off-topic-tech

[sand saddle]

I mean still depends on the quality of the mobo itself as well.
The z790 apex encore should do 8000 fairly reliabily with 14th gen

But the voltages are quite high at that point

[wanton orchid]

35 at 7200 is 17 at 3600
9 at 1800
it basically means we didn't lost any latency to ddr3
but with insanely more stability bandwidth and tuning

and I find it's soon from a ddr generation

[twin dew]

Igor tested hundreds of CPUs and of the best clocking ones only small minority could do 8000MT/s on Asus Maximus Z790 Apex
www.igorslab.de/en/r-batches-13900kss-and-imc-regressions-intel-core-14th-gen-binning-results-from-almost-600-cpus/7/

Yes, more might have been able to do it by manually tweaking the IMC side voltages.
And not just the DRAM ones.

DDR5-8000 can only be achieved by 2 of the 15 CPUs, and only on the Apex mainboard. One of the CPUs doesn’t even manage DDR5-7000 stably on the Hero

[sand saddle]

They had all IMC voltages on auto?

[twin dew]

Which usually means "Way too much" on Asus.

With high DRAM frequency setting.

[sand saddle]

Yeah

[twin dew]

And yes, they were auto except VDIMM equivalent at 1.6V.

Same MB, same RAM, same settings, 2 of 15 could do it.

[edgy hazel]

Imo there is a big difference for consumers between "It can run 8000MHz" and "It can run 8000MHz with some tweaking and stability testing"

[twin dew]

Yeah, those were the ones that were stable in specific y-cruncher test.

Booting at 8000MT/s is pretty useless if it isn't actually stable enough to be usable.

But again goes into the Intel vs. AMD marketing.
Intel markets what the golden samples can do on special MBs.
AMD markets what almost any CPU can do on almost any MB.

[gilded helm]

AMD is frankly leaving money on the table by not having some equivalent to KS CPUs.

[twin dew]

And Intel doesn't even try to test the quality of the IMC in CPUs when it bins them for various SKUs.

[sand saddle]

Certainly, but hopefully noone is buying a $600 OC board to run it at xmp

[twin dew]

So buying 14900K doesn't get on average any better IMC than bottom end 14th gen, and possibly no change from 13th gen on the IMC side.

[gilded helm]

Is there any testing from 25W to 100W to see if laptops can expect anything?

[sand saddle]

Well, you need a -k cpu to have unlocked VCCSA

[twin dew]

But doesn't change the silicon quality.
As Intel didn't bin for IMC quality, just core quality.

[sand saddle]

Yes

[twin dew]

Which aren't directly related.

[sand saddle]

But as the voltage is locked, it's directly related to how well your ram will perform

[narrow folio]

how long would it take intel or AMD to bin the chips during production?

does the bin process start with the finished silicon, on the interposer or after the assembly of everything?

[twin dew]

They already do.
But Intel doesn't do IMC binning.

No idea if AMD does IOD binning to match those to best CCDs.

[narrow folio]

Oh, Intel just doesn't care about the memory controller? Bummer

[twin dew]

Yup, they bin for core clock max.
But don't test for IMC, so you can get 14900K with absolute crap IMC.

Because the die had great cores.

[wanton orchid]

marketing target "your cpu is good because the numbers are high on the core" "if bad ram number then blame ram"

[urban tulip]

I would still get new amd cpu if the numbers are close

Because doing a good intel system costs

Well i have amd currently because i didnt want to buy water cooling to have no fan noise

[twin dew]

Mostly with AMD you either need to select good gaming performance (7800X3D), good multi-thread performance (7950X), or both with need for manual tuning at "insane" price (7950X3D).
While on Intel you get almost as good performance in both from same CPU, with "only" the E-core shenanigans causing issues at times.

[dire igloo]

12600KF is $140 rn, great value for a lot of builds.
But beyond that, you'd have to have a usecase for multicore loads to justify getting Intel over AMD (with AMD still being a strong contender)

Also, you don't have to buy a Z-series chipset, B is often fine

[urban tulip]

I meant to beat 7800x3d

[dire igloo]

Yeah no, ain't happening

Not for gaming anyway

[urban tulip]

well it can happen

[dire igloo]

For multicore, maybe - but then there's 7950X3D

[urban tulip]

You just need to try hard enough with ram tunning and maybe core oc

that ram really scales in games

[dire igloo]

You could try the same on AMD

[urban tulip]

Amd has a little problem

That the performance relies more on fclk

I need to see benchmarks from gear 2 oc

[dire igloo]

Intel has the problem that their CPUs run hot as shit

[urban tulip]

Thats why you watercool it

[twin dew]

Why would you need extreme RAM OC?
And not just 6000-6400MT/s?

[urban tulip]

Because it scales in games

[twin dew]

When the actual performance is about the same in most games.

[urban tulip]

Especially tuning ram timings

[urban tulip]

Or is it

[twin dew]

The tests I have seen only had few outliers that benefitted from 7200+MT/s RAM on 7800X3D.
Vs. 1:1 6000 RAM.
Both on tuned timings.

And in others it was the other way around.

[dire igloo]

Adds cost.
Also, contact frame. More cost.

[twin dew]

If the game benefitted from latency or bandwidth.

[wanton orchid]

1:1 6000 ??
3000fckl?

[twin dew]

Think the comparison was to 7600MT/s RAM.

[urban tulip]

You know, x3d is a prime example that cpus in general are memory limited

[twin dew]

UCLK, MCLK, not FCLK.

[dire igloo]

I only have a few screenshots from this video, never watched it tho
https://youtu.be/MOatIQuQo3s

[dire igloo]

Only in some scenarios

[edgy hazel]

Yeah let's argue with the dude who does nothing but overclock RAM all day

[dire igloo]

In others, extra cache does jack shit

[urban tulip]

Not really some scenarios as it uplifts the average a lot

[dire igloo]

That's how an average works

[twin dew]

Doesn't include the now possible 7400+ MT/s on AMD.

[wanton orchid]

game *

[twin dew]

Just compares 6000MT/s RAM on different settings.

[dire igloo]

If it doesn't harm your performance and causes big uplifts in some scenarios, the average goes up

[twin dew]

And then untuned 6400.

[dire igloo]

Doesn't mean that it's better for all usecases

[twin dew]

And getting the high RAM frequencies needs much higher voltages, which then will cause issues later when the silicon degrades with time and voltage.

Because silicon always degrades with time, heat, voltage and usage.
How fast depends on various factors but higher each of those is then more there is, just the magnitude changes.

So being able to make it stable now, doesn't mean it will still be stable in 2 years.

[wanton orchid]

and AMD runs lower on power and voltage when ocing

[twin dew]

But both have their points.
But "On Intel you can get higher RAM frequency" isn't one.

Because that just isn't true at this time.

And higher RAM frequency doesn't always mean better performance, even with both being hand-tightened to the max.

[narrow folio]

when you win the IMC+Board+Ram lottery you can get 8000 on intel...

but I would not hope for that

[dire igloo]

And if you are hand-tightening to the max, you're getting a JEDEC kit anyway which costs much less

[twin dew]

And the board absolutely needs to be one of the 1 slot per channel OC boards.

When AMD can semi-reliably do 8000MT/s on 2 slot per channel boards with good memory layout.

Not all CPUs, but high percentage of them.
If the RAM can do it.

With AGESA 1.0.0.7b and later with the NITRO training.

IIRC.

Or at least on the one slot per channel.
With higher percentage than Intel.

[narrow folio]

I hope to see HBM stacks next to the CPU in the near future...

probably too expensive

[twin dew]

www.servethehome.com/intel-xeon-max-cpu-is-the-sapphire-rapids-hbm-line/

[narrow folio]

2-4 CCD, 1 IOD, 1-2 HBM v3(e)

[twin dew]

Exists for servers from Intel.

[narrow folio]

Oh, nice

64GB HBM2e as either the only memory, combined with the memory (needs code changes) or as L4 cache

[twin dew]

But seems that V-cache Genoa-X EPYC runs around those in the L4 cache mode?

When the program hasn't been optimized to manually use the HBM.

But that might be bad take, haven't found the source comparison yet.

https://www.phoronix.com/review/amd-epyc-9684x-benchmarks

[narrow folio]

either way, both options would be prohibitively expensive (for me)
will take years to maybe come to the desktop

[twin dew]

But anyways, ordered the two TP-LINK TL-SM5110-SR SFP+ modules and 15m OM3 fiber cable with LC/UPC dual-connectors on each end.

[narrow folio]

Hope it works out

[twin dew]

Would have wanted to wait in some way for money reasons, but could order inside Finland now with the modules being on 5e discount per right now.

[twin dew]

Shouldn't have problems.
The TP-Links seem to be standards compliant, and the switches seem to take any standards compliant modules.
Didn't get the little cheaper Ubiquiti ones as couldn't find any mention of using them in anything but Ubiquiti switches.

