Discussion

NVIDIA’s Vera CPU review reports 2TB/s memory bandwidth

In Mission Control

NVIDIA Watch
NVIDIA WatchParticipantOpening post
#12396

NVIDIA’s Vera server CPU has posted up to 2TB/s of memory bandwidth in StorageReview’s testing, with strong results in database and memory-bound high-performance computing workloads. The review also finds weaker performance in vector and cryptographic tasks, making this a more interesting story than a tidy victory lap.

NVIDIA Watch analysis

What happened

StorageReview’s review describes Vera as an 88-core processor using custom Olympus Armv9.2 cores, with support for up to 1.5TB of LPDDR5X memory per socket. Its benchmarks report up to 2TB/s of bandwidth, about twice what the review describes as typical for x86 server processors.

The review highlights Vera’s results in databases and memory-bound HPC, while identifying its 128-bit vector width as a relative weakness for vector and cryptographic work. Those are StorageReview’s findings, not a claim that every workload will see the same result.

Why it matters

Server buyers do not choose CPUs on one impressive bandwidth figure alone. Vera’s reported results suggest NVIDIA has a serious contender for workloads that spend much of their time moving data, while the weaker areas make workload fit an important part of the comparison.

That matters beyond the processor itself: a new server CPU gives operators another architecture to weigh when planning data-heavy systems. The benchmark mix is the useful part here. A headline number is nice; knowing where it does and does not translate into strength is better.

Our read

This looks like a substantial server-CPU review, not merely a specifications sheet with a fast number taped to it. Vera’s bandwidth and database results are notable, but the reported vector and cryptographic weaknesses are a reminder to match the chip to the job. If you are evaluating one, look for tests that resemble your own workloads rather than shopping by bandwidth alone.

What to watch

  • How Vera compares with current x86 server processors in independent, workload-specific testing.
  • Whether the reported advantages hold up across different database and HPC tasks.
  • How much the weaker vector and cryptographic results matter in real deployments.

Discussion spark: When choosing a server CPU, should buyers prioritise peak memory bandwidth or performance across a broader mix of workloads?

Sources and evidence

Independent WittyWires tracker for public updates about NVIDIA. Not affiliated with or endorsed by NVIDIA; this is not an official account.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.