For those of you who don’t know, AI and high performance computing is big business. There’s a reason Nvidia’s market cap exceeds his $1 trillion. The reason is explosive demand for the company’s enterprise offerings, including the powerful H100 Hopper GPU. Yes, this monster processor can cost over $30,000 and shares much of its DNA with humble GeForce gaming graphics cards.
As a fun experiment, YouTuber Geekawan (via tom’s hardware) tried one of these monster cards in some gaming benchmarks. The results were interesting, but good enough for a good laugh in the end.
The H100 variant used by Geekerwan was the PCIe version. It has 80GB of HBM2e memory, 14,592 CUDA cores and a 350W TDP. Compare this to the RTX 4090 with 24Gb GDDR6X, 16,384 CUDA Cores and 450W TDP. The H100 isn’t supposed to be a slouch, right?
In fact, the card is very weak for gaming, with a graphics score of just 2,681 in 3DMark Time Spy. That’s less than the Radeon 680M integrated graphics. In Red Dead Redemption 2, the card couldn’t even hit 30 FPS at 1080p.
Taken seriously, these results are not surprising. The H100 is a very powerful card, but it was not designed for graphics applications. In fact, it doesn’t even have a display output. The system needed a secondary GPU to provide the display. It’s also missing some other fixed hardware that’s essential for gaming.
Additionally, there is the fact that the drivers are not fully optimized for gaming. During our gaming tests, we found that the GPU was consuming less than 100W and was severely underutilized.
So if you have $30,000 to spare, buy a car. Or buy an RTX 4090 system and a car with your leftover money. The powerful H100 card is not meant to be used to power the ultimate gaming rig.
