a big chunk of being able to achieve the goal

[wanton orchid]

yea without the bloat, arduino can run random people desktop task for administration and payement processing basically

[soft bloom]

One thing to know what will do X
Another thing is to make that stuff

[wanton orchid]

I said lot of means, i.e people space and materials, so you can set it up

[soft bloom]

Well, fair.
At least we wouldn't try to run JS on 16gig ram machines to ask chatgpt how much is 2+2

[wanton orchid]

^

that is a good summary of what pisss me off these days

[twin dew]

Also, even building modern tolerance manual milling machine or lathe from scratch would need lot of time and production machine generations to get back to same tolerance level.

[wanton orchid]

people are forced way too overingeenered stuff because software stack is trash

[twin dew]

Just on the purely mechanical side.

Sorry, not CNC, but...

[soft bloom]

How much of llm bloat is from every search engine trying to give ai summary?

[soft bloom]

Oh, and standard units aka meter, gram, second...

[wanton orchid]

I mean on good boostrap condition it can get on pretty fast, but certainly not for public production readiness, laboratory level
then doing the scaling thing later on

by pretty fast I mean a bit less than a decade

that is still a decade

you can skip a great part of it by repairing some old stuff that still barely works, but its tricky

[twin dew]

Also, with that 60GHz WLAN, and all other signaling stuff at high frequencies, there is usually purely analog modulator/demulator stage(s) where you convert the frequency up and then back down via constructive/deconstructive interference.
It is just the carrier wave that is 60GHz frequency, not the data signal.

Still sine wave signal generator at very high frequency in there to feed that process.

So I intentionally "cheated" with that.

[wanton orchid]

and also last stage / first stages buffers/deserializers has to be very reliable and extremely fast

[soft bloom]

So, after updating arm it displays 35%charge
But still a006 error too...
No idea how that's possible
But dsp wouldn't update - process hangs inside app

[wanton orchid]

maybe 35 is supposed to be default

at least it's effectively uncalibrated

[slow flint]

New server for the rack what do you think

[mental oriole]

[soft bloom]

I thought that too, but it did drop below as it was discharging

[soft bloom]

So the best summary of the progress I can give, without chronology, is:

At some point it allowed update DSP
Drawed power through dc
Got charged to above 23v, with individual pairs going slightly above 3.2v
It either was charging itself through bms, or manually evened out...

Showed a063 bms related error
But now displays 0%

Sometimes I wish there was smart lock with basic tests like "are you going to write down every step taken after open this box?"

What's the worst out of this is that even if it works in the end there's no way correct conclusions will be made

[wanton orchid]

did you fully power cycle after updates ?

[winged valley]

I mean this is what I use to test 14" MBP screens

The copper braid is a """"""button""""""

[soft bloom]

aside from doing the "hold 2 buttons to reset" and attempting to discharge and charge I am not sure what that would be given that it immediately reacts to any button press

[cursive fable]

Anyone have a good source they use to test and setup a software Equalizer? (testing out a new soundbar for a friend 😏)

[wanton orchid]

I meant disconnecting the cells temporarily

[languid gulch]

i feel like we need a livestream of this attempt at being Electroboom

[bronze jasper]

Power buttons usually don't have much voltage. My PC cases power button is busted rn so I've been turning on my PC by using a paperclip and bridging the power button pins on the motherboard

It's just a signal circuit

[visual tree]

Someone called?

[winged valley]

Whos being electroboom?

Someone is messing with batteries I think? Convo is hard to follow

[soft bloom]

Dude every time I see him touch board my eye twitches

[midnight osprey]

hey guys, which is the best Anti Aliasing method pls?

TSR in my case was set by default

using a 6950xt

my monitor is 100hz ... I always seam to need to turn on Vertical Sync ... is it related to AA?

