On September 19, IT Home reported that NVIDIA founder and CEO Jensen Huang held a joint live stream with Intel CEO Pat Gelsinger early this morning to discuss the collaboration between the two companies.
Intel announced yesterday that it would invest $5 billion in NVIDIA (IT Home note: approximately 35.531 billion RMB at current exchange rates), and the CEOs of both companies stated that they would combine their strengths to jointly develop multiple generations of custom data center and PC-related products.
During the live stream, Jensen Huang mentioned that NVIDIA can currently achieve NVL72 configurations in Arm architecture CPU rack-level products, while x86 architecture CPU products are limited to NVL8 due to the use of PCIe structure. The focus of this collaboration is to introduce NVLink into Intel’s data center CPUs, allowing NVIDIA to provide products related to both Arm and x86 architectures.
Huang also revealed, “Currently, the x86 architecture ecosystem cannot utilize NVL72 level products, and our collaboration aims to directly integrate x86 CPUs into the NVLink ecosystem, creating rack-level AI supercomputers.”
When asked why the new chips would not be manufactured by Intel, Huang responded that NVIDIA has indeed collaborated with Intel, but the importance of TSMC cannot be overlooked, and currently, Intel’s IFS has relatively limited integration space.
Huang commented, “Both Pat Gelsinger and I acknowledge that TSMC is a world-class foundry, and we are both major customers of TSMC; their capabilities are beyond doubt. However, our dialogue and collaboration today have yielded results.”