Tech Lab forum - General Discussions!

[jade scaffold]

like if you are gaming and playing GTA 5 or somthing dedicate most of the power to the GPU side and GTA is only gonna hit 2 cores

which is a pretty good way to have a lower power part compete with higher power parts

the only issue you have then is if something turbo demanding like cyber punk pops up hits all the core and all the GPU at once and it has to evenly split power load

[mossy gull]

Depends on how powerful the CPU part is.

[mossy gull]

My laptop can powershift if needed, but I rarely, very very rarely do I see the powershift happening

[cerulean helm]

https://www.youtube.com/watch?v=sSyjZFtZjbY

[mossy gull]

https://youtu.be/2_Y3E631ro8

[cinder lagoon]

I think this is really cool - granted the backplate could be made of metal instead of acrylic and therefore wayy thinner but the concept is great, would install myself

[balmy flicker]

that set up is asking for leaks there are some quick conects in the auto world that work like that and they often leak granted I am sure the car is running higher pressure and temps but at the same time the car can handle said leak

[mossy gull]

You actually do not want it to be from metals, if you do you can spread the heat on the back of the board.

[mossy gull]

https://youtu.be/1cfKc8U4iwg

[cinder lagoon]

oh totally didnt consider that but theres gotta be something more durable than acrylic and non-conductive

[tender plank]

I don't get what they're trying to do with this video.

Like, is it just supposed to be "lol, look how expensive Nvidia cards are!!!"?

[mossy gull]

Just a dumb video I got recommended

[jade scaffold]

https://www.youtube.com/watch?v=HGaYIZR9wDE&ab_channel=CNET

Sony released new ear buds

and honestly i really hate how they look compared to the 4th gen versions

[surreal moss]

Sony airPods

[mossy gull]

I prefer the more rounder looks

[jade scaffold]

i dont mind the shape of them

its the glossy plastic

somehow glossy plastic manages to look cheaper than recycled matt plastic

[pearl moon]

This is why i'm glad my grandparents have a massive box of cables at their house, it means I only keep the more obscure cables.

[mossy gull]

I got some Lexip Ceramic feet for my mouse and damn they're smoof

[mossy gull]

https://youtu.be/uryvVFB2tOk

When AMD destroys the competition

[cinder lagoon]

I hat the product name with every fiber of my body tho xD

[tender plank]

I have a 3' X 2' X 6" drawer filled about 2/3ds full with various cables I've collected over the years.

[jade scaffold]

Finally 3D V-Cache on mobile

Still hoping to see it on an apu

[jade scaffold]

Just found out this is gonna be exclusive to 1 laptop

Sad

[sonic meadow]

wow, how rude 😦

[mossy gull]

For now, the 6800M was also launched on 1 laptop, but now it's on a few, Notable of the Asus ROG G15 Advantage Edition and Corsair Voyager

[jade scaffold]

not this one

according to AMD and Asus the ROG Strix Scar 17 will be the only laptop for the forseeable future with it in

heres hoping for some lower end chips to get 3D V-Cache for mobile

[mossy gull]

Maybe 1st gen limited run

[cerulean helm]

Someone made an app for a sensor panel using HWinfo on the Microsoft Store. This looks really cool.
https://apps.microsoft.com/store/detail/infopanel/XPFP7C8H5446ZD

[cinder lagoon]

Huh pretty cool 7-Segment display style panel

[cerulean helm]

And its constantly being tweaked and updated. I just ordered a freestanding 7 inch display to use as a sensor panel. Will report back once I get it up and running.

[cinder lagoon]

It would be great if you could transmit the Hardewaremon data via the network or something so that you could use an rPi so Windows does not have a second monitor

[cinder lagoon]

But that is probably not on the list because it would require linux compat which is basically a rewrite

[balmy flicker]

what might cause heavily modded mine craft to hang on a rig with 16GB @2133MHz ram and an I5 12600k

the game files are stored on an HDD but that should only affect load times

[mossy gull]

It does effect for other people as well.
Especially when they jump around a lot.

[balmy flicker]

not on a server so IDK

it hapens when moveing around my base and when menuing but not when exploring at mach 2

the mod pack is stone factory and it only realy happens in my factrory

normaly it only happens onces about 5-10 min after loading the world

[jade scaffold]

thats just minecraft being the horrifically unstable mess that it is

for some reason minecraft and ONLY minecraft causes my PSU to trip

even when using things like optifine and limiting frame rate

do you have your minecraft set up properly to use your available RAM

because you have to actually tell it to use more than 2GB of ram

[pearl moon]

Optifine is notorious for causing issues with other mods and otherwise being unoptimized itself, it's typically recommended to go with one of the other mods does what optifine is supposed to and more without issues.

[jade scaffold]

I didn't see how optifine can trip ocp on the PSU but okay

[pearl moon]

I personally use the Sodium mod

I'm just saying in general optifine is known to cause issues.

[jade scaffold]

Also sodium isnt compatible with forge is it though

[pearl moon]

https://www.curseforge.com/minecraft/mc-mods/sodium-reforged

[jade scaffold]

Ah

Might help with spirits issue then

[balmy flicker]

it gets and uses 8+GB

ill have to look at the optimisation mod they use

[cinder lagoon]

memory leak - what kind of hang are we talking about specifically?
keep F3 open and look at the top right when it hangs, chances are that the garbage collector is clearing up stuff.

you can also set -Xms (minimum memory size) - it sometimes helps the garbage collector run in intervals that don't cause hangs. set it to about 50-75% of the -Xmx setting

also: what minecraft version?

[balmy flicker]

it is on 1.16.5
it happens once a play sesion early on so i kida doubt memory leak but it is java...

ill look into those setting it might help the power consumption

[cinder lagoon]

It will not help power consumption but it might mitigate any memory stuff.
I was talking about how long the hang was

[cerulean helm]

Getting the new sensor panel setup using InfoPanel and HWinfo. So far I'm liking it.

Didn't realize HWinfo pulled my UPS sensors. So added that to the panel. Still tweaking the setup but it works for now.

[balmy flicker]

a about .5-1.5 sec it is not consistent

[jade scaffold]

1.16.5 is known to have some weird stutter issues when modded

[balmy flicker]

figures and that mod pack is noto the light side (looks at the 5-10 min load time)

[jade scaffold]

yea i dont know what it is about 1.16.5 but every time me and my friends try make a modlist for it its just not particulary stable
1.12.2 I think is the most stabel we've tried

though i dont really risk playing MC anymore

[balmy flicker]

1.12 was used for a long time as it was very stable and well optimize (for mine craft)

[amber fable]

Never in my life have I actually wanted an ultra wide monitor. Now I'm doing three way diffs all day and suddenly...

This damn thing is too narrow.

[mossy gull]

People always call me crazy for having a 3:2 display below my 21:9, funny enough I always have the vertical headroom on the 3:2.

[cinder lagoon]

The Surface book aspect ratio really is comfortable for laptop working

[amber fable]

Normally, I would prefer a taller monitor for work, but for merging wider is better.

And right now my life is one huge continuous merge commit.

[mossy gull]

Who said it was a Surface Book?
My Surface Book is in my bag or other room majority of the time

[sacred seal]

I have been doing a lot more stuff with AI recently, and have been thinking about getting a use p40. Problem is I already have a 6900xt in my rig, so fitting both could be somewhat, shall we say, "interesting".
I'd also need a new PSU. Anyone know of something that would let me mount it externally in its own enclosure?

There are lots of 1x to 16x adapters that use a USB cable, but I was wondering if there is an equivilant for 16x

[sonic meadow]

There are external pcie enclosures that run over a... 4x link I believe over usb-c/thunderbolt
You could also just relocate it with a 16x riser cable assuming your case has space (a miniatx board in an atx case for example)

[sacred seal]

I technically have the space for it in my case, but its going to be right up against the intake for my 6900xt

Basically I would like something like this, but with a x4 (or better) interface
https://www.amazon.ca/Graphics-Buckle-USB3-0-Extender-Adapter/dp/B09Q5YP7MF/ref=sr_1_58?crid=158WMU2IKTD8N&keywords=external+16x+pcie+enclosure+-nvme&qid=1691040382&sprefix=external+16x+pcie+enclosure+-nvm%2Caps%2C215&sr=8-58

[mossy gull]

You also can use a normal flexible pci-riser

[sacred seal]

Would have to be a very long one, haven't seen one that would be long enough

[mossy gull]

https://lian-li.com/product/600mm-pci-e-4-0-riser-cable/

I get mine always from Lian-Li

[sacred seal]

Hmm, that could work, though that costs half as much as a used server would

[mossy gull]

PCIe risers are sadly expensive, but they are 4.0

[cinder lagoon]

Ohh I just thought Microsoft was the only one with 3000x2000 Screens🤠

[mossy gull]

Nope, but it ain't a Surface Book, just a Surface Pro

[jade scaffold]

Don't get a used p40 unless you can get it for well and truly dirt cheap

[sacred seal]

They seem to be going for 300 cad on ebay. For 3090 levels of performance in some AI tasks, I think that is a decent trade off

The 24GBs of vram alone is nice

Just bieng able to put the entire model on the card speeds things up significantly

[jade scaffold]

3090 levels of performance is literally only because the 3090 lacks certain features so has to brute force them

[sacred seal]

I know, but its still 1/3rd the cost of a 3090

And I am betting these server cards are probably in better condition then some random mining gpu

[jade scaffold]

A RTX a4000 should obliterate a p40

Can get them for 400 used

[sacred seal]

Only 16Gbs of ram, and the cheapest I am seeing them is for 800 cad

If the model doesn't fit on the card, performance drops by like 70%

If it runs at all

[jade scaffold]

Right that's a shame because the p40 is a pile of shit for ai by modern standards

It actually loses to the a2000 at a almost comical degree if not leveraging it's extra memory

Double check there's no Tesla v100s/quadro GV 100s floating around at a reasonable price

[sacred seal]

Tesla V100s go for about 1.1k, and the quadros go for 4500+

Or well the v100s with 32GBs of vram

If it wasn't a massive pain in the rear to game on, I would concider selling my 6900xt and getting a v100

My other option was to get 4 k80s and stick them in a server. Would cost a bit more then the p40, but I would get 64GBs of vram (apparently most AI tasks let you pool vram with multi gpu)

[jade scaffold]

You can't game on V100 it literally lacks the shit needed to game on unless you wanna try Jerry rig GV100s kernel and shit into it

And yea ai scaled really well off multi GPU

[sacred seal]

I think people have gotten it to work by running it through a second GPU, but I may be mixing up my cards

[jade scaffold]

If you can source yourself a smx2 mainboard for cheap

You can get the smx2 v100s for far cheaper

Which will give you a scalable solution where you can just socket more GPUs in as you need them and get pretty much the best link between them possible

[sacred seal]

Yea, those seems to start in the 4k range

Though granted those also come with GPUs

Or not...