And they like to vendor lock in some things AFAIK.

Also got Cat6a shielded extension block.
Going to try to just shorten the current cable going between the desk and the closet in suitable place, put new connector and continue with new cable, to reuse the already installed cable.
Two sections of shielded Cat7 cable and shielded "Cat7" chinese RJ45 connectors and the shielded Cat6a extension, for just 1Gbit/s speed.
The cables I have done work fine on 10Gbit/s so far.

[edgy hazel]

4 weeks till I get my metal fans 😦

[pure karma]

everyone hates slow shipping

[sharp oasis]

Random question. Would using Vram in phones be a idea?

Hold on

[pure karma]

also how is it allready almost december my birthday is in 12 days allready bruh...

[sharp oasis]

I'm prolly wack

[dusk rivet]

amazon done spoiled me with overnight or same day shipping

now 1 week shipping feels like an eternity

[jagged snow]

High power consumption

[sharp oasis]

Yeah I see

[jagged snow]

I can barely get anything under two sometimes

[sharp oasis]

Now at least

[sharp oasis]

Give or take a week at least for me

[dusk rivet]

ordered some clothes, says 4-8 day shipping, i hate that it has a wide window

[sand saddle]

It took me a week to realize that my payment did go through lol

[sharp oasis]

Tho is it possible for power efficent Vram. Or is that just not a thing

[jagged snow]

It is

And phones don't really need it

[sharp oasis]

As I'm thinking just for the sake of thinking. Is the lpddr5 in my phone is roughly half the bandwidth of my Dram on my desktop

And I remember the ps5 uses Vram as a "unified ram"

So to speak

[dusk rivet]

apple’s architecture is confusing as hell

[sharp oasis]

So my brain went why not apply the same concept to phones

[sharp oasis]

I mean it in the sence the ps5 uses Vram as ram and Vram at the same time

[dusk rivet]

i thought i had a pretty decent grasp on computers, then looked at apple specs and was like “wtf is this”

[dusk rivet]

yeah ik, the unified memory just reminded me of apple’s stuff

[jagged snow]

The gpu is on the same package as the cpu and memory

So not much benefit to having two sets of memory on that die

[sharp oasis]

Yeah ofc

I mean it is probably unreasonable to need excess of 50gb/s of bandwidth on a phone

[jagged snow]

And power consumption would be too high with higher bandwidth memory at this time

[sharp oasis]

Tho I wonder what could the difference be if somehow we soldered GDDR6 chips to a phone

I know it's completely incompatible

[twin dew]

Non-G DDR is lower latency (better for CPU)
GDDR is higher bandwidth (easier to get high enough bandwidth for GPU).

[dusk rivet]

apple just variably assigns memory to ram or vram right?

[twin dew]

If you had same bandwidth DDR5, it would be better than GDDR6.

The GDDR as the only memory in consoles gimps the CPU performance in many cases.

[sharp oasis]

Tho can ddr5 be at the same bandwidth as the GDDR6?

[twin dew]

Yes, just need more channels.

[sharp oasis]

Fair

[dusk rivet]

does the motherboard dictate dual channel vs quad channel or is it the cpu, or both

[twin dew]

Remember that normal DDR5 currently is 128bit.

[sharp oasis]

Tho my question lies in Lpddr5 I'm assuming LP stands for low profile

[sharp oasis]

Oh shit forgot abt thay

[twin dew]

CPU has to have enough channels, and then MB of course needs to have the traces and slots.

[dusk rivet]

so both would have to support quad channel

[twin dew]

So CPU is the first limit, and then some MBs on the server chips with up to 12 channels need to have the channels implemented.

[twin dew]

Low Power

[sharp oasis]

Apologies

[dusk rivet]

are there gaming motherboards that support quad channel, and are there gaming cpu’s that support it

[twin dew]

Lower power per bit transmitted, but needs to be soldered to keep the signal integrity.
The RAM slot has too variable imperance etc.

[sharp oasis]

Ahh

Thus making phones that have interchangeable ram impossible

[twin dew]

No, because only Threadripper etc. HEDT up support more than 2 channels.

[twin dew]

Usually the soldered on RAM on thin&light laptops is also LPDDR.

[dusk rivet]

got it, so we’re stuck with dual channel

[twin dew]

Because more channels needs more die area, and bigger socket etc.

[sharp oasis]

I've seen chime books with it

[twin dew]

Memory channel needs LOT of pins.

[sharp oasis]

Chromebooks*

[twin dew]

About 100 per channel IIRC.

[dusk rivet]

is it a possibility for the future?

[twin dew]

72 just for the data lanes.

Ok, more than 100, think it was closer to 150.

[sharp oasis]

Do phones benifit from memory latency

[twin dew]

Up to the CPU manufacturers.

[dusk rivet]

threadripper is single cpu only right?

[twin dew]

We did go from one channel to two long ago.

And there have been 3 channel, 4 channel, 6 channel, 8 channel and 12 channel CPUs in the HEDT, workstation and server markets so far.

Also for the LPDDR, there might be CAMM standard variant coming with LPDDR support.

[sharp oasis]

Tho a mobile cpu has quad channel ram.

I never figured as such

But damn that's kinda cool ig

[twin dew]

4x 32bit, same as current DDR5 CPUs.

[dusk rivet]

guess i’ll be sticking with the 2x32gb ram i picked out then

[twin dew]

Because DDR4 was 64 bit, DDR5 is 2x32bit subchannels per DIMM.

[sharp oasis]

Ahh

[twin dew]

So 4 channels on phone is 128 bit.

Been that way forever for LPDDR.

[sharp oasis]

I know nothing about what goes into phones lol

[dusk rivet]

i know how to use my phone, that’s about it

[sharp oasis]

Well now I know something

[dusk rivet]

phone hardware be weird compared to desktops

[sharp oasis]

Idk if I'd call it wierd, just very diffrent

[dusk rivet]

desktop hardware is easier cause i can literally point at my pc and go “that’s X component”

[twin dew]

But nothing magical about GDDR, just optimized for bandwidth with less complex memory controller (less die area per memory bus width)

With much higher latency instead at same frequency.

[sharp oasis]

Ahh

Idk I consider the pixel 6 to be semi modern, but it makes sence that it's got about 1/6th the power of my desktop cpu

And whenever my sister plays her games on it VS my desktop some of them don't load properly

Roblox games are sometimes really poorly optimized for mobile

[dusk rivet]

we’ve kinda reached a point where it’s hard to downsize hardware to fit more powerful stuff in the same or smaller space

much smaller and you have quantum tunneling issues

[sharp oasis]

Not really. No, mediatek has crammed Hardware accelerated RT into a mobile chip

The dymensity 9300 iirc

[twin dew]

And how is that relevant to Irohs statement?

[sharp oasis]

Still in a phone

And based on benchmarks rivals low end desktops

Hold on ill go grab a benchmark

[twin dew]

So?
You can add anything if you are willing to pay the die size penalty.
Up to the reticle limits of the process.

The Nvidias top end dies are hitting that.

Anything below is just question of how much you are willing to pay per chip.

[sharp oasis]

Fair

Tho to be fair this is quite remarkable

A little behind top bench

[dusk rivet]

maybe quantum computers will eventually help, but that’s a long ways off

best quantum computers only have ~100 qbits last i heard

and they’re the size of rooms

[sharp oasis]

I think snapdragon 8g3 is ever so slightly faster

[twin dew]

And need cooling close to absolute zero IIRC.

[dusk rivet]

mhm

[twin dew]

Reason for the size.

Yup.
The actual "core" isn't that large, but the cooling requirements make the whole thing huge, with most of that hidden in other rooms.

So the "room sized" isn't even the whole thing in a way.

[dusk rivet]

traditional computers started out that big, hopefully they’ll find a way to downsize

[twin dew]

Problem is preventing random flips just from thermal noise.

[dusk rivet]

qbits are super fascinating to me

[twin dew]

Some diamond substrate based thing that is working for few qubits in room temperature?

All others basically need liquid helium cooling.

With liquid nitrogen precoolers.

[dusk rivet]

helium has no solid form doesn’t it?

[twin dew]

In closed loop systems.

[twin dew]

According to our current understanding, no, at least not anywhere near "normal" pressures.

[edgy hazel]

Not unless I'm involved

[twin dew]

With pressure, yes:
https://en.wikipedia.org/wiki/Helium#Solid_phases

[charred pewter]

so.... that 'new' replacement enclosure that Sabrent sent me ... the one that has capacitors and an older serial than my 'good' enclosure ......... well. its crap too. it doesnt seem to have as many mounting issues, but it did randomly drop the drive during a backup once, which corrupted the entire backup on the drive.

then it had issues with corrupting a second backup on it ...

so i went back to my old usb-a to sata plug cable adaptor, and the backups are working just fine to the drive (after reformatting)

yeah, im done with Sabrent

3 fucking pieces of crap... 1 good one (must be a fluke i got a good one LOL)

[wanton orchid]

tbh sabrent is not specialized in usb devices AT ALL

[charred pewter]

clearly

[edgy hazel]

I think he noticed

[wanton orchid]

the usb conversion layer is probably at best very brittle

[jagged snow]

Evidently so

[sharp oasis]

@twin dew you were saying that gddr has bandwidth at expense of latency. Does my gpu benifit at all by using the tightened timings option in the adrenaline software panel

[twin dew]

Probably to some extent.