[jagged snow]

No, AA is just smoothing edges

MSAA will give you the best quality at the biggest performance hit

FXAA is "traditional" aa and tends to look pretty good but worse than TAA or MSAA

TSR is temporal super resolution, it's the platform-agnostic upscaling technology built into UE5

It tends to have low quality and bad motion artifacts in my experience

[midnight osprey]

oooooh

will switch to MSAA then

[jagged snow]

It works by rendering at your native resolution, upscaling, and then downscaling again

[midnight osprey]

for people with lesser GPU's?

[jagged snow]

Not really, it's an alternative to fsr/dlss/xess

[torpid saffron]

So I have a question: my laptop is obviously pre built with Nvidia and Intel stuffing and is several years old but is running win 11. It has the Nvidia app, but refuses to update drivers from the app. The Microsoft driver update tool doesn't update the gpx driver, so I have to manually do it straight from nvidias website, which does work. Can anyone tell me why?

[jagged snow]

Windows update uses its own driver registry, it's often very out of date. What error do you get when trying to install with the nividia app? Have you tried a full ddu and reinstall to see if it fixed the issue?

[glossy glacier]

might just be an End-of-Support GPU?

[bronze jasper]

if the laptop is running windows 11, it's got tpm 2.0 and can't be that old could it?

[soft bloom]

I wonder what are use cases for ast libs in languages
Aside from static analysis for reformatting and type annotations ofcourse

[glossy glacier]

~Intel 8th Gen/AMD Zen+ and Nvidia 9 - 20 series
could also be running on not officialy supported HW

[winged valley]

Believe it or not, not always

[bronze jasper]

oldest gpu's that are supported is 10** series anyways

[glossy glacier]

no, 10 series stopped receiving support back in Oct 2025

[bronze jasper]

they get some updates for another 2 years still. not game ready though

[pure karma]

theres a thing called a bypass

[bronze jasper]

yeah but i thought people don't want to switch to windows 11

[pure karma]

🤷‍♂️

just people talking out there ass and not actually willing to make the change

linux is still at only 4% after all

[bronze jasper]

yeah i know i'm one of em lol. i been using win 11 since i got my 4080

i was a linux user before that because amd drivers (on windows) for the vega 64 succccked

[pure karma]

i like linux but the app suport is just not there

until its 99%+ of apps work it will always stay worst for me

[bronze jasper]

i learned all the proton tricks that are put together. nothing i wanted to do i couldn't

[pure karma]

rn at most you can game and anything else you need to just install windows through linux so at that poitn why bother

[bronze jasper]

wine is very robust. it can run a lot. and where it fails, thats where proton steps in with it's pals, dxvk and vkd3d

if you do it right, the only time you see those compatibility layers is at install. then it feels more or less like a real app in that environment

keep in mind theres very level over head. it's x86 compiled code running on x86 compiled hardware. it's just the linux kernel here is wearing a hat that says "i'm totally windows rn"

then code is stupid and it falls for it

[slow flint]

My friends keep saying "I don't want to switch to windows 11, I'll try Linux when SteamOS comes out"

I tell them to install Bazzite because its the same thing basically and more up to date

But its not SteamOS so they won't bother

[bronze jasper]

i switched to linux when the steam deck was announced, because i had used it before and wanted to get my chops back up. thats when i figured out that my vega 64 performed way better on linux with their open source graphics stack

[glossy glacier]

SteamOS is already available though?
In a "here is the iso, have fun, don't expect support" way but it is available

[slow flint]

They're not super tech Savy so they're more referring to a more general release build made to work on all systems
That's why I just recommend Bazzite as its exactly that

[bronze jasper]

most of what makes steam os, steam os, is that steam big picture is running on arch. It's not much more than that. You install steam onto any linux distro and you're more or less good to go.

actual steam os is immuteable and would probably annoy most full PC users

[cyan crescent]

Can a 7900xtx use a ful 4.0x4 slot? Not for gaming

[glossy glacier]

Electrical x4? Sure, you're just bandwidth limited then
Physical x4? Only if open ended

[cyan crescent]

Its only electrical 4

[soft bloom]

windows wtf?!?!?!?
I didn't even visit it

[sharp matrix]

so windows is activating advertising for these social media sites now, thats cool

also how do I opt out of this bullshit

[mental oriole]

Uninstall windows

[bronze jasper]

lol you're complaining about windows but you've never paid for it. dude you're clowning yourself. just use linux.

