Quantum #84

Issue #84 of the weekly HPC newsletter by HMx Labs. Oak Ridge to get a new AMD supercomputer, Google’ spend on hardware placed in context, more benchmarking tips and don’t forget to sign up for HPC Club.

Quantum #84

This is going to be a rather brief introduction to this week’s HPC News as I need to get my head stuck back into preparing my presentation for Quant Minds. I procrastinated and selected slide layouts and fonts for most of the day yesterday rather than writing the actual content and that was when I wasn’t building Ikea furniture… Supercomputers are just another kind of flat packed furniture right? 🤣

I still have the output from about 3000hours of benchmark runs across over 1000 VM instances to analyse for this presentation so I shall leave it there for today and let you peruse the best of what happened last week in the world of HPC below while I work out which CPU ISA is best for running your workloads.

Oh, and don’t forget to sign up for HPC Club! See you then 😄


In The News

Updates from the big three clouds on all things HPC.

HPC Cloud Updates WE 02 Nov 2025
Updates to AWS, Azure & GCP in the last week relevant for HPC practitioners. Surprisingly quiet week for being this close SC25 and re:Invent.

Oak Ridge National Laboratory announces a new AMD powered supercomputer

HPE to Build AMD-Powered Exascale and AI Supercomputers for Oak Ridge
HPE and AMD today announced they have been selected to build two systems for Oak Ridge National Laboratory, a successor to the Frontier exascale-class HPC system, along with an AI supercomputing .... - Read more from Inside HPC & AI News.

I thought I had included this before but I must have forgotten… kinda interesting along with some of the other financial services use cases that are emerging for AI

https://qz.com/openai-project-mercury-automate-wall-street-investment-banking

This provides some context to the relatively insane amounts of compute that we seem to be trying to put together as a species at the moment.

Google Spends More On Servers Than The Whole World Used To
Here is a funny number to chew on. Sometime in the early part of 2026, if current trends persist, Google will have a spending rate on servers that is in

HPC Wire talks about a new AI to make the 90% of science we throw away more accessible…. But I do wonder if we should just be doing less of the throwaway science in the first place (a la Sabine Hossenfelder et al).

https://www.hpcwire.com/2025/10/31/science-loses-90-of-its-data-a-new-ai-approach-could-change-that/


From HMx Labs

Benchmarks can be tricky things. Not only is it important to pick the right one that represents your workloads, you also need to make sure you’re interpreting its output correctly

Interpreting Benchmarks: Throughput or Latency?
Benchmarks are fun but making sure you pick the right one and interpret the results correctly can be tricky

And a bit of Halloween fun

Slurm of the Dead
The bubble popped but the jobs keep running. Just some fun and giggles for a Friday on Halloween

Lastly, if you haven’t booked your spot for HPC Club sign up before its full

HPC Club #5
Reservations for the next HPC Club in London are now open.

Know someone else who might like to read this newsletter? Forward this on to them or even better, ask them to sign up here: https://cloudhpc.news