[jagged snow]

As long as it doesn't create instability, yes

It's an optional setting because it's not guaranteed to work

[sharp oasis]

It's stable to what I need but it's a one click option.

I was wondering if it made any diff with It on or off

[jagged snow]

It's just a slightly more aggressive timing preset compared to the defaults

[edgy hazel]

https://tenor.com/view/just-try-it-alex-engvid-give-it-a-shot-make-an-attempt-gif-26267862

[sharp oasis]

Ahh

[twin dew]

But the Sabrent USB thingie was probably just standard Chinese design, with their logo in the original version and now not even that.
Same thing bit Fractal Design in ass with the Torrent fan controller.
And the NZXT H1 PCIe riser.

Outsourced design in both latter cases to factory related design teams from China.

With some standard USB-SATA chip.

Which might then have gotten Chinese reverse engineered copy that is then used in the later variants.

Yes, counterfeit chips do exist, and are big problem in some areas.
Almost compatible copies with fake markings etc.

And if the production line doesn't have Sabrent QC people on it all the time, doing random sampling from every batch, the chinese contract factory WILL change the parts.

[charred pewter]

right

i even noticed the 'new' models used a 25mhz chip where the old was 30mhz .... a tiny little change

[twin dew]

Pretty standard operating procedure, produce what was ordered exactly for the first batch.
And later ones will be whatever if they think they can get away with it.

And keep eye out at all times if the orderer QC is starting to slip as "unneeded".

[maiden coyote]

It does.. at absolute zero.

[twin dew]

Nah, like I linked, needs multiple atmospheres of pressure even then.

Or maybe actual exact absolute zero would work.
But what we can get doesn't do it yet.

Lowest so far seems to be 38 picoKelvin.

[maiden coyote]

exact absolute zero will work.. it pretty much has to because that's what that means.. no thermal energy

but that's also impossible to reach.

you can get closer and closer, but no matter how you do it, you'll always be .01 above zero

[charred pewter]

talking ONI again?

[edgy hazel]

brother it's snowing outside

[charred pewter]

aye lol ...

[dire igloo]

My immediate reaction leaving work today: "Why's the rain falling so slow?"
Turned out is was this weird snow-rain mix

[edgy hazel]

It's full snow here

And it's expected to snow all night

And I need to use public transport tomorrow morning

I guess I won't

[twin dew]

I have forgotten to put the radiator covers into place on my car and it was too cold today for the engine to heat up properly.
Too efficient cooling, would need to check the engine coolant thermostat at some point.

Always forget to do that during summer.

[dusk rivet]

that’s called sleet

[edgy hazel]

what u call my mom?

[dusk rivet]

large gravitational mass?

[stray badger]

But it is still autumn is it not?

[edgy hazel]

the first snow marks the end of autumn sry

I don't make the rule

[dusk rivet]

it would forever be autumn where i’m at then

we don’t really get snow

[edgy hazel]

just wait a few years

climate change will bring ya snow

[dusk rivet]

gotta wait for the earth to leave this ice age, then start another one, then we’ll get snow

should only take a couple hundred thousand years

maybe million, i don’t remember how long the cycle is

[edgy hazel]

we're not in an ice age aren't we?

[dusk rivet]

iirc we’re in the tail end of an ice age

something about we still have glacial ice?

yeah we’re still in an ice age

[edgy hazel]

I'm trying to find something proper, but between US sites blocking the EU and climate change deniers it's kinda hard to find something real

yeah we're in one

[dusk rivet]

in a “warm interglacial” period

[edgy hazel]

bro I don't wanna go into normal age then

this shit is already too hot

[dusk rivet]

toasty

[dire igloo]

Meteorologists consider 01.12. the start of winter - not too far off, margin of error

[dire igloo]

Watch us any% speedrun this shit

[edgy hazel]

talking about dying, love Australias official Fursona

[dapper cedar]

Hi somebody wanna play Satisfactory?

[low scaffold]

Hello there totally awesome user! You can checkout #looking-for-group-old and check the pins on how to find folks to play with. <3

[edgy hazel]

you failed 😦

gl next time

[dire igloo]

Their WHAT?

[edgy hazel]

yeah they comissioned it for the Olympics but didn't know what furries are and the artist they got did

yeah they did it for the 2020 olympics

with trading cards, coins and all

[dire igloo]

Nice

[abstract hawk]

@jagged snow my friend responeded and I basically got "simply for thermal reasons" which makes sense and was my thought I just thought there was more to it

[jagged snow]

There is definitely more to it than that

[abstract hawk]

ah

id love to hear if you have time

[jagged snow]

Yeah, give me a minute

[abstract hawk]

sounds good thx

[dire igloo]

@twin dew what's the current "guaranteed" stable maximum speed for AM5 DDR5?
As in: for someone who doesn't want to bother with OC or tuning, what would you pick?

[jagged snow]

It's actually heavily dependent on the board

For the most part, 6000/6200 are a good bet

[dire igloo]

Yeah but new AGESA

I wanna know if we can safely recommend 6400 or 7200 to people who are minmaxing

Wtf am I talking about, minmaxers get recommended to tune

[jagged snow]

Probably not as a "no hassle" solution

@abstract hawk wattage by itself isn't really an indicator of performance, but as a general trend the performance per watt increases with each generation, and lately the wattage has also increased. Power consumption pretty much directly translates to thermal energy, and so more power consumption does also equal more net heat. In the case of a laptop, the lower power consumption is limited by the fact that it's not very easy to move that much heat in such a small area, as well as the fact that you're already down to half an hour of battery life at that level of power draw.

However, if you take the same piece of hardware and limit it to half the power draw you'll still get more than half of the performance since you can drop power draw significantly with a slight reduction in voltage and clock speeds.

[sand saddle]

We can't as 6000+ xmp won't set the rest of the settings correctly

[jagged snow]

Furthermore, laptop hardware that shares a name with its desktop counterpart is not actually the same hardware. A mobile 4070 is a completely different gpu die than the desktop version, so there's very little performance correlation

[abstract hawk]

i thought it would be the same die... that is quite intresting to me. i thought it would be the sorroundings that would be different but similar core components

that is very intresting. thank you

[jagged snow]

Yeah, for example desktop 4070 uses the ad104 and the mobile version is the ad106

[abstract hawk]

hmm

dont know what those exact numbers mean unfortunately

[sharp oasis]

Higher = worse die @abstract hawk iirc

[abstract hawk]

hm ok

thx

[jagged snow]

That's the same die as the desktop 4060ti(and apparently some versions of the 4060, although most use the ad107)

[abstract hawk]

ahh ok

[gilded helm]

Laptop naming gets even more ridiculous at the top end

[narrow folio]

8>16 = apple

[edgy hazel]

isn't the RAM integrated into their processor?

[gilded helm]

I wonder if Apple will consider CAMM? There are valid performance reasons it would be good for laptops over SO-DIMMs, especially for megaAPUs. But Apple probably doens't care.

[edgy hazel]

bro

every company is

"they just want your money" no fucking shit sherlock

[gilded helm]

Yeah, there's some amount of markup I'd be willing to pay on SSDs, RAM and whatnot when buying things. I'm pretty lazy, and don't have a wife to nag me about stuff like that. But damned if I don't immediately shout F THAT when browsing pretty much any tech products from any company and see how much that stuff costs to upgrade.

Not if you want to have it already inside the machine when it arrives

Also, it's hilarious to me how bad Lenovo's SSD options still are for their laptops. Still only 1TB per M.2 slot.

Validate some 2TB and 4TB options and upcharge people. What are they waiting for?

It's definitely a great time to build.

Yeah, 4TB SSD prices fell through the floor

I think the bottom I've seen on good 4TB drives was around $140?

Whatever the bottleneck is on making 8TB cost effective, it doesn't look like that's happening any time soon.

[dire igloo]

$152 is current pricing, so we're close

Temagroup MP34

[gilded helm]

5? How does it divide up the lanes?

[edgy hazel]

Magic

[gilded helm]

Once we get PCIe 6 I can imagine SSDs just staying at gen 5, and anyone who needs more will just go to those full x16 AIC SSDs.

[languid gulch]

ok, super dumb question time

[gilded helm]

I kinda hope with PCIe 7 or some we can get to a point where all ports on a laptop are the same.