I dont know how that edge popup happened to you, but that's not an ad. That's edge does for popular websites when you visit them. i'd consider checking your PC for a RAT.

"i want to use windows so bad i'll pirate it. but then also i hate it and will complain about it every chance i get"

[sharp matrix]

funny microsoft is shooting themselves in the foot so to speak with windows 11, because of all of the telemetry, more countries/companies are exiting usage of windows all together because of all of the data collection it has

[bronze jasper]

https://store.steampowered.com/hwsurvey/Steam-Hardware-Software-Survey-Welcome-to-Steam the hardware survey does not indicate people are leaving windows in droves. windows 11 has majority stake and growing, among steam users

[sharp matrix]

cites a biased data set as proof of such, good job👍

meanwhile medium.com/@averageguymedianow/the-great-operating-system-revolt-how-china-and-europe-are-challenging-windows-dominance-ff15dce150b4

china and europe, to name a few are already in the process of exiting usage of windows, because of all of the data collection

[bronze jasper]

how is it biased? Valve has their own linux distro. Valve is the largest Linux gaming hardware manufacturer

[sharp matrix]

but i get it, not alot of people are following this news, but corporations and countries, are basically having enough of windows and all of its data collection, windows 11 being the main drive, because the amount of data collection it does actually violates privacy laws, when you look into it anyways

[visual tree]

*Me who still uses 4 Windows 10/11 product keys I got while studying in college

[sharp matrix]

and worse of all because of how much windows 11 sucks as far as disabling any of it, no choice exists inside of windows besides going to linux, which is currently what is happening

[visual tree]

You could say I got a lifetime windows by enroling in college lol

[bronze jasper]

medium isn't news. It's opinions written by anyone that get paid for page views

[sharp matrix]

but not alot of windows users care about things like privacy, so i guess we have even more telemetry to look forward to in the future i guess

[bronze jasper]

That article even talks about mobile OS's like windows is even a thing in that category

[glossy glacier]

Almost every last Windows Update had at least one mayor issue
Many new windows feature come with mostly no way to configure them as sysadmin

[sharp matrix]

neither is citing a biased source like steam reviews to, if im to be honest

[bronze jasper]

The only way to prevent telemetry is to disconnect. Consider if you will a world where linux is the dominant OS, and windows is the 3%er. Do you think companies will stop making programs that collect telemetry?

[sharp matrix]

that is known to be heavily biased towards windows after all

[bronze jasper]

Steam reviews? I linked the STEAM HARDWARE SURVEY. Are you familiar with that ?

[sharp matrix]

sorry thats what i meant

[bronze jasper]

every month Steam does a census on their users. With permission each time data is collected. It's a random selection of users each month and we get a qualified look at the hardware / software ecosystem of PC gamers. The vast majority of who are using Steam

It's the best look we got at pc gaming trends. It's real objective data

If there is a bias in the data, that's because there is a bias in the market.

[sharp matrix]

real objective data, sure keep believing that

[bronze jasper]

Do you have any proof that Valve is manipulating the hardware survey to show windows 11 as the dominant OS?

[sharp matrix]

dont get me wrong i use steam, but objective, noway

[bronze jasper]

Yes. It is objective data. Nothing indicates it isn't

[glossy glacier]

The survey is objective
The Users are not

[sharp matrix]

only within at best the steam market, but it says nothing more about the larger pc market as a whole

[bronze jasper]

I get that you're offended by the survey's results but you have to bite this bullet. Among steam users, Windows 11 is growing in adoption

[sharp matrix]

as i said, its biased, your getting a very limited perspective only looking at that

but you want to think your getting a full picture of the pc space just by looking at that, go right ahead

[bronze jasper]

PC gamers and the larger pc market as a whole are two very different things. However, the vast majority of gaming PCs will have Steam. Any PC with a dedicated GPU surely will.

