Tech Lab forum - General Discussions!

[pearl moon]

Memory cards are 8MB, games could be several GB, whatever they could fit onto a DVD-ROM at the time.

[cinder lagoon]

Bruh I totally misread that yeah

[pearl moon]

and even then, the memory cards could hold a whole bunch of saves, as long as you didn't play many games with huge save files like Gran Turismo 4 with it's 2MB save files.

with the homebrew functionality i could store saves on the HDD, but alot of games aren't compatible and otherwise there is a higher risk of corruption.

for context, Monster Hunter World save file size is only 11MB

and that's an xb1/ps4 game

[balmy flicker]

done thanks now lest see if it does the other stupid thing again

[jade scaffold]

im disapointed that this atrocity isnt a consumer chip

its an APU btw

more of an SOC really

its got a CPU a GPU its got Onboard HBM3 Memory

3TB/s speeds and because HBM is so close to the chip itself the access latency is super super low probably equivalent access latency to L3/L4 Cache

[mossy gull]

Still can be an APU

SoC is what your phone has and usually also has the OS on the die

[jade scaffold]

SoCs rarely have permeant storage on die

very rarely

[cinder lagoon]

Nope, The OS is almost always on a separate eMMC flash chip!

[jade scaffold]

its actually not all that common that they have RAM On the SOC itself and not on the main board the things attached to

but that chip there is everything except for permanent storage on one chip which is simply amazing

[mossy gull]

So far I can read an SoC has the CPU, GPU, modems and the whole shenanigans on it.
An APU is basically just a CPU with a "dedicated" GPU

[cinder lagoon]

This however is getting more and more common. Think M1, also many phones.

[jade scaffold]

M1 really on the SOC

its very very close to the SOC tho

[mossy gull]

Apple M are SoCs

[jade scaffold]

we know

[cinder lagoon]

https://www.golem.de/2012/153043-253470-253469_rc.jpg
M1 die shot. DDR is on chip

[mossy gull]

And also the network modems

[jade scaffold]

thoes are some really small DRAM modules

[mossy gull]

Or RAM chips are unnecessarily huge

[jade scaffold]

no ram chips arent unnecessarilly huge

they do have a lil protective layer over them but that only adds like a mm in all directions if that

[mossy gull]

If I compare to HBM, then RAM is unnecessarily huge

[jade scaffold]

thats because HBM isnt DRAM

HBM is totally different

like TOTALLY different

[mossy gull]

I know

[jade scaffold]

also i'd like to see a 256gb HBM chip

they like dont exsist

[cinder lagoon]

https://www.techinsights.com/sites/default/files/2019-04/samsung-18-nm-dram-analysis-1.jpg
Samsung memory die. Not much wasted space.

[mossy gull]

I wish more GPUs or memory existed of HBM

[jade scaffold]

you can get like 96gb of HBM with 6 modules i think it is

[jade scaffold]

same most GPUs wouldnt benefit from it

but the space savings would be nice

what the hell are these chips on the M2 then

they cant be SSD can they

no they arent

[mossy gull]

It will be amazing for smaller PCs, like now we need to humongous systems, but with HBM RAM and VRAM the system can be a lot lot smaller.

[tender plank]

L1 cache?

[jade scaffold]

thoes arent L1 cache

they are too far away from the cores

[sonic meadow]

l1 and l2 cache are on the same physical bit of silicon as the cpu

[jade scaffold]

L1 has to practically be touching the core

[sonic meadow]

would be too slow anywhere else

[cinder lagoon]

M2 die shot
https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F69502024-a55e-4ceb-bbd4-0cf1193153d8_1024x421.jpeg

[jade scaffold]

infact L1 im pretty sure is literally ON the processor

[jade scaffold]

Thoes DRAM modules i dont think are actual DRAM

[cinder lagoon]

This on top looks like regular DDR

[jade scaffold]

i think they are just where they connectors are

i dont see much point in putting DDR on chip then also putting it off chip because that negates any advantage of putting it on chip

[sonic meadow]

yeah, literally in the cpu (zen2 die shot)

also wow the branch predictor has a lot of storage

[jade scaffold]

unless its just a very very tiny amount of DRAM they are using as cache

[jade scaffold]

thats not how i meant

[sonic meadow]

well if that wasnt what you meant, then you were quite likely wrong

[jade scaffold]

i meant on the Proccessor not on the CPU die as in very much physically touching the CPU Core itself

[cinder lagoon]

https://cdn.wccftech.com/wp-content/uploads/2020/11/AMD-Ryzen-5000-Zen-3-Desktop-CPU_Vermeer_Die-Shot_1-scaled.jpg
higher res M2

[jade scaffold]

yea L1 cache inside the Core itself

[cinder lagoon]

Also at the top: "Global memory interconnect"

[jade scaffold]

thought so L1 cache access has to be as close to instant as physically possible

[mossy gull]

I wouldn't be surprised if Apple has found a way to put massive amounts of DRAM on their SoCs

[cinder lagoon]

It is memory on the M2 because putting so much RAM on die just isn't feasible
(as you would probably not scale down memory because it doesn't gain you anything)

[jade scaffold]

apple havent found a way to do shit

apple have done nothing that no one else haven't already done all apple have done is implement it

[cinder lagoon]

They have found a way to have 2 separate GPUs acting as literally one, no driver or whatever support needed!

[jade scaffold]

the issue with DRAM is DRAM Cells are nigh impossible to shrink

[jade scaffold]

NVidia already did that one first too

[sonic meadow]

DRAM cells are already the smallest bit of storage anyway

its a transistor. 1 transistor per bit

[cinder lagoon]

You can just put them on a smaller node but it won't gain you anything and it gets less error tolerable

[sonic meadow]

faster storage is typically bulkier, as it relies on things like flip flops instead of semiconductor implementation details (DRAM is technically a bug 😆 )

[cinder lagoon]

Nope, also they have always needed special software / driver support!

[jade scaffold]

on consumer GPUs yea

[sonic meadow]

has apple figured out how to do it, or just their drivers make it seamless to the end user?

[jade scaffold]

their Data center ones just go through a chip set on their carrier boards and get immediate unified memory

[sonic meadow]

since they control the platform from top to bottom

[cinder lagoon]

Radeon 3870X2
ATI/AMD was the first company to put dual-GPUs on a single PCB, something Nvidia only attempted with the second version of the GTX295

[sonic meadow]

unified memory is not the same thing as two gpus acting as one

[jade scaffold]

correct

but they also all act as one big GPU

they all do the same Work

and yea AMD had the first dual GPU i believe

but its somthing apple had nothing to do with

[cinder lagoon]

Afaik they have literally used that infinity fabric thing to connect to another GPU and it is not seamless because of drivers

[sonic meadow]

so you are saying that what you said isnt correct then?

[jade scaffold]

they did what they still do have a controler, PCIE Chip thingy what they call them changes every generation that then can split out load between the 2 seperate GPUs

or in data centres cases 8 GPUs

both AMD and NVidias Data Centre GPUs can do this

a notable Dual GPU tho is the Dual Vega 7 that amd made for apple

[sonic meadow]

for non-gaming tasks, having them just be a giant threadpool is going to be a lot easier. you dont have the silly latency issues getting in the way

[jade scaffold]

which is a crazy lil thing

[cinder lagoon]

What I mean is some ULTRA MAX PRO chip that has just a second unit fused to it but it does not need special driver support afaik

[cinder lagoon]

No, not this

[jade scaffold]

you mean they did a Chiplet design?

[sonic meadow]

or being apple, does it just already have the drivers installed?

[jade scaffold]

because thats also been a thing for a while now

[cinder lagoon]

https://youtu.be/D24ueW8G0-w?t=886
Here

14:46

[jade scaffold]

Yea thats not a new thing

Apple didnt invent that

both NVidia and AMD have been considering using such a design, I think actually Next gen thats what they are going to be Chiplet GPUs

[sonic meadow]

so its a modern AMD cpu, but a gpu instead

[jade scaffold]

yea

[sonic meadow]

a dual core GPU 😆

we have already had seamless 4-way gpus though 😛 https://www.youtube.com/watch?v=-aMiEszQBik

[jade scaffold]

also as linus mentions the main issue with Multi GPU is Bandwidth somthing data centres GPUs dont have an issue with because they have specific hard ware made especially for effectively gluing 8 4090+ calibure GPUs into 1 big GPU

[sonic meadow]

its an interconnect issue. if "because datacentres", then it would have been doable via SLI/crossfire bridges previously

[jade scaffold]

the SLI and Crossfire bridges were extreemly low bandwidth

[sonic meadow]

yes, because getting the bandwidth necessary (which makes 16 lane PCIe look like PCI) isnt an easy task

[jade scaffold]

yea which makes it really exspensive which is why NVLink took so long to get to consumer GPUs and even then it was a stripped back version of the thing

[sonic meadow]

for reference, a 4070 has memory bandwidth of approximately 500GB/s, where a pcie gen5 x16 link is approximately 500Gb/s

[jade scaffold]

yea

NVLink3 which is what the A100 uses can do 600GB/s

which isnt anywhere near as fast as the GPUs memory

but in most cases its definatly fast enough for the GPUs to function and it also does not include the PCIE Lanes which i think on an A100 system would be Gen 4 not 5 i think

we do actually have SMX5 now which has an NVLink Speed of 900GB/s advertised by NVidia though HBM3 is like 3TB/s speeds because 5120-bit bus go brrr

[sonic meadow]

it does slightly surprise me that vram bus width hasnt really gotten any wider. the rtx 4000 series has basically the same range of bus widths as the... geforce 8 series cards

At an introductory price of US$300, the GeForce 7950 GT...
/cries

[jade scaffold]

