Compute GPUs like Nvidia’s H100 officially fall into the category of graphics processing units, but they don’t have enough dedicated hardware to render graphics. His $30,000+ card, Nvidia’s H100, was found to perform worse than the integrated GPU in benchmarks like 3DMark and Red Dead Redemption 2. Geekawan.
Nvidia’s H100 card is based on the company’s GH100 processor with 14,592 CUDA cores that support a variety of data formats used for AI and HPC workloads, including FP64, TF32, FP32, FP16, INT8, and FP8. I’m here. In contrast, his Nvidia consumer GPUs, such as Nvidia’s AD102, only properly support FP32. The GH100, on the other hand, has only 24 raster operating (ROP) units and has no display engine or display output. Additionally, Nvidia has not optimized the Hopper driver for gaming applications.
But apparently it’s still possible to render graphics and support ray tracing on Nvidia’s H100. Graphics rendering is pretty slow. His 3DMark Time Spy score for a single H100 board is 2681 points, which is even slower than AMD’s integrated Radeon 680M score of 2710.
But running games on a $30,000+ card doesn’t make much sense, and Nvidia didn’t design the GH100 to render graphics. His GH100 from Nvidia has graphics-only hardware built-in, but it’s not built to provide substantial performance in games, which is why it’s slower than AMD’s integrated Radeon 680M.
Nvidia’s flagship compute GPU isn’t just for graphics, but it outperforms everything in data center AI and HPC applications, and it’s exactly what it’s made for.
