AMD AI Chip Challenges Nvidia’s H100 as Rivalry Enters New Era
Advanced Micro Devices (AMD) has launched its new MI300X artificial intelligence (AI) computer processor to challenge Nvidia’s dominance of the AI hardware market. AMD’s chip will compete with Nvidia’s H100, continuing a public spat over who could make the most powerful Graphics Processing Units (GPUs).
AMD’s chief executive, Lisa Su, said the MI300X is “the most advanced AI accelerator” in the industry. Experts confirmed the MI300X outranks Nvidia’s H100 in raw specs.
AMD and Nvidia Share AI Spoils
AMD also launched its ROCm6 platform to support device programming. OpenAI will add support for programming the MI300X in its Triton AI software. Microsoft and Meta are already using the chip for their AI workloads.
Companies processing AI algorithms for clients are diversifying their hardware. Amazon Web Services, which recently struck a deal with OpenAI rival Anthropic, will use its own Trainium2 AI chip and Nvidia’s new H200 for AI workloads. Microsoft is also expanding its hardware to include the H200 and its own Maia 100.
The MI300X continues the decades-long rivalry between Nvidia and AMD. While AMD struggled to compete with the likes of Intel for central processing unit market share, the gap between the performances of its graphics processing units (GPUs) and Nvidia’s became narrower as time went on. While AMD cards trail Nvidia in ray-tracing and creative performance, AMD offers comparable performance when running games in 4K.
The Future of Crypto Mining GPUs
Before the launch of OpenAI’s ChatGPT bot, some of Nvidia’s GPUs were being used to mine cryptocurrencies. Before the network transitioned to proof-of-stake, Ethereum relied heavily on GPUs to finalize transactions.
Read more: What is Cryptocurrency Mining?
Since Ethereum transitioned, GPU miners of other coins are likely to only break even after years of mining. They may be better off joining mining pools, as application-specific integrated circuit (ASIC) machines have become the preferred option for large mining companies.
Read more: Top Cryptocurrency Mining Pools To Join 2023
Some companies, like Vancouver-based Hive, use a combination of ASICs and GPUs. Other companies, like Northern Data and Iris Energy, recently bought GPUs to offer AI services to supplement mining revenues.