XD

i guess they dont really NEED the wider bus with the type of data they handle

[sonic meadow]

yeah, and chips got faster. waaaay faster

[jade scaffold]

yea

like the Speed of GDDR now is crazy

[sonic meadow]

but they arent really using more chips, which is one of those things that would give rise to higher buswidths

[jade scaffold]

the only reason its data transfer rate is lower than HBM is because HBM a crazy ass bus width

[sonic meadow]

yeah. hbm is silly wide

embarrassingly wide, for those embarrassingly parallel tasks

[jade scaffold]

but anyway back to AMDs goofy SOC APU thingy

the CPU and GPU appear to share memory

[sonic meadow]

is that the new EPYC chip? or something else

[jade scaffold]

they both appear to share the HMB

[jade scaffold]

kinda

i think it might be the same socket

is that Bigger than EPYC?

i really cant tell

the Die area sure as hell is though jeeez never seen so much die on one chip

[sonic meadow]

SP5 is 75*72mm

[jade scaffold]

lets play count the die

i think im counting 20 dies there

the chips called MI300 btw

there is some older versions that have a similar concept

Mi100 and Mi200

[sonic meadow]

13 chiplets, 24 cores, graphics, hbm3

[jade scaffold]

its also like a turbo stacked thing

there are chiplets ontop of other chiplets

[sonic meadow]

yeah, IO is under everything else 😆

[jade scaffold]

just like burry it under there hope to god the probably 500w worth of Compute sat ontop of it doesnt incinerate the poor thing

[cinder lagoon]

...interesting to see that apparently the Type-C port of my motherboard shares power rails with all other ones... And has no diodes in it!
I plugged in my phone for some charging and the USB OTG system checks for devices.

Ignore the crumbs I was eating chips 🤤

[jade scaffold]

is it possible to split 1 16x slot into 2 shared bandwidth 16x PCIE Slots

[sonic meadow]

Depends on the motherboard. Look up pcie bifurcation (you would end dup with two X8 slots though)
For two x16 slots you would need a pcie switch (PLX). But those are silly expensive. They also work on any motherboard

[mossy gull]

Depends really on the motherboard, but they won't stay x16, you'll get 2 x8.

[jade scaffold]

I can probs aquire a pcie switch for a reasonable price somewhere

[mossy gull]

I would not skimp on that, but your motherboard has to support it

[jade scaffold]

im not saying skimp on it

and yea i think the high end server mobo probably supports it

[mossy gull]

Check with the manufacturer to be sure

[jade scaffold]

infact defiantly supports it because it otherwise only has its M.2 and U.2 ports for other expansion

other than the 1 16x slot it has

i know there are cards that have both U.2 ports and 10GB Ethernet on them but im not sure how easy they will be to nab for reasonable prices

[sonic meadow]

I don't think you need motherboard support for pcie switches, though I might be mistaken (or just some boards don't work for .. reasons)

[jade scaffold]

its a server board so even if you need bifurcation for some dumb reason which for a switch you shouldnt atall need it has it because servers NEED to be able to alocate their PCIE lanes

[cinder lagoon]

So you want 2 pcie devices to share the same bandwidth because you know that only 1 will be active at any given time?
Not possible. PCI protocol needs a constant link and does not support differentiating between two devices.
Bifurcation only works because each has their own connections which don't get interrupted.

[cinder lagoon]

If at all, this is it

[jade scaffold]

yes it require a seperate chipset i know

what im asking for is a device that can take that 16x slot and run 2 seperate devices through it

[sonic meadow]

something like this: https://c-payne.com/collections/pcie-packet-switch-plx-pcbs

[pearl moon]

Holy crap, that's more expensive than most consumer mobos

[mossy gull]

Half the price of a motherboard overhere

[jade scaffold]

Yea for server stuff tho that's pretty cheap

[sonic meadow]

I also saw server targeted stuff. looked to be starting around the $1200 mark

plus I think it was part of a larger solution (ie you need to buy the board with the PLX switch on it, and then buy the daughterboards with the slots themselves, plus cabling, etc)

the big issue is that the PLX chip is expensive, last I saw a price it was something like $100. and that was likely in the pcie 2/3 days

[jade scaffold]

i only need PCIE 3 my servers not got PCIE 4 or 5

has an old boi CPU

On another note

had another instance of Something knocking out 12V power on my PCs

and only 12V

as RGB and Fans all still worked and they are all 5V and 3.3V if i remember correctly

[sonic meadow]

12v going down would knock out the entire system

[jade scaffold]

well the CPU and GPU arent getting any power but the RGB and Fans were still getting power

[sonic meadow]

fans should be 12v too

rgb is system dependent (some use 5, some use 12)

[jade scaffold]

not all fans are 12V

None PWM Ones almost always are

[sonic meadow]

pc standard fans are all 12v

[jade scaffold]

but PWM 4 pin fans use 5v typicaly

[sonic meadow]

nope, still 12v. the pwm pin might be 5v. but the actual fan spinning power is 12

[jade scaffold]

then not sure why power is cut to PCIE and CPU

and nothing else

thats just weird

[sonic meadow]

very

[jade scaffold]

its cut exactly the same on both boards as well

which is even more odd

[sonic meadow]

is your psu split or single rail (for 12v)?

[jade scaffold]

single i believe

not 100% sure

[sonic meadow]

how do you know you are losing 12v, and that its not some other issue?

[jade scaffold]

Sound card looses power completely the only place it can get power is the PCI Power lanes the system also doesnt respond atall to the power button being held down

at some point down the line PCIE Power is totally cut

had the issue once as well and the servers heat sink went cold

[sonic meadow]

have you shoved a multimeter across the 12v pins after its gone weird?

or are you just guessing its a 12v issue?

[jade scaffold]

im guessing Power IS Cut to the CPU and PCIE Lanes which is fact something has to be causing that and i dont know what

best guess sofar is OCP though OCP usually shuts off the entire system not just power to the CPU and PCIE Slots

[sonic meadow]

OCP would be on the PSU. which would either shut down (taking everything with it). or shutting down the rail (which would have the same effect). there is typically only a single 12v rail going from the PSU to the motherboard/other. and even in cases where they have multiple rails (which isnt common anymore), the entire motherboard bar the CPU is all off the same one. that means the fans and the pcie slots

[jade scaffold]

yea which is why i didnt think it was the PCU

However

the exact same thing happens to 2 systems

both loose power to the CPU and both then keep their fans running

which is just Odd

it happening to both systems makes me think the PSU is reacting to somthing

i might try speaking to phanteks and getting the PSU replaced the thing has a 12 year warrantee though it could be an external stimulus like a Power spike or somthing causing it though why this would cause a partial shut down when i ran a 550w PSU for years without such an issue off the same sockets and surge protected plugs as before is beyond me

[jade scaffold]

Next time it happens ill go poke the Test headers and some of the 12V Pins,
What i really want to do tho is find a way to recreate the occurrence see if its an external stimulus or an internal one
it only happens very rarely so if its an eternal one it might be very hard to pin down

[sacred seal]

https://www.youtube.com/watch?v=xJKzKbqxa0A&ab_channel=GamersNexus Steve is still mad at Asus

Though it seems like the issue is mostly resolved at this point

[sacred seal]

Also Asus apparently bricked a bunch of their routers for two days by having them download a courupted config file, even on devices where automatic updates were turned off.

[jade scaffold]

asus taking AMDs place of being so bad at making software their devices are sometimes unuseable

[surreal moss]

Not sure where else to stick this:

https://youtu.be/9DwcerOzLJw

@sudden pike and @olive fiber

[mossy gull]

Been ages ago I seen that vid, also you tagged the wrong Equinox I think

[olive fiber]

Still annoyed my ryzen CPU sits there for 15 seconds checking ram each time I turn it on

Just to keep you guys updated on what annoys me.

[surreal moss]

Oops

[cinder lagoon]

Haven't had that yet... is fastboot turned on in the BIOS? Do you unplug it over night? Is the BIOS battery still good?

[sudden pike]

Did you ping the wrong Equinox lmao

[surreal moss]

I did. Enjoy? 😬

[sudden pike]

xD

Ur good lol

Im at work but Ill watch it later

[olive fiber]

Brand new machine, I do not unplug it, and yes, I think fastboot was enabled. I'd have to double check.

I googled it and found others had the same issue. Power on, and it sits for 15 seconds before it gives any indication that it's POSTing

Then you get the normal 'Hit F2 to enter bios' prompt and whatever

Sounded like it was normal for AMD 7xxx

Just annoying since the entire rest of the boot sequence is like, 5 seconds to windows desktop. 😛

[cinder lagoon]

Afaik this behaviour happens without fast boot. But I'm not sure. Mine also takes its sweet time but its 5800X3D with x470

[olive fiber]

I got the 7900X3D. Dunno if it's the 3D cache that's the issue?

I'll look again for the fast boot setting

Yeah, fast boot is enabled. I also ready someone comment about having automatic ram timings enabled causes this? I don't know. Wasn't easy to find people that seemed to know what they were talking about

[cinder lagoon]

It should be in the "boot" tab - I would just say that you probably can't do much about it.
It could be that the motherboard does some additional memory training if the XMP (or whatever the new fkn name is) is enabled

[sonic meadow]

The am5 ryzens do ram training on every POST rather than on config changes, that's why it's taking so much longer

[cinder lagoon]

AHA!!