[wanton orchid]

why go to full aic when you could just go to pcie 6 x4
it would blow any storage need anyway already

[languid gulch]

need a bit of help decoding my

www.gigabyte.com/us/Motherboard/B450-AORUS-PRO-WIFI-rev-1x#kf

& what would be the best option for adding in a 2nd NVMe drive. top NVMe is currently occupied, along with 3 of the 4 sata spots. would just slotting in the 2nd NVMe right to the board be the right move, or should i grab an NVMe-to-SATA adapter & swapping out one of the smaller capacity 2.5" SSDs with it be the better option so that i'm not splitting PCIe lanes

[wanton orchid]

splitting pcie lanes ?
usually splitting is a static thing

[gilded helm]

I would trying benching it with the lanes split and see how much IOPS and 4k QD1 stuff suffers.

[wanton orchid]

4k will not suffer

it's latency based
you don't need 4 lanes for 4k qd1

[languid gulch]

it'd be mainly as a game drive, not OS or anything that needs to be super high speed, just better than spinning rust

my case has 2x 2.5" cages on the back, so i've been kinda tempted to grab NVMe/SATA adapters since it looks like they're not going any bigger than 4TB with 2.5" SSDs now

[gilded helm]

Hm. Game loads are primarily CPU bottlenecked anyway. I think split lanes will be fine.

[wanton orchid]

why would you want to go sata if you can go nvme though

[languid gulch]

just for extra capacity that's not too slow

[wanton orchid]

latest motherboard's M.2 ports don't support SATA mode anymore

it was only meant for easy M.2 SSD marketing for storage manufacturers

[gilded helm]

Bolt EV 2LT looking very good, btw. And I have multiple dealers in the area to pit against each other, so nobody is even trying to get away with markup.

[languid gulch]

nice. hopefully it'll be a good match for what you need

[gilded helm]

There's a chance I'll be one of the people who sell their EV for more than they paid, thanks to subsidies.

Your crystal ball is as good as mine, but the resale prices of Bolts look pretty good.

Honestly I just wish it had better suspension. Buit it's not a sportscar.

But it even throws in stuff like a heated steering wheel, it's pretty feature complete.

Alt fuels like hydrogen are probably only going to work if someone like a trucking company builds out the whole system for themselves.

Garbage trucks around here tout they're running on clean burning natural gas in big letters on their sides, but that's never trickled down to consumer vehicles.

[edgy hazel]

alt fuels are too late compared to electric unless we discover something insane soon

[languid gulch]

kk, NVMe drive ordered 🙂 thanks for the help. Crucial P3, duplicate of what i already have 🤣 so i could maybe raid them if i wanted to in the future

[edgy hazel]

skrrr

[languid gulch]

somehow the black friday sale price on it is more than i paid for the first one 🤣

[gilded helm]

Solid state batteries feel real enough at this point that hydrogen and the like aren't going anywhere in consumer vehicles. Compared to fusion power and other perpetually "soon" techs.

There are solid state batteries that work, it's just a question of making them at scale. Not like fusion power and self driving which have never been made to work 100%

I've lost access to the fuel efficient car I was mostly borrowing, so now I'm both putting miles and burning tons of gas in my old convertible. So I'm that much more motivated to get into a Bolt. Or something. I don't rule out a different car, but the price of the Bolt is so tantalizing.

I saw someone with a limecat profile pic the other day, and I was amused by how many people didn't recognize it.

[narrow folio]

www.servethehome.com/ayar-labs-teraphy-is-terrifyingly-quick-with-fiber-directly-plugging-into-chips/

4Tbps bidirektional optic. Nice

at only 10W per port too

[stray badger]

They fuel up at the landfill, which is where the natural gas comes from

[abstract hawk]

Clean natural gas simply means less pollution in the form of nitrides and stuff(that might be the wrong word) Still produces co2. Clean natural gas is not saving our planet

[gilded helm]

Right. "Clean burning" is almost always just green washing.

Same thing when they were trying to tout "clean coal"

[charred pewter]

i prefer to white wash my coal generator so it is clean burning

[abstract hawk]

Yep. Iirc it was a process that turned coal into methane and burned that, correct?

[gilded helm]

I never looked into it beyond immediately rolling my eyes

[abstract hawk]

Fair enough

[gilded helm]

But people eat that nonsense up unfortunately

[abstract hawk]

I wish there was less negativity against nuclear especially in countries that have the cash flow required

Or renewables I don’t care

Anything

[gilded helm]

Pretty much. I don't think nuclear will ever get the mind share. Ukraine probably set things back another decade, and as a result people will keep judging nuclear power based on decades old designs.

And you still have people like John Oliver putting out sensational nonsense about nuclear.

[abstract hawk]

https://tenor.com/view/emoji-gif-24664106

Even in Europe there is negativity especially in Germany I think they have zero nuclear

[winged valley]

Next week Im going to attempt to revive a 32GB DDR4 DIMM that's throwing a whole bunch of errors in memtest (across 2 different machines)

Im going to do this by replacing the packages, theoretically that should fix it and is much cheaper than replacing the whole DIMM

[gilded helm]

It'd probably take someone like Amazon deciding to vertically integrate their power infrastructure for data centers to get new nuclear going.

[abstract hawk]

Maybe it’s just me nerding out but nuclear is the coolest energy production besides fusion

[gilded helm]

Which is also why Satis nuclear is kinda sad to me. Because it's probably the only thing in the game that's represented worse than it is in real life.

All the other metallurgy and manufacturing is fantastically good because of futuretech.

[abstract hawk]

For me I don’t even think about it that much but you are kinda right. Those big towers are not where the reaction happens, there is not constant deadly radiation, but then again jt is a game and game design

I just haven’t done nuclear because I just make basic allu production and also nuclear seems complicated

[gilded helm]

Well, you don't need nuclear. The game gives you plenty of oil. You'll come out with a similar building count per unit of power, but oil based power is vastly simpler.

Mostly my issue is the game introduces recycling spent fuel, but then stops short on casking the waste as an end point.

[abstract hawk]

People who think nuclear waste is a bigger issue than co2 are insane imo

[gilded helm]

Three phase nuclear that introduces mox fuel rods as a third step to make things even more complicated could be fun, and then in the end you can cask the waste. Waste casks being permanent would be fine.

[jagged snow]

If anything, I feel like satis has a semi-reasonable take on nuclear

I'm glat there's not a meltdown system like so many people clamor for

[abstract hawk]

Unfortunately I still wouldn’t do this since just the fact that the items are permanent bothers me

[gilded helm]

RIght, that was a hilarious oversight of John Oliver's nuclear video. He sensationally explained how all the nuclear waste ever made in the USA could fill a few stadiums I think? Without explaining why that's a bad thing, because frankly that's not even a lot of space, and we don't even do recycling.

[jagged snow]

As you said, the biggest misrepresentation is that they present the cooling tower as being the location of the reactiona nd the fact that there's so much radiation everywhere

[jagged snow]

The us does do a small amount of recycling for arms manufacturing, but other than that yeah

It really wouldn't take an insane amount of space

[abstract hawk]

lmao

do you mean recycling in general?

or specifically nuclear recycling and using heavy water based nuclear reactors

[jagged snow]

Specifically, I meant reuse of depleted uranium

[gilded helm]

It's hard to say because a lot of designs are theoretical, but my understanding is regulating nuclear power in such a way that maximal recycling were required, you'd end up with trivial amounts of waste. Potentially no waste if we keep finding new ways to recycle it.

[abstract hawk]

ah ok i missunderstood

[gilded helm]

https://www.amazon.com/dp/B08T5QN2TR

I'm looking at picking up a few of those unless anyone has spotted something similar for less?

[rustic panther]

I think the radiation leaking makes sense

I don't think Ficsit really cares much about that when there's costs to be cut

[narrow folio]

Have two of those, working great so far

In hindsight I maybe should have changed one to a 100W version to charge my laptop with more power, but if I game with it I can always use the brick that came with it

and the 240W versions of those chargers that a next gen gaming laptop could maybe support are still quite expensive

[gilded helm]

Well, the 240W one is cool, but it's gigantic

[narrow folio]

yeah, the 715 has nano in its name for a reason 😁

it all depends on the requirements. the single USB-C output may be too limited for some use cases

[gilded helm]

I like the 65W one because it can probably fit over any outlet without blocking anything else. If I'm going to go for something big liek the 240W one, I'd also probably want it to do wireless charging.

I guess maybe upping one to the 100W would be good though

This one? https://www.amazon.com/Anker-Charger-Compact-MacBook-Pixelbook/dp/B0C4DGBHY2/

[narrow folio]

looks like the right one to me, amazing how Amazon often fails to write the correct product code into its offerings

Ok, my bad. even anker themselves don't write a usual 3 digit number on their own website.

www.anker.com/eu-en/products/a2343-100w-usb-c-charger

[gilded helm]

Does anyone have that 100W Anker? How does it work with power strips that have more standard spacing?