[jade scaffold]

You have to look real hard for them

But I nearly got one cos I found one for 50 quid

[jade scaffold]

Just thought of something worth considering Google collab gives GPUs for a monthly a couple years back it was 10 USD a month for a p100 and 50 for a v100 not sure what it is now but there's a chance it's now V100 for 10 and A100 for 50

[sacred seal]

I use the free colabs for some things, have not been terribly impressed with them. Too many funky problems for me to concider spending money on them

[jade scaffold]

Yea but think of the money saved

I'll have a lil look when I get home to see if I can find any ballin deals on sxm2 NVlink boards that are reasonably shippable to Canada

[jade scaffold]

this atrocity of an adapter card exists too

[sonic meadow]

when you said atrocity I was expecting something worse, with like rgb or some shit. not something that looks pure function and nothing else

two six pins rather than one 8 pin is an interesting choice though

especially without the spacing to allow you to use 8 pin plugs (ones without the breakaway sense pins)

[cinder lagoon]

What does this even do?😂 I legit can't tell

[sonic meadow]

its an smx2 to pcie adaptor

[jade scaffold]

Yep

It's super jank

But with a bit of knowhow you can make it work

[jade scaffold]

Also have checked the Tesla gpus support open GL

[sacred seal]

Everything I am working with supports cuda

[jade scaffold]

But as far as I can tell quite literally fuckall else other than open cl so gaming is probs a no-go like 90% of the time

[sacred seal]

https://www.youtube.com/watch?v=AYDnGeqvwAA&ab_channel=Level1Techs

[amber fable]

If anyone ever asks you to work on a modified version of Telegram for them, say no.

The Android version has Java classes that are routinely over 20k lines, some are over 30k lines.

The iOS version is reportedly "harder to work with" than the Android version.

I like the name of this build task

[cinder lagoon]

Someone would think they'd follow the java programming guidelines and keep classes purpose-locked xD

I bet they are but the purpose is just too big

[amber fable]

"Method call has 36 parameters, expected 37"

[amber fable]

Oh how truly exciting. The method of doom strikes again.

Now it has grown to 38

Thankfully, it seems there is a big block of functionality related to this parameter, but it is guarded by if thingy != 0

So thingy is now set to 0 and git rebase --continue

[mossy gull]

Okay, why are Intel and NVidia fanboys the absolute worst?
I am suppose to build a 4K gaming machine for 1600.-.
I can easily build 1 with the 7800X3D and 7900XTX.
But I got a speech on how AMD never has stable performance and always crashes.

[jade scaffold]

AMDs current gen is fine

their last gen was mostly fine too did go 8 months without any dirver updates but nothing was immensly broken like the RX5000 gen was

[mossy gull]

They believe the 3070 paired with an i5 13600 will out perform the AMD build I originally wanted to build.

[jade scaffold]

thats an immense ammount of brain death

[mossy gull]

You don't say

[jade scaffold]

even if you downgrade the cpu a bit to a 7700 as the 7800x3D is overkill for 7900xtx or 4080
you can only really do a 4080 build for about 1900

[mossy gull]

7700X is 320.-, and 7800X3D is 350.- overhere.

[jade scaffold]

you're going off euros?

[mossy gull]

Yes

[jade scaffold]

where you getting 7800x3Ds for that price from

[mossy gull]

Local store, but they come without a box though

Most of the time they are combo deals with an Asus Strix B650 or TUF B650 board and 16GB RAM for around 600.-

The Strix B650 usually goes for like 250.- to 300.- overhere.

[cinder lagoon]

average Userbenchmark Moderator

[mossy gull]

Sounds like it

[jade scaffold]

The amount of copium userbenchmarks reviews of Ryzen has is insane

[mossy gull]

The person who wants the PC told me that AMD builds budget components and is a copycat manufacturer, and they want the "best of the best".

[jade scaffold]

While the 4080 is superior to the 7900XT if only because of NVidias extra features. But in gaming none raytracing your be a fool not pick the xtx over the 4080

[mossy gull]

He picks the 3070 over the 7900XTX

He doesn't believe me the 7900XTX can run 4K no problem

He also thinks FSR is from NVidia because it works on his older 1660

[sacred seal]

There is ceratin level of irony in calling AMD the copycat, they made x64 which is the foundation of all modern desktops. See how they like running with just 4GBs of ram

[sonic meadow]

I mean AMD literally copied Intel's chips early on, and then started making their own

You know, those x86 chips that both companies are now so famous for. That x86 which got expanded to make amd64 (which is the basis of the x86-64 extensions to x86)

And the limit 4GB of ram on 32bit x86 was always incredibly soft, even before the first 64bit chips you could have more than that. Its not like the 16bit chips were limited to 64KB of ram either.
Heck, the whole 3.25GB limit from Windows was entirely artificial too, xp (no service pack) would happily let you use 4GB, it was only with sp1 that they introduced the 3.25GB limit, and people have hacked the kernel in vista 32bit to allow more than 4GB. The issue arises that in 32bit windows by default an application can't use more than half the total system memory (without both a boot flag set in Windows, and a flag set in the application)

[mossy gull]

They all copy something from each-other.
But thinking that NVidia and Intel are the only companies inventing shit for PCs, disregarding what AMD is doing and calling them a budget company that copies everything is just ignorant.
Without AMD we would probably still use 32-bit, we still would rock 1-core with 2-thread CPUs, we still would play pixelated games because NVidia would still use max 32MB VRAM and more.

It's just that Intel/NVidia fanboys are the most annoying to deal with in my opinion because they're blind in who truly develops shit.
Like saying FSR is from NVidia because it runs on NVidia cards.

[jade scaffold]

It's not even really copying from each other one makes an innovation like for example tesselation and it's actually super useful and cool so everyone starts making use of it and then the rest have no option but to implement it too to keep competitive

[mossy gull]

https://youtu.be/5KZ-jY_OqUA

[tender plank]

So, TIL, do not close your browser when you have multiple windows open and one of them is full screen. It REALLY doesn't like being brought back from the dead like that.

Here's some of what was involved in trying to get things back to normal.

[jade scaffold]

XD

[cinder lagoon]

I always have some form of a Tab backup extension running.
What was the problem with it being fullscreen?

[mossy gull]

My Edge never had a problem with it.
It just boots up in Full-screen again

[ionic tusk]

I hate printers

Had printer refresh today

Friday special something went wrong

Hahaha one printer I printed to before it got swapped out couldn’t see it through IP turns out wait for the printer technician lol

Turns out was somehow the Ethernet in the wall somehow broke one wire F

[mossy gull]

Corsair, isn't that mic too sensitive?
https://youtube.com/shorts/s0du39PJ1Sg?feature=share

[tender plank]

The issue was that it would open as full screen, but when I went to leave full screen the window would cease to exist. Except it wasn't actually closing the window, it was just removing it since the tabs were still open. Although, part of the issue was probably trying to load 130+ tabs at once.

[sonic meadow]

oh god yes, it must be

[mossy gull]

I think Corsair's going crazy, although I like the ideas they have.
https://youtube.com/shorts/02xc9CjzRg0?feature=share

[sonic meadow]

cute. and I dont mean that in a "this is at all useful" kind of way

[mossy gull]

What about, walk with stick, aim with Tobii, and shoot with a mouse like a gun.

[sonic meadow]

you have just reinvented the hosas setup 😆

[mossy gull]

A HOSATAM😃

[sonic meadow]

kill it with fire

[mossy gull]

No

[tender plank]

Just because it can pick up like that doesn't mean it's too sensitive, it just means it has excellent gain control. A shotgun mic can work very well from 1 foot away and from 20 feet away, it's just a matter of adjusting the gain appropriately.

[pearl moon]

So i'm just making a sanity check, but a 12gb 3080 that a friend is selling is 100% a buy no questions asked right?

Especially when my current gpu is a 2070 super

[mossy gull]

Depends on the price

[pearl moon]

Oh i forgot to say lmao, $350

[mossy gull]

I held a shotgun mic for 100m distance, it's amazing

[mossy gull]

Bargain

[pearl moon]

Yeah i figured lmao

Especially for DCS VR which really wants a higher end gpu, and it really wants more vram

[mossy gull]

12GB is what I call currently the minimum required for optimal experience.

Especially for 1440p

[tender plank]

Yeah, those ones are more problematic at closer ranges, you start getting some severe presence effect within a couple feet of them.

[mossy gull]

They're best between 50m and 90m

[tender plank]

Wait... 100m? Dafaq? I read that initially as 100 feet which is possible but requires extremely good management of the noise in the area and specialized mics. At 100m, you'd be better off with plants or hidden lavs if you can.

[mossy gull]

They usually have a dish at the back to mute the noise around them

[tender plank]

Oh, that kind of mic. That's not a shotgun, that's a parabolic mic or something like that. The shape of a shotgun is to dampen off axis noise which is what the dish would be doing there.

[mossy gull]

The one that owns it told me it was a long range shotgun

[tender plank]

¯_(ツ)_/¯ It doesn't fit the definition for a shotgun mic that I'm aware of, there may be more to it when it comes to heavily specialized tools like that.

[mossy gull]

I am not really specialised with "earth based" mics

[pearl moon]

VR is quite a different workload, especially oculus vr as half the GPU resources are being tied up encoding video, but either way it'll be a huge improvement to the 2070 super i'm currently running, i might need to get a 5800x3d or something to see a proper performance increase, but i figure at $350 it's worth getting for the futureproofing alone if nothing else.