[bronze jasper]

I've recognized it's tied to steam since the start. What you're doing here is building a straw man

[sharp matrix]

citing the biased perspective here, is far from straw man

[bronze jasper]

you're too much dude. I cited the STEAM HARDWARE SURVEY and you're calling it my limited perspective.

Some people just can't admit they're wrong. Trust me when i tell you, that this year is also not the year of the linux desktop

🫳 🎤

[sharp matrix]

wrong about what, when i said what i said i wasnt strictly talking about the gaming pc market, which seems to be your core argument, when i was talking about the pc market as a whole, which you yourself even said are two different things

and then you say im building some kind of straw man argument, when i was never strictly discussing the same thing your only focused on here

sure i can understand if i was only talking about pc gaming, but i never was to begin with

hence why i said, your biased perspective doesnt tell much....

meanwhile techhq.com/news/open-source-china-linux-kylin-kernel-desktop-de-microsoft/ as i said, countries are already moving away from windows 11, because of how bad the data collection has become

but i get your going to ignore the news, because it doesnt align with whatever this narrative your trying to paint here

[bronze jasper]

kylin is just ubuntu repurposed for government offices in china. It's not like china is adopting linux across the board.

outside of desktop pc use, laptops and desktop pcs' running a desktop environment i mean, window isn't the dominant OS at all. Linux rules servers, android and apple rule mobile.

kylin is not going to replace the majority of windows users in china

gaming pcs' are the best window into the desktop market. And steam hardware survey is the objective truth on that matter. It's funny to me that you're calling it bias because it doesn't show the results you're claiming

[bronze jasper]

china has 1.5 billion people. rolling out kylin into government buildings will replace 30-50m pcs. Windows is still going to maintain it's 80-85% dominance and linux will probably take a 5% market share

Ubuntu has been around for decades and hasn't disrupted that market yet. Kylin isn't changing anything with it other than branding and what apt servers it defaults too. I dont see people adopting it on their own.

[pure karma]

i havent payed for any installs i still have active but i atleast had the decency to activate them lol

[bronze jasper]

i dont care about microsoft getting the money or not. they have tons. it's just, why bother with windows if you have to pirate it and still hate it?

<@&387163995947270144> spam bot here

[pure karma]

OOB support for stuff

not everyone and definitely not myself wants to have to go through loopholes to have certain software work

[bronze jasper]

i love effective hacks

[pure karma]

to me linux is just a cool change of look for a pc only doing basic browsing

[bronze jasper]

if someone tells me i can't get something going in wine or proton, i'm gonna try anyways

[bronze jasper]

i do love how the windows will wibble wobble when you move them

[dire igloo]

i have to use windows because linux doesn't support the software I need for my voluntary work
so I have all the right in the world to complain about it

[dire igloo]

do you know how many Steam users have chosen their OS themselves?
from what I remember, a large portion of Steam HW survey entries are from prebuild owners

it's not biased by who is collecting the data, but what data it's being collected from

[pure karma]

problem with the steam HW survey is that it has tons of acient hardware and prebuilts yea

so its not really a good representation of whats happening NOW and neither is it a representation of OS choice

[bronze jasper]

you can put linux on a prebuilt still. The point i'm making is that 2026 is not going to be the year of desktop linux, again. The reasons behind this are many and you're touching on one right now.

[bronze jasper]

this is a weird claim. They don't sample every single steam user. Just the ones that log on while they're conducting the survey, and the person has to click "yes i'll participate"

I see no reason why Valve would bias the numbers to show windows 11 is dominant OS. That makes no sense

[pure karma]

yea ofc but just because someones decade old pc is on the survey now the numbers dont represent now but instead esentially an average of the last 10 years by example

[bronze jasper]

the decade old PC is going to bias the numbers towards windows 11?

