cross-posted from: https://ani.social/post/16779655

GPU VRAM Price (€) Bandwidth (TB/s) TFLOP16 €/GB €/TB/s €/TFLOP16
NVIDIA H200 NVL 141GB 36284 4.89 1671 257 7423 21
NVIDIA RTX PRO 6000 Blackwell 96GB 8450 1.79 126.0 88 4720 67
NVIDIA RTX 5090 32GB 2299 1.79 104.8 71 1284 22
AMD RADEON 9070XT 16GB 665 0.6446 97.32 41 1031 7
AMD RADEON 9070 16GB 619 0.6446 72.25 38 960 8.5
AMD RADEON 9060XT 16GB 382 0.3223 51.28 23 1186 7.45

This post is part “hear me out” and part asking for advice.

Looking at the table above AI gpus are a pure scam, and it would make much more sense to (atleast looking at this) to use gaming gpus instead, either trough a frankenstein of pcie switches or high bandwith network.

so my question is if somebody has build a similar setup and what their experience has been. And what the expected overhead performance hit is and if it can be made up for by having just way more raw peformance for the same price.

  • brucethemoose@lemmy.worldEnglish
    4·
    10 hours ago

    They are a scam.

    But:

    • Using gaming GPUs in datacenters is in breach of Nvidia’s license for the GPU.

    …So, yes, they are a scam. But:

    • Datacenter is all about batched LLM performance, as the vram pools are bigger than models. In reality, one can get better parallel token/s on an H100 than you can on 2x RTX Pros or a few 5090s, especially with bigger models that take advantage of NVLink.