[mossy gull]

https://www.reddit.com/r/Amd/comments/13hygsi/the_true_limits_of_the_7900_xtx_unlimited_power/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button

Damn, AMD, why not give us that card

[cinder lagoon]

Needing a small reactor at home to be able to run that thing would certainly be a factor xD

Pretty sure they created the limit so that pretty much no where in the world, you'd need two separate circuits to run that machine

[sonic meadow]

its a *hey, we arent doing an Nvidia and letting our cards draw silly amounts of power" limit. or thats how it smelled to me.
the person got a ~15% performance uplift (which is solid) for a... 50% power increase?

[cinder lagoon]

That ratio has been pretty much the same on AMD cards since like forever (at least talking from personalnexperience)

So kinda checks out

[sonic meadow]

tbh I was expecting the cards to be further along the efficiency curve than that stock, they could have raised it a bit more for still noticeable gains.
however, the card in the link has three pcie power connectors. this give the card a power limit over them of <drumroll> 450w, plus another 75 over the slot for a total of 525w. this suggests they were wanting to stay within the limits of the three 8 pin pcie connectors since it has a power slider limit of 430w

[cinder lagoon]

Seems reasonable (ahem 12VHP fires ahem)

[barren juniper]

apparently hp did this recently with some printers for some reason.

also, not sure if this was posted here or not, but https://www.youtube.com/watch?v=3XlxEIIF22I rx 7600 is 269$

apparently 77% performance of 4060ti for 67% price so not much better

[jade scaffold]

this generation of gpus is really shit

[sonic meadow]

yeah. most of the 40 series has been... weirdly placed in the lineup

the 4060 is (outside of framegen, and other "software" bits), functionally the same as a 3060 (or was it the 3060ti?). which is bizarre

[jade scaffold]

the most anoying part especially on NVidias part they could have made vastly more powerful GPUs this gen, and were 100% capable of just obliterating anything AMD Offered but they didnt do that they made 1 actual Good gpu then just went the rout of 2000 series where the cards are just overclocked last gen with a smaller node size

[sonic meadow]

the vram available yoyos up and down as you go up the lineup, which is also fun

[jade scaffold]

the xx70 cards should be faster than the previous gen 80ti cards and in 2000 and 4000 they just fucking arent

[sonic meadow]

I still wish I got a 2070s instead of the 2060s I got. silly not currently in stock products... (I could have waited, but there was something I needed urgently so piggybacked my GPU order off that)

[jade scaffold]

the 2060s honestly is the only card from 2000 i would ever recomend

[tender plank]

https://youtu.be/MCxYfXe1DAA

[jade scaffold]

the 1080ti is just vastly superior to the 2070s and you could get it for arround the same price at the time of release

[sonic meadow]

not sure how the prices were here at the time. wouldnt have surprised me if the 1080ti was still over 1k when I got the 2060s, and I do remember the 2070s being sub-1k

[cinder lagoon]

Pcpartpicker has price history

[sonic meadow]

the 20 series also has RT cores, which the 10 series doesnt

now you could argue that the RT cores in most of the 20 series werent really worth it, but they were something that distinguished them from the previous gen

[jade scaffold]

1K???????

i bought my 1080ti a month or 2 after the 2070 Super launched for £400the 2070s was £440ish iirc

the RT Cores on 2000 are barely functional as well

performance in raytracing on the 2070s is just flat out aweful

[sonic meadow]

400 quid is 800nzd (currently)

[jade scaffold]

3000 and 4000 raytracing is much better its actually what i would consider useable too bad raytracing is wasted on AAA Games which dont hardly benefit from it anyway

[sonic meadow]

I have done a bit of RT stuff on my 2060s without too many issues. performance wasnt great obviously, but the few things I tried have been playable

[jade scaffold]

depends on resolution i guess

but the 2070s cant do 1440p Raytracing at 60fps most of the time the RT Cores in 2000 were just sadly bad

not to mention the fact that the 2070s can in some cases be upto 15% slower than the 1080ti nevermind overclocked 1080tis or liquid cooled ones

the 4070ti has a similar issue on paper its about as fast as the 3090 but it has nowhere near as much headroom as the 3090 as such a well cooled 3090 can out perform the 4070ti quite significantly

[sonic meadow]

doesnt go back far enough 😦
did some skimming on pricespy. and from what is still there, none of the 1080ti variants looked to go under $1k at all, while the 2070s went down to about 750. I think the one I was looking at was about 8-900 when I got the 2060s

[jade scaffold]

the thing thats really annoying about 4000 is that the 4090 is 82% Faster than the 3090 why the hell are the rest of the GPUs not 82% faster than last gen
Hell i would have took 50% faster or even 30% faster but nooooo we get marginal improvements for the rest of them at best

[sonic meadow]

yeah, most of the 40 series feels like its functionally 1 number higher than it should be

which is mostly exemplified by the 4060/ti and 4070(non ti)

it does sadden me that the xx70 cards are now considered by many to be mid-range, when they used to be high-end cards. but that was back when nvidia actually sold low end cards

[jade scaffold]

yup

Nvidias Jetson Mini computers annoyingly enough fit perfectly as a low end to NVidias current gpus showing they do have the ability to make really cool low power high performance GPUs and they just DONT

with the efficiency of the 4090 and it is a dam efficient card you could make a 30w card that shit stomps a 1070 imagine how awesome that'd be for a ultra thin and light laptop

[sonic meadow]

to be fair, AMD and Intels integrated have been "good enough" for a few generations now, and unless you need extra display outputs then the low end stuff is basically not necessary

at which point a 1030 or something is still perfectly fine

an actual low end gaming card that supports RT-adjacent stuff like DLSS/etc, and was available in low profile (pleeeeeease) would be amazing

[jade scaffold]

the Jetsons have tensor cores and like not even the limited ones the Gforce cards have they are fully enabled tensor cores DLSS on a tiny little 15w Chip for somthing like a steam deck would be epic

[inland dove]

I don't really keep track of this stuff.

Is the GPU market really bad still like it was during the pandemic?

[jade scaffold]

its not as bad price wise especially if you wanna go second hand

and i really would recomend going second hand ATM NVidias 4000 serise dont deserve any money and AMD arent doing amazingly better at the lower ends

[inland dove]

Oh no.

That wouldn't be in my price range.

And that'd be way overkill for the games I play.

[barren juniper]

ahh see that hadnt come out yet. i will watch now

but yeah this current gen cards out side of the 4090 and xtx has been a let down

[barren juniper]

6700xt is like less than 300? get that.

its the best value card we will have for a while

if you cant afford that, go down the rx 6000 stack until you can, but if you cant afford an rx 6600 then go used, as anything below it is not really worth it unless you need the small form factor of the 6400

[barren juniper]

there is one. its the rtx 4000 sff. it costs the same price as a 4090 and uses 70 watts i believe. it pretty much performs as well as the current 4060ti

[jade scaffold]

do you mean the A4000?

[barren juniper]

yeah that thing

rtx ada a 4000 sff or whatever it is called

i almost considered getting one because of its form factor but not for that price i wont

[jade scaffold]

yea they are Quadro GPUs

not made for normal users

[barren juniper]

which is the main reason why nvidia restricts its vram on gamer cards

[jade scaffold]

made for professional applications and such

[barren juniper]

and is why the 3060 series sold so well

[jade scaffold]

its actually not

VRAM isnt the main feature of the Quadro type cards

their main selling point is their drivers and expanded feature set

[barren juniper]

why sell someone a card for 600 that they can sell for 3k

its why the 3070 only had 8

[jade scaffold]

no thats literally just NVidia being cheap

like they have been with the 4060ti

with its 128bit memory bus

[barren juniper]

well i guess them doing segmentation like that could be considering them being cheap

[jade scaffold]

8gb of GDDR6x costs nvidia like $30 or somthing like that

[barren juniper]

yup

and its finally biting them in the ass

[jade scaffold]

so they just like dont put it on the gamer cards and upcharge

[jade scaffold]

except its not

because no one is actually competing with them to any reasonable level

so they can get away with it

[barren juniper]

"We fully expect zero sales in some of our stores this time around." Microcenter execs referring to interest in the 4060ti

[jade scaffold]

consumer GPUs aren't NVidias main customer

never mind the low end

[barren juniper]

60 series are where they usually make the most sales, at least in the past, and their launches have been big events. not this time

[jade scaffold]

Sales =/= proffit

[barren juniper]

but i guess they are an ai company now

[jade scaffold]

selling 1000 4060tis is less profitable than 500 4080s

and NVidia have been an AI Company since Maxwell 2.0

arguably since Kepler

the H100 is basically a 4090 with HBM Memory and High end NVLink and it'll cost nearly 7x as much per GPU Data centres and super computers will buy hundereds if not thousends of them NVidia couldn't give less of a shit about low end

NVidia have also developed a poor tendancy to make a really good generation then make a nearly worthless generation then make a really good one again then a next to worthless one again See 1000 and 2000 then see 3000 and 4000

[barren juniper]

yeah outside of the 4090 this generation has been completely pointless for gamers

though part of that i think is because of the over supply of last gen cards

also according to Wendle from level 1 techs, the rx 7600 is a plug and play for linux, though I don't think that would effect many in the gaming department

[jade scaffold]

Amds cards do play really nice with Linux generally

Nvidias consumer ones less so

[cinder lagoon]

Nvidias consumer ones less so

[mossy gull]

Because "gamers don't play on Linux"

[jade scaffold]

Their server and workstation gpus have proper Linux drivers that work

Their server ones almost flat out don't work on Windows i don't think

Because what crazy mf using windows server on a data centre or super computer

[surreal moss]

Oh they’re out there. Normally because the software only runs on windows

[cinder lagoon]

I'd consider it "work™️"

[jade scaffold]

it is extremely rare that a data centre would Use windows for anything as theres nothing Windows can do in the data centre that linux cant do substantially better
most things you'd do with a super computer as well couldnt really be done on windows as im pretty sure windows has no support for splitting the workload of multiple machines

