Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Nvidia announces next-gen RTX 5090 and RTX 5080 GPUs

theverge.com

Post reply on HN
Print view View original

771–776 of 776 posts

Re: Nvidia announces next-gen RTX 5090 and RTX 5080 GPUs

#771

Post by menaerus »

janwas wrote: post #765
ryao wrote: post #764

Earlier quoted context omitted.

The purpose of L1 cache is to avoid long round trips to memory. What you describe is L1 cache doing what it is intended to do. Unfortunately, I do not have your code, so it is not clear to me that it benefits from doing 2 AVX-512 loads per cycle. I am also not sure what CPU this is. On recent AMD processors at the very least, it should be impossible to get FMA throughput that is 50 times higher from L1 cache bandwidt…

The code is unfortunately not (yet) open source. The CPU with 50x is an SKX Gold, and it is similar for Zen4. I compute this ratio as #FMA * 4 / total system memory bandwidth. We are indeed not fully memBW bound :)

I'd be curious if you measured 50x on a single core implementation or is the algorithm distributed to multiple cores?

I ask because you say that the results are similar to Zen4 so this would sorta imply that you run and measure single-core implementation? Intel in multi-core load-store looses a lot of bandwidth when compared to Zen3/4/5 since there's a lot of contention going on due to Intel cache architecture.

Re: Nvidia announces next-gen RTX 5090 and RTX 5080 GPUs

#772

Post by snvzz »

SirMaster wrote: post #732
snvzz wrote: post #626

Earlier quoted context omitted.

I have seen AMD's PR materials for RDNA4, and as far as I can tell, they do not say anywhere anything like that. People read too much into "designed for RDNA4".

https://cdn.videocardz.com/1/2025/01/AMD-FSR4-9070.jpg Why would they write that on their marketing slides?

Because it only works on these cards right now.

Further elaborated by their GPU marketing people on interviews. To summarize "RDNA4 for now" and "we're looking into supporting older...".

Re: Nvidia announces next-gen RTX 5090 and RTX 5080 GPUs

#773

Post by officeplant »

brokenmachine wrote: post #635
officeplant wrote: post #446

Earlier quoted context omitted.

Cheaper options like the Steam Deck are definitely a boon to the industry. Especially the idea of "good enough" gaming at lower resolutions on smaller screens. Personally, I just don't like that its attached to steam. Which is why I can be hesitant to suggest consoles as well now that they have soft killed their physical game options. Unless you go out of your way to get the add-on drive for PS5, etc Its been nice to…

>Unless you go out of your way to get the add-on drive for PS5 Just out of interest, if I bought a PS5 with the drive, and a physical game, would that work forever (just for single-player games)? Like you, I like to own the things I pay for, so it's a non-starter for me if it doesn't.

In my experience it varies on a game by game basis. Some games have limitations (ie. Gran Turismo 7 having only Arcade mode offline)

Re: Nvidia announces next-gen RTX 5090 and RTX 5080 GPUs

#774

Post by duffyjp »

brokenmachine wrote: post #633
duffyjp wrote: post #239

Earlier quoted context omitted.

I got 4K TVs for both of my kids, they're dirt cheap-- sub $200. I'm surprised the Steam hardware survey doesn't show more. A lot of my friends also set their kids up on TVs, and you can't hardly buy a 1080P TV anymore.

Does Steam hardware survey show the resolution of your usual desktop, or your gaming resolution? eg I run at 4k in Windows normally, but quite often run games at 1080p.

I'd bet it's either the native display resolution or whatever you had for your desktop when submitted. They're able to gather all kinds of hardware specs so I'd lean to the native resolution as the most likely answer.

Re: Nvidia announces next-gen RTX 5090 and RTX 5080 GPUs

#775

Post by nahnahno »

vonneumannstan wrote: post #680
nahnahno wrote: post #321

Earlier quoted context omitted.

Yes there should be. We don’t want to pay literal 10x markup because the card is suddenly “enterprise”.

Totally unreasonable expectation. Sry. The cards are literally built for gamers for gaming. That they work for ML is a happy coincidence.

You can’t possibly be naive enough to believe that Nvidia’s Titan class cards were designed exclusively for gamers.

Re: Nvidia announces next-gen RTX 5090 and RTX 5080 GPUs

#776

Post by berbec »

berbec wrote: post #743
williamDafoe wrote: post #399

Earlier quoted context omitted.

The 3090 + 5900x is a mistake. The 5900x is 2 x 5600x CPUs. So therefore, when the games asks for 8 cores, it will get 6 good cores and 2 very slow cores across the infinity switching fabric. What's more, NVidia GPUs take MUCH MORE CPU than AMD GPUs. You should either buy an AMD GPU or upgrade/downgrade to ANYTHING OTHER THAN 5900x with 8+ cores (5800x, 5800, 5700, 5700x3d, 5950x, 5900xt, anything really ...)

I've been looking hungrily at a 5700x3d, but just can't justify spending more money on AM4. I'm saving up for a 9800x3d. I've considered just disabling a CCD to see how that works.

disabled once CCD. got 10fps in Limsa

bought a 5700x3d. got 30fps in Limsa

3D V-cache is magic.

Post reply on HN