[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

2022-05-12: Ghost posting is now globally disabled. 2022: Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/3/ - 3DCG

View post   
View page     

[ Toggle deleted replies ]
File: 2.36 MB, 720x720, nvidia-geforce-rtx-40.webm [View same] [iqdb] [saucenao] [google]
918920 No.918920 [Reply] [Original]

How are we feeling about this, /3/?


>> No.918925

>only 24GB
>only 16GB
>only 12GB

>> No.918928


>we feeling Who the fuck cares about feelings are you a woman?

>> No.918929

Not even paper launch, straight up vaporware that's how i feel. Fuck electronics.

>> No.918962

not /3/

fuck off

>> No.918963

very underwhelming

>> No.918964

>hardware for accelerating 3D rendering
>utilized in 3D rendering applications
>not /3/

>> No.918966

>brings absolutely nothing new to 3D, just more computing power and processing speed
Not /3/

>> No.918970

I don't even know when I'm gonna update my PC nor what GPU to buy anymore. I've been using my laptop for years at this point, I've got gtx 960 in my desktop lol. I can't even use Nanite in Unreal, performance is not the best either. And I realy don't want to wait ANOTHER 2 years for 5xxx series.

>> No.918972

take it to /g/

closing this thread.

>> No.918978

we typically discuss cut down castrated bricks for goymers with "rtx" fad cores in /v/ or /g/

>> No.918979

They had to gimp the 4080 12gb didn't they.

>> No.918983

Incoming 4070 with 2 variants, 10 GB and 8 GB
$749 for the 10 GB and $649 for the 8GB.

Such AI, much models.

>> No.919001

>compares baseline without DLSS vs new DLSS
>refuses to elaborate further

to the trash it goes

>> No.919008

this: >>919001. probably a 15% uplift for rendering and general compute at best

i think the 4080 12gb is just a last minute rebrand of what was supposed to be a 4070

>> No.919024


>> No.919126

Between getting a 3090 and a 4090, is it worth the price difference? Does DLSS 3 benefit us in something?

>> No.919162

t bh 2x 3090 + nvlink might actually be the secret smart move depending on your memory requirements.

>> No.919166

This. This is what I've done. 24GB is such a large limitation even with the performance a 4090 could give you.

I bought a second 3090 2nd hand as soon as I heard they had dropped NV-Link for the 4000 series

>> No.919173

NV llink doesnt really do anything for multi gpu rendering.

>> No.919195

Despite having 3x more transistors and paper teraflops, raw performance seems to be barely higher than Ampere. Looks like all those transistors went to DLSS 3.0. They get "4x more performance" by rendering every 4th frame and interpolating the rest. Not helpful to us in any way.

>> No.919213

I'm >>919126
what would you say then?

>> No.919215

iirc scenes need to be loaded completely onto each gpu so you'd get capped at 24 gigs despite having multiple gpus in the box. nvlink would pool the memory and allow larger scenes. there is a perf cost i believe, but it's better than out of core.

go ask somewhere smarter than here like the discord community of your renderer of choice.

otoy's making some big claims about perf but they shill pretty hard. it'll take awhile before the renderers update anyway.

>> No.919218

even without nv link each gpu gets things loaded into it. what nvlink used to do was to pool it together true, but now its pretty much a gimmick.
You would get about 1-2% increased efficiency with NVlink - data gets loaded into each card either way.

currently You can get 3090TI for 1000 bucks if not less if used.

I can only speculate but it is not worth to spend 600+ bucks more for 10-15% increased workspeed. you might as well get 2x 3090's.

>> No.919229

>Does DLSS 3 benefit us in something?
DLSS 3.0 uses motion vectors to create blend frames, if you want 1 frame performance for super fast renders this wont do shit for you, and even in animation rendering there is nothing to be gained

>> No.919235
File: 32 KB, 1281x608, studio-ada-perf-chart-3d-blog-1280w-dark.png [View same] [iqdb] [saucenao] [google]

>> No.919243

>>3.5 slots
it's a chunky boi

>> No.919341

>nvidia provided comparisons
I would be very careful anon.

>> No.919366

Good point, but at least they can't cheat with DLSS here.

>> No.919369
File: 1.28 MB, 720x900, Nvidyas_dancve.webm [View same] [iqdb] [saucenao] [google]

Hopefully there's an affordable 4070 with 16gb for offline render. Maybe a sub $1000 3090 would be best bang for buck with both offline and UE5?

>> No.919379

I doubt it considering the base 4080 is only 12gb. We will have to wait for the TI variants of these, which will probably be 2 years away, seeing how the market is going.

>> No.919432

watch it them comparing the image clarity for long render without AI denoiser
comparable quality with new AI denosier.

most of the time their AI denoiser is only good for stills and not animation/motion (see glitches during gameplay)

>> No.920052

>they had dropped NV-Link for the 4000 series
So.. if I have space in my motherboard for 2 GPUs then it's pointless if I have 4 series?

I can't buy two 4090 and make a pool of VRAM?

>> No.920064

It was always pointles buy A6000 goy

>> No.920092


>> No.920114

You are retarded. The biggest reason to pay a premium for those is on-demand driver support, which only enterprises need.

>> No.920136

>buy A6000

Delete posts
Password [?]Password used for file deletion.