[jade scaffold]

video encoding isnt very demanding atall

[pearl moon]

Tell that to task manager showing 50% of my 2070 used purely for realtime video encoding to be sent to the oculus to decode with it's onboard processing

[jade scaffold]

A task manager is rarely right B video encoding is handled by a separate part of the GPU to 3d rendering

It's why separate stream pcs are mostly dead aside from those who want their stream to never go down even if their main pc crashes

[mossy gull]

A 2nd pc that records also means less impact

[jade scaffold]

By maybe 2%

If your on an Nvidia GPU anyway

Maybe just maybe if you have 4 monitors while streaming watching your own stream playing a game with a guide up you might loose 7%

The dedicated encoders and decoders on gpus now are ridiculously good

[cinder lagoon]

This might be absolutely right - depending on the encoder settings used, even the hardware encoding could take up a significant part of the ressources.
Testing game performance with and without Recording yields a drop of somewhere around 5-10% on the newest GPUs

I don't actually know of any tools for reading out the actual percentage of GPU used by encoding

[jade scaffold]

I'm not even sure if you can for Nvidia gpus could be one of those things locked

[cinder lagoon]

The applications using it might have some control over it

Like maximum latency for one frame and maximum bandwidth

[sonic meadow]

Task manager can give you a graph of the percentage of the various engines on the GPU in use, one of them being it's encoder load

[cinder lagoon]

I do not trust these views tho

[sonic meadow]

Though that's not as a % of total GPU resource usage

[cinder lagoon]

They seem really inconsistent and weird

[sonic meadow]

It's seemed pretty consistent from some testing i did a while ago

Stuff like that is going to be being reported from the drivers

I think hwinfo can show you too

[cinder lagoon]

Nah some things go on some engines while other things (that should be exactly the same) go on other engines

[jade scaffold]

Yea the only reason you tend to loose performance when encoding is because of the power usage of the encoders which isn't very high but it's there

[sonic meadow]

heck. I lost performance on my old 660ti by running the fan at 100% (it used part of the boards power budget)

[amber fable]

https://www.youtube.com/watch?v=FGW3TPytTjc

[sacred seal]

Glad to see a detailed breakdown of all the bad data LTT has put out

Interesting that both HW unboxed and GN did not attend LTT expo

[sacred seal]

And the fact that they stole hardware to sell to someone else is beyond despicable

[inland dove]

Linus is a bad guy now?

[sacred seal]

His personal ethics have never been great, and LTT has never had a particularly high level of quality, but this is a new low.

[amber fable]

TBH, Billet Labs probably has a viable legal case against LTT.

But it would probably cost them more to sue LTT than they could ever recover.

Not for the "review", but for the sale of the prototype.

[sacred seal]

Oh, they certainly have a valid legal case.
I defiantly think they could sue and make back more then they could recover, the potential damages could be massive (If a competitor managed to get a hold of it they could basically claim damages for the entire development cost, plus a quantum for lost development time, and future profits) and LTT would be able to pay it. However its likely to cost Billet labs everything they have to fund the lawsuit

Heck, there even may be a viable criminal charges against linus for corporate espionage, though that gets messy

Canada has no law against it, so I don't know if exerdition would apply

And the fact that this situation did not immediately resolve itself with LTT buying back the monoblock and sending it straight back to billet is beyond bizzar to me.

[inland dove]

I guess they lived long enough to see themselves become the villain.

[jade scaffold]

you allways hear linus on this high and mighty shit about company ethics if he actually belives what he preaches he should have actually used it properly before judging it and then he should really be covering the cost of rebuilding this exspensive ass prototype

[sacred seal]

Well just ethics in general. Adblock is piracy comes to mind

[jade scaffold]

isn't adblock banned on all LTT work computers

[inland dove]

Well if that's the case.

I'm going to use adblock even more.

[cinder lagoon]

yeah I was actually also unhappy with a good handful of the last couple vids... I've always seen the discontinuity between what was said, what was fact and what an editor put on screen willy nilly but it really is kind of a low point in terms of quality...

I would cut some slack tho for having variable quality testing in some cases because testing processes are not done for many of the categories.
On the other hand:
If I were them, I would've waited on using unfinished processes and rather do it manually until everything is actually set up with proven accuracies

selling the block is actually just plain careless tho - I will definitely vote with my views in the coming weeks

[cerulean helm]

The theft of the prototype block and selling it at LTX wasn't careless, that was illegal. And beyond unprofessional.

[cerulean helm]

So apatently Linus made a comment on his own forum regarding the video (this is via someone else so take that as you will), and it isn't good.

[amber fable]

I don't know what inspired me to check the LTT reddit just before bed, but it is on fire

And yeah, the Linus Response™️ is Not Good™️

But it follows past patterns of behavior, namely the dude has no ability to take criticism.

Oh my.

[cinder lagoon]

I do not see why GN did not reach out privately first but but not even addressing on the podcast shits gonna be on fire...

[amber fable]

Oh, and the top comments on today's LTT video are all pointing out issues with the video and the "test" "methodology" used.

I hope this doesn't calm down for a looooong while, they need to get serious about actually being tech and not empty entertainment.

[jade scaffold]

And linus again needs to do what he preaches he caused that small water block company to stall he needs to do the right thing and compensate them

[mossy gull]

I don't know if it's real or a elaborate stunt between channels

[jade scaffold]

What lmao

[sonic meadow]

I would be highly surprised if Steve (Gamers Nexus) would pull a stunt like that, it would massively hurt his journalistic integrity

[mossy gull]

YouTube been wack lately, so that's why I have my suspicions.

[sacred seal]

With the beating that LTT is taking, and GNs upstanding history I would be shocked if this was manufactured

[mossy gull]

I'll have to see in the long term, currently I don't trust much that happens on YouTube

[cinder lagoon]

LMG Disaster view count influence:

   7 vid Avg  | 48h avg | delta %
LTT   1.410M  | 0.694M  | -51%
SC    0.188M  | ------- | -------
TL    0.482M  | 0.248M  | -49%
GL    0.336M  | ------- | -------

...will definitely be noticed.

[sacred seal]

What percentage of views are generated in the first 48 hours? I'm guessing not enough to account for the drop off, but that delta is probably not a true value

[cinder lagoon]

I've taken the last 7 videos from before the Incident ™️ and averaged them
A huge portion of youtube views (at least for LMG) is generated in the first 48 hours

They said something akin that somewhere during WAN

[sacred seal]

Yea, its high, but I am guessing more like 80-90%

[cinder lagoon]

I'll keep an eye on the numbers the next couple days and repost updated tables

[sacred seal]

What you would have to do is get the 7 day average as it was on the 13th to get a better idea

And a 7 day average after

That would account for recency bias and weekly viewing habits

[cinder lagoon]

True - taking values ONLY after 7 days respectively would be more accurate

[sacred seal]

Still, what you have here is really interesting

I would not have thought to log it

Looking at there social blade it does not seem like they have been impacted very much

[cerulean helm]

Why? Why is that necessary? Who says they LMG hasn't been made aware of the issues for a long time (they have). GN has a pretty good track record of proper journalism when dealing with shady companies. LMG is no different in that regard.

[sacred seal]

GN answered this in todays HW news video

Basically GN did not want LMG to be able to get out in front of it and confuse the messaging

Even now LMG is lying. They said they had reached an agreement to pay for the loss, but in reality they only sent bitlabs a offer to give some sum of money, and they only did this after GNs video went live

https://youtu.be/X3byz3txpso?t=226

Steve talks about not reaching out at 6:25

[cerulean helm]

Watching this video currently. And yeah its sad watching Steve just look exhausted and disappointed in Linus and LMG in general.

[sacred seal]

You got to love Louis's take
From yesterday

And today

It will be interesting to see if other youtubers start calling out LTT. Right now its mostly being passed off as a GN vs LTT thing

[cerulean helm]

I saw Jay at least make a comment on Twitter that clearly showed support to GN for the video, And Brian who ran BPS Customs made a lengthily post highlighting he has had issues with LMG for awhile as well.

[cerulean helm]

Well, this may be pertinent to the Discord channel.
https://www.msn.com/en-us/news/technology/discordio-suffers-massive-data-breach-announces-closure/ar-AA1fiqJ4

[sacred seal]

How the heck does someone make a venerability in webcode that lets you download the entire database?

[amber fable]

Spicy.
https://threadreaderapp.com/thread/1691693740254228741.html

[amber fable]

Lots of ways. Most of them really, really dumb.

[sacred seal]

That's what I thought...

[cinder lagoon]

I've heard that the company has not had a great track record with female (public facing) figures but I gave everyone involved the benefit of the doubt (also madison)

[sacred seal]

While I am sure there is a kernel of truth to madison's thread, I'm not inclined to believe most of it. There were not that many people working at LMG when she first joined as I recall, the fact she was answering to multiple managers does not make much sense to me

And the fact that she only refers to them as managers, not by there actual names

Or I guess coworkers

[cinder lagoon]

LMG Disaster view count influence:

   7 vid Avg  | 48h avg | delta %
 LTT   1.410M  | 0.694M  | -51%
 SC    0.188M  | ------- | -------
 TL    0.482M  | 0.248M  | -49%
 GL    0.336M  |*0.088M* |*-75%*
*TQ    0.249M* |*0.123M* |*-51%*

[sacred seal]

Also her whole non compete stuff does not make a load of sense considering the multiple channels that were run by LMG members at the time. Terran had one, Nick had one, I think brandon had one,

[cinder lagoon]

Nah that just was because she already had some kind of community

[amber fable]

Last I knew they were getting close to 2k lost floatplane subs.

[jade scaffold]

https://www.youtube.com/watch?v=0cTpTMl8kFY&ab_channel=LinusTechTips

overall a very well put together response video

[cinder lagoon]

I feel like this is an OK response but I still feel pissed so I'm kinda gonna hold off for a little longer

[tender plank]

It feels like an apology video where they're saying "I'm sorry for getting caught". Hopefully it turns out to be sincere and it was just the result of gradually growing issues. Hopefully they'll work out a week per year or some kind of process where they can take some time to do a company wide assessment of their processes and what may need to change or be updated.

[cinder lagoon]

A week per year is not the right approach - taking time when it is necessary is the correct one

To be clear:
now it wasn't "necessary", it was "overdue"

[tender plank]

Yeah, it'd be better to do it more frequently, but a single day or something here or there may not be sufficient to properly consider alterations to the processes as people talk about things, bring them up with management, talk about changes, interact between teams, etc. It's a 120 person company meaning formal or informal meetings will almost entirely include smaller groups of people other than maybe an all-hands once they're ready to wrap up where they go over major planned changes and look for suggestions from there.

It was similar for me when we were developing a new schedule a few years back. We could go back and forth between the different people on shift, send details to the manager, get his feedback and assess with the people on shift again. It took a week or so for that, although it probably would have been faster if we had been able to get everyone on the team in at once.

