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

321–330 of 776 posts

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

#321

Post by nahnahno »

vonneumannstan wrote: post #304
blixt wrote: post #156

Pretty interesting watching their tech explainers on YouTube about the changes in their AI solutions. Apparently they switched from CNNs to transformers for upscaling (with ray tracing support) if I understood correctly though for frame generation makes even more sense to me. 32 GB VRAM on the highest end GPU seems almost small after running LLMs with 128 GB RAM on the M3 Max, but the speed will most likely more than…

If you want to run LLMs buy their H100/GB100/etc grade cards. There should be no expectation that consumer grade gaming cards will be optimal for ML use.

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

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

#322

Post by shadowpho »

simondotau wrote: post #73
ryao wrote: post #61

Earlier quoted context omitted.

Leaks indicate that the PCB has 14 layers with a 512-bit memory bus. It also has 32GB of GDDR7 memory and the die size is expected to be huge. This is all expensive. Would you prefer that they had not made the card and instead made a lesser card that was cheaper to make to avoid the higher price? That is the AMD strategy and they have lower prices.

That PCB is probably a few dollars per unit. The die is probably the same as the one in the 5070. I've no doubt it's an expensive product to build, but that doesn't mean the price is cost plus markup.

>That PCB is probably a few dollars per unit.

It’s not. 14L PCB are expensive. When I looked at Apple cost for their PCB it was probably closer to $50, and they have smaller area

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

#323

Post by kllrnohj »

Havoc wrote: post #227
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…

> regular consumers will have trouble buying 5090s. They’re not really supposed to either judging by how they priced this. For non AI uses the 5080 is infinitely better positioned

> For non AI uses the 5080 is infinitely better positioned

...and also slower than a 4090. Only the 5090 got a gen/gen upgrade in shader counts. Will have to wait for benchmarks of course, but the rest of the 5xxx lineup looks like a dud

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

#324

Post by Terretta »

logankeenan wrote: post #264
terhechte wrote: post #179

Earlier quoted context omitted.

If you have 128gb ram, try running MoE models, they're a far better fit for Apple's hardware because they trade memory for inference performance. using something like Wizard2 8x22b requires a huge amount of memory to host the 176b model, but only one 22b slice has to be active at a time so you get the token speed of a 22b model.

Do you have any recommendations on models to try?

Mixtral 8x22b https://mistral.ai/news/mixtral-8x22b/

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

#325

Post by izacus »

alkonaut wrote: post #136
thefz wrote: post #128

> GeForce RTX 5070 Ti: 2X Faster Than The GeForce RTX 4070 Ti 2x faster in DLSS . If we look at the 1:1 resolution performance, the increase is likely 1.2x.

That's what I'm wondering. What's the actual raw render/compute difference in performance, if we take a game that predates DLSS?

Based on non-DLSS tests, it seems like a respectable ~25%.

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

#326

Post by MisoRamen »

themaninthedark wrote: post #236
sigmoid10 wrote: post #218

Earlier quoted context omitted.

Nvidia also doesn't make the "core" (i.e. the actual chip). TSMC and Samsung make those. Nvidia designs the chip and (usually) creates a reference PCB to show how to make an actual working GPU using that chip you got from e.g. TSMC. Sometimes (especially in more recent years) they also sell that design as "founders" edition. But they don't sell most of their hardware directly to average consumers. Of course they also…

It seems that they have been tightening what they allow their partners to do, which caused EVGA to break away as they were not allowed to deviate too much from the reference design.

It is an ever uphill battle to compete with Nvidia as a AIB partner.

Nvidia has internal access to the new card way ahead of time, has aerodynamic and thermodynamic simulators, custom engineered boards full of sensors, plus a team of very talented and well paid engineers for months in order to optimize cooler design.

Meanwhile AIB partners is pretty much kept in the blind until a few months in advance. It is basically impossible for a company like EVGA to exist as they pride themselves in their customer support - the finances just does not make sense.

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

#327

Post by iandanforth »

jbarrow wrote: post #288

The increasing TDP trend is going crazy for the top-tier consumer cards: 3090 - 350W 3090 Ti - 450W 4090 - 450W 5090 - 575W 3x3090 (1050W) is less than 2x5090 (1150W), plus you get 72GB of VRAM instead of 64GB, if you can find a motherboard that supports 3 massive cards or good enough risers (apparently near impossible?).

Sounds like you might be more the target for the $3k 128GB DIGITS machine.

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

#328

Post by robotnikman »

paxys wrote: post #41
codespin wrote: post #20

Earlier quoted context omitted.

At what point do we stop calling them graphics cards?

Nvidia literally markets H100 as a "GPU" ( https://www.nvidia.com/en-us/data-center/h100/ ) even though it wasn't built for graphics and I doubt there's a single person or company using one to render any kind of graphics. GPU is just a recognizable term for the product category, and will keep being used.

The Amazon reviews for the H100 are amusing https://www.amazon.com/NVIDIA-Hopper-Graphics-5120-Bit-Learn...

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

#329

Post by izacus »

popcalc wrote: post #314
jmyeet wrote: post #46

The interesting part to me was that Nvidia claim the new 5070 will have 4090 level performance for a much lower price ($549). Less memory however. If that holds up in the benchmarks, this is a nice jump for a generation. I agree with others that more memory would've been nice, but it's clear Nvidia are trying to segment their SKUs into AI and non-AI models and using RAM to do it. That might not be such a bad outcome…

Was surprised to relearn the GTX 980 premiered at $549 a decade ago.

Which is 750$ in 2024 adjusted for inflation and you got a card that's providing 1/3 of performance of a 4070Ti at equal price range. 1/4 with 5070Ti probably.

3x the FPS at same cost (ignoring AI cores, encoders, resolutions, etc.) is a decent performance track record. With DLSS enabled the difference is significantly bigger.

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

#330

Post by cogman10 »

jbarrow wrote: post #288

The increasing TDP trend is going crazy for the top-tier consumer cards: 3090 - 350W 3090 Ti - 450W 4090 - 450W 5090 - 575W 3x3090 (1050W) is less than 2x5090 (1150W), plus you get 72GB of VRAM instead of 64GB, if you can find a motherboard that supports 3 massive cards or good enough risers (apparently near impossible?).

What I really don't like about it is low power GPUs appear to be a thing of the past essentially. An APU is the closest you'll come to that which is really somewhat unfortunate as the thermal budget for an APU is much tighter than it has to be for a GPU. There is no 75W modern GPU on the market.
Post reply on HN