The first performance test of the NVIDIA Grace CPU has emerged through the Geekbench benchmarking software, revealing the potential of this high-end CPU based on Arm architecture, designed for high-performance computing in servers and data centers.
NVIDIA has developed a Superchip configuration that combines two of these processors, resulting in a total of 144 cores based on Armv9 architecture. However, the recent performance test evaluated a single chip with 72 cores, providing an impressive insight into its capabilities.
Performance on Geekbench
The CPU NVIDIA Grace Hopper Superchip, which combines a 72-core Arm Grace CPU with a Hopper H200 GPU, scored 1,636 points in single-core performance and 74,440 points in multi-core performance on Geekbench. This latter figure significantly stands out, surpassing the best Intel Xeon CPU and ranking just behind AMD’s most advanced CPU, the Threadripper 7995WX.
For context, the AMD Threadripper 7995WX, with 96 cores at 5.10 GHz and a power consumption of 350W, scores 81,408 points in the same benchmark. In comparison, the NVIDIA CPU, operating at 3.35 GHz, achieves 74,440 points with a power consumption of approximately 200W. The complete Superchip configuration (CPU + GPU + LPDDR5X memory) consumes a total of 500W. Although AMD’s CPU is 9% faster, its energy consumption is 75% higher.
On the other hand, the Intel Xeon W9-3495X, with 56 cores and cooled with liquid nitrogen to reach 5.30 GHz, scores 72,786 points, falling below the NVIDIA Grace CPU and exceeding 1,000W of energy consumption.
NVIDIA’s Bet on Arm
These results demonstrate why NVIDIA has decisively bet on Arm architecture. The company is not only implementing this technology in its CPU NVIDIA Grace but also in sectors like vehicles, laptops, and consoles, in collaboration with MediaTek to develop SoCs with Arm processors and GeForce RTX graphics.
NVIDIA Grace represents only the beginning of what the company has planned. By 2026, NVIDIA will launch its new generation of CPU, called Vera, which will incorporate an innovative architecture, a more advanced manufacturing process, a higher number of cores, and a next-generation memory to replace LPDDR5x. This advancement promises to revolutionize the professional CPU sector, an area where NVIDIA already leads in the realm of GPUs.
The debut of the CPU NVIDIA Grace on Geekbench has highlighted its potential to rival the most advanced offerings from AMD and Intel, not only in terms of performance but also in energy efficiency. This evolution underscores the growing importance of Arm architecture in the future of high-performance computing, placing NVIDIA at the forefront of this technological transformation.