[cinder lagoon]

There's gotta be something like the garbace collector in Java:
some meter where you can read off how rough things are going or the frequency and aplitude of errors being brought in by the community

Everyone working there has access to this meter and if it's full, the time required will be taken by everyone to work together on the problems

[tender plank]

Agreeing on a week or maybe a 2-3 day period 2-3 times per year where they basically shut everything down and look into major changes that need to be done would probably be for the best. They generally have a backlog of videos, so 2-3 days of downtime may not even be noticed all that much.

[cinder lagoon]

Something like this but with the meter and reserving 3 days, taking more if needed

[tender plank]

It would also permit planned maintenance of infrastructure to be done with less impact.

[amber fable]

Reddit seems to think the video contains a few errors... An unredacted email for example.

I haven't watched it yet, but reportedly they plug LTTStore a few times as well which is just... What. The. F.

Time to watch it I guess.

[sacred seal]

The video is pure PR spin. The fact they launced the video monetized speaks volumes. I just clicked through but they never really seem to admit that they were wrong. There is nothing about any of the more serious accusations.

[amber fable]

My RAM should get here in a hour or so \o/ I'm upgrading from 32gb to 64gb because modern games take so damn many resources ||lol, no. Actually it is because I'm having issues with running our of RAM and having stuff randomly get killed by the OOMK when running large C++ compile jobs.||

As a side note, I'm glad I stubbornly ignored all the advice that a 3950x was stupid for gaming. It really comes in handy for work.

[cinder lagoon]

• They addressed the Billet labs situation
• Linus is sorry about "answering emotionally"
• They've addressed the wrong data and what they plan on fixing this issue in the future, aswell as removing it from old videos or removing said videos entirely.
• They did not have a sponsor (tho joking about having one) for this video despite having had offers for one
  Video monetization is a very minor source of income on LTT - it does not speak volumes imho

skipping through misses much tho please enlighten me, what you're missing

TBF the Store plug really did not have to be in an apology video but I understand it at atleast one point (the store manager section)

[cinder lagoon]

Given that the water block is seemingly gone - they can only offer money imo

Also they called the water block a GPU accidentally

[sacred seal]

Actually they found the water bloc, but billet labs did not want it back opposed to the money as they had already spent a lot of money making a new one, and they don't trust LMG to get it back in a timely manner
Billet labs sent them a GPU as well as the block

[cinder lagoon]

Ohh I did not know this, thought it was block only

[sacred seal]

Which makes the situation where they used the wrong card even more rediculous

[cinder lagoon]

Thats true

[sacred seal]

In regards to billet labs, they only addressed what happened after it was auctioned. Not why it was there for multiple weeks before, or how it ended up auctioned in the first place (other then colton messing up)
Linus provided factually misleading information about the timeline. That is not answering emotionally
I'm not mad that they had monitization on. I am upset that they thought the community would not mind if they did not have monetization on. If you are truly sorry, you act sorry. There should be zero glimps of impropriataty.

When I say that they never admit that they are wrong, what I mean is instead of appoligizing for what they did, they only really talk about how they are going to fix it. There is no reflection, no we are going to have to work hard to earn your trust back, just here is why you can trust us now

[cinder lagoon]

Isn't "colton messing up" a reason enough how it ended up there?

I'll give you a solid point on the second paragraph tho

[sacred seal]

How was Colton able to mess that up? Supposedly he is the person in charge of all the communications between LMG and billet, and from the sounds of things also the person who authorized the sale. How does someone mess that up, and it not be malicious to some extent?

If someone loans you a car, tells you that they loaned you a car, how do you accedentally sell the car?

There are reasonable explanations. There was a miscomunication as to what item was bieng talked about for example

[cinder lagoon]

As far as I understood it the block (idk about the card) got accidentally put to other video props

Which would've been eligible for auction

[sacred seal]

Colton says he messed up. So he told them to put it with props in this situation?

The issue isin't so much what happened, its the fact that we have to guess what happened

[amber fable]

RAM upgrade is done, peak usage hit 65% while compiling Telegram Desktop (current project is updating a custom set of Telegram clients for a customer)

That is a pretty solid amount above the "something is gonna die" line.

[sacred seal]

Nice, I upgraded my ram to 64 GBs not that long ago so I could run LLMs without hitting page files

[amber fable]

TBH, I was very sure that 32 would be all I would ever need on this system when I built it.

But damn is C++ a resource hog.

LLMs are probably worse though.

[cinder lagoon]

https://youtu.be/0cTpTMl8kFY?t=788
"our process has failed when I was selecting items to include in the charity auction"
And he forgot to put the Billet labs contact as an addressant in the email, pretty shortly after having recieved notice of the situation"

[amber fable]

After having watched that video, I am not even slightly reassured that LMG is going to improve.

[sacred seal]

Depends on the size of the LLM, there are ones that can run off of 4GBs of ram. But for 30b+ models you need like 24+GBs of ram

[amber fable]

It was a giant non-apology full of "we don't really get it" red flags.

[cinder lagoon]

Apparently I don't get it either - most of my concerns have been addressed

Granted tho - i do not care much about "i'm sorry" as I'm REALLY not resentful

If the video quality keeps dropping I'm naturally gonna not want to watch the videos anymore

[sacred seal]

"our process has failed when I was selecting items to include in the charity auction" What process? Why do you even have a process for something so niche? What they should of said is "There was a communication break down and ___ happened because of ____ which caused ____". If you are actually sorry you don't hide behind vague terms
And he forgot to put the Billet labs contact as an addressant in the email, pretty shortly after having recieved notice of the situation Why is he making a new email instead of replying directly to billet labs. Why is he not handling this personally at this point?

The email thing is forgivable, if beyond stupid

[amber fable]

https://www.reddit.com/r/LinusTechTips/comments/15spr86/due_to_popular_demand_here_is_a_complete/

[cinder lagoon]

Habits - I could imagine how the direct "reply" feature is just not used in some situations that could reasonably apply here
Colton did handle this personally

[sacred seal]

No, he went and told procurement to go do something.

[cinder lagoon]

Did LTX not have charity auctions for some time?
If you plan on doing something like selling old video props, you're gonna create a "process" or "plan" what to sell and how to sell it, rihgt?

[sacred seal]

Ok, fair. I did not read it as such, but that interpretation makes sense

However that does not diswade my critisizim that they are blaming the process, not the person

[cinder lagoon]

Procurement is in CC as to further handle the process but this screenshot indicates that Colton was directly involved in that.
I have to also say tho: the Subject reads "Re: Hi from" but apparently it does not have a recipient set

[cinder lagoon]

I believe that they have a company policy that forbids pointing at a simple person as "shit flows uphill"

Pointing at a single person (which I think this comes really close to) does not help imo

[cinder lagoon]

Making jokes (or keeping the video forcefully "funny") was innappropriate imo - i can see why they would've tried tho

[sacred seal]

Directly involved, not handling it personally. This is after they just sold the item, I don't even know what procurement would even have to do with this (unless its about the GPU, in which case you go down there personally, pack the card yourself and don't send any thing until you have mailed it to them)
To be clear, this is a small nitpick, but you have stacked error after error after error here. You miss multiple shipping dates, then you sell the item. And you somehow don't realize this error until multiple business days later, and the only reason you do is there is a huge scandle?

This is a problem with how people are behaving, not with procedure. As such the appoligy should reflect that. You don't have to name the employee in question, but you specify what they did wrong

[cinder lagoon]

To be clear: LTX was on 29th and 30th July, when the mail was from 10th August
It really was not an apology video, the title reflects that

[amber fable]

Want to see something funny?

[sacred seal]

Yea, that is my point. They don't believe they have done anything wrong here. They are just placating there audience with this.

Yea, I saw that

Which explains why they did not get the block back instantly at least

They had no clue who bought it

[cinder lagoon]

Lying about the reason is appropriate because it could very well send off a shitstorm

Its a PR decicion alright

[amber fable]

Trying to placate their audience. However, then Madison flew over on a bombing run and...

[cinder lagoon]

Whats with madison now? Shes been gone for like half a year

[amber fable]

Someone has been deleting a lot of comments.

Allegations of workplace harassment.

[sacred seal]

I don't think that the Madison thing is going to affect them. I don't see GN any of the large youtubers picking that up, and I doubt that anyone is going to be like "I diddnt care about the GN stuff, but now I hate LMG"

[cinder lagoon]

Yes I've heard about that but we had that convo some hours ago

[sacred seal]

UDF tech might, given that they have been doing coverage of the thing, but I don't think that GN would care about it without more substantiating facts.

[amber fable]

Well of course GN won't. They are all about hard data. This is interpersonal drama.

[cinder lagoon]

Yes

And as long as I don't see any similar factual accusations of workplace harrasment that is rooted in the company structure, I'm not going to act on any of it

[sacred seal]

Well they are about holding companies accountable, and they just hit LTT. Its not completely outside there purview. That said I would be disapointed if they do cover it

[cinder lagoon]

I do not believe that they were trying to profit off of billet labs in any shape - given that it was a charity auction aswell - but I expect to not hear similar stories again
their gain would have been so miniscule, unless actually being paid by a big company sponsor that would be able to mass-produce such a thing but that gets into the conspiracy territory really fast

[sacred seal]

If they make a response to LTTs response video (which I don't think they will), it would be useful evedence, if only in the capacity to show that LMG rushed out there appoligy video and is unwilling to properly address the concerns of the community

[cinder lagoon]

Agree

[sacred seal]

I doubt that they were trying to profit (they would get a tax write off, but I doubt the block sold for more then a few hundred), but then again how do you screw up so badly? It should not be the case that LMG doing something illegal makes about as much sense as any legitimate answer

[cinder lagoon]

I blame murphy's law until proven otherwise

[surreal moss]

Is this the only time something was sold by accident or do other instances exist?

[cinder lagoon]

I believe this is the only time but I have no informations from any subreddits or forums

[surreal moss]

If it’s a one off that strikes me as a legitimate mistake

[cinder lagoon]

Granted, I do not remember them having access to any other prototype product

[sacred seal]

Company X offers money to LMG for the prototype. After the review LMG sends it off to them, where they spend a month+ reverse engineering it. They then send it back to LMG who put it up for auction, thereby creating a credible reason for the leak, deflecting some blame in case something happens. Colton purposely does not include Billet in the email so that they can point to it as evidence. GN exposes the plan too quickly, so they quickly loose the auction info, buying time to "resolve" things with billet labs.

