Nvidia says architecture-specific tuning unlocks more Vera CPU performance
Nvidia said at SC26 that software optimizations tailored to its proprietary Olympus cores can improve Vera CPU performance by nearly 39% in some computationally intensive workloads. In its miniKernelGEMM benchmark, an Olympus-optimized kernel raised Vera to 7.53 instructions per clock and 21.75 FP64 operations per cycle, nearly 67% above Grace on a per-core, per-clock basis. The figures are vendor-generated results from a narrow matrix-multiplication workload and have not been independently validated.
Vera is positioned as both a standalone data-center CPU and a host for Nvidia’s Vera Rubin NVL72 servers, so its ability to run existing software without architecture-specific work is relevant to AI infrastructure adoption.
Nvidia's Vera CPU, designed as both a standalone data-center part and a host for Nvidia's Vera Rubin NVL72 servers, offers impressive performance thanks to substantially higher execution resources per core than its predecessor, Grace. However, Nvidia admits at Supercomputing 2026 (SC26) that existing software cannot always take full advantage of Vera's resources without adaptation for the architecture.
The company shared that architecture-specific software optimizations can boost Vera's performance by nearly 39% in certain computationally intensive workloads. But despite offering unbeatable single-thread performance, the CPU is not positioned for general-purpose workloads.
Get the Tom's Hardware App

Get the full Tom's Hardware experience on your device with our brand-new app. Bookmark articles, use our smart search feature, personalize your feed, and get push notifications so you never miss a beat.
#Nvidia#Nvidia Vera CPU#Nvidia Grace CPU#Olympus microarchitecture#Vera Rubin NVL72#Arm Neoverse V2#OpenBLAS#miniKernelGEMM