[jade scaffold]

they work pretty flawlessly, the reason AMD Cards work so well even at their consumer level is because AMDs GPU Software stuffs is all open source

NVidias Kernel on their very high end enterprise stuff are open source but things like their CUDA and Drivers aren't

[mossy gull]

I am waiting till AMD finishes their emulation layer for Cuda applications

[jade scaffold]

its actually out on some of their enterprise stuffs i do believe

however its gonna be pretty meh

it sure as hell will be nice to have

[mossy gull]

They also want to release it on their consumer cards

[jade scaffold]

but if an application supports both CUDA and OpenCL its not gonna be worth an AMD GPU trying to brute force CUDA even if CUDA is typically alot faster

[surreal moss]

I am willing to be that every datacenter has at least one windows server in it

[sonic meadow]

probably a considerable number in some of them

[mossy gull]

My dad's company servers run on Unix

[sonic meadow]

https://tenor.com/view/shocked-i-know-this-unix-system-gif-14433570

[surreal moss]

The datacenter I administer have both. Thankfully I only have to focus on the Linux side

Mostly

[sonic meadow]

'nix isnt too different. usually

[mossy gull]

I never worked solely with Unix

[jade scaffold]

im thinking of doing somthing really dumb im Eying up Audeze Euclids

[cinder lagoon]

...need to change my dynamic DNS IP cuz some russian ass keeps trying to log into my router 🙄

[jade scaffold]

russain moment

[balmy flicker]

fun

[cinder lagoon]

Ye

[barren juniper]

so my local microcenter has 70 rx 7600s in stock, while 67 4060tis. i guess they were right when they said they wont sell well

i believe this is the one on manhattan too

[jade scaffold]

yea seems about right

the last gen 3060ti was just too good

[mossy gull]

Funny enough, in EU the RX 7600XT has sold almost 4 times as much as the 4060👔

[ornate geyser]

Hey guys 🙂

This may be a super long shot but I love playing star citizen and would like to upgrade my pc to boost my performance for less than 300$ do you guys have any recommendations? My specs are down below.

GPU: AMD Radeon RX 5700 XT
CPU: Intel(R) Core(TM) i5-9600KF @ 3.70GHz 6 Cores
Motherboard: ASRock Z390 Phantom Gaming 4S-IB
RAM: 16gb
SSD: 225gb
HDD: 1tb

[sacred seal]

I presume SC is on your SSD already?

Not much you can do for 300. Possibly get a better cooler and overclock your CPU

[ornate geyser]

yes, and thank you

[sacred seal]

Depending on the quality of the SSD, you might see some benifits from upgrading it.

[mossy gull]

You can get the RX 7600XT

[mossy gull]

I believe you can get it around 300.-

[cinder lagoon]

Yeah, the 5700XT was never particularly fast

[mossy gull]

This is on AMD EU
https://shop-eu-en.amd.com/amd-radeon-rx-7600-graphics/

[cinder lagoon]

Not sure if they realize that they currently have this massive price gap...
how about a medium tier GPU for like 500?

[mossy gull]

The problem is the 6900XT currently kinda fills that gap

There's webshops that sell them for 500.- to 800.-

[cinder lagoon]

Yea but they're old and gulp power

[mossy gull]

But they sell them for that low, then they need a 7700XT that's more powerful than the 6900XT for around the same price, that'll be pretty hard to do

[jade scaffold]

the 5700XT at the time was pretty good it was 2070 Super level which isnt too bad especially when it was priced like a 2060

obviously missing some of the fun features of NVidias RTX cards but at the time nothing could even use thoes features and the games that came out with said features used the later gen varients of them which NVidia arbitrarily blocked from 2000

[jade scaffold]

also yea its a bit of an issue they need a 500 and 700 priced cards but then they would have a bit of a weird and missleading performance skew when compared to NVidia so they really need to be careful with how they make them

[cinder lagoon]

Huh apparently GN ranked the 5700XT best of 2019 (thought they were kinda like last gen's 'meh' )

[jade scaffold]

yea no the 5700XT after the first 6 months was by far the best price to performance gpu arround

[sonic meadow]

the issue with the radeon 5xxx series at the time was that a bunch of people were having driver issues when using them

[jade scaffold]

oh yea

[cinder lagoon]

Ahhh yes I remember actually

[jade scaffold]

at launch you would be better off with a 1050ti than a 5700xt

my friend personally verified that one for me

[cinder lagoon]

Also OpenGL performance still sucked

Like really

[jade scaffold]

also just dont get the reference version of the 5700xt

it would overheat real easily

[cinder lagoon]

...where were we?
Ahh yea $300 PC upgrade for SC

[jade scaffold]

you'd need a card with more VRAM me thinks

though im struggling to think of somthing the for $300 that has more than 8GB of RAM and is actually a decent improvement over the 5700xt in performance

not sure how the 7600 stacks up?

the 7600 is only slightly better not a worth while upgrade atall

perhaps if you're willing to go second hand you might be able to find something half decent for that price

you can get 3070s for just over 300 where i live

[cinder lagoon]

Also this

Yea but then a 12G variant, pls

[mossy gull]

The 5700XT is more or less on-par with the 3060.
While the 7600XT is on-par with the 4060 Ti, so there's an performance jump

[jade scaffold]

literally just watched a comparison video worst case the 7600 was about 5% slower best case it was 15% faster on average they 7600 is about 5% faster 300 for a 5% performance improvement is not worth it in any world that exsists

the only option to get a card that is actually reasonably better will be at the moment to buy a second hand one

[mossy gull]

Benchmarks don't say much, especially not with AMD

[jade scaffold]

what lmao

people playing games and recording the averages an minimums of 2 cards while playing a game isnt representative of their gaming performance?

lmao

been looking arround ebay odly enough the 3070 is going for less on average here than the 6800xt is or the 6700xt is which is odd i woulda thought the AMD cards being worse off feature wise might have a lower price tag

i also coulda swore they were a lil cheaper at launch

[cinder lagoon]

Nope cuz still nobody really cares about RTX and DLSS3

[jade scaffold]

just kinda surprised the 6700xt is going for more than alot of 3070s tho

even though they basically perform the same in alot of things

was kinda hoping to see some really good 6700xt deals its got more VRAM than the 3070 and is far faster than the 5700xt and the Extra VRAM would be nice for star citizen

[sonic meadow]

Noctua finally coming back in to the game with updated stuff... soon™️ https://www.youtube.com/watch?v=Cng_J4sBaNs

[jade scaffold]

thats some funky fan geometry

[sonic meadow]

aero/fluid dynamics is weird

[cinder lagoon]

Ye seen the toroidal fans yet?

They are far more quiet for the same flow

[jade scaffold]

i know they work really well in water

im not sure they are capable of high static pressure in air though

[cinder lagoon]

Not so much static pressure but quiet high flow

[jade scaffold]

yea

[cinder lagoon]

Those are basically prototypes tho

[jade scaffold]

which might be good for a air cooled case fans

[mossy gull]

When you need VRAM the 3070 loses a lot, the 3070 has a problem where it won't use all 8GB VRAM and already started using shared memory around 7.5GB

[jade scaffold]

all GPUs will do that

[mossy gull]

My 6700XT will go to full 12GB before using shared memory

[jade scaffold]

it shouldnt do

otherwise you loose even more performance on shared memory

[mossy gull]

Also, in Cyberpunk 2077, the 6700XT wins over the 3070 in 1440p

[jade scaffold]

yea it will do

the 3070 is conventionally speking more powerfull but NVidia being the forever cheapskates that they are cheaped out on VRAM so in games that can use alot of VRAM it struggles

[mossy gull]

My little brother had the 3070, and it lost in a lot of games vs my 6700XT in 3440x1440, it's kinda sad considering his 3070 was more expensive

[jade scaffold]

yea

what NVidia really need to do is allow DLSS 2 to work on everything and anything

not 3 tho DLSS 3 can gargle my balls its so shit

[mossy gull]

Fun fact: DLSS 3 even works on AMD cards, it ain't optimal but it still works.

But NVidia keeps saying it keeds the Tensor cores of that generation card to work.

[jade scaffold]

yea but DLSS 3 looks like ASSS

any yea as long as your card can do or brute force tensor flow operations it can run DLSS

or any AI thing

just like RTX voice works on anything

[mossy gull]

DLSS or any upscaling looks like ass to me, I always prefer native rendering

[jade scaffold]

but on cards without tesor core has a small peroformance impact

DLSS 2 cant really be told apart from native

not unless you go full peformance crazy with it

like upscaling 640p to 4k

quality mode basically gives you a little more FPS and vastly superior AA

[mossy gull]

I can easily see artefacting.
So anything that looks out of place or over sharpened I notice it.

[jade scaffold]

so you play all games with AA and Post processing turned off

[mossy gull]

I use AMD Super Resolution instead of any AA

[jade scaffold]

because both of thoes create some pretty goofy artifacts

[mossy gull]

Yes

[jade scaffold]

so why dont you just play all your games on a low end card

because everything you play is gonna look streight out of the Xbox 360 anyway

[mossy gull]

Or you play everything on higher resolution than intended

[jade scaffold]

Post processing is responsible for lighting

its what makes game lighting not look turbo flat

but it also leads to artifacting especially when screen space effects are involved

[mossy gull]

On most games it barely makes a difference.

[jade scaffold]

yes it really does

[mossy gull]

I had post processing make lighting look jarring, like pixelated lighting, so I usually turn it off.

[jade scaffold]

what?