The fact that this does not sound 100% insane compared to the actual chain of events is concerning

To be clear, I do not believe that this is what happened

[cinder lagoon]

I would need some more information to believe that like a sudden rise in income or whatever, some form of payment that cannot be attributed to a regular sponsor

[sacred seal]

What if the sponsor gives 10% extra on every sponsorship for a year? Or just buys out a bunch of ad inventory, thereby inflating the cost of ads?
Your never going to be able to prove this, but the fact that you don't instantly go "That is a completely insane conspiracy theory" is what is concerning. Out trust in LMG is that low

[cerulean helm]

Looks like LTT has lost at least 300k subs on Youtube by my rough estimate.

I honestly don't care what LMG has to say after Linus's insanely dumb forum post where he doubled down and then lied.

Video was only made after GN's HW news that has a 12 minute clip spliced in with an update to the LTT fallout where Steve produced the receipts showing Linus straight up lied about compensating Billet Labs.

[amber fable]

This is the single most cursed deepfake I have ever seen.

https://cdn.discordapp.com/attachments/783929255985676298/1141724401981739078/linus_toxic_gossip_train_1.mp4

[sonic meadow]

that is gloriously cursed

[inland dove]

Technology has gone too far.

[sacred seal]

You compare that to the deepfake video that LTT did like three years ago, its amazing how much tech has progressed

[inland dove]

You can't trust anything to be real anymore.

[sonic meadow]

am I real?

[inland dove]

No.

I've decided that you aren't real.

[mossy gull]

That isn't the right question, what is real

[sonic meadow]

oh no, not again

[surreal moss]

Took me a second to realize that is an italicized I and not a /

[sonic meadow]

am / real?

[surreal moss]

I
/
I
/

@fleet wasp would be abusing this all over the place

[fleet wasp]

What? / would NEVER

[surreal moss]

Iike you have no history…?

[fleet wasp]

l am a perfectly respectable citizen.

Hold on, got something to test...

Oh hey it's Thursday...

Convenient

[mossy gull]

https://overclock3d.net/news/gpu_displays/amd_s_aiming_to_make_starfield_their_debut_fsr_3_game/1

If AMD Fluid Motion is as good as they say, it can be a huge kick in NVidia's nuts.
Especially if it works on older GPUs as well

[cinder lagoon]

Can't wait to play my 380+ mod Minecraft modpack at more than 20 FPS 😂

[mossy gull]

It's rumoured that AMD Fluid Motion can generate up to 3 or 4 extra frames, if this is true, although I don't think it'll be on release of FSR3, then you could play your Minecraft at at least 40fps😃🙃

[jade scaffold]

its almost certainly not true

and even if it was the results of doing such a thing would lead to a heavilly distorted and incorrect image

theres a reason not many people like DLSS3

because human movements are some what random they are impossible to predict meaning in anything where stuff is moving fast it just eats shit which isnt an issue DLSS2 had

if you want frame generation to be actually accurate you have to already be running at obsene Frame rates

[sacred seal]

Well, it technically would depend on the initial framerate. The higher the frame rate, the less time there is between each frame, and therefore the lower the latency cost. However the higher the initial framerate, the less useful a tech like this would be

[jade scaffold]

it would also be more accurate at higher frame rates

the lower the FPS the worse Frame generation does

which kinda negates the point

again an issue DLSS2 and FSR2 didnt have as they simply improve an already existing frame rather than try and predict the movement of some dipshit within a 16ms time frame

[cinder lagoon]

Generating any more than 1 frame between frames is of absolutely no use imo because it just gets inaccurate af

[jade scaffold]

generating 1 frame is already inaccurate af its about passable in games where shit isnt moving super fast but the second shit gets moving fast everything gets all messed up

also im looking for a 14-15inch OLED display
can be from any device as long as a control board can be found for it or if a EDP - mini HDMI adaptor exsists for it
also it cant have a chin as the plan is to make a thing with symetrical bezzels on all sides

also needs to be touch screen if anyones got any ideas for devices to rob such screens from that'd be great

[mossy gull]

https://wccftech.com/amd-fsr-3-might-generate-up-to-4-interpolated-frames-be-enabled-on-driver-side/#:~:text=The code states that AMD's FSR 3 will,interpolated frame for every single frame when enabled

[jade scaffold]

yea thats likely a value option added to test to see what happens if you try that

its legitimatly not possible to predict that many frames with current tech

not even close to accuratly anyway

and having only 2/5ths of your frames being accurate isnt gonna make for a great experience

also for frame generation you need to know the depth of objects meaning FSR3 wont be able to be slaped onto what ever people like sadly

[mossy gull]

Well, maybe AMD has the tech, Fluid Motion has been in use by professionals for year apparently.
So we'll have to see.

[jade scaffold]

yea i doubt that

what is a possibility is combining the upscaling of FSR 2 and frame generation of FSR3 which could actually yield an enormous increase in performance with only relativly small loss in quality

[mossy gull]

It's AMD, they do all kinds of weird shite without the consumers knowing it.

[jade scaffold]

which with most of AMDs stuff being open source could be the real kick in the balls to NVidia

[mossy gull]

If FSR2 is easily upgradable to FSR3, just like when they went from 1 to 2.
Then it'll be a massive kick in NVidia's nuts.
That could mean that AMD has a 200+ game library head start, not even 50 games support DLSS3.

[jade scaffold]

FSR 3 appears to be driver side tho as your article points out

which if its only on AMD GPUs

it might not end up as widely supported due to amds relativly smaller share of the market

[mossy gull]

That's the rumour though, AMD could've changed it in the past months.

[jade scaffold]

the reason FSR2 is on like everything is because it works on Anything

its not quite as accurate as DLSS2 but its still damn good which breaths alot of life into some older GPUs

intels open source one too is a nice one to have in there too im seeing that poping up in quite a few places recently

[mossy gull]

It's nice, but doesn't work as flawless as FSR.
I had quite the trouble with XeSS in some games

[jade scaffold]

one reason FSR3 could be locked to AMD GPUs could actually be that it makes use of AI hardware on AMD GPUs

which would make some sense for generating whole ass frames

[mossy gull]

What if it could work with and without the AI hardware?

[sacred seal]

Generating more frames basically means you are looking further into the future. 1 frame is 50%, 2 frames is 66%, 3 is 75, so on and so forth. Not only does that mean that the frame is less likely to be accurate, but it also means that the "snap back" effect would be increased

[jade scaffold]

i see the latency addition without ai hardware being far too significant

[mossy gull]

I mean FSR1 & 2 don't use AI hardware while DLSS does.
So

[jade scaffold]

upscaling is easy enough

creating a whole new frame is an entirely seperate beast

[sacred seal]

Now an alternative would be to add in a one frame latency, so instead of looking into the future, you already have that future frame generated and you just need to generate the frames between two known points

[jade scaffold]

yea

which is possible

[mossy gull]

XeSS can work with and without Intel's AI hardware, AMD can pull the same approach with FSR3.

[jade scaffold]

but also still has accuracy issues

XeSS has 2 completely different versions

the AI one works far better than the none ai one

and again

[mossy gull]

But it works and both versions are in a game.

[jade scaffold]

Upscaling is comparitivly easy

[sacred seal]

Yea, you will run into the acceleration effect.

Where the inbetween frames do not account for non liniar movement of objects

The way around that is to add two frames worth of latency

[jade scaffold]

like we've been able to upscale things for years and years now its only recently AMD and NVidia have decided to leverage that as a way to increase performance

Upscaling can be done with just algorithms

you dont NEED AI
AI Helps for certain details but its not needed

[sacred seal]

There is upscaling, then there is upscaling.

[jade scaffold]

if you are trying to predict an object moving along a changing curve with only visual data with an algortithm you are going to have a bad fucking time

[sacred seal]

AI has the advantage of knowing what an object is supposed to look like, so when you upscale it can add in lost (but potentially inaccurate) data

[mossy gull]

Still, FSR3 isn't released yet and we don't know how it works, but I have the feeling it can work both with and without AI accelerators just to keep its accessibility.

[jade scaffold]

i bet they'll try

but i bet the none ai version will just make soup out of an image

especially in anthing where things are moving fast as even AI has issues predicting that

[mossy gull]

But here's the thing, many games don't fully utilise the CPU, that extra CPU headroom you also can use for AI.

[jade scaffold]

no you really cant

[sacred seal]

Not really

[jade scaffold]

CPUs have piss poor AI performance

unless they have dedicated AI accelerators

which intels do have

Amds dont though i belive

and regardless that would limit you to new gen cpus

negating the main advantage of this kinda tech

[mossy gull]

It's still something though.
How good it'll be we'll see.

[jade scaffold]

it may as well be nothing

[sacred seal]

I have been playing around with LLMs a fair bit recently. A generation of 200 (ish) tokens on a 13b model takes around a minute on my 6900xt. The generation time doubles if it has to use the CPU at all

[jade scaffold]

i dont think you understand how bad CPUs are at AI

[sacred seal]

AKA I run out of vram

[mossy gull]

I mean you can't keep older CPUs in mind forever, eventually they have to be phased out by both hardware and software.

[jade scaffold]

CPUs are good at fetching data for AI but thats only really useful when trying to feed a GPUs Data

[mossy gull]

That's always fun when that happens

[sacred seal]

I have never tried running it pure CPU, but generating a 30b model with half on my CPU took on average 15 minutes to generate. (typically 30b models take 3x longer then 13b models to generate)

[jade scaffold]

point is AI benefits from running lots o stuff in parallel
trying to use a device that can only do 16 things at a time thats already trying to run a game is just wasting enormous ammounts of power

[mossy gull]

I have models from my job that can even make the 4090 cry.
We have roughly 128GB of VRAM for a single simulation project.

