What are you pursuing NVIDIA and AMD with? Intel Reveals Latest Progress in "AI Chip Battle"

Cailian News Agency, May 23 (Editor Shi Zhengcheng)On Monday local time, at the High Performance Computing Exhibition held in Hamburg, Germany, Intel, which is undergoing a strategic transformation period, revealed the latest details of the company's future AI computing power strategy deployment.

(Source: Intel)

On Monday, Intel provided a series of parameter forecasts for the next generation Max series GPU chip FalconShores, which is the most concerned in the market:The high bandwidth HBM3 memory specification will reach 288GB and support 8bit floating-point operations, with a total bandwidth of 9.8TB/s

Of course, compared to the computing powerhouse NVIDIA, Intel's AI chip business is in a state where there is almost no market share at all. As a traditional chip manufacturer, although Intel had already demonstrated its flagship data center GPU codenamed "PonteVecchio" as early as 2021, the continuous delay in actual delivery also missed the opportunity for this round of AI explosion.

Intel stated on Monday that it is nearing completion of delivering the PonteVecchio based Aurora supercomputer to Argonne National Laboratory in the United States.Intel also emphasizes that this system will perform better than HPC using NVIDIA H100 chips.

The Argonne National Laboratory, which has made significant contributions to the Manhattan Project, also stated that,We will use this supercomputer to develop a series of generative AI models for the scientific research community

But for Intel's AI chip strategy,The biggest uncertainty is that FalconShores will enter the market no earlier than 2025At that time, NVIDIA is likely to come up with a stronger competitor, and AMD, another competitor eager to try the AI track, will also launch the MI300 chip by the end of this year.

Intel Vice President and General Manager of the Supercomputing Business Unit, Jeff McVeigh, stated on Monday that the company is taking time to redesign chips after abandoning its previous strategy of combining CPU and GPU. Although the company aspires to have the strongest CPUs and GPUs in the market, it is difficult to say that the same supplier can find the best combination of both at the same time. On the other hand, if you have independent products, you can compete in platform level choices.