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

181–190 of 776 posts

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

#181

Post by FuriouslyAdrift »

smallmancontrov wrote: post #25
kcb wrote: post #23

Earlier quoted context omitted.

Yup, the days of the value high end card are dead it seems like. I thought we would see a cut down 4090 at some point last generation but it never happened. Surely there's a market gap somewhere between 5090 and 5080.

Yes, but Nvidia thinks enough of them get pushed up to the 5090 to make the gap worthwhile. Only way to fix this is for AMD to decide it likes money. I'm not holding my breath.

AMD announced they aren't making a top tier card for the next generation and is focusing on mid-tier.

Next generation, the are finally reversing course and unifying their AI and GPU architectures (just like nVidia).

2026 is the big year for AMD.

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

#182

Post by munchbunny »

zeroonetwothree wrote: post #30
ericfrederich wrote: post #18

Earlier quoted context omitted.

4k gaming is dumb. I watched a LTT video that came out today where Linus said he primarily uses gaming monitors and doesn't mess with 4k.

Yep. I have both 4k and 1440p monitors and I can’t tell the difference in quality so I always use the latter for better frames. I use the 4k for reading text though, it’s noticeably better.

That's why I also finally went from 1920x1200 to 4k about half a year ago. It was mostly for reading text and programming, not gaming.

I can tell the difference in games if I go looking for it, but in the middle of a tense shootout I honestly don't notice that I have double the DPI.

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

#183

Post by FuriouslyAdrift »

blitzar wrote: post #133
kaibee wrote: post #47

Earlier quoted context omitted.

Don't necessarily count Intel out.

Intel hate making money even more than AMD.

Intel's Arc B580 budget card is selling like hotcakes... https://www.pcworld.com/article/2553897/intel-arc-b580-revie...

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

#184

Post by vel0city »

perching_aix wrote: post #120
dagmx wrote: post #48

Earlier quoted context omitted.

It’s 3 extrapolated frames not interpolated. So would be reduced lag at the expense of greater pop-in. There’s also the new reflex 2 which uses reprojection based on mouse motion to generate frames that should also help, but likely has the same drawback.

> It’s 3 extrapolated frames not interpolated. Do you have a source for this? Doesn't sound like a very good idea. Nor do I think there's additional latency mind you, but not because it's not interpolation.

Interpolation means you have frame 1 and frame 2, now compute the interstitial steps between these two.

Extrapolation means you have frame 1, and sometime in the future you'll get a frame 2. But until then, take the training data and the current frame and "guess" what the next few frames will be.

Interpolation requires you to have the final state between the added frames, extrapolation means you don't yet know what the final state will be but you'll keep drawing until you get there.

You shouldn't get additional latency from generating, assuming it's not slowing down the traditional render generation pipeline.

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

#185

Post by rfoo »

boroboro4 wrote: post #172
lemoncookiechip wrote: post #159

I have a feeling regular consumers will have trouble buying 5090s. RTX 5090: 32 GB GDDR7, ~1.8 TB/s bandwidth. H100 (SXM5): 80 GB HBM3, ~3+ TB/s bandwidth. RTX 5090: ~318 TFLOPS in ray tracing, ~3,352 AI TOPS. H100: Optimized for matrix and tensor computations, with ~1,000 TFLOPS for AI workloads (using Tensor Cores). RTX 5090: 575W, higher for enthusiast-class performance. H100 (PCIe): 350W, efficient for data cente…

H100 has 3958 TFLOPS sparse fp8 compute. I’m pretty sure listed tflops for 5090 are sparse (and probably) fp4/int4.

Yes, that's the case. Check the (partial) spec of 5090 D, which is the nerfed version for export to China. It is marketed as having 2375 "AI TOPS".

BIS demands it to be less than $4800 TOPS \times Bit-Width$, and the most plausible explanation for the number is - 2375 sparse fp4/int4 TOPS, which means 1187.5 dense TOPS for 4 bit, or $4750 TOPS \times Bit-Width$.

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

#186

Post by bee_rider »

lemoncookiechip wrote: post #159

I have a feeling regular consumers will have trouble buying 5090s. RTX 5090: 32 GB GDDR7, ~1.8 TB/s bandwidth. H100 (SXM5): 80 GB HBM3, ~3+ TB/s bandwidth. RTX 5090: ~318 TFLOPS in ray tracing, ~3,352 AI TOPS. H100: Optimized for matrix and tensor computations, with ~1,000 TFLOPS for AI workloads (using Tensor Cores). RTX 5090: 575W, higher for enthusiast-class performance. H100 (PCIe): 350W, efficient for data cente…

How well do these models do at parallelizing across multiple GPUs? Is spending $4k on the 5090 a good idea for training, slightly better performance for much cheaper? Or a bad idea, 0x as good performance because you can’t fit your 60GB model on the thing?

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

#187

Post by robin_reala »

MuffinFlavored wrote: post #144
codespin wrote: post #20

Earlier quoted context omitted.

At what point do we stop calling them graphics cards?

How long until a "PC" isn't CPU + GPU but just a GPU? I know CPUs are good for some things that GPUs aren't and vice versa but... it really kind of makes you wonder. Press the power button, boot the GPU? Surely a terrible idea, and I know system-on-a-chip makes this more confusing/complicated (like Apple Silicon, etc.)

“Press the power button, boot the GPU” describes the Raspberry Pi.

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

#188

Post by FuriouslyAdrift »

sliken wrote: post #122
fulafel wrote: post #118

Earlier quoted context omitted.

Doesn't it always. (Except sometimes on some hw you can't turn it off)

I believe the cards that are intended for compute instead of GPU default to ECC being on and report memory performance with the overheads included.

Anything with DDR5 or above has built in limited ECC... it's required by the spec. https://www.corsair.com/us/en/explorer/diy-builder/memory/is...

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

#189

Post by Cumpiler69 »

KMnO4 wrote: post #174
ryao wrote: post #97

Earlier quoted context omitted.

If I recall correctly, the 3090, 3090 Ti and 4090 were supposed to replace the Titan cards that had been Nvidia's top gaming cards, but were never meant for gaming.

Someone very clever at Nvidia realized that if they rename their professional card (Titan) to be part of their "gaming" line, you can convince adults with too much disposable income that they need it to play Elden Ring. I didn't know of anyone who used the Titan cards (which were actually priced cheaper than their respective xx90 cards at release) for gaming, but somehow people were happy spending >$2000 when the 309…

>but somehow people were happy spending >$2000 when the 3090 came out

Of course they did, the 3090 came out at the height of the pandemic and crypto boom in 2020, when people were locked indoors with plenty of free time and money to spare, what else where they gonna spend it on?

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

#190

Post by Hilift »

jsheard wrote: post #4

32GB of GDDR7 at 1.8TB/sec for $2000, best of luck to the gamers trying to buy one of those while AI people are buying them by the truckload. Presumably the pro hardware based on the same silicon will have 64GB, they usually double whatever the gaming cards have.

100% you will be able to buy them. And receive a rock in the package from Amazon.
Post reply on HN