[mossy gull]

That was when I still used my 1060

[jade scaffold]

if you turn it off you practically dont have propper lighting

you get old school lighting

which to say the least is aweful

and you arent gonna use any kinda raytracing to off set that because your so artifact sensative

[mossy gull]

When Raytracing is available, I turn it on usually

[jade scaffold]

so you can see Quality DLSS artifacts

but not the vastly more obvious RT ones

[mossy gull]

I don't use DLSS because I have a fancy dandy 6700XT

[jade scaffold]

literally you saying you think it looks like ass followed by you saying you can see the artifacting pretty much implying that you wouldnt use it because of that
But the static effect that raytracing will throw on things if your card isn't insanely fast at raytracing doesn't bother you?

[mossy gull]

I rather have Raytracing over shit that looks blurry or over sharpened in the distance.

[jade scaffold]

but you'll use AMD super resolution which has the exact same problems?

[mossy gull]

I don't mean FSR, I mean Super Resolution, which basically lets you for example play 4K on a 1080p panel

[sonic meadow]

aka MSAA

[mossy gull]

But without the MSAA oddities

[sonic meadow]

oops, meant SSAA. MSAA is a subset of that 😆

[mossy gull]

Except Super Resolution isn't AA, it funny enough tasks the card a tad less compared to any kinda AA

[sonic meadow]

you cant easily test it against msaa or ssaa since games havent offered that as an option for a long time (it doesnt mix as well with deferred rendering)

[mossy gull]

Most games just look at the resolution of your panel these days, funny enough my games think it's higher than it actually is.
It's also not used very often, only by idiots like me

I still play games that don't go above 1600x900,and a few that are stuck at 1280x1024.

[jade scaffold]

https://www.youtube.com/watch?v=vdD0yMS40a0&ab_channel=LinusTechTips

:o

Hyper thin hand held PC with these fans?

[mossy gull]

Or full blown high powered laptops

[jade scaffold]

im more excited for hyper thinn devices with these

[mossy gull]

Hyper thin laptops with tons of power 😃

[jade scaffold]

they arent that amazing at cooling but like i could see someone lining like 5 of them along the back side of a laptop as well as using the shell for some passive dissipation on a 45w laptop

[sacred seal]

If I am remembering correctly they basically just are good for thin and lights. They can't scale it up to handle more airflow

[jade scaffold]

you can just attach a bunch of them to a big vapor chamber tho

[sacred seal]

Yea, but how big of a vapor chamber would you need to cool something like a 3070

[jade scaffold]

Depends on power draw

[sacred seal]

You would need like 25 modules if my math is correct

[jade scaffold]

Laptop 3070s can go as low as 45w iirc

[sacred seal]

Yea, but at that point are they really useful as 3070s?

[jade scaffold]

He showed 2 of them could do like 15w so in theory 6 of them could do a lower end laptop 3070

[sacred seal]

The spec I found was 115 watt for the laptop variant

[jade scaffold]

115 is for the real high ends

Thoes are for desktop replacement devices

But you could do like a razor blade or somthing with these things

Or even better a steam deck esque device

[sacred seal]

I was responding to Little's comment about a device with tones of power

[jade scaffold]

Or any actual thin and light

[sacred seal]

Yea, for thin and lights these things could be amazing.

[jade scaffold]

The only way you are doing somthing really really high power is lining the edge of a 17 inch laptop with the things

[sacred seal]

Seems like the base power rating for a 3070 in a laptop is 80 watts?

[jade scaffold]

thats deffos not right

[sacred seal]

I have not found anything with lower power

I presume that TGP is supposed to read as TDP

[jade scaffold]

its the same thing pretty much

just means total graphics power

[mossy gull]

For now, maybe soon they'll be good enough for gaming machines

[sacred seal]

The tech is fundamentally limited. You need to make them exponentially larger to get more cooling

[jade scaffold]

not necessarily

you can just move the air quicker

[sacred seal]

That's what they said in an interview

They were asked if they could make a larger unit for higher power devices and they basically said it wasn't possible if I am rembering correct (this was a few months ago)

[jade scaffold]

they are saying that higher frequency's and higher amplitudes could vastly improve performance and that its something they are working towards

[sacred seal]

Improve, yes. How much of an improvement do they need to be viable for higher power draw applications

[jade scaffold]

a 17 inch laptop could likely do 100w at the moment if you just lined the back side of it with them

[mossy gull]

I mean, Intel couldnt get past 14nm for ages, yet TSMC got down to 7nm back then

[sacred seal]

Couldn't do it economically

[mossy gull]

The thing is, technology ain't standing still, it's only a matter of time it's possible

[jade scaffold]

100w is a fair bit if they can even get a 20% improvement thats 120W and with GPUs getting vastly more efficient its feasable to get a 70 class laptop cooled by these things

[sacred seal]

They still have to deal with physics

[jade scaffold]

these things also see the same Manufacturing issues TVs and Processors have meaning in theory these can in mass production be extremely cheap to build

[mossy gull]

Even if you go for a hybrid solution, 100w of these flat flippers and your fans to cool if it gets too much for them.
Then you have a hybrid solution

[jade scaffold]

you can also do some passive cooling if you are willing to make your laptop out of not shit materials

[sacred seal]

Passive cooling gets difficult when dealing with higher power draw systems. You don't really want the laptop to start cooking the users hands if the active cooling fails

[jade scaffold]

if active cooling fail it'll just you know

[sacred seal]

Thats my understanding why its not really used on modern devices

[jade scaffold]

overheat and turn off

long before it can saturate the chasty

[mossy gull]

For example, I have an Asus G15 Advantage Edition, even if it had these tiny flip floppers the system could cool passively when not gaming, and turn the fans on when gaming.

[sacred seal]

You still need to fit a larger cooling solution into the laptop

[mossy gull]

There's enough space between the cooler and the housing though

[sacred seal]

For the required heat pipes and the modules?

[mossy gull]

It has a vapor chamber, but the lower spec has the thicker heatpipes, they basically share the housing

[jade scaffold]

the performance of these things relative to other cooling methods depends on scale

but the fans and heatsink in a standard thin and lights get pretty brutally shit on by these things at the moment

a much bigger fan will obviously do better as a larger fan can be scaled in 3 dimensions unlike these lil things which can only really be scaled in 2

while you wont see any full fat like 100+W TDP GPUs in laptops with these you will be seeing the lower end skew 70 type cards in laptops with these in them

[sacred seal]

Fans get exponentially better the larger they are. These can only be scaled linearly, and I believe only in one dimension (making them deeper gives limited benefit)

[jade scaffold]

do you know what exponentially means

[mossy gull]

I am just waiting for an hybrid solution with these flipper floppers

[sacred seal]

Yes. You double the size of a fan you get far more then double the airflow output

[jade scaffold]

thats because you arent doubling the fan size

you are increasing the size by 8x if you double all dimensions

[sacred seal]

If I am not mistaken you double the volume of a fan you still get far more then double its output

Due to the fact that the further from the center the blades are, the more air they move for a given RPM

[jade scaffold]

ehhhhhh

kinda

its next to impossible to compare a small fan to a big one

because blade geometry has to totally change based on size

otherwise that extra speed you get on the edges will only get you more noise

[sacred seal]

Yes, its not exactly a 1:1, but there is a reason why people buy 240mm fans instead of 4 120s

[mossy gull]

Reminds me instantly of this
https://youtu.be/GIGlQBx4tys

[jade scaffold]

yea larger fans by nature have to spin slower

unless you give 0 fucks about sound

also you get to a point where your blades just hit the air and cause cavitation in the air which is less than ideal as now the outer edge of your big spinny thing is doing a whole lot of absolutly fuck all

[mossy gull]

The problem with spinning a fan faster is also turbulence.
Sometimes a fan is steady flow at lower speeds, but turbulence at higher speeds.

[sacred seal]

Fans have to be optimized for some given speed

Unless they have variable pitch

[mossy gull]

I noticed this on those cheap fans you can buy around summer.
At full speed most just feel way too turbulent and are very loud, moving air pretty badly, they basically move the air just as much as at lower speeds.

[jade scaffold]

the main advantage i see with these new blowy majigs not right to call them a fan really i dont think is the Enormous static pressure they can produce this means they can pull air past things like small holes in the case of a laptop keeping the surface of it very cool and because they bypass the issue of boundy layers with good ol brute force they can pretty much instantly heat their output to nearly the same temperature as the cooler. the cooling ability of these fans is entirely based on the surface area they can cover pretty much so so while possibly more limited than big fans we have already seen for a given volume they are far better than small fans

im also pretty sure you could have a 2 sided vapour chamber and within 1cm thick fit like 20 of them along the back edge of a laptop

though the ammount of air that would move out the back i think would probably be annoying to someone because the output of air is at like 100mph now imagine 20 of them stacked 2x10

[mossy gull]

I think it was Alienware that actually placed fans in front in their laptops to cool the laptop better, these flippy floppers can take over that function

[sacred seal]

They deffinatly could do that. Each module looks to be about 1" wide, so 10"s total

[jade scaffold]

then with a double sided vapor chamber make a cooling sandwich

if each can do lets say 10w 20 of them could do 200w

[sacred seal]

The minis are only rated for 5 watts

[jade scaffold]

you know what high power laptops might not be as unfeasible as i thought

they have slightly longer ones

[mossy gull]

Hybride cooling

[jade scaffold]

which are more efficient

[sacred seal]

I wouldent say slightly

[jade scaffold]

but are basically 2 minis long

just under 2 long overall

also they are Really efficient

i think one large fan would be more efficient

but for the compactness

[sacred seal]

I think the larger ones are good for 15 watts?