[dusk rivet]

i might go sff for the upcoming pc build
https://pcpartpicker.com/user/irohlikestea/saved/yKcJwP

[gilded helm]

Why SFF?

[maiden coyote]

you definitely don't need any of those noctua fans if going sff...

you'll have 6 fans.. 3 on the gpu, 2 on the aio and 1 in the psu.. that is all the room that case has for fans.

[urban tulip]

modern powerplants cant even physically explode

[pulsar garden]

they can if operated wrong enough or if something breaks

ie: water pumps, control rod control, valves, coolant leak

[urban tulip]

they for sure can still meltdown

[pulsar garden]

or an electrical failure

[pulsar garden]

entirey untrue

a reactor always could meltdown

its just extremely unlikely with the technology we have now

[urban tulip]

what im saying that its extremaly unlikely that it can explode like chernobyl

[pulsar garden]

fun fact: some people simulate nuclear meltdowns irl

[urban tulip]

chernobyl was a walking design flaw, operated like diy project

[pulsar garden]

yes

[pulsar garden]

its so they can create a substance called corium

its a byproduct of nuclear meltdowns, and is roughly half the temperature of the sun

[urban tulip]

are the thorium reactors real? I lost track of the progress on them

[urban tulip]

what do they need it for

[pulsar garden]

corium has only been created 4 times accidentally

[pulsar garden]

for shits and giggles

[pulsar garden]

twice at chernobyl, once at fukushima and one more at another meltdown

[orchid verge]

anyone well versed in audio tech by chance? looking at changing out the wave xlr as it seems to have tons of issues. currently i have sm7b into a CL1 into the wave xlr. then headphone out from wave xlr into my bose companion 20 speakers.

looking for either a wave xlr alternative, or a USB powered speaker setup + wave xlr alt. etc

am a big fan of the wave link software for the wave xlr with the audio routing and easy ness of it. but open to swapping to something more robust if its more stable

[sharp oasis]

Yes ish

[twin dew]

No, 6000MT/s is max you can use with "just enable XMP".
6200MT/s and you also need to manually force that UCLK:MCLK 1:1.
And 6400MT/s isn't still guaranteed even with the new longer more complete training.

The 1:2:s work now on XMP now, but the training times are much higher.
So wouldn't recommend for that reason.

[dire igloo]

Alright, thx

[twin dew]

And 7200MT/s is too low to be better than the 6000-6400.
Need 7600MT/s for that IIRC.

With tight timings on both.

The 7200 loses still because of the 1:2.

So 6000 is for everyone, 6200 needs one extra tune, 6400 and up is more of hobby territory, and when going above 6400, you need to aim for 7600 or better as next step.

[languid gulch]

when the US was first starting up its nuclear program, the department of energy made 2 proposals: thorium & uranium. thorium basically can't melt down, but "takes up more space" & needs some weird chemistry. uranium was something that was basically ready to be built but could melt down fairly easily without an absolute crapton of preventative measures. then the defense department found out that uranium reactors could be used to make materials for nuclear weapons & put their thumb on the scale

[twin dew]

Need to find if I can buy one of the left side type, instead of the RJ45 type on right...

Only up to Cat6, so maybe I will just do the extra work then.

[twin dew]

Connecting RJ45 is much slower than doing punch block.

[gilded helm]

Ended up getting two 100W AC warts and a small AC/USB hub.
www.anker.com/products/a2343-100w-usb-c-charger
www.anker.com/products/a9128-6-in-1-charging-station

My wish would have been for the AC/USB hub to have four AC (well spaced!) and wireless charging, even by adding some bulk. But so it goes.

[narrow folio]

I can recommend a Focusrite Scarlett Solo or 2i2 from my experience with the Scarlett series, but maybe the Vocaster series may be an option for you.
Arturia also has nice small interfaces, have the Minifuse 2 and it's working like a charm

I can't say anything good about USB powered speakers... never heard one sounding good (enough for me)

I use Nubert A-200 1st gen, no problems since years, but cost is a thing...

a friend has some cheap logitech USB speakers and those are okish for system sounds and bit of youtube background listening.
it depends on your expectations (how good it should sound), requirements (connections), budget and space (you have to place the speakers)

[dire igloo]

Some audio guy I talked to once said that Focusrite interfaces are overpriced and mediocre, he recommended M-Track Solo

Edifier M4 or any 1280 variant are his speaker recs (or creative Pebble)

[narrow folio]

going away from USB power opens up the whole studio monitor space and IMO that is way too big to get an unbiased opinion

but the Edifier M4 are an often seen rec because of the price/perf ratio is really nice

on the other hand you get a PreSonus Eris E5 XT for about the same price

oh, wait.. my bad, no you don't. price was for a single speaker 😁

just some speakers I have heard and may be an option:
KRK Rokit RP5 G4
Adam T5V
Yamaha HS5

[narrow folio]

talking about audio routing it usually comes down to Voicemeeter

if you go for a soundinterface you usually get an intro edition of a DAW, often Ableton Live Lite.
but it is overkill to have a DAW running just for routing sound and maybe some slight FX like comp/gate😁

[twin dew]

@surreal sun Any specific reason for planning 7900X3D?
When 7800X3D should be better and easier to use if you just play games.
And 7950X3D is better if you also do multithreaded work on the CPU instead of just games?

[surreal sun]

yeah, sorry. It was a mistake. I’m gonna get the 7800X3D. My apologies. If your interested in my list it’s here https://pcpartpicker.com/list/2gsrGP

[surreal sun]

I may update to 64GB RAM.

[twin dew]

Only matters if you need more than 32GB.
There isn't real performance difference between 1R and 2R sticks on DDR5 anymore, like there was in DDR4.

[surreal sun]

yeah, i did read on that how there really isn’t any difference

does my build look solid?

anything you think i could change to make better for the 1500 USD $ price range

[dusk rivet]

i was like “how is this build so cheap” then realized half the parts don’t have a price tag

[twin dew]

I haven't looked and won't at least today.
Might be more people interested, but mainly the 7900X3D you originally mentioned seemed so weird selection.

[dusk rivet]

i’d take a look at the h6 flow @surreal sun

[surreal sun]

yeah lol! Idk why Partpicker isn’t picking them up or something

[dusk rivet]

bit smaller, bit cheaper

[dire igloo]

Meh

[dusk rivet]

still dual chamber tho

i personally prefer it to the h9, but it depends on what you need

[dire igloo]

Phanteks NV5

[surreal sun]

one of the reasons i picked H9 flow due to it comes with 4 fans. where as the H6 comes with 3. so i thought there was a advantage there.

[dire igloo]

That's because they're out of stock at many places

[dire igloo]

Those fans are shit tho

No advantage really

[dusk rivet]

yeah never use nzxt fans

[surreal sun]

Okay, so i just am gonna switch that out to the H6 flow.

A lot of people (not here), told me to get the 4070 TI but i’m tbh really glad i didn’t get it,

Idk if these fans were good, but i’m considering going with them. https://www.amazon.com/gp/aw/d/B09X1G8BDC/ref=ox_sc_saved_image_3?smid=A25D34XLVTWT9M&psc=1

and later on, i’ma get a AIO cooler for CPU

[dire igloo]

NV5 is better

You should replace stock fans on Nzxt cases anyway

[dire igloo]

Overpriced.
Consider TL-C12CW-S or TL-S12W-S or TL-S12W

[dusk rivet]

i imagine they just like the look of the nzxt case

almost went with it for the same reason

[dire igloo]

Can still make a suggestion based on hardware quality

[dusk rivet]

true true

nzxt’s primary allure is that aesthetic aspect

and they do make beautiful cases

formd t1 v2 tho

so smol

[dire igloo]

Honestly, imo NV5 looks better than both H6 and H9.
It's just a matter of aesthetics

[dire igloo]

Jonsbo TK-1

Super cute, looks great

Like. Look at this

[dusk rivet]

looks easier to build in actually

and look at that lil guy

[orchid verge]

@narrow folio ty. what i meant by usb was audio over usb for the speakers. mainly because the PC is in another room so I wont have access to the 3.5mm ports directly unless I run a cable. I tried an m-audio before I got the wavexlr and the issue was the microphone output came through the headphone out on it so I could not separate it. I have a pair of krk's in my closet, they are just too bulky for my desk.

[dire igloo]

GPU clearance sucks, but hey IT IS CUTE

[dusk rivet]

can it fit a 313mm gpu?

[dire igloo]

Idk

[twin dew]

4-pin to 2x 3-pin adapters are a thing?
If you mean that speaker output and mic input were on same 4-pin?

[dire igloo]

If I ever build a PC for my gf (if I ever get a gamer gf), it's probably that case.
Maybe the white version

[dusk rivet]

given its total depth is 310mm, gonna say no

the gpu would literally be longer than the case

[dire igloo]

Is there a white version? There better be a white version.