[sacred seal]

And there is a significant cost to just the effect of having anything run on the CPU

[jade scaffold]

and the AI Accellerators in newer CPUs arent anything close to matching the just raw performance of a GPU ignoring GPUs AI hardware

like the good ol K80 is better than any modern CPU for AI

and the K80 is wank at ai compared to modern GPUs

its like 10 years old

[mossy gull]

What about Xeon-Phi?
I have 1 laying around 🙃

[jade scaffold]

Xeon Phi also are terrible at AI

they just dont have the cores

[sacred seal]

The nice thing about accelerators for CPUs is they can leverage the system ram

[jade scaffold]

and Accellerators in CPUs are more so made to run little things

like face recognition
they help out with chatbot stuff which is much more power efficient than slamming the normal cores with that work load and also faster

[mossy gull]

They were amazing for the time we had for simulations.
Like engine burn in Mars's atmosphere 😃

[sacred seal]

I've been thinking about building out a server using a butload of k80s, depending on what happens with my job in the next few weeks

[jade scaffold]

K80s cant chain together how you expect them to

they cant pool memory

[sacred seal]

For LLMs they can

[jade scaffold]

1 K80 is actually 2 12gb GPUs

[sacred seal]

Each AI is assigned a different chunk to generate

I don't understand it, but apparently it works

[jade scaffold]

yes but then you are going through PCIE and i belive K80s cant just direct communicate they have to share data Through the CPU

[sacred seal]

Each k80 is only working on part of the model, so it works the same way that I can run half a model on my GPU and half on my CPU

[mossy gull]

There's not many GPUs that can stack the VRAM to my knowledge.
Like the Vega FE was able to, but the Vega 64 wasn't, even though they were mostly the same.
And I think same goes for even current market workstation/server GPUs.

[jade scaffold]

all AI ones made since pascal can

the advantage of SMX type GPU arrays is the cpu has nothing todo with them really
they have a PCIE switch in effect and they just shoot data between eachother at like GDDR6X Speeds 900gb/s which allows them to actually behave as one big unified GPU

though SMX systems are hard to get your hands on

though if you can get a mainboard at a reasonable price the rest is easy to do

its getting the main boards that is the main issue

[sacred seal]

I'm debating between k80s and p40s. I can get 3 k80s for the cost of one p40

[jade scaffold]

i think one P40 should be massivly faster than 3K80s

i seem to remmber reading somthing about this

ill see if i can find the article comparing a bunch of K80s to 1 P40

[mossy gull]

I know from the Vega FE that it use to use the unused PCIe lanes for the VRAM stack.
But you needed a board that had at least 2 full PCIe 3.0 X16 slots.
Which weren't many of.

[jade scaffold]

the Tesla P40 can be upto 26x faster than a K80 according to toms hardware

[sacred seal]

If you could that would be appreceated. Of course the extra vram of the k80 solution is nice, but the p40 has newer archetecture, and means I can do future scaling

[jade scaffold]

the K80 only supports FP32 is the issue

the P40 can do INT8

[sacred seal]

Ah, yea, that explains it

[jade scaffold]

making it depending on the AI load just comically faster

[sacred seal]

K, well I am getting a p40 then

[jade scaffold]

even modern GPUs now all Do FP32

[mossy gull]

They just need to come with boards with a dedicated bridge for multiple GPU support.

Lets stack 4090s

[jade scaffold]

you cant really properly stack through PCIE

its fast bandwidth wise

has latency issues tho when trying to stack a bunch of GPUs though

[mossy gull]

Still, it sounds cool to have 128GB VRAM

[jade scaffold]

which is part of the reason quadro NVLink machines didnt scale liniarly

SMX Systems scale a bit closer to Linearly its not liniar yet but they are much closer to that perfect scaling than PCIE Based NV Bridge solutions

[mossy gull]

I rather have too much VRAM with less stable performance over not enough VRAM and even worse performance.
But then I ain't competitive in games.

Currently I miss having 32GB VRAM since 1 of my Vega FE's died.

[jade scaffold]

also yea no wonder the P40 is so much faster its clock speed alone is 2x higher its got 2.5x the memory bandwidth 1.5x the core count 50% higher IPC even in an identical load the P40 would obliterate 3 K80s i belive

i got 24gb so im all good

just waiting on the next gen GPUs now

[mossy gull]

Which isn't enough for some of my projects, I kinda made them tad too VRAM hungry😃

[jade scaffold]

5090/8900xtx or what ever AMD decide to call it

[sacred seal]

AMD is steping down from high end GPUs for the next few gens I believe

[jade scaffold]

thats good

thats the strat for getting market share

most of the market is the low to mid end

not the ultra high end

[sacred seal]

They were having difficulties putting 13+ chiplets on a single die

[jade scaffold]

xD

[mossy gull]

Which so far is a rumour though.
They still can release a 8900XTX or whatever they want to call it.

[jade scaffold]

i see

[sacred seal]

If they can get that figured out though

[jade scaffold]

NVidias next gen is also gonna be chiplet i believe

[mossy gull]

50 series is rumoured to be monolethic dies.

[jade scaffold]

also possible

Jenson was talking alot about chiplets tho last we saw him

im hoping for chiplets

[sacred seal]

I'd guess monolithic, they don't have amd's prior experince with chiplets

[jade scaffold]

this is true

however they've been holding off on chiplets for a while despite saying they are the future

[sacred seal]

Trying to figure that out from the ground up takes time. How long was zen in development for? Like 6 years or something?

[jade scaffold]

yea but think how long its been since 2000

[mossy gull]

AMD has been putting shit together for ages now.
Even though not true MCM design, HBM GPUs were AMD's thing back in the day.

[jade scaffold]

by 5000 it will have been 6 years

hbm was mega dumb on like 90% of what AMD put it on tho xD

HBM2 on a vega is the goofiest flex ever

[mossy gull]

They still did it and proved it sort of worked.

[sacred seal]

back in the day Fury was a fairly recent development by the component history standards

[jade scaffold]

near 0 performance advantage

all the compactness tho

the Vega PCBs were smoll

[sacred seal]

The fury nano was the only real advantage of the tech

[mossy gull]

16GB HBM2 on the Vega FE, I loved it, I had 2😃

[jade scaffold]

yea its alot of memory

but you werent getting any beifit from the HBM

the Vegas were just too slow to leverage it and too limited in certain tasks that would have benifitted from it

[sacred seal]

Well the HBM allowed that much memory to be added at a lower cost to power consumption

[mossy gull]

Vega FE was also 1 of the fewer cards that was able to play Resident Evil Remake in 4K ultra settings just thanks to the 16GB HBM2.

[jade scaffold]

that wasnt anything todo with 16gb the 1080ti managed that fine

and the 1080 ti only had 11

[mossy gull]

Vegas were workstation cards and should've stayed that way

[jade scaffold]

yea AMD kicked the vegas in the knees in terms of workstation performance

the only one that wasnt kicked in the knees was the frontier edition

which just didnt have the raw power needed to keep up with the 1080ti and titan XPP XP? (tomany Titans with the leters X and P in their names)

[sacred seal]

Vega was a archetecture, not a card class

[mossy gull]

The Vega FE was the Vega 64 workstation alternative with double the HBM2 and proper "Crossfire" support

It also had a gold shroud instead of silver

So it didn't fit in any build

[jade scaffold]

the Vega dedicated cards consisted of 3 cards

none of which were really relevant at the time

cool cards

got hard boddied by pascal

the Vegas were weirdly inefficient too for what they were

like im sorry 300W for the Vega 64 you fuckin wot

[mossy gull]

I just hate NVidia's software with a passion.
Reason why I actually don't want to get their cards

[jade scaffold]

yea Nvidias software is meh

[mossy gull]

And now nobody complaints about 4090's power usage😃🙃

[jade scaffold]

the video and picture taking stuff is nice

[jade scaffold]

the 4090 is one of the most efficient GPUs ever made

the Vega 64 was 35% slower than a 1080ti with a 28% higher power draw

which is INSANE

[sacred seal]

Well the 1080ti came out a fair bit after if I recall

[mossy gull]

The problem I have is it still looks like it was made with W2000 in mind and I don't want to log-in.
Not to mention everytime the NVidia driver crashed on me I had to reboot the entire system to make it work again.
While with AMD and even Intel I could just reboot the driver no problem.

[jade scaffold]

1 year after

the 1080ti though was identical to the titan

which released at the beginning of the generation

quite literally the only difference was the titan could do FP16 and had 1gb more ram

[sacred seal]

Was it? I thought they used different dies

[jade scaffold]

nope same die afaik the 1080ti same core count and everything

there was also 2 titans in pascal

the first one to release was actually about 3% slower than the 1080ti because of its lower clocks and memory speed and the 2nd one was like 3% faster and i think was node shrunk or somthing

[mossy gull]

But doesn't sound half as cool.
1080Ti doesn't really roll of the tongue that easily.

[jade scaffold]

tru

[mossy gull]

Same that Quadro sounds so much cooler than A6000

[jade scaffold]

infact the 1080ti GPU was the same di used in the Titan Xp the Titan X the Quadro P600 and the Tesla P40

fuck me they got their use out of that fucking die

[mossy gull]

Damn, where did all the cool naming went to.

Vega, Titan, Quadro, FirePro, Tesla

[jade scaffold]

i loved the Tesla naming scheem i everyone still calls the new data centre GPUs Teslas

not officially their name but everyone calls them that

[sacred seal]

Those have nothing on the real old names. AMD had the Rage series

[mossy gull]

I still call the A series Quadro.
I call AMD GPUs Navi

[mossy gull]

I mean it sounds cool.
AMD Rage😃

Same that the Xbox Series X should've been called Xbox Scarlet.
Just keep their names and stop with the dumb numbering

[jade scaffold]

also the old Teslas looked way cooler

[mossy gull]

Don't tell others, but I think blower cards are fucking cool

[jade scaffold]

uhhhhhhh thats a good fuckin price

Pascal was peak blower card design

then they made 2000

and ewww

like really ewwww 2000 looked so crap

they fixed it with 3000 tho 3000 had fuckin awesome heatsinks

[mossy gull]

I don't like the fans that push into the card, I always have the feeling it heats up the PC instead of it blowing straight out of the case.

[jade scaffold]

they actually dont though

they depending on your fan set up actually improove CPU Temps

which is odd

but they do

[mossy gull]

I know, but still, I just have that feeling which makes me feel uneasy.

[jade scaffold]

also they keep real cool like my 3090 would only get to 70C which aint bad for a stock cooler

kinda wish i put the cursed water cooling on it rather than my proper water cooling

theres somthing funny about a CPU AIO hanging off a GPU that makes me happy

[mossy gull]

You mean as cursed as the Vega FE LC?

