281–290 of 776 posts
Translation: No significant actual upgrade.
Sounds like we're continuing the trend of newer generations being beaten on fps/$ by the previous generations while hardly pushing the envelope at the top end.
A 3090 is $1000 right now.
Earlier quoted context omitted.
That's what I'm wondering. What's the actual raw render/compute difference in performance, if we take a game that predates DLSS?
We shall wait for real world benchmarks to address the raster performance increase. The bold claim "5070 is like a 4090 at 549$" is quite different if we factor in that it's basically in DLSS only.
Ooo, that means its probably time for me to get a used 2080, or maybe even a 3080 if I'm feeling special
Even though they are all marketed as gaming cards, Nvidia is now very clearly differentiating between 5070/5070 Ti/5080 for mid-high end gaming and 5090 for consumer/entry-level AI. The gap between xx80 and xx90 is going to be too wide for regular gamers to cross this generation.
The only difference is scalar. That isn't differentiating, that's segregation. It won't stop crypto and LLM peeps from buying everything (one assumes TDP is proportional too). Gamers not being able to find an affordable option is still a problem.
Earlier quoted context omitted.
From all the communication I’ve had with Nvidia, the prevailing sentiment was that the 4090 was an 8K card, that happened to be good for AI due to vram requirements from 8K gaming. However, I’m a AAA gamedev CTO and they might have been telling me what the card means to me .
Why does 8K gaming require more VRAM? I think the textures and geometry would have the same resolution (or is that not the case? but in 4K if you walk closer to the wall you'd want higher texture resolution as well anyway, if the graphics artists have made the assets at that resolution anyway) 8K screen resolution requires 132 megabytes of memory to store the pixels (for 32-bit color), that doesn't explain gigabytes…
Ooo, that means its probably time for me to get a used 2080, or maybe even a 3080 if I'm feeling special
Why not go for AMD? I just got a 7900XTX for 850 euros, it runs ollama or comfyUI via WSl2 quite nicely.
I'm also trying to tie together different hardware specs to model performance, whether that's training or inference. Like how does memory, VRAM, memory bandwidth, GPU cores, etc. all play into this. Know of any good resources? Oddly enough I might be best off asking an LLM.
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?
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?).
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…
They are intentionally keeping the VRAM small on these cards to force people to buy their larger, more expensive offerings.
This is maybe a dumb question, but why is it so hard to buy Nvidia GPUs? I can understand lack of supply, but why can't I go on nvidia.com and buy something the same way I go on apple.com and buy hardware? I'm looking for GPUs and navigating all these different resellers with wildly different prices and confusing names (on top of the already confusing set of available cards).
One way to look at is that the third party GPU packagers have a different set of expertise. They generally build motherboards, GPU holder boards, RAM, and often monitors and mice as well. All of these product PCBs are cheaply made and don't depend on the performance of the latest TSMC node the way the GPU chips do, more about ticking feature boxes at the lowest cost. So nvidia wouldn't have the connections or skillse…