[dusk rivet]

so uh yeah, not gonna work with my setup

there isn’t a single 7900xt that small afaik

wait no there is

they’re just expensive

7900xt 276mm, but 900usd

standard 7900xt is like 750usd

[dire igloo]

Yeah, I think that and 4070 ti are the best GPUs that fit

[orchid verge]

i mean treating the audio out and mic in as 2 completely separate loops and control it via PC. in my case, I daisy chain my speakers out of my wave xlr headphone output making it all one chain instead of something like mic -> interface -> pc & pc -> usb sound output -> speakers

[dusk rivet]

gonna stick with the formd t1

gonna go noctua fans, cause i don’t need rgb for em anyways

plus i’m prolly gonna need that fan performance

p12 redux 1700rpm iirc

[dire igloo]

Phanteks T30, Unifans P28, TL-C12C

Much better options either for performance or for value

Noctua just isn't worth it nowadays

[dire igloo]

Could consider Jonsbo D31 Mesh

[sand saddle]

Ping me once you're ready to buy

Or fireworker if I was/am too toxic

[dusk rivet]

i want performance, quiet, and not super expensive

given it’s basically entirely hidden in the formd t1 looks isn’t a priority

[stray badger]

Arctic

[dire igloo]

Fuck toxicity, it's bullshit. Stupid fucking toxic assholes don't have a place here

[dusk rivet]

just a touch ironic

[dire igloo]

P28 is the king of noise normalized performance rn

And Noctua is super expensive for what you get

TL-C12C are also great for value

$14 for a three pack, that's a steal

[sand saddle]

Too many people overestimate their requirements and effectiveness of case fans

You want quiet, get good coolers and make silent fan curves

[dire igloo]

For value, yes. Tied or trades blows with TL-C12C

[sand saddle]

The cpu cooler and gpu model matters a lot more than case fans

[dusk rivet]

going 240mm aio for the cpu, ek nucleus

[sand saddle]

Case fans aren't even 2nd, I'd take a non-restrictive mesh case with shitty fans over a sealed off glass front with T30 fans

[dusk rivet]

formd t1 has no glass

solid front panel, then mesh for like every other panel

[sand saddle]

I'd be happy to discuss your choices once you're actually buying and then I'll argue over your choices

[sand saddle]

I was still on the topic of the irrelevance of case fans performance

[dire igloo]

Bruh, if you want maximum NNP, 360

[dusk rivet]

360 legit won’t fit in the t1

240 is the max, maybe a 280

[dire igloo]

Sucks I guess

D31 should fit

[sand saddle]

It's 3x bigger

[dusk rivet]

really like the sff of the t1

[sand saddle]

Fireworker, a bit excessive with the up arrows eh?

[dusk rivet]

extra desk space is gonna be nice

[maiden coyote]

I'm still confused as to why you want order noctua fans when I'm pretty sure the case can't fit them

you're right about it being small.

[dire igloo]

Better?

[dire igloo]

SWP4

Best 25mm fans out there rn (for NNP)

[narrow folio]

I don't know the interface, but seems like you had direct monitoring on

usually a sound interface doesn't send the input to the output on its own except there is a direct monitoring option

depends on the interface, some don't have the option at all, some have a hardware switch and some can switch it with software

reduces the lag to basically zero, useful i.e. when live recording yourself

[edgy hazel]

Behringer DACs fte

[narrow folio]

I must say that USB Audio to the Monitors is not my recommendation because you can't do much without the PC.
You need a cable to and between the loudspeakers anyway. So I'd rather go the analog way

Behringer interfaces are usable, just don't expect much regarding latency

[dire igloo]

Me is parrot.
Me say M-Track Solo

[narrow folio]

fte? For the emperor?

[dire igloo]

probably typo'd and wanted to say FTD - from the depths

[narrow folio]

Uli makes some very affordable gear👍

[jagged snow]

Interesting how China has recently made big jumps in silicon manufacturing https://arstechnica.com/security/2023/11/hackers-spent-2-years-looting-secrets-of-chipmaker-nxp-before-being-detected/

[dire igloo]

https://vxtwitter.com/SwammyXO/status/1728885232491983194

[surreal sun]

Question. What do you all think. Is the 7900XT better than the 4070 TI?! i’ve had people tell me a lot of mixed opinions and just wanna know what you all think.

[twin dew]

Depends if you need CUDA or RT performance is the absolute must.

Or just want good raster performance.

For first: Nvidia, for second: AMD

[dire igloo]

It entirely depends on your usecase

[dire igloo]

Lemme throw in Nvenc and OptiX

[surreal sun]

I’m gonna be streaming while gaming, so don’t know if that makes a big factor. but i’m gonna play satisfactory, Hogwarts legacy, Farming Simulator, games like that.

I feel like streaming while gaming can make a big factor.

[dire igloo]

Hm. Probably 4070 Ti then for the hardware encoder

[twin dew]

AMD has made significant improvements to the H264 HW encoder AFAIK.

But not sure if enough.
But the B-frame etc. supports are now back in.

[dire igloo]

It's literally what made me recommend a GPU that'd give you less fps but better stream quality

[jagged snow]

Yeah, to my knowledge they're pretty evenly matched this gen

@surreal sun I'd probably reccomend the 7900xtx in this case

[dire igloo]

Sent you a dm with a test stream I did recently

[twin dew]

The AMD H264 used to be crap, but AFAIK not anymore.
And their H265 is AFAIK better.
And AV-1 is about same, if the specific software supports both.

[dire igloo]

For AV1 yes. For the rest, ehhh.
AMD is still inferior but good enough if you don't care about best quality

[twin dew]

But I don't have the eyes for such comparison.
And might need manual tuning of the settings.
IIRC it still doesn't use the B-frame support by default.

Needs to be selected.

[surreal sun]

yeah. it’s not like a stream crazy thing, or have crazy overlays.

[twin dew]

But better to check better sources on that.

[surreal sun]

referring to #off-topic-tech message

[dire igloo]

That doesn't matter

[jagged snow]

They're "close enough that I'd probably still reccomend it. If you really want that quality you'd be software encoding anyways

[twin dew]

Just that it isn't as bad as it used to be.

[dire igloo]

Stream quality doesn't degrade significantly with more complex overlays etc.
It's a matter of good encoder vs acceptable encoder

[dire igloo]

Basically this. Still not great. Still not something I'd recommend to someone streaming regularly.
But good enough if you want to test it out

[jagged snow]

Oh interesting

[surreal sun]

the XTX version is out of my budget. i was just gonna get the XT version. Shoot the XTX is 1000 USD $ where i live, i’ve found 4090’s on sale for that.

[jagged snow]

Oh I see

Then I'd probably still reccomend the amd card here

[dire igloo]

Software encoding takes up significant resources and therefore introduces a giant tail of potential other issues.
Unless you use a second PC for streaming/encoding, I'd stick to a hardware enc

[surreal sun]

Okay, so the 7900 XT great! thanks guys!

[jagged snow]

That's what I was referring to

[dire igloo]

Well, how serious are you about streaming?
How regularly do you stream?

[dire igloo]

That's like top 0.5% of streamers tho

Also much more complex setup, not worth it

Unless you make a living off of streaming

[jagged snow]

Unless you're a very big streamer the quality difference is small enough that it won't matter here

[surreal sun]

max 3 times a week. Usually 2.

and mainly i’m just streaming a game from my ipad, connected to my computer capturing the game.

so mainly the game isn’t even running on my PC, it’s on my ipad and going to my pc and my pc is capturing the gameplay.

so the 1 other time a week, i will stream it will be the game playing on my PC and streaming at the same time.

[dire igloo]

Definitely use software encoding for that

[dire igloo]

I'd say that still makes 4070 Ti valid.

How big is your audience?

[jagged snow]

Really?

[dire igloo]

Depends on the viewership, but I'd say yes.
Sounds like minmaxing the stream is more important than minmaxing game fps

The stream quality hit from the 7900xt is much more significant than the fps hit from the 4070 ti

[surreal sun]

it’s not a lot tbh. i do edit little like 10-30 min videos though. I think the most i’ve had on stream was 75 viewers.

[surreal sun]

Really?! 😳 4070 TI is valid woah 🤯

[jagged snow]

Well, looks like you're back where you started lol.
I don't necessarily agree with the 4070ti but he's also not wrong with the comparison

[dire igloo]

Video editing makes an even bigger point for Nvidia.
I reckon Premiere?

[surreal sun]

I’m sorry what? are you talking about Adobe Premiere? i haven’t decided yet, this is my first windows PC and i’m coming from mac which i was using there video editing IMOVIE

Sorry, i should’ve said it all in one paragraph but that’s everything i do, stream, play games, and edit some little 10-30 min videos. nothing insane.

