GoKawiilNVIDIA's Vera data center CPU isn't ramping up until later this year but I recently had the opportunity to try out this new ARM-based CPU designed for agentic AI workloads. NVIDIA's Vera CPU with its in-house-designed Olympus CPU cores ends up packing a heavy-hitting punch with competitiveness to Intel/AMD x86_64 CPUs that I have never seen out of any other ARM or non-x86_64 processors. Continue on with these early benchmarks of the NVIDIA Vera CPU on Linux.
Vera is NVIDIA's next-gen data center CPU designed for agentic AI and similar modern data center workloads. Vera most notably will be found with NVIDIA NVL72 Vera Rubin as the host CPU for powering these powerful racks for AI while it will also be found standalone for CPU racks. Unlike NVIDIA's Grace that uses Arm Neoverse-V2 cores, Vera makes use of NVIDIA's in-house "Olympus" core design.
Vera features 88 Olympus cores that claim to deliver 2x the performance of its predecessor with leading energy efficiency too. Olympus is compatible with the Armv9.2 ISA and supports FP8 precision, 176 threads in total via spatial multi-threading, and is paired with LPDDR5X memory for delivering up to 1.2TB/s of memory bandwidth. Over Grace, Vera also has double the L2 cache at 2MB per core, a larger unified L3 cache at 164MB, and supports PCIe Gen 6 as well as CXL 3.1 connectivity.
The Vera CPU as tested for this initial benchmarking had a peak 450 Watt socket TDP. With the LPDDR5X memory it's around 50 Watts or less of power consumption.
NVIDIA Vera data center CPUs remain on track for shipping in the second half of the year, but ahead of the ramp, NVIDIA invited me to their Santa Clara headquarters to run some of the first public benchmarks of this new CPU with their Olympus cores. In this article are those very initial results. But before talking performance, it's first important to note the level of Linux support. With Vera not officially out there yet I really didn't know what to expect for the upstream Linux kernel support and the like or what their plans are among the major ARM64 Linux distributions, etc. Fortunately, NVIDIA Vera is already in good shape with upstream Linux kernel support. Linux 7.1+ has the key driver support in place and Vera should work on ARM64 server Linux distributions like Ubuntu, Fedora, etc. NVIDIA will also continue to provide Base OS as their modified version of Ubuntu with all the Vera patches ready. ACPI is relied upon and no needing to deal with frustrating Device Trees or other headaches for NVIDIA Vera on Linux.
With Vera being compatible with Armv9.2 and complying with Arm's Server Base System Architecture (SBSA), it ends up making use of many of the common ARM Linux drivers for support, as part of the reason it's more difficult pre-launch to track the upstream kernel support status in advance of hardware availability. Among that common ARM Linux code being used is the ongoing work around Arm Confidential Compute (CCA) for confidential computing with VMs to be supported by Vera.
Last year GCC and LLVM Clang added support for the Olympus cores. That means GCC 16.1+ or LLVM Clang 21+ for building optimized binaries for Vera. It was great and ambitious of NVIDIA introducing this Olympus compiler support so early and they can be applauded for that initiative. For comparison, while NVIDIA upstreamed their Olympus support in March of 2025, it was only in December when the AMD Zen 6 (znver6) support was added to GCC and this February when it appeared in LLVM/Clang. NVIDIA's Olympus compiler upstreaming is akin to the well-in-advance support we have been used to seeing from Intel over the years. Great seeing NVIDIA take similar initiative with their Vera upstreaming for compilers and at large. Hopefully this continues with future NVIDIA CPU generations.
I wasn't able to test loading different Linux distributions on NVIDIA Vera or the like, but from everything told while at NVIDIA, the upstream open-source Linux support is in good shape already for Vera. My testing was done on Ubuntu 24.04 LTS with their base OS configuration of a patched Linux 6.18 LTS kernel paired with GCC 16.1.
The NVIDIA Vera with its Olympus cores were working nicely on Linux. One caveat is some power management tuning still in the process of being upstreamed. Recently I covered on NVIDIA working on ACPI CPPC v4 support for Linux and this appears to be one of the related bits. Due to that power management tuning still happening, unfortunately, NVIDIA asked that CPU power consumption monitoring not be engaged during this initial round of benchmarking. Similarly, CPU frequency monitoring was not allowed either during this first round of testing. Additionally, the NVIDIA Vera benchmarking was done from one of their pre-production, open-platform system where as it will be more relevant for looking at power and frequency in an actual production, enclosed server chassis designs later in 2026.
NVIDIA also requested only specific workloads relevant to the intended workloads/domains that Vera is catering to in the data center be tested. So this first round of Vera benchmarking isn't too comprehensive across the spectrum of possible workloads but limited to the benchmarks that were permitted based on what they feel were most relevant -- plus the fact I was only spending one day at NVIDIA's offices. For these initial NVIDIA Vera benchmarks they preferred the scope of benchmarks be limited to target use-cases they feel most relevant for their modern data center customers. This isn't a sponsored article but I obliged to their requests in order to run these initial Vera CPU benchmarks. Hopefully in future rounds of Vera testing over the months ahead there will be a more widespread set of benchmarks for those curious about the Olympus CPU core performance at large. Similarly, in being able to report on the power efficiency and performance-per-Watt once their power management code is tuned.