2025 Wrapped: HMx Labs Edition
A little recap of both your and our favourite bits of 2025 in the world of HPC and HMx Labs.
2025 is now well and truly wrapped. Here’s the HMx Labs Edition.
I quite liked the idea of doing a 2025 wrapped but recapping things from memory was never going to work well. I thought I’d try a few AI tools to do it, but they were all distinctly mid. At least they managed to give me the raw data I needed to write my own.
From a purely social media analytics point of view the top three short form posts (also posted on LinkedIn) were about vibe coding HPC tools, numerical stability in financial analytics and my recent thoughts on Nvidia’s acquisition of SchedMD.
Longer form content on the blog was topped at the start of the year with my non predictions (but hopes) for HPC in 2025 followed by the first two chapters of an Introduction to Supercomputing and then our paper on optimising HPC financial services workloads on AMD CPUs.
Which interestingly is a fairly good correlation with my own highlights for HMx Labs (HPC division) this year (and of course I’m limiting all of this to the work we can talk about publicly and isn’t covered by NDAs).
For me, speaking at Quant Minds and all the analysis that went into that presentation and our paper on optimising financial services workloads were highlights. As was chapter two of an Introduction to Supercomputing.
To that list though I’d add the HPC Club events we ran through the course of the year. Not only did we manage to get speakers from AWS, IBM, Nivida, AMD but we did our first non-London event with Simon McIntosh-Smith speaking about Isambard-AI, the UK’s fastest supercomputer.
We also put out no less than 200 posts and 52 issues of the weekly HPC focused newsletter totalling over 40,000 words. That’s not including the book or paper. If you read all that I apologise for taking up around 3 hours of your life this year with my inane rambling! 😁
Not everything was perfect though. We had some misses too. The scheduler selection tool still isn’t live (but it’s so close). And while we can certainly produce great research using our tooling to run benchmarks at scale, we still haven’t managed to get that tool in front of anyone outside of HMx Labs. Both were goals for this year. I’m hopeful we’ll hit them early in 2026.
When I released chapter 1 of an Introduction to Supercomputing I had intended to release one chapter a month. At the time, that felt attainable. I’ve managed two in six months. Not as planned but I think the twist to involve third party collaborators definitely improves things, even if it takes longer. Its taking a while but I haven’t given up on the book. There will be more to come.
I’ve had fun this year, all thanks to not only the amazing team at HMx Labs, but also everyone we’ve collaborated with throughout the year and the clients we’ve managed to work alongside.
See you in 2026 for not only more of the same but bigger and better things too. Stay tuned!