[narrow folio]

just out of interest, are the en/decoders the same on the 40xx or 7xxx?

as within a series

so a 4050 has the same coders as a 4090?

[twin dew]

Usually yes.
So all AD### dies should have the 8th gen NVENC and same version of NVDEC.
In some cases there are differences of how many times the blocks are there, so how many simultaneous encodes or decodes there can be.

All Navi 3#:s should have VCN4.0 on AMD side.

[surreal sun]

So i’m guessing you’d still say the 4070 TI?

[gleaming shadow]

Humz... crashes.. how do read the crash log to identify the crash?

[jagged snow]

Depends on the game

[gleaming shadow]

Satisfactory

[jagged snow]

Then this is the wrong channel

[gleaming shadow]

i know 😛

but i want to learn to read the log

thats not right on topic is it?

[jagged snow]

It's definitely satisfactory related

Talk about nerd stuff here like science and programming and hardware stuffs. Not Satisfactory stuffs, there are dedicated channels for that.

[gleaming shadow]

fine, its a crash log of a random UE5 engine game 😉

[jagged snow]

But mostly understanding just comes through experience

[twin dew]

In general with stack traces, upper lines are the closest to crash.
And then going down each line is what called the previous line.

And then you just guess based on the file and function names.

[gleaming shadow]

i thought i had it in cluboard but it aint... so cant share right now have to wait for a new one, had 2 in a row

now its fine, did read couple of lines of EOS, thats the epic online service

[twin dew]

But at that point make #1038092680493801533 thread and specifically ask to explain about the crash log.

[gleaming shadow]

aight will do 🙂

[sharp oasis]

@jagged snow

Why am I getting non ping notif about servers discussing topics

It's the discord notif

[jagged snow]

That's really weird

[sharp oasis]

I was not pinged

And yet I'm seeing stuff like that

[jagged snow]

Look at that summary

That's fascinating

[sharp oasis]

It's really specific

[gleaming shadow]

clash of clans?

[jagged snow]

That seems llm generated

[gleaming shadow]

notification settings probly maybe an update?

[sharp oasis]

Could it be Ai being inigrated to discord to watch over convos based on likes and dislikes

[jagged snow]

I would guess it's something like that

[sharp oasis]

As it's not just here

[gleaming shadow]

what brand phone?

[sharp oasis]

Pixel

[jagged snow]

That doesn't seem like a phone specific faeture

[gleaming shadow]

could be Samsungg does crap like that

[sharp oasis]

Other servers are also letting me know of people discussing bdays, gaming on Linux machines, topics of interest

[jagged snow]

I normally keep them off but I'm going to turn on discord notifs and see if I can reproduce

[sharp oasis]

I wonder if it's an update or someth

[jagged snow]

That's really interesting

[sharp oasis]

Tho @jagged snow its spot on with the notif

[jagged snow]

Yeah, it's absolutely correct about the convo

I'm fairly impressed

[sharp oasis]

Interesting

[jagged snow]

Especially impressed that it picked that topic to reccomend to you

Although it does feel pretty wrong to have discord be algorithm-driven

[sharp oasis]

Could discord be attempting to parse user message history and feed it data for an agorithim

[jagged snow]

It looks like that's exactly what's happening

And it's been known for a while that discord has been training llms on user messages

They've released a few different features lately using the dataset

[sharp oasis]

Hmm

Interesting to say the least

[jagged snow]

This is just a change because it's a user-facing algorithm-driven engagement system

[sharp oasis]

I'm midly disturbed that discord knows my likes and dislikes

[jagged snow]

Yeah, that's very interesting

Since it chose that specific convo to reccomend to you and you talk about gpus more than anything else in here

[sharp oasis]

And even if I didn't update discord

[jagged snow]

Do you have beta/experimental enabled?

[sharp oasis]

It probably has this data

[sharp oasis]

I only have developer options enabled

That's all

I only noticed this yesterday with one topic notif about Linux gaming in another server I'm in

[jagged snow]

I'm sure that I've given out enough data on discord that an llm-driven service could pretty accurately make some rather frightening conclusions about myself and my real-life identity

[sharp oasis]

A bit worried about that ngl

[jagged snow]

Considering I've written the equivalent of several novels

[tough owl]

www.androidauthority.com/google-drive-missing-files-3389552/

[jagged snow]

Good one google

[jagged snow]

You'll appreciate this #off-topic-tech message

[sharp oasis]

I wonder if this will expand into other discords as well as I get more message historys

As it seems to need like 3-4k messages at minimum @jagged snow before it starts sending me notifs of discords

[jagged snow]

Huh

[sharp oasis]

As basing some evidence so far.

The only other server I'm receiving notifs for in I have 4.4k messages there

In any other server I'm not having the same effect

So whether it's message count or activity based

As I am pretty regular in here and there

[tough owl]

light work

[sharp oasis]

I have 8k messages here

[open root]

how can i fix a "outgoing reliable buffer overflow" error?

[dire igloo]

Yes

[dire igloo]

I'm lacking a lot of context for this

Ah nvm, got the context now

[dire igloo]

15,880 for me

[sharp oasis]

I'm guessing there probably an activity threshold

[sharp oasis]

So how many messages is that daily I assume

Or I'd probably better assume it's a weekly avr

I wonder if there's patch noted for this

[tough owl]

I have used two accounts to message here 🙂

[jagged snow]

1154 days

So just shy of 30 messages/day

[sharp oasis]

I doubt it's a daily average that determines it

I assume it's using past data to make a activity threshold

[jagged snow]

Although it is worth noting that I was very active in the server for a time, then left it muted for almost a year

[sharp oasis]

Quite clever not gonna lie

So it's probably understaning activity and engagement

Damn.

I mean this has been a part of partnered servers for a long time

With limited metrics

[dire igloo]

Yes, there is. Slowmode, 5s message timer

[sharp oasis]

No not that

The notifs

I assume it's using past data to determine inactivity VS activity in servers

[tough owl]

wccftech.com/china-loongson-3a6000-cpu-reaches-core-i5-14600k-performance-clock-to-clock-test-overclocked-to-3-ghz/

[winged valley]

Google isn't very helpful on this, is memory density the capacity of an individual package?

[edgy hazel]

Very fun work day

[charred pewter]

not fun work day

beating down fraudsters left and right ... holidays suck

[edgy hazel]

Retail?

[charred pewter]

pretty much

online retail, which is worse

[edgy hazel]

Yeah get out of there homie

[dire igloo]

Same

[edgy hazel]

I had a fun 7-8 day

[visual tree]

I still can't forget that email when a customer demanded a "fiber optic connection to 5G internet"

Will haunt me in my dreams

[winged valley]

They just wanted a 5 gigabit line

what retailer do you work for?

[visual tree]

I work for an ISP and had to record customer addresses who wanted fiber optic connection through EU-funded project

Glad I don't have to deal with that anymore lol

I was recording so-called "white addresses" - addresses where fixed broadband speed is less than 30 Mbit's and there is no commercial interest to invest in NGA network

[dusk rivet]

i feel like im missing something cause that doesn’t sound unreasonable

[visual tree]

Some addresses were covered by 4G/5G instead of fiber though because the cost to pull 1 km fiber to a small village with few houses is just too much

[sand saddle]

Oh shit

[winged valley]

Where do i buy memory packages

Theres gotta be some super secret store out there

Aliexpress is sketchy

[sand saddle]

Ram?

[winged valley]

Chips

Yes

[winged valley]

need to buy ram chips (ddr4, lpddr5, gddr6)

[sand saddle]

Why?

[winged valley]

upgrading a laptop and reviving a dimm both for practice

[sand saddle]

Can't help you there, don't know about it

[winged valley]

you got me excited

Wait nvm digikey sells them i'm just dumb

[charred pewter]

you want ram, or ram chips ... ?

or ram sticks ......

i mean, some repair people want the chips themselves... not the sticks ...