[jade scaffold]

Also Shadow dont know where you are from but these might be of interest to you
https://www.ebay.co.uk/itm/275926388476?_trkparms=amclksrc%3DITM%26aid%3D1110006%26algo%3DHOMESPLICE.SIM%26ao%3D1%26asc%3D20200818143230%26meid%3D8c3d79f404f447f5930d690edee76151%26pid%3D101224%26rk%3D4%26rkt%3D5%26sd%3D394613634419%26itm%3D275926388476%26pmt%3D0%26noa%3D1%26pg%3D2047675%26algv%3DDefaultOrganicWebV9BertRefreshRanker%26brand%3DNVIDIA&_trksid=p2047675.c101224.m-1

£160 each if you buy 3 its £150 each

they do come from china tho so you might be waiting a hot minute for them to come

[mossy gull]

What doesn't come from China these days.....

[jade scaffold]

so you see the part that drops down under the screen there is that part usually flixible or is it usually sold because ideally i need it to not be there

[sacred seal]

Thanks for the suggestion. I'm in Canada so that isin't an amazing deal, but it is better then what I would typically find

[mossy gull]

I am kinda waiting till the A6000 drops in price.
I have been wanting to build a strong mini-ITX PC, but GPU manufacturers in their infinite wisdom decided we need 5 slot GPUs.
And the A6000 is 2 slot.

[jade scaffold]

the price in £ wont directly translate over to canada theres a chance it might actually be more reasonable

[sacred seal]

225$s + shipping vs 280 cad

[jade scaffold]

just like underclock a normal card and make it small or water cool

water cooled ITX systems are the best

[mossy gull]

Lets try and watercool in the NZXT H1 V2😃

[jade scaffold]

why do all these screens have the chin

its annoying me now

[mossy gull]

Because

With my stacked displays I have the upper 1 hanging upside down so the bezel doesn't interrupt me.

Which looks very odd

But also fun

[jade scaffold]

i just wanna make a goofilly large tablet

let me make my goofilly large symetrical bezel tablet

[mossy gull]

Okay, by my opinion tablets need bezels to hold on to.
But manufacturers in their infinite wisdom want to remove all bezels to make them look "better".
What's your opinion on a tablet that you can barely hold?

[jade scaffold]

you can still have thinn bezzels and good palm rejection works

and like

this is gonna be a 14-17 inch tablet

holding it might break your arms

https://www.ajparts.co.uk/replacement-for-nv160wum-nx3-16-0-led-lcd-screen-fhd-165hz-display-panel.html?srsltid=AfmBOoqhmbW2G5dGjB5PQaywh6xHfOhpi18ZaH_5NvtijDN2DmeW4e5gpGM

OOO

DAMIT

Its IPS

[mossy gull]

But it'll be hard to hold onto.
I noticed with my Surface Book 2, which has thicc bezels, that when holding it I sometimes still touch the display.
And modern tablets want to look like edgeless phones

[jade scaffold]

yea but bezzels look ugly

so ill take the inconvinience for the style points

[mossy gull]

You're weird

says the weird 1

But the Surface Book 15 inch top part looks damn sexy, even with the bezels though.
I personally dislike edgeless displays, like 90% of phones have it and they're uncomfortable to hold.

[jade scaffold]

i like a bit of edge

my nothing phone 2 has bezzels but they are thin

and symetrical all the way arround

which looks really good

no stupid chin

[mossy gull]

I mean, look at Samsung with their dumb curved displays.
Holding them firmly made you send your not so charming pictures to your friends.

I miss the time where I could hold my phone firmly.

[jade scaffold]

i had a Huwei phone which had curved edges honestly not a huge fan of that kinda design

as much as i hate apple

their phones do look gud

they got that style down

but then nothing went and did it better for a reasonable price and then gave me a studant discount

so like fuck you again apple

[mossy gull]

Especially with the 14, I absolutely dislike the Notch, I prefer a hole punch

[jade scaffold]

i cant decide on which i prefer it varies from moment to moment

[mossy gull]

Hole punch is more versatile in my opinion as everything can be displayed around it or in case of the Galaxy Fold be hidden entirely.
While a notch is always visible

[mossy gull]

I have an iPhone, just to prove Apple fans that the lightning port is dumb and that Apple should go to USB-C 🙃

[jade scaffold]

next gen they are going to USB C finally

bet they will still keep with crappy 15w charging tho

[mossy gull]

I know, I hope they also get a file browser so I can attach an SSD onto the phone

[mossy gull]

By EU law, Apple has to comply with fast charge standards and isn't allowed to throttle 3rd party fast chargers that aren't Apple Certified.

[jade scaffold]

oh no i didnt mean that

i meant they just wont have fast charging atall

[mossy gull]

My iPhone 12 has fast charging

[jade scaffold]

i refuse to call 20W charging fast charging

its bearly faster than wireless

[mossy gull]

Fair enough

[jade scaffold]

i feel like 35w is probs the lowest you can really get away with calling Fast charging

45w though should be standard

cant wait for more 100w phones to be a thin

full 3 days charge in 10 mins yes please

[mossy gull]

What about the Poco Phone's 33w

[mossy gull]

I personally prefer a thicker phone if that also means it'll survive 2 days instead of half a day

Also, many companies have to adapt their designs for the EU laws, where batteries have to be easily user replaceable.

So there's a chance phones get thicker again

[jade scaffold]

i doubt it

you can make batteries repairable without making phones bigger

the whole everythings glued to make it smaller thing is BS

its just to make it harder to repair

theres a phone thats like the framework of phones super repairable

its prety normal thickness

[mossy gull]

Easily replaceable, which means the back probably has to be opened easily without heatguns or whatever you need these days

[jade scaffold]

yea

easy

its called screws

[mossy gull]

Without screws

[jade scaffold]

doesnt have to be without screws

screws make it easilly accessible

anyone can undo 4 screws and toss the battery out

we arent going back to the days of dropping phones and the back and battery fly off

[mossy gull]

https://www.androidauthority.com/replaceable-batteries-eu-2023-3335711/#:~:text=European Parliament passed a change to a law,This could fundamentally change how smartphones are designed

[jade scaffold]

thats fucking stupid

now about 10000 more phones are going to become ewaste

how do you screw up basic shit EU

you coulda just said using easilly accessible tools such as screwdrivers

but no you had to make ever phone near water a fucking explosion hazard again

[mossy gull]

The problem is with "non-removable batteries" is most people throw away their phones if the battery dies quickly because it costs too much to replace the battery, they simply don't know it is replaceable or it is too complex for themselves.

By allowing the average person to be able to easily swap the battery a phone will get much older and the only thing that gets thrown out are the batties, which are recycled for 95%.

[jade scaffold]

not what i meant

having snap off backs makes waterproofing impossible

fire not much of a concern with modern phones you can dunk most modern phones in a pool and they'll be dandy

having a 4 screw system would allow for just as much accessibility to batteries

with near equivilent water tightness to modern phones

providing the best of both worlds

[mossy gull]

No it doesn't, the Galaxy Fold is water proof and has massive gaps

[jade scaffold]

because it doesnt have gaps

everywhere where theres a gap theres a membrain

and shockingly because of its design its a shit show to take apart

[mossy gull]

I am talking about the Galaxy Fold, which is made to be water proof, unlike the water tight phones.

[jade scaffold]

yea it has a rating of IPX8

and has a membrain in all the joints which prevent water or dust from getting to the actual internals

and its one of the most obnoxious phones to dismantel

[mossy gull]

https://www.androidcentral.com/phones/is-the-samsung-galaxy-z-fold-4-water-resistant

Literally water gets in, but it's treated in a way with a coating that makes it resist water for a while

With an easily removable back you can do the exact same as the Galaxy Fold

It's not magic, older rugged phones did it as well

If you want to you can also make a laptop this way.

[jade scaffold]

then you gotta worry about dust particulate getting in it

and also if your phone gets filled with water and stays filled with water that coating wont do shit

[mossy gull]

What if you seal the phone except the battery compartment?

[jade scaffold]

or if you god forbid spill somthing slightly acidic

[jade scaffold]

would make phones substantially bigger to do so

and people quite like their thin phones

i dont see why the EU would just go fuck total waterproofing

[mossy gull]

If you do that even current phones will die.
The water tightness or resistance is tested in fresh water, not salt water, not liquid acid.

[jade scaffold]

no they wont

[mossy gull]

Because it doesn't fuck with waterproofing

[jade scaffold]

because they are actually totally water tight

[jade scaffold]

if you have a snap on back you arent gonna be able to make the phone water proof

you can coat it but that wont help against some coca cola or orange juice

[mossy gull]

https://www.euronews.com/green/2023/06/22/new-eu-law-to-force-smartphone-makers-to-build-easily-replaceable-batteries

[jade scaffold]

yea and

there are ways you can make things easy to reapir

[mossy gull]

I would love to repair my phone, but it's cheaper to buy 1

[jade scaffold]

with just basic tools everyone has

[mossy gull]

And everyone can work with micro electronics?

[jade scaffold]

why would they have to

if they are just replacing batteries

4 screws out

back off batery out new one in

back on screws

literally 2 mins

[mossy gull]

I know enough people who can't even screw out the screws of an iPhone because they're too small for them

[jade scaffold]

no electronics knowlage needed

i phone screws are proprietary

they have apples weird security thing

[mossy gull]

Nope, not here in mainland Europe

[jade scaffold]

putting 4 philips screws in the back anyone can manage

and if you cant you arent competant enough to snap the back off without breaking it anyway

[mossy gull]

You over estimate the average joe that can't even plug in a phone in 1 go

[jade scaffold]

i think you'll find they can

and i think you'll find that most people dont even look at their phone when plugging it in

unless they got a micro b phone

but thoes pretty much arent made anymore

[mossy gull]

What's so bad about easily user replaceable batteries?
Anything that requires tools will be seen as "too complicated".

[jade scaffold]

nothings wrong with user replacable batteries

but if you cant undo 4 philips screws

you arent gonna snap the back off a phone

to replace the battery are you

[surreal moss]

A lack of proper tools doesn’t make someone incompetent

Just ill equipped

[mossy gull]

Not everyone is precise enough for tiny ass screws

[jade scaffold]

they dont have to be tiny

you can make they pretty reasonably sized

[mossy gull]

The flatter the device the smaller the screws

[jade scaffold]

and if you cant manage it then ask someone with basic competance to do it

[jade scaffold]

not true

you can have massivly wide screw heads on super tiny thread screws if you want

[mossy gull]

Fine, put a T3 screw in the Galaxy S23
T3 is what people consider very small'

[jade scaffold]

im against making phones more fragile for the sake of the ignorant and incompetant
taking 4 screws out is somthing i could teach a 3 year old

why not a philips head

with just a wide head or a flat head

[mossy gull]

Actually our flat phones are more fragile than phones back then when they had easily replaceable backs

[jade scaffold]

only the screens are

current phones are vastly more watertight their main boddys are substantially stronger too

[mossy gull]

You can't just make the head wider, you can but then you need a specialised head that's meant for flat screws

[jade scaffold]

no you dont

flat head screwdrivers exsist for a reason

[mossy gull]

Which belong to 1 of the most hated screws because you barely have grip on them.

