Nvidia’s H200 is a next-level generative AI and HPC accelerator that carries a “holy $H!T” $30,000 price tag—and Linus finally got one in his lab (since Nvidia wouldn’t send it for review). He rips it open, peeks at the custom silicon and cooling design, then runs it through the wringer against a dual-Epyc 9965 server and the flagship RTX 5090 to see where it really shines.
The benchmarks highlight jaw-dropping AI inference performance and hefty power draw, proving the H200 is a beast for datacenter workloads—but also a terrifying commitment if you’re not running at scale. If you’re curious about Nvidia’s top-tier gear (and have deep pockets), this teardown and comparison is a must-watch.
Watch on YouTube
Top comments (0)