Although it is not yet the deadline, the probability that NVIDIA will fail to complete its plan to acquire Arm is higher. However, for the entry of CPU business, NVIDIA seems to have accelerated its pace accordingly. Recently there was news that NVIDIA was ready to recruit hundreds of engineers to create a CPU design and engineering team in Israel. Perhaps seeing the competitiveness of Intel's return to the discrete graphics business, NVIDIA will continue to expand its leading position and business scope in data center/high-performance computing.
The biggest reason for choosing to establish a CPU R&D team in Israel may be that after the acquisition of Mellanox, NVIDIA's team size in Israel has increased to about 2800 employees, and the CPU team should also work with its high-level network development, HPC technology and data processing teams.
It is reported that after the creation of the new CPU team, NVIDIA will recruit hundreds of additional people, including hardware, software and architecture talents to enter the related research and development work.
The creation of the CPU team by NVIDIA should have been in its layout long ago. However, it has not been revealed whether NVIDIA took the lead in developing the x86 architecture or the Arm architecture product design. In fact, both of these pros and cons exist for NVIDIA, so we can only wait for the good news.
If the development of the CPU team is not yet understood, but the thing that offloads some low-level tasks from the CPU to the GPU to improve performance, NVIDIA is already doing.
NVIDIA has been hiding a feature in its enterprise and consumer GPUs since the Turing generation, which is to offload driver responsibilities from the CPU to the GPU through a piece of system processor called GSP or GPU to improve performance and efficiency. Currently, it is officially unlocked and is enabled in the latest NVIDIA driver 510.39.01 or later. GPU support ranges from turing-based Tesla T4 all the way to A series such as the A100, A2, and A40.
It is worth mentioning that although this feature is currently only available for enterprise solutions, there is news that NVIDIA has announced that it will also turn on the function in key components of Max-Q laptops in 2022 and beyond.
For NVIDIA to enter the CPU market, I believe it is an inevitable choice for chip giants. At present, Intel is regaining the independent graphics business, and AMD can be a dual A platform for components, and the debut of apple's self-developed M1 series chips is the peak, I believe that everyone is more in the eyes, remembered. The formation of a benign and complete ecological chain is the vision of the major chip giants, for consumers, of course, it is hoped that more players will pour in to stir up the market and compete with each other to avoid the emergence of the idea that the dominant manufacturer wants to "lie and win".