[pure karma]

obviously if a pc that dosent support win 11 is gonna drop the percentage

so it shouldent be included in a choice survey

[dire igloo]

you can also build a PC yourself, yet prebuilds exist. the point is: the majority of gamers are not into tech, they're into gaming.
they buy a prebuild which comes with an activated OS installed (most commonly Windows) and they don't change it cuz they're not interested in making changes, they're interested in playing games

[bronze jasper]

you're kind of supporting my position that 2026 will again not be the year of the linux desktop

[dire igloo]

it's called dataset bias or selection bias.
Steam hardware survey treats all current steam users equally, so it's a rather poor tool to be using for market trends

[bronze jasper]

it's only sampling people who log on during the survey. there is no selection bias. it's entirely random selection from those who log on in a given week

[bronze jasper]

it's entire purpose is to show market trends. thats the only reason valve hosts it.

[dire igloo]

my cousin's PC which he bought used in 2019 and never touched again will have the same weight as my setup which was last changed in November

[bronze jasper]

why would they weight pc's? that would be biasing the data

[dire igloo]

I'm supporting neither position, because both of you are making good points and shit ones

I agree that more people who pick their OS themselves will be choosing Linux now

[bronze jasper]

lol desktop linux zealots are really trying to tear down one of the best market analysis tools we've had in the PC space

it really kills the narrative that everybody is leaving windows 11

[dire igloo]

however, I also think that Windows will continue to dominate the consumer market because of convenience, network effect and incompatibility

[dire igloo]

it is one of the best PC market analysis tools we have, I agree.
but you treat it as ultimate infallible wisdom which it absolutely is not

[bronze jasper]

that's why android and ios took over mobile markets. one day windows may lose the desktop market but that's not this year

[dire igloo]

the selection bias is in "which people use Steam" and not "which people does Steam pick for its survey"

the survey done by the speaker treats everyone in the room the same

[pure karma]

which is bias towards the majority... which is just about everywhere

[dire igloo]

you can use it to discover market trends, but it takes a lot more work than you saying "windows 11 has majority stake and growing"
you're implying causality without proper analysis

[dire igloo]

if you want to know what the majority thinks, then that's the way to go.
but you have to make sure that the data you're sampling from is actually representative - which is where that bias comes from

[bronze jasper]

except in steam's case, it's a room full of every gaming pc on the market. I would think that gaming pc's that don't have steam on them are an outlier and not common at all. That's just the nature of the market right now.

[dire igloo]

"Steam HW survey shows the Win11 share growing" → "people on Win10 migrate to Win11" or "Linux users quit gaming" or "new gamers are less likely to be choosing Linux for their starting OS"

meanwhile the claim of "people are leaving Windows" is an implicit statement about people who make conscious choices about their OS

both of your positions can be correct, you both just suck at discussing for the sake of understanding. you're discussing for the sake of being right

[dire igloo]

1. not every PC is a gaming PC. in fact, the initial claim was about countries and companies - which is the exact opposite of gaming
2. you are still neglecting the fact that "people who make conscious choices about their OS" only make up a very small subgroup of "people who use Steam"

[bronze jasper]

lol you're still trying to tear down the accuracy of the steam hardware survey. Keep going. it's fun.

I assure you , after all is said and done, 2026 will not be the year of desktop linux again.

your 2nd point actually lends towards my point. That's exactly what i'm trying to say. Most are not switching to linux from windows 10. They're just going to keep using old unsupported win 10 or win 11

[dire igloo]

let me say it again: I agree that the steam hardware survey accurately represents the data it intends to represent, being the current PC setups of active Steam users.
However, I don't think that that is the data which is relevant to this discussion

[bronze jasper]

You kind of came into this conversation late so maybe thats why you have no idea what my point is

[dire igloo]

.

[bronze jasper]

Yup. You agree with me, which is weird, because you're still arguing. Discord moment i guess

[dire igloo]

I've read the entire convo and you're basically just stuck in "steam hw survey biased" vs "no, it's infallible objective truth" and you're both wrong

[bronze jasper]

ok now this is a gross misrepresentation. i'll ask you to drop it now. argue your little heart out at the strawman you've built. I gotta go

[dire igloo]

just because I agree with some of your conclusion doesn't mean I 100% agree with everything you brought forth

[jagged snow]

People watch a youtube video on logical fallacies and think accusing everyone of using them is a free win in any argument 💀

You're just arguing in bad faith here and it's exhausting.