[jade scaffold]

i'd say thats a worth while trade off

[jade scaffold]

wair really

i thought it was 10

20x15 is like 300w

jeezus christ

it wouldnt translate perfectly to that

[sacred seal]

Yea, but then you have a very large cooling assemblage

[jade scaffold]

yea its long

very long

but also very thinnn

and if its a gaming laptop you can like forgo the track pad and cram it on where the num keys are

[sacred seal]

Basically 3-4x larger then you would find on a equivalent laptop

[jade scaffold]

yes

[cinder lagoon]

They plan on practically doubling the cooling capacity per area about every two years

[jade scaffold]

500w thin and light laptop when

[cinder lagoon]

So we jus gotta wait a lil to get high-power laptops.

[cinder lagoon]

Ayoo think about them batteries, alright? 🤠

[sacred seal]

I guess I am misremembering the interview... I thought they were looking to do like 20% every 2 years

[cinder lagoon]

Just saw the LTT vid, think they said 50% "like moores law"

[mossy gull]

The problem is the American Aviation shenanigans only allow power bricks up to 330w.

Who have been holding back stuff like powerful laptops and powerful batteries.

[cinder lagoon]

The problem is that they only allow 99Wh batteries...

The bricks themselves are not entirely the problem if you max tier laptop only lasts 30mins during idle away from the wall

[mossy gull]

Most powerful laptop bricks are and have been for over 10 years around 330w.

[cinder lagoon]

Yes, and if more is needed there are two.

330W should be more than enough

For most people at least

[sacred seal]

Well even if they did not limit the size of the batteries, there is a limit to how big of a battery you could physically fit

A 1000 watt laptop is sort of pointless if it dies in 30 seconds

[mossy gull]

But what if people want 1 brick?
Like I prefer 1 brick over 2, because 2 just adds unnecessary bulk.
I basically want a decent and efficient enough laptop on the go, but a true powerhouse at home, but I dislike having 2+ machines but have no choice because of these dumb limitations.

[cinder lagoon]

So a portable workstation

They already exist but are pretty much useless without 1 or even both bricks

[sacred seal]

Its called having a EGPU

[jade scaffold]

a max teir laptop isnt really meant to be a laptop tho its what you would call a desktop replacement

though dual processor laptops would be cool

like a lil arm CPU just to run the os and browser and music apps and a big chunky boi processor to game/video edit or what ever

[mossy gull]

Then give me ePCIe 8x or Thunderbolt that supports PCIe 5.0, but nooooo, we're still stuck with Thunderbolt with PCIe 3.0 4x

[jade scaffold]

thounderbolt can do PCIE 4.0

[mossy gull]

MacBook

[jade scaffold]

MacBooks are single processors

they just have 2 processor dies linked to eachother

i more so meant dual architecture

[cinder lagoon]

Thunderbolt is getting to its physical limits.

[jade scaffold]

though i imagine that would require some serious driver tomfoolery

[cinder lagoon]

You mean like intel GPU plus nVidea but for CPUs?

[jade scaffold]

ehhh more so multiple architectures built on one actual CPU

[cinder lagoon]

Technically, linux and android can already do that, it's CPU hotplug.

[jade scaffold]

like intels big little architecture

but with much more efficient ARM Processors as the little ones

[mossy gull]

Most laptops still carry Thunderbolt 3, which is PCIe 3.0

[cinder lagoon]

ARMs*

[jade scaffold]

rather than the just reduced X86 that intels little processors use

[jade scaffold]

everyone has done it

it doesnt belong to everyone intels just the most prominant advertiser of the thing

[cinder lagoon]

Please google big.LITTLE architecture and look what it exactly is

They are more efficient because they don't need to be as compatible as the "big" ones. They are also clocked down for energy saving.

[jade scaffold]

YES

i know

intels little processors though are still ifaik X86

they are just the bare bones varient of X86

[mossy gull]

1 thing I dislike about big.LITTLE is that MS won't update W10 to support it.
And W11 is bullpoop

[jade scaffold]

where as fully fleged processors have several other extra instructions on them relating to FPUs and other accelerators

[mossy gull]

Like MS wants Windows to be more tablet like.
But where's my goddamn full-screen start menu

Or easily changeable themes

[jade scaffold]

AMDs current gen EPYC can actually go up to 128 cores but it has to loose some of thoes nice to have features to do it but what loosing these features also does is increase efficiency as they take power to run and actually take up physical space

[cinder lagoon]

You want Windows 8.

[sacred seal]

Or windows 10 in tablet mode

[mossy gull]

I have W10 with full-screen start in desktop mode
It's an option actually.

[cinder lagoon]

Nop, they are a reduced instruction set.

[jade scaffold]

they are listed as exclusivly x86

i cant find anything about the little processors being anything other than X86 anywhere

[cinder lagoon]

They crashed when the windows scheduler didn't know about certain missing features

[mossy gull]

Intel's e cores are still X86, they just lack hyper threading and other instruction sets like AVX-512

[cinder lagoon]

So "reduced" as in the regular cores are just bloated?

[jade scaffold]

no

the regular core have vastly more feature support

these extra features are often physical constructs on the CPU core that support things like little said of AVX-512 or even some of the FPUs
the E Cores are actually just X86 no extra instructions or feature which makes they VERY efficient

[cinder lagoon]

Ahh hmm

Tho x86 is already pretty bloated, intel even wants to revise it to remove unused legacy crap

[jade scaffold]

yea

X86 itself is fine

but theres alot of extra features in modern processors

which just arent used

or can be now emulated with minimal performance loss

Some people will miss these features but in intels and AMDs eyes they can get fucked

i coulda swore intel nuked AVX512 on their 12 gen CPUs

[mossy gull]

They actually want to completely remove native 16-bit and 32-bit support and only use 64-bit support

[jade scaffold]

but like the physical Cores for the AVX 512 was still there

[mossy gull]

They did

[jade scaffold]

tbf you can emulate 16 bit and 32bit without any issue atall

[mossy gull]

That's why they want to nuke the native support

It's being used less and less

[jade scaffold]

like the performance loss is near none existent for doing so and the CPUs will likely be able through drivers to emulate without any extra steps

[mossy gull]

Funny enough, they said that removing native support for 32-bit and 16-bit increases 64-bit performance

[jade scaffold]

i think when our current method of designing processors eventually gets stuck all the peformance improvements we will see will come from these kinda little optimisations as well as the addition of new acellerators

[mossy gull]

When you say "accelerators" it reminds me instantly of Xeon Phi

[jade scaffold]

are thoes the weird PCIE Xeon CPUS

[mossy gull]

You mean, bunch of Atom cores smacked onto a PCIe card

[jade scaffold]

pretty much

[mossy gull]

Yeah, funny enough you could game on them though

With some trickery

[jade scaffold]

you could also take the CPUs out of them

and they would go in the standard Xeon socket

the 3647 one at the time i do beleive

[mossy gull]

The ones I held were cooked onto the card

[jade scaffold]

some of them ere yea

like this one

but some looked like this

and despite their weird nubbin they did actually fit into and work in normal Xeon sockets

[mossy gull]

I always loved how wack PC shit was back then

[jade scaffold]

theres also this

what im going to call poor mans hbm

[mossy gull]

I miss HBM consumer GPUs

[jade scaffold]

from one of the weird nubbin havvers

the only reason to put HBM on a consumer GPU is Compactness

and its real expensive so they dont

but i'd like a comically tiny 4090

however i have some potentially good news for you

AMDs MI300 serise should be out soonish

which are full on APUs with high end GPUs and CPUs and 128gb of HBM3 Memory on them

not consumer but probably the best use of HBM memory ever

[mossy gull]

Whoohoo, powerful tiny PCs

[jade scaffold]

maybe

gotta wait to see if ASRack make any waky tiny mobos for it

[mossy gull]

I dislike how big GPUs have become.
I want smaller top of the line cards again.

[jade scaffold]

as thats where all the wacky tiny mobos that support stupid server CPUs come from

their coolers are big but the PCBs on GPUs have actually gotten smaller

like the 3080s PCB is the same size of the R9 furry xs PCB its silly small

the 3090s was a bit taller sadly but only because it had to mount such a massive cooler

i really wanna desodder a Super computers GPU core

[mossy gull]

I am just tired of their oversizeness and the inefficiency of having the VRAM so far away from the GPU.
I remember the Vega being the "VRAM" king for so many freaking years. 16GB of HBM2 memory, and people would look weird at you for having 1, yet only you could play RE1 remake at highest possible settings.

[jade scaffold]

because they have HBM they look enormours real cool looking things

[jade scaffold]

theres no disadvantage other than size to keeping VRAM away from the GPU

the things a consumer would use a gpu for just do not benifit from HBM or ultra low latency

[mossy gull]

I am not the average consumer though, nor do I care about highest gaming performance, as long my simulations run fast.

[jade scaffold]

your sims are unlikely to benifit

[mossy gull]

But they suck tons and tons of VRAM 😃

[jade scaffold]

the things that do benifit are enormous learning models such as AI or Data center level Data management where you are trying to move data at several TB/s

[jade scaffold]

yea not nearly as much as HBM will benifit to you

[mossy gull]

You know the simulations I run are dynamic?
Or well they are when I am at work, but not when I am at home

[jade scaffold]

for HBM3 to really be viable you need calculations that will eat well over 100GB sometimes several hundred GB

otherwise GDDR is more than fast enough

so for memory amounts under really 46gb HBM3 will offer next to no advantages because you just arnt moving data fast enough

when you get to massive simulations like simulating accurately re-entery on a planet thats when you need these enormous GPUs with 128gb of HBM3 and then you need 8 of them linked together through high speed NVLink each connection way is 900GB/s thats when GDDR just isnt fast enough

[mossy gull]

All I know is going from 32GB HBM2 to 12GB GDDR6 was a mistake for my simulations

[jade scaffold]

yea from 32gb to 12 probs wasnt a good switch up

but i bet if you picked up a quadro with 48gb you'd be dandy

i think the new quadros might actually do 96gb

[mossy gull]

Well, lets say, I had 2 working Vega FE, then 1 died, so I went from 32GB to 16GB, sadly AMD stopped supporting them so I was forced to move on, sadly me being a poor sod, I got the 6700XT, damn that was the biggest mistake in a while.