[jade scaffold]

and flat head screws dont need any tools atall

a knife works fine

[jade scaffold]

only a problem if you mega torque the screws down which doesnt need to be done

[mossy gull]

Good job, now you told someone to scratch their phone which can void their warranty.

[jade scaffold]

scratching your phone cant void the warentee

and popping the back pannel off is just as likely to scratch it anyway

[mossy gull]

If it's scratched too badly it does

[jade scaffold]

no

no it doesnt

[mossy gull]

Scratch your phone, send it up for warranty to replace the screen, you bet they'll ask for 350.-

[jade scaffold]

no knife or screwdriver is gashing a scratch big enough in a phone for a company to thin you are intentionally breaking your phone for a refund

otherwise keys with your phone would pretty much insta void your warentee

which they dont

[mossy gull]

Okay, I am outside, I quickly need to replace the battery, let me find my screwdriver to replace the battery.

Or I could easily pop off the back and replace the battery

[jade scaffold]

what lmao

what scenario do you need to pop a battery outside

who carries an unshielded bomb in their pocket just to hotswap their battery

carry a power bank

[mossy gull]

Also, screws can get damaged, which also isn't ideal, knowing how most people deal with screws, I don't want to know what happens if they have to do it multiple times.

[jade scaffold]

if the phone has a plasic back it will break from popping it off

or even if its thinn aluminium it could break

depending how its mounted you might need a pick to pry it off anyway

[mossy gull]

Oh wait, my phone is 3 years old, battery degradation is a thing, my phone won't hold it the entire day but I am on the road the entire day.

[jade scaffold]

what

thats stupid

immensly so

what you didnt have time to pop it in when it arrived at your house

phone batteries you know arent just gonna pop up in shops

phones wont be useing standardised batteries

[mossy gull]

Not stupid, I am away for 12 hours, am I suppose to do nothing while travelling by train?
I have to use my phone the entire day.

[jade scaffold]

POWER BANK

you arent going to carry a fully charged lithium ion battery that has no case just to hotswap a battery

thats immensly dangerous to you and everyone arround you

[mossy gull]

LG had a phone with hotswappable battery that didn't have the flimsy back panel, something you can improve upon.

[mossy gull]

And a batterybank isn't?
Those burn like a motherf×cker

[jade scaffold]

no a battery bank isnt

because it has a case

so the chance of somthing puncturing it or bending it is low

[inland dove]

Are you arguing against removable batteries?

[jade scaffold]

no

not in the slightest

[surreal moss]

That might be an improvement from screws

[jade scaffold]

im arguing that forcing tooless swaps makes phones less survivable especially around water or drinks

which is true

[mossy gull]

Not ideal, was also 1st gen product.

[jade scaffold]

oh yea that thing

so much of the batteries volume was armour that it has piss poor batterlife

[inland dove]

Don't spill water on your phone/drop it in water?

[jade scaffold]

sometimes its unavoidable

[surreal moss]

Whoa whoa whoa whoa careful with your logic in here

[jade scaffold]

and like

[mossy gull]

Most phones don't even have a water tightness rating and dropping them in water or spilling drinks will kill them.

[jade scaffold]

it rains

[jade scaffold]

thats not true atall

most phones do have some kind of IP rating

[mossy gull]

Which killed my old Motorola

[jade scaffold]

usually atleast IP54

which makes them splash resistant

so resistant to RAIN

[surreal moss]

Gotta insult the the competency of the general population instead

[mossy gull]

You know that the IP rating isn't really viable anymore after 2 years of use?

[inland dove]

I had an LG fortune 2 that had a snap on plastic backplate and removable battery and had zero issues with it.

I had that phone for a couple years and even swapped the battery out on it once to extend it's life span.

[jade scaffold]

if your phone battery is dying going through 4 screws isnt going to be a challenge for 99% of people and it improves survivability of phones in the face of liquid and dust

if someone cant use a screwdriver they likely know someone who can. if they dont then they likely should never have their hands on somthing as dangerous as a unshielded lithium ion battery

[mossy gull]

The fact is, the EU wants that even 80 year old grandpas with shaky hands can remove the battery.

[jade scaffold]

yea but an 80 year old isnt going to do that anyway

[surreal moss]

Met plenty that can and do despite any shakes

[jade scaffold]

like have you met old people they either want nothing todo with tech and hardware or they are really savy with atleast the hardware

[mossy gull]

I have a grandpa who still rocks an old Android smartphone because they can easily replace the battery without the use of tools

[surreal moss]

Remove the battery or maintain a device? Using screws or other attachments to bind a device together aren’t automatically blockers. I suspect the majority of the language is targeted against proprietary screwheads or fasteners that artificially interfere with the ability to maintain a adevice

[jade scaffold]

use a philips head or a flat head it would be fine

everyone has one

and if not they are like pennys

[surreal moss]

And also likely targets the average individual. An 80 year old with disabilities wouldn’t qualify

[surreal moss]

And that’s a choice for that individual. Not an industry standard requirement

[jade scaffold]

if the law was just. Make batteries easy to replace without specialist tools or proprietary tools it'd be fine

[mossy gull]

They want it to be doable for even the least tech savvy people.
And tiny screws aren't that easy.

[jade scaffold]

yea they are

[surreal moss]

That’s your opinion

About the screws

Also depends on how serviceability is defined

However, you two have already done this loop and keep talking in circles about it

[mossy gull]

Okay fine, I have repaired over 20 iPhone 3G(S), which had normal + screws, yet nobody I know wanted to do it themselves because they were scared it would go wrong.
Which was 1 of the easiest devices to work on for me, but for others too complex.

Not everyone is as tech savvy as us

[sacred seal]

I don't see why toolless should preclude waterproofness. I have a insta360 one rs which is a modular camera, and its waterproof

[mossy gull]

The EU said themselves it doesn't change waterproofness, and multiple tech journalists say the same

[jade scaffold]

on the i phone 3G you had to do 2 screws then pry the scren then mind the ribbon cables the 3G for its time was not a simple disassembly in the slightest compared to modern phones it is but compared to the phones of the time i could see why someone wouldnt wan to tak it apart

[sacred seal]

If it has a entire battery module you can remove without tools, I presume it is perfectly possible to make a phone battery that is the same

[mossy gull]

Majority of phones have funny ass ribbon cables that make it scary for the average joe.

[jade scaffold]

yea they did

but on alot of phones the first thing you had access to was the battery

so you didnt have to go that far

[surreal moss]

It certainly is. However they do tend to be a bit bulkier

[sacred seal]

The only thing that would be difficult to waterproof is the contact points

[jade scaffold]

oh no thats easy

[sacred seal]

It seems to me like that is a solved problem at this point

[jade scaffold]

th issue is if you want your phone to be small you cant armour up the battery

[sacred seal]

You don't need to armor up the battery. You just need it waterproof, and that can be a thin membrain

The phone itself is the armor for it

[jade scaffold]

yea it is

thats not what i meant

[mossy gull]

Back in the day, and I think 1 of the last Galaxy XCover phones still had a replaceable battery and was fully water resistant.

[jade scaffold]

it wasnt

the first galaxy that was actually fully waterproof was the S7

other ones were water resistant

[mossy gull]

The 1st Galaxy XCover has IP67

[jade scaffold]

but they didnt have the crazy water resistance that modern ones do

[mossy gull]

Nope, but the 1st XCover was released in 2012, so that was 11 years ago

[sacred seal]

Basically you would have something like this, where the phone itself is a waterproof module, the battery is waterproof, and you just connect them before putting the back on. The plastic that goes between the back and the phone itself could be super thin

[jade scaffold]

yea you can do that

but it takes up alota space

its possible

i just like my phones small with massive batteries

[sacred seal]

Shouldent, the waterproofing plastic does not need to be structural

[jade scaffold]

it needs to cover everything though

[sacred seal]

You are adding a later that's mabey .1mm thick

[mossy gull]

@jade scaffold the Samsung XCover 6 Pro has an easily replaceable battery with removable back and has IP68, same as the Galaxy S23, which is completely sealed.

[jade scaffold]

yea and its 3mm thicker it does also have a slightly larger battery to be fair

[sacred seal]

THis is the iphone 14. instead of having the body adhear to the back to give waterproofing, you would have it adhere to a thin plastic membrane, and make the cover detachable

[jade scaffold]

how would you then repair everything else

[sacred seal]

Law only requires that the battery be replacable, no?

[jade scaffold]

theres a pretty scarece chance you are taking such a thin and well fitted rubber membrain off without damaging it

so make batteries more replacable and sacrifice replacing all the other stuff that can go wrong?

:I

sure

oke

[mossy gull]

With IP×8 you basically put a thin water resistant layer over your components to make them water resistant, which is possible with any bloody phone.
And to be honest, I seen iPhones filled with dust, and they were suppose to be dust proof.

[mossy gull]

What about in the style of the LG G5 and make it more rugged.

[sacred seal]

Never said it would have to be rubber, could be a thin plastic and a disposable part

[jade scaffold]

while i like the idea of the G5s lil hotswap battery

[sacred seal]

It would not be a worse solution to what we have currently at least

[jade scaffold]

its got appauling battery capacity

[sacred seal]

Though I agree that basic tools should be allowed

[mossy gull]

That was back in the day, same size batteries currently have 4000mah

[mossy gull]

Precision tools can be considered specialised tools though

[sacred seal]

Yea, so you are not allowed to require precision tools

Most watches are waterproof and user servicable

Most digital*

[jade scaffold]

also battery density hasnt really been increasing technically

we have just gotten alot more ballsy

[mossy gull]

I almost want to say, my Galaxy Watch Classic isn't

[jade scaffold]

like theres been some advancements but alot of it is battery management is beter so big batteries dont self destruct super violently anymore typically