AMD 3rd Gen EPYC Milan Review: A Peak vs Per Core Performance Balance
by Dr. Ian Cutress & Andrei Frumusanu on March 15, 2021 11:00 AM ESTDisclaimer June 25th: The benchmark figures in this review have been superseded by our second follow-up Milan review article, where we observe improved performance figures on a production platform compared to AMD’s reference system in this piece.
SPEC - Single-Threaded Performance
Single-thread performance of server CPUs usually isn’t the most important metric for most scale-out workloads, but there are use-cases such as EDA tools which are pretty much single-thread performance bound.
Power envelopes here usually don’t matter, and what is actually the performance factor that comes at play here is simply the boost clocks of the CPUs as well as the IPC improvement, and memory latency of the cores. We’re also testing the results here in NPS1 mode as if you have single-threaded bound workloads, you should prefer to use the systems in a single NUMA node mode.
Generationally, the new Zen3-based 7763 improves performance quite significantly over the 7742, even though I noted that both parts boosted almost equally to around 3400MHz in single-threaded scenarios. The uplifts here average over a geomean of +25%, with individual increases from +15 to +50%, with a median of +22%.
The Milan part also now more clearly competes against the best of the competition, even though it’s not a single-threaded optimised part as the 75F3 – we’ll see those scores a bit later.
In SPECfp, the Zen3 based Milan chip also does extremely well, measuring an average geomean boost of +14.2% and a median of +18%.
The new 7763 takes a notable lead in single-threaded performance amongst the large core count SKUs in the market right now. More notably, the 75F3 further increases this lead through the higher 4GHz boost clock this frequency optimised part enables.
Facebook
Twitter
RSS
120 Comments
View All Comments
MenhirMike - Monday, March 15, 2021 - link
Ampere Altra *Server* that is. I'd love to get a system with the CPU, but priced in the realm of "Let's tinker with it and try it out" along with "Let's not cool it with 15000+rpm 40mm fans".kgardas - Monday, March 15, 2021 - link
Avantek provides some workstation as a more silent solution: https://www.avantek.co.uk/ampere-emag-arm-workstat... -- I'll leave price options to you...MenhirMike - Tuesday, March 16, 2021 - link
Yeah, but that Avantek is old tech: https://www.anandtech.com/show/15733/ampere-emag-s...Calin - Tuesday, March 16, 2021 - link
"ARM on the server is actually gaining foothold"They have won some niches and are expanding from there.
I don't think they have enough fab capacity to build all the processors they could sell (especially as AMD is capacity-limited and Intel is - apparently - yield limited).
Spunjji - Friday, March 19, 2021 - link
In the intervening 7 years, it has only become more obvious as an eventuality. Unless you're denying the existence of AWS' serious investment into that ecosystem...Wilco1 - Sunday, March 21, 2021 - link
Yes, Graviton is already 14% of AWS and still growing fast.prisonerX - Monday, March 15, 2021 - link
ARM prediction is probably good, but not with NVIDIA, they're unlikely to be approved.Crazyeyeskillah - Monday, March 15, 2021 - link
Nvidia will have no impact on arm improvements. They merely seek to take Intel and AMD out of the equation by pairing Custom Arm servers with their gpus.Yojimbo - Monday, March 15, 2021 - link
NVIDIA can have servers with custom ARM chips without buying ARM.Yojimbo - Monday, March 15, 2021 - link
And by pointing this out I mean that NVIDIA have no intention of taking Intel or AMD out of the equation. They want their GPUs to be used anywhere with any CPU. The problem is Intel and AMD potentially taking NVIDIA's GPUs out of the equation.