[dire igloo]

this is where the discussion about steam hw survey started.
Slechtvalk hadn't contributed anything meaningful before that, so this is what I'll consider the start.
and everything from there on was you defending the legitimacy of the steam hw survey while others were criticizing it

[dire igloo]

fallacy fallacy: invalidating the opposing position solely based on the fact it involved a logical fallacy

[bronze jasper]

yeah i said "among steam users" if you notice. Stop pinging me.

[dire igloo]

you're using gamer statistics do disprove an argument made about government administration and enterprise IT
and I can assure you from first hand experience, they absolutely are looking to move away from Microsoft products

[bronze jasper]

Yes i've said linux dominates server already. Stop pinging me

You came to this conversation late. Find a different mark.

[jagged snow]

Linux desktop is also gaining ground in many environments.

[dire igloo]

you introduced an argument to disprove a point nobody made.
I'd call that a textbook strawman.

However, I don't care about that, because I'd rather have you acknowledge the shortcomings of steam's hw survey than get stuck in an endless loop of accusing each other of bad faith arguments.

so please, try to actually argue my points. steam's hw survey is an accurate enough representation of the systems used by gamers.
however, the data is by its very nature not accurate enough to make any claims about consumer choices

[bronze jasper]

tldr. stop pinging me

[dire igloo]

ah yes, the two types of computer systems: servers and gaming PCs
conveniently forgetting about workstations, laptops, office desktops, supercomputers, mainframes and probably a bunch of others

[bronze jasper]

oh i never said any of that. stop pinging me

[dire igloo]

if you think that I'm off the mark, provide the appropriate context.
nothing you said after introducing steam's hw survey gave me any reason to believe that I'm missing context

[bronze jasper]

i owe you nothing. stop pinging me. i guess you don't actually read what i'm saying.

[dire igloo]

an argument made about government administration and enterprise IT

i've said linux dominates server already

implication: the only type of government or enterprice computer system that isn't covered by steam hw survey are servers.
and since you yourself just admitted that that's bogus (which I am 100% certain you knew before replying to me about servers), I can't help but come to the conclusion that you're arguing in bad faith.

I trust you to be tech-savvy enough to know that there's more than gaming PCs and servers, so you mentioning only servers in response to me is either a massive blunder or intentional diversion.

[dire igloo]

I've read everything you said from the point you told Slechtvalk to stop complaining about Windows

also, I've stopped pinging you several messages ago.
so stop making up bullshit just so you can be mad at me

[wanton orchid]

thank you chat for not getting at fireworker with the "they said they are arguing in bad faith with proof but that's ad hominem !" argument fallacy

[jagged snow]

Thanks twitter
No embeds with tracker protection on, eh?

[visual tree]

Why are they still selling this crap when you can buy a switch for few dollars more 🤦‍♂️

Some example on amazon costs 7 dollars. I wouldn't pay even 2 dollars for this

[dire igloo]

I don't understand what you mean

[twin dew]

Well, that is some kind of PoE powered 4-port switch.

Unless that output port on the main device has 24+ pins.

[visual tree]

The guy mentioned his speed dropped when he connects multiple cables so it's probably just a splitter

[twin dew]

Cannot be without 24+ pins.
More likely that it has extra HW to bypass the switch with just one connector in use.

[visual tree]

Either way, shitty equipment

A solid gigabit switch is like only 15 euros (the one with 4 or 5 ports)

*Unmanaged one, that is

[twin dew]

And that is working as 1Gbit/s when all in use, and 10Gbit/s when just one is in use?

4-port 10Gbit/s switch isn't cheap.

[visual tree]

Total is 1 gbit for all ports but that is enough for my case

[twin dew]

The person in pic complains it drops from 10Gbit/s when multiple cables are connected.