Could AMD’s MI300X take a chunk from Nvidia’s AI market share?
AMD seems to be rapidly approaching on Nvidia in terms of AI processing power.
Share this article...
WePC is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Prices subject to change. Learn more
AMD announced a potential titan in the AI-specific hardware industry at its recent Data Center & AI Technology Premiere. The Instinct MI300X is said to dominate as a purpose-built AI accelerator, and it could be coming to take down Nvidia’s H100. Could AMD’s MI300X take a chunk from Nvidia’s AI market share?
What is the AMD Instinct MI300X?
The Instinct MI300X GPU is an advanced AI processing chip, designed by AMD, that improves upon the MI300A. It combines multiple GPU “chiplets” within a single unit, connected through shared memory and networking links. These chiplets, part of the CDNA 3 family, are designed specifically for AI and HPC tasks.
Unlike the MI300A, which included CPU chiplets alongside the GPU chiplets, the MI300X is a “GPU-only” version. AMD replaced the CPUs with two additional CDNA 3 chiplets, resulting in significant improvements to the chip’s capabilities.
Instinct MI300X vs Nvidia’s H100 (Hopper)
There isn’t a great deal of information out there yet, as we only have what was announced at the recent AMD event. However, when comparing the MI300X to NVIDIA’s H100, there are some important factors to consider.
Firstly, the NVIDIA H100 is already available in large quantities, while the MI300X is not yet released. NVIDIA also has a vast ecosystem of software and researchers in the AI industry, giving them an advantage.
Secondly, the new MI300X offers high-density HBM chipsets with 192GB of memory, which may not be matched by Nvidia in terms of memory capacity and timing anytime soon. This means there could be a significant memory advantage for the MI300X.
Thirdly, the MI300X lacks a Transformer engine like the H100, which can greatly enhance the performance of popular AI models like LLM. This could result in longer training times and delayed market entry for those using the MI300X.
Lastly, AMD has not provided any benchmarks yet, which is understandable since the MI300X hasn’t been launched. However, performance in training and running AI models depends not only on the GPU but also on the system design. It will be important to see direct comparisons between the MI300X and H100 to get a clearer picture of their performance.
What are the pros?
This may seem like the H100 is vastly superior to the Mi300H, but not at all, there are some areas in which the MI300X shines. During the keynote at the Data Center & AI Technology Premiere. AMD CEO DR. Lisa, Su announced that the MI300X offers 2.4 times the memory density and 1.6 times the memory bandwidth of Nvidia’s H100 “Hopper” GPU.
This is great for storing and accessing data in the chip’s memory, making for faster read times over the Hopper H100, but will it be enough to get ahead?
If you think this is cool, you should see what AMD did with its Genoa-X CPUs
AMD is a little late to the party, but does it matter?
With Nvidia’s H100 GPUs already on the market, and Nvidia having a much larger software portfolio under their belts, it’s going to be a bit of a slog to get AMD up and running in the industry, however, they are off to a flying start.
During the keynote. DR. Lisa Su announced partnerships with Giants in the industry, such as Hugging Bear, and Amazon. Sometimes getting your foot in the door is half the battle.
AMD’s Instinct MI300X is a significant achievement, advancing the AI industry with its remarkable capabilities. It enables larger and more complex AI models, pushing the boundaries of artificial intelligence. The future of AI is here, powered by AMD.
Should Nvidia be worried about AMD?
Absolutely, Nvidia once dominated the GPU industry but, in recent years, AMD slowly began taking away Nvidia’s majority. Thanks to the greed of Nvidia and the massive improvements AMD has managed to make to its GPUs, drivers, and software components. AMD has gained a lot of ground in the GPU department.
Referencing the Steam Hardware Survey, 14% of gamers now own an AMD GPU, whilst the remainder is mostly Nvidia, it represents a point that AMD has gained a lot of ground over the last decade. Whilst this is a dip over the 20% share AMD held around 2020, we strongly believe that there’s going to be a boom in AMD GPUs in the coming years. Thanks to the points we previously mentioned.
It could always swing the other way, and it’s hard to tell. Nvidia is definitely an industry leader in the AI space, but it definitely shouldn’t discount AMD as a threat just yet. Could AMD’s MI300X take a chunk from Nvidia’s AI market share?
If you missed the AMD data center and AI technology premiere, you can watch it on Youtube.