[winged valley]

I said chips already

[edgy hazel]

i love chips

[visual tree]

Chio memory chips

[edgy hazel]

nom nom eats your 7900xtx

[sterile plinth]

anyone on I got an html problem!!

as you can see im doing getElementsByClassName

but its literally giving me an element with a different class

[charred pewter]

thats....

wtf

ohhh

you are getting by either or class name, and those have one or the other in it

messageContent__21e69 is the matching class name

HOWEVER that said ..... its not how javascript should be working .... document.getElementsByClassName("class1 class2"); should only match elements that have BOTH .class1 and .class2 on it

im doing some manual tests myself in console on our site... and js is behaving properly... if i put two class names in it, it only matches elements that have BOTH the class

although what i cant seem to do, is get my output to format like yours with the html .... (im using Chrome for the above)

it may be that you are 'flipping open' a matched element, and seeing whats inside it, instead of just looking at the top level of match items? (console can be weird sometimes lol)

[sharp oasis]

that has to be decent lol

[charred pewter]

sure? 🤷 ... 😉

[sharp oasis]

this is just my 5600 and 6700xt tho

im confused

[charred pewter]

doesnt it rank you among 'like hardware reports' ? not against 'the whole world of hardware'

[sharp oasis]

idk

i know its midgrade hardware

[charred pewter]

i think 3dmark does against similar hardware ...... not sure what passmark does

[sharp oasis]

mabey it puts you agaisnt what it has for database

[charred pewter]

seems i dont even have passmark on my machine anywhere lol

[edgy hazel]

passmark is

not optimal

scrolling up and down text in a window really fast isn't really a good benchmark

[charred pewter]

lol

neither is 3d jets made from the 1980s

[edgy hazel]

fuck it i'll run it too

[charred pewter]

😄 yeah, im curious what my result will be

[edgy hazel]

It made the fans spin up

[charred pewter]

now the big box of falling balls was cool to see

apparently a 6700xt is WAY WAY better than a 3080 12gb

cause initial results on my gpu scores are not even half of what connors image shows

[edgy hazel]

The spinning woman 😍

[charred pewter]

passmark is biased on AMD gpus ...

it didnt even get my gpu near 50% usage during any test

[edgy hazel]

The test are like 5 seconds long

[charred pewter]

pathetic 2d score lol

pathetic 3d score

pathetic overall

[edgy hazel]

AYO

[charred pewter]

thats what a 3080 gets you ... jack shit

my cpu/ram/disk are all superior

[edgy hazel]

The fractal thing just gave me the craziest coil whine

[charred pewter]

heh!

[edgy hazel]

Like a laser gun

THE WIGGLE BALL TOO WTF

Wiuwiuwiuwiuwiuwiu

[charred pewter]

WEEEOOOWWEEEEOOOWWEEEEE

im convinced passmark has no fucken clue what to do with my nvidia gpu

[edgy hazel]

Yeah RAM is where I fall apart

[charred pewter]

cause theres no way my gpu is less than HALF performance of a 6700xt

[edgy hazel]

3000mhz DDR4

[charred pewter]

my 3dmark ...

[sharp oasis]

ehst id ut cpu

[sharp oasis]

what

[sharp oasis]

HOW AM I HIGHER

99th precentile

could be old data ig

[edgy hazel]

you have literally a lower score

what are you saying

[sharp oasis]

lower data

[charred pewter]

seriously, i dunno what crap my passmark took to do the 2d/3d tests

[sharp oasis]

my passmark data is old

[edgy hazel]

it's not a good benchmark

[sharp oasis]

tryed 3dmark

my pc isint even considered a normal gaming pc lol

[charred pewter]

heh

[sharp oasis]

Its just about as good as the avreage gaming laptop of 2023

[charred pewter]

my 3dmark and passmark scores directly conflict with each other... so SOMEONE is wrong

[sharp oasis]

sadge

[sharp oasis]

passmark

prolly

[charred pewter]

13% on 2d score... i reran it twice now

its so far into the red its like 'are you on a pentium?'

[sharp oasis]

does nvidia not like what passmark uses for 2d rendering

[charred pewter]

i dunno...

i know its using my gpu, cause the 'usage' goes up to 28% ... fps pins to 144hz/fps ...

i wonder since i have a global fps limiter in the nvidia control panel... its reading results wrong

(because to make gsync/vsync work the best, you set that limiter)

[sharp oasis]

ahh mabey

[charred pewter]

actually, im not finding that limiter in the nvidia panel now ...

i have max frame rate as 'off'

[sharp oasis]

huh

odd

[charred pewter]

OH WELL... maybe this is why i had deleted passmark from my machine 😄

[sharp oasis]

lol

[edgy hazel]

Was about to run a new 3d mark but it's too stupid to scan my system info

[charred pewter]

yeah i dont get it... reinstalled it... turned off all my monitoring afterburner stuff ... it just wont go above 185 on 2d score

[edgy hazel]

So here is my old

[charred pewter]

thing is a broken piece of ass ... 😄

[sharp oasis]

yeah

[gilded helm]

I'm surprised they haven't made a couple more parts of 3DMark free by now.

[sharp oasis]

i guess im on par for my lower budget prefomance

[edgy hazel]

I got it on sale for 3€

[gilded helm]

Some day I'll impulse buy it

There'd probably need to be a sale while I'm testing something I just bought

[edgy hazel]

It's just worth having it

Okay 81C Hotspot at 400W and 2840MHz

[gilded helm]

My only use case for it is making sure the hardware I buy falls in line with everyone else

As long as I can run the Cyberpunk ingame benchmark and whatever combination of other things I already have or are free, I'll probably skip paying for 3Dmark.

Not even because it's a lot of money, I'm just lazy.

[edgy hazel]

good thhing I tried again

skrrrr

How to overclock amd: everything up, voltage down and pray

[narrow folio]

would be interesting where this small spike on the left side comes from.
quite a bunch of 'weak' results would be needed in that range. maybe high temps in the case?

[gilded helm]

https://www.amazon.com/StarTech-com-USB-Type-Barrel-Cable/dp/B009C3CJG4?th=1

shop.usa.yamaha.com/en/p/pro-audio/pro-audio-mixers/analog-mixers/mg06-6-channel-analog-mixer

Something like this is perfectly fine, right? Finding USB to barrel connectors so I can get rid of the single unit power adapters for my devices?

I just need to match volts/amps

[dusk rivet]

spec sheet says it'll fit

Top or bottom 2x120mm

side 2x120mm or 2x140mm based on components

1x 240mm radiator

[winged valley]

this company wants me to pay $573 for 32GB of LPDDR5

nvm, $320. But that's still uh, wow? Feels more expensive than it probably should be

At that point I might as well just pull chips off of a regular DIMM and reball them

[sharp oasis]

I'm guessing bc it's low power ddr5

[winged valley]

Nah I've found other distributors selling similar models (same package type, just small variations in design) selling for as low as $25 a piece

(Ofc they won't sell to me as a consumer tho, since I do this stuff for work anyways boss might be fine with me using our company name to be able to buy this stuff)

[jagged snow]

Must be apple

[winged valley]

Unfortunately its some chinese chip supplier

Not apple

Its 4 64Gbit packages (or, 32GB usable RAM)

Trying to upgrade my laptop

[jagged snow]

Oh nice

If you want to just rip the ic off a module, corsair sells a 32gb 4800 sodimm for 75

[winged valley]

That won't work

I specifically need 64gbit packages

Laptop only has 4 sets of pads for ram

Although all this being said, I'm starting to understand why laptop manufacturers charge out the ass for ram lol

Its still probably too much, but considering they have to use denser than normal packages due to space limitations, it makes sense

[jagged snow]

Ah

I'm not seeing much with packages that large, probably why they're so expensive

All the consumer stuff I'm seeing has 32gbit ics

[winged valley]

Yep

Most laptops I've seen with soldered ram eiher have 8 packages or 4 plus a slot

This just have 4 and its annoying

Now, 16GB is fine, I'm not multitasking as heavily on my laptop as I do on my desktop anyways. I am going to be upgrading it from 4 to 8GB of VRAM tho and that price makes more sense ($87 on digikey)

[jagged snow]

Oh nice

[winged valley]

I look up 1Rx8 64GB LPDDR5 and this is pretty much the only result https://www.newegg.com/hpe-64gb/p/0RN-000Z-00VD2

Its not even LPDDR5, and that is a lot more than 8 packages, which means they aren't 64Gbit

also $1k no thanks

If I knew which packages MSI was using for the 32GB model I wouldn't be going through all this headache lol

[twin dew]

Or the peaks for stock clock and OC cards?

[twin dew]

LPDDR5 and DDR5 aren't pin compatible.

[winged valley]

LPDDR5 SODIMM

Don't think I've seen it in a desktop stick

[twin dew]

Doesn't exist. LPDDR5 cannot be on anything but soldered on board.
SODIMMs are normal DDR5 too.

[winged valley]

Gotcha, sounds like my life got harder lol

Not that I would've found a reasonably priced dimm with what I needed to begin with

[twin dew]

But the problem is that LPDDR5 is bough in bulk.
And anyone selling it "retail" has needed to buy at minimum thousand chips to sell.

And most using it in production buy it direct from manufacturer.

Or in those chinese sellers case, can be cannibalized, or might be production rejects etc.

[twin dew]

Because that specific LPDDR5 chips are marked as "CS" status in Hynix data, so not in full production yet.

[winged valley]

Ooooohhhh okay that makes a lot of sense

I assumed MP meant mass production, couldnt figure out what CS was

Commercial Sample?

[twin dew]

Probably.

Would fit the other manufacturers statuses.

[winged valley]

You seem to be knowledgable in this

You buy a lot of chips?

[twin dew]

No.

[winged valley]

Oh

[twin dew]

Just general knowledge from everywhere.

[winged valley]

Gotcha

[languid gulch]

LPDDR6 when