[jade scaffold]

alot of the old Radeon Pro GPUs are going real cheap

[mossy gull]

Depends which 1 will be more powerful the coming years between AMD and NVidia.
Considering it barely makes a difference for the programs I use, as long the drivers work properly.

[jade scaffold]

there was a point where the Pro varient of the 6900xt GPU was actually going for less than just look arround big company

NVidia will be more powerful

[mossy gull]

I know, I need to find a good deal

[jade scaffold]

the 4090 isnt even close to their best GPU

[mossy gull]

But 4 times the price as well

[jade scaffold]

its actually kinda scary how much power the H100 pulls

yes

its like a 500w chip in the SMX5 socket varient

and thats at lower clocks

[mossy gull]

20% performance for 4 times the price is already a no-no for me, As long I got the VRAM I am good.

[jade scaffold]

not sure on the exact specs that power draw might be from having absurd ammounts of tensor cores

hey

maybe the MI300 from AMD will be good for you

[mossy gull]

If I can find 1 for cheap I'll beat my little brother for years

[jade scaffold]

128gb of HBM3 24 CPU cores the GPU can likely be virtualised

though we dont know the exact specs of the GPU

yet

but its likely actually gonna be quite good

[mossy gull]

We'll see, I just currently hate waiting for each dumb sim I do

[jade scaffold]

the GPU Die looks almost as big as the ones on their actual gaming GPUs

[mossy gull]

It's 1 of the reasons my 3D models stay unfinished, I keep running out of VRAM during rendering and sim stuff.

Lets wait 1 hour for 1 bloody frame and lets replace the SSD each half year.

@jade scaffold I found W6800 for 1K😃

Although I may can wait for the next Titan, it should have 32GB or 48GB VRAM

[jade scaffold]

thats not bad

[mossy gull]

It's the same price of the 7900XTX

[jade scaffold]

but with more VRAM i thinks right

[mossy gull]

Yes, what if, just hear me out, what if you could trick the old Crossfire to stretch the VRAM to 64GB

[jade scaffold]

im not ure it'll let you

i dont think crossfire can achive the bandwidth needed for making a big pool of memory

[mossy gull]

It won't, reason I didn't get the 3rd Vega FE after mine broke

And AMD basically stopped supporting it after NVidia stopped supporting their solution on anything but server GPUs

[jade scaffold]

for Nvidias NV link to do that it needs to do 900GB/s which is a interconnect as fast as GDDR6X

i mean the older NVlinks werent that fast

but they were more so meant for 2 GPUs not 8 at the time

[mossy gull]

I mean, NVidia removed NVlink on anything that ain't some fancy dandy high performing AI card

[jade scaffold]

yea

[mossy gull]

I miss having the amazingness of stacked GPUs🥲

The times when you could show off your quad GPU system🥲

[jade scaffold]

yea

3090s double stacked looked REALLY cool

[mossy gull]

Lets get a bunch of 2nd hand Vega FE and have a bronze stack in your PC

Is it me or do I like colours otjers hate

[jade scaffold]

i liked the weird water cooled gold ones

[mossy gull]

I have 2 of them, currently in their box

Their radiators were in the front of my case.

1 had Conductonaut on it and the other Kryonaut, the Kryonaut 1 died sadly enough

[jade scaffold]

just thought of something

you know the weird fans we were talking about before

in theory right shouldnt they pretty much last for ever bar changing out their dust filters

because i know the big versions of them or rather the things they are based off can like live forever pretty much

[mossy gull]

Depends on the motor, you know most fans still use a brushed motor instead of brushless

[jade scaffold]

Not actual fans

The little cooling thingies that blast air at like 120mph or what ever it was

[cinder lagoon]

Piezo-electric vibrating fan-blades

[pearl moon]

Idk if the others mentioned this, as they went on several side tangents that I didn't wanna read through, but you could also upgrade to 32gb ram. It won't be an immediately noticable difference, but it makes the game run smoother and somewhat reduces stutters and such.

[cerulean helm]

Oh, the New ICUE update fixed a long standing issue, the software bloat. No longer does it require 3+ GB of space for the program as it only installs the modules you need for your system. As in it doesn't store hoards of data for devices you don't have installed.

[pearl moon]

nice

bruh i was on v3, and it had me install v4 before it would install v5

[ornate geyser]

What kind of performance increase should I see. And would this be a simple install?

[surreal moss]

Gpu would be a simple install but I don’t think you’ll see much performance improvement. Looking at your specs I want to ask which drive you have SC installed on?

If SC is on you hdd and you don’t have room in the ssd then you can buy another 1 tb ssd to add to your system. Google for me says they start at around $60.

The other thing that sticks out to me is 16 gb ram. Like play3rxthr33 suggested you could look into upgrading/adding additional sticks there as well.

[cinder lagoon]

We have outlined the increase further above but I remember it being somewhere around wacky 15%.

[mossy gull]

Plus it's more efficient if you want a more efficient GPU

[jade scaffold]

Honestly I'd recommend waiting for the 7700xt to drop and just saving up money till then

[cinder lagoon]

Ye

[jade scaffold]

That I hope will have like 16gb of vram and will be substantially faster than your old gpu

[cerulean helm]

Uh, I would have done a fun uninstall (including deleting the Corsair App Data folder) before updating as you may have lingering files over from the old version.

[signal star]

Got 16GB of 3600 CL 16 ram, is there a faster set of RAM and more of it? 64GBs?

[mossy gull]

There's more and there is faster.
But with faster RAM your motherboard and CPU have to be able to support it properly or else you'll have an unstable system

[cinder lagoon]

Faster? Not really. Everything above comes with higher latencies at the benefit of faster transferspeed.
More? Certainly. 2x16G would be the ideal configuration for gaming.

[jade scaffold]

More fun server issues my main switch that I was using decided to behave weirdly because that's what u get for being jank so I went and got my cheap un managed switch as a temporary replacement

My server refuses to even acknowledge it's exsistance

No flashing lights and troubleshooting yeilds the blatantly incorrect statement that there is no ethernet cable plugged in

The server has to have some kind of block I'm not aware off on unmanaged connections

[mossy gull]

Can also be a router/modem problem for not assigning an IP over the unmanaged switch

[jade scaffold]

dont know why it would only not assign an ip to the Server

[mossy gull]

Ask your modem, some modems also have to be reset to understand the change

[jade scaffold]

So it could assign ips to 5 other devices but not the server

The other computers and TV are all good just not the server

[mossy gull]

Then I don't know, I know static IPs can get buggy at times

[jade scaffold]

the only reason i went to use the shittle little unmanaged switch was because the Old router i was using as a switch decided it wanted to try Eat the fucking network

[mossy gull]

Then check how everything's set-up

[cinder lagoon]

Discord now supports basic Markdown:

• Headings
• unordered lists
• Hyper Links (supposedly... seemingly not.)

https://www.markdownguide.org/tools/discord/

[mossy gull]

https://www.pcgamer.com/if-youve-got-a-modern-gigabyte-motherboard-theres-a-bios-setting-you-need-to-disable-to-avoid-pcs-latest-security-calamity/

[cinder lagoon]

Afaik there is no setting and it's basically a root kit

But the fixes should have automatically installed themselves given the root kit :topkek:

[sacred seal]

The article is confusing, is the issue with the motherboard or there stupid sofware suite?

[sonic meadow]

(about half way down) it sounds like its about the auto bios updater in their software suite

but the option to disable it is apparently in the bios?

[cinder lagoon]

The motherboard has something embedded in the BIOS that auto installs to System32 which is basically an auto BIOS updater but the webrequests are catastrophically insecure

And it apparently does not have a way to turn it off

[sonic meadow]

and mine isnt on the list. it could do with a bios update though, currently running a late 2020 one 😆

[cinder lagoon]

The issue exists on like 270 models of Motherboards

[sacred seal]

Yea, basically anything 400 series or newer.
Automatic bios update seems like a very dumb idea, given the difficulty and risk involved in updating a bios normally

[sonic meadow]

I think that was how I bricked my last motherboard...

either that or it wanted to take half an hour for ram training /shrug

[jade scaffold]

not necessarily you can build protections to prevent any kind of corruption while updating not sure if gigbyte ones have such protections though

[cinder lagoon]

They made it a safe process with the online exec tho

[jade scaffold]

Opinions on Apples Vission pro? Both technologically and possible future implications for VR/AR headset development as a whole and any Hopes for the Device.

Personally other than the goofy front facing screen I actually really like the look of it
the the panels and lenses as well as the tracking are likely gonna be the some of the best ever seen on VR Headsets
in addition i love the use of actual good materials on the headset and it actually looks super comfy and actually suitable to lay down with.
My main hope i think is support for Open source standards as well as standard PC support but this is apple we are talking about so it very well could be locked into their own ecosystem which would be a real shame i think. I really hope the release of this headset as well as the Meta quest pro at its current pricing cause valve to get their heads out of their ass and build a more modern headset with new lenses and screens.

