NVIDIA Vera SMT Performance Benchmarks
Recently I was provided remote access to a NVIDIA Vera server with the dual socket Vera CPU Superchip providing 176 Olympus cores plus with NVIDIA's Spatial Multithreading "SMT" implementation providing 352 total threads. With this benchmarking was free rein for testing, including CPU power and frequency monitoring. Over the next few weeks will be a number of articles looking at different areas…
NVIDIA Vera, a new server with a dual socket Superchip CPU featuring 176 Olympus cores and 352 total threads through NVIDIA's Spatial Multithreading (SMT) implementation, was recently benchmarked. The goal was to evaluate the performance impact of SMT on NVIDIA Vera, which differs from traditional Simultaneous Multi-Threading (SMT) implementations from AMD and Intel. Vera SMT relies on physical partitioning of core hardware resources, providing two threads per core, similar to AMD and Intel processors.
Researchers found that NVIDIA's SMT implementation promises improved utilization and more predictable behavior, particularly for agentic AI workloads. To investigate this, a series of benchmarks were conducted on NVIDIA Vera, comparing the performance with SMT enabled and disabled. The benchmarks were carried out on a NVIDIA Olympus setup running modified Ubuntu 24.04, Linux 6.17 64K kernel, and GCC 15.3 compiler.
Throughout the testing, peak CPU frequency and power consumption were closely monitored to analyze the impact of Spatial Multithreading on NVIDIA Vera.
Written by urgent.news from Phoronix's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