Another thing i love about it thats a huge upside when compared to certain headsets like the meta quest pro is apple show it being very modular in the way of the head strap speakers and face cushion which i think is a thing the Meta Quest Pro Should have really embraced more as not much other than the cussions on the pro can be swapped out which is a shame really

[pearl moon]

I agree, it needs PC support, not just Mac support, and some way of modifying it like people do now with Quest 2s.

and not be $1000 for no reason

[jade scaffold]

apple showed it being modular as far as head band and face cushion go which is really good that way people can get the fit they want on it

but also i dont think alot of people will do that unless they want like a head mounted battery pack

because i gotta be honest that headstrap with how its built looks really dam comfy

apple are also boasting a 12ms response time from doing somthing to it getting to your eyes

which is AMAZING

i really hope this sparks some good competition and causes Valve to really strive to build an absolute baller VR Headset

[pearl moon]

That shit is $3500

I'll pass

[jade scaffold]

oh yea its expensive

[pearl moon]

I can guarantee $1500 of that is Apple name alone

[jade scaffold]

but its also substantially better than the previous holder of the title "Best VR Headset" while somehow being cheaper than it

so while no doubt it has that apple tax on it

[pearl moon]

we need competition in the $500ish range, not at the top

[jade scaffold]

oh yea

but i mean more in the 1000 range for competition for the still somewhat available to your average consumer if they want to save up for it

and the quest 3 and PSVR2 pretty much have a choke hold on the $500 range

theres very little that can beat them

[pearl moon]

Does PSVR2 work on PC?

[jade scaffold]

in theory you could make somthing with the same specs as the Quest 3 minus the whole ass computer thats in it and sell that as a PC VR headset which would imidiatly obliterate the Indexes Value propesition past its controlers

[jade scaffold]

some modders have made a steam tool that works i believe

[pearl moon]

Oh 100%, they could easily take $100+ off the pricetag if they removed all the guts and just made it a PC only headset.

[jade scaffold]

yea

and the Quest 3 the head set part of it atleast

is just superior to the index in every way but refresh rate

and like do you really care about 24hz

or rather is sacrificing 24hz worth vastly superior visual clarity

i just wish the Quest controlers were a little more like the valve knuckle controlers

[pearl moon]

It's probably more like the system they have inside it couldn't handle 24 more fps

[jade scaffold]

its not that its just they use 120hz screens

the 120hz mode is made mainly for PCVR use

[pearl moon]

ah

[jade scaffold]

and the quest 3 at $500

vs the Indexs $1000

[pearl moon]

Index headset only is $500

[jade scaffold]

yea but a quest 3 will be 500 including controlers

and you dont need base stations for it to work

for a fully functional index system you need to spend atleast $1000

[pearl moon]

yeah, but atleast for me i'd be almost exclusively be headset only for DCS World and Racing Sims

[jade scaffold]

yes but you still need base stations

and you need 2 of them for the headset to track

which is another $300? ish

[pearl moon]

I didn't think you did for the small amount of movement a seated position has.

[jade scaffold]

they have gyros but trust me if you loose headset tracking and it has to rely on gyros all racing sims and whats not will cause you to drift a bit

also it makes you kinda motion sick

[pearl moon]

Though one thing i would be willing to spend money on is some vr gloves

[jade scaffold]

and for just the headset alone $500 for the index is dreadfull value

[pearl moon]

oh yeah i'm not saying index is good at $500

[jade scaffold]

its using Older frenell lenses and its screen resolution... oof

the res on the index is super low compared to even the quest 2

[pearl moon]

ah yes, me and Fresnel go way back

[jade scaffold]

1440x1600 per eye for a combined 2880x1600 overall

[pearl moon]

I do technical theatre, and Fresnel is a family of light fixures

[pearl moon]

On which one?

[jade scaffold]

thats the Index

[pearl moon]

what's Quest 3?

and Quest 2?

All this VR talk makes me want to ask my brother if I could borrow his Quest 2 he has lying in a box, probably buried in his closet somewhere.

[jade scaffold]

im not sure we know im seeing some numbers but the Quest 2 is 1832x1920 per eye for 3664x1920 overall, the Quest 3 is rumored to be 2064 x 2208 per eye or 4128x2208 overall

hey if you can test drive a VR headset before you buy your own thats probably the best way to do it

that way you know if VR is right for you

[pearl moon]

Yeah

[jade scaffold]

personally i started on a Oculus CV1

my friend sold it to me for £70

even came with an extra sensor which like thats a bargin

now i run a Quest Pro and the screens on the Quest pro despite not actually being that high resolution actuall loosing 32 pixles horizontally per eye over the Quest 2 look awesome full array local dimming supported both in stand alone and Link mode leads to some truly Sexy visuals in games

[pearl moon]

The main thing that will decide if i wanna get a quest 3 is if/when the community figures out how to bypass the need to login to facebook on it

[mossy gull]

Before you go crazy on the price.
The Apple Vision Pro is in the same ballpark as the Microsoft Hololens 2.
Then the displays Apple use are more or less 1 of a kind and there's no VR headset with displays of the same quality.
Yes it's overpriced, but it's only competition literally stopped producing for consumers.

Besides, the Hololens 2 is 3500.- and the Industrial version is 5000.-

[jade scaffold]

the Holo Lense is also a very different type of device

the Apple Vision can do full VR as well as AR much like the Quest pro

it just has substantially better build quality and an ungodly ammount of cameras and sensors

though meta need to real quick coppy apples menu function they are using on the Vision pro

on the vision pro you just look at things and tap your finger and thumb together

and like the Quest pro Could 100% do that and that would make navigating menus so insanely fast

[jade scaffold]

just make a burner facebook account

im honestly not too sure why anyone cares

"but oh my data" valve and literally everyone else do the EXACT same shit as Meta do, meta are just more well known for doing so

[mossy gull]

I tried 1 a few times and felt the Quest Pro to feel cheap and it's still ungodly huge.

Not to mention, everywhere I look, the inside out tracking of the Quest Pro is in black and white and looks pretty bad.
I don't call that a proper AR.

[jade scaffold]

Then you havn't tried a quest Pro?

The quest Pro is is full colour

[mossy gull]

Doesn't feel much different to me, still feels laggy

[jade scaffold]

And the pro is tiny what are you on

[mossy gull]

I still feel its big and bulky

[jade scaffold]

It's smaller than a quest 2

Which is already pretty small when compared to am index

And tracking latency lower than the index

I think who ever quest Pro you tried was busted af

[mossy gull]

What all VR headsets have in common for me is they feel bulky and make me feel nauseous from motion sickness

[jade scaffold]

Yea that's never gonna change then

[mossy gull]

I have worked with the Hololens, even though bulky, I like playing around with holograms that don't feel all jittery with each movement I make.

[jade scaffold]

I think you may be talking out your ass the hollo lens was natoriously jittery

[mossy gull]

Oh I know, VR headsets haven't changed much, but with Hololens it more or less was less visible to me because it "blends" in with the surroundings.

[jade scaffold]

I really don't see how you can try a headset that has full colour tracking then think it's black and white

That's not a mistake that someone who's actually tried a product should be able to make

[mossy gull]

It wasn't my own headset, I don't know how someone set something up.
Nor is communication easy when you can't properly see what you're typing on your phone

[jade scaffold]

I think you might be mixing up the pro with a different headset

[sonic meadow]

I have vague recollections of the passthrough on my vive pro being rather washed out, but that wasnt intended to really be used to actually look at or do stuff

[mossy gull]

I think it was the Meta Pro, glossy front with 2 weird eyes, battery pack in the back

I so far only use the Hololens 2 extensively because that's the only AR headset not making me nauseous.

Keep in mind Ibif anything Apple says its true, the AR feature would be better than any headset currently on the market

[jade scaffold]

The pro is built for vr and ar on my pro I an use my monitor just fine while wearing it

It's cams do really blow out room lighting tho for some reason even though it can adapt fine to monitors

[mossy gull]

Maybe it's my phone where I rarely crank up the display brightness that make it appear (near) black

@jade scaffold do VR headsets still use LCD displays?

I rarely look into them but that can cause my nausea the most

[jade scaffold]

It depends

The quest Pro is micro led

The psvr 2 is OLED

The new apple one is micro OLED

[mossy gull]

Micro LED, aka LCD with full-array local dimming

PSVR, I never tried it because I lack the PlayStation

Was it the Vive that had an OLED?
Like the original Vive

[jade scaffold]

I don't think so

Oleds look pretty bad with frenell lenses

Though if you use pancake lenses with OLEDs you need much higher resolutions which is why the vision pro is micro OLED with a 4k res per eye

Well over 4k resolution

[sonic meadow]

I think it was the index which had an OLED display

[mossy gull]

I see LCDs flicker, I don't have that problem with OLEDs.

[jade scaffold]

Nope

[sonic meadow]

led backlit lcds? ccfl backlit lcds? all of the above?

[jade scaffold]

Index is lcd

[sonic meadow]

oh yeah. it used ips panels when other stuff wasnt

[jade scaffold]

Yee

[sonic meadow]

people were worried it would have a more noticeable screendoor effect

[jade scaffold]

Micro OLED +pancake lenses is the dream

[mossy gull]

Yes, I see all LCDs flicker.
With IPS and TN panels being really bad for me.

[sonic meadow]

huh. weird. then again i know some people are way more susceptible to flicker than others

had never heard it being panel related though, usually backlight

with some techs (like bfi) being substantially worse for them

[jade scaffold]

Yea also oleds strobe and flicker well the good ones do

[mossy gull]

They do, but not like LCDs

[jade scaffold]

If oleds don't strobe then they are susceptible to burn in which might not be a huge problem with a vr headset

Oleds strobe weirdly they act like sideways crts

[mossy gull]

LG uses active cooling to prevent burn-in

[sonic meadow]

but burn-in isnt heat related

its wear essentially