Andrew Jones (hpcnotes)

You know when you watched a recording of a meeting back at 2x speed?

Anyone else feel like their whole week is being played at 2x speed and many days/meetings are just racing past?

Wasn't that one :-)

Probably giving away too much info if I say publicly which it was ...

Yes, 4 years ago (even 2) your statement would have been accurate but much less so now and into the future.

Proclaiming that #AI is a bubble / is useless / is not real etc. because you had recent poor experiences with an #LLM/chatbot is a bit like saying electricity will never be useful because your electric toaster burnt your bread this morning.

As someone involved in planning future AI infra and DCs, I'm not entirely convinced your statement is wholly accurate :-)

Yes and no.

At some level, yes, DCs would be built out anyway for other workloads.

But the design points for DCs primarily serving AI (especially training) are meaningfully different than traditional hyperscaler DCs.

Who, me? 😁

If it trains on less than 100,000 #GPUs, is it even real #AI?

If it's not written in #Fortran77, is it even real #HPC?

If it doesn't use tapes, is it even real #archive #storage?

If it doesn't recycle at least 3 previously discarded ideas, is it even credible as "the new game-changer AI chip"?

HPCwire Supercomputing news and information focused on emerging HPC applications in science, engineering, financial modeling, virtual reality, databases and other compute intensive tasks

HPCwire.com is posting the plenary talk summaries and the TPC25.org site will have full video recodings of all the talks.

Yup. The runway/growth for #AI is only just starting.

People who mistakenly assume AI is going nowhere because of their limited experience with some chatbots or specific LLM are missing the (much) bigger picture.

When you need to quickly check a spec on some #HPC tech … and @glennklockwood.com’s blog comes higher up the search results page than the vendor’s own website(s) - and presents the required info simply and quickly too.

I’m sure we can arrange someone to sell you some corporate copilot licences 😁

Would a be fun twist if @bsky.app untags themselves from this thread 😁

From Supercomputers to the Frontlines of AI Inference: Glenn Lockwood on Infrastructure That Lasts | Shared Everything In this episode of Shared Everything, Glenn Lockwood, just named Principal Technical Strategist at VAST Data, shares what decades at the bleeding edge of large-scale systems design have taught him abo...

Podcast: “Supercomputing used to be about going fast in one direction. In AI, that just drives you off a cliff faster.” — @glennklockwood.com Lockwood on why flexibility, not just performance, defines next-gen infrastructure. Now at VAST. shared-everything.simplecast.com/episodes/fro...

#HPC Leadership scores so far

Q1: 97% of people got it correct
Q2: 76%
Q3: 73%
Q4: 84%
Q5: 89%
Q6: 88%
Q7: 65%
Q8: 78%
Q9: 90%
Q10: 70%

The most selected wrong answer (33% of people) is for Q7.

Nearly 20% of people got 10/10.

Still nobody who says they don't need the course got 10/10.

So far, none of the people who ticked the Q11 box for “I am already experienced enough, I don't need this training” have scored 10/10.

Of those who have scored 10/10, all of them also marked one of the Q11 options saying they would like to attend (even if not possible this year).

Sounds sensible to me 😁

Bidding on known-lost causes is really about the opportunity cost. If an RFP requires a team to focus for weeks or months, don’t bother if you know the upside isn’t there.

Some of the other options on that question too …

😁

Well, three days of content covers a lot more than those ten questions!

It shows you your score at the end.

Different question perhaps is what should be an acceptable pass mark / score? 😁

Some #HPC fun for the week ahead:

a quiz on HPC leadership topics

forms.office.com/r/vCcEX690h2

(and no mention of #ai or #quantum!)

My judgment is questioned then 😁

Sunset over the beach

Lessons learned from three years in cloud supercomputing I recently decided to leave Microsoft after having spent just over three years there, first as a storage product manager, then as a compute ...

In the few days I have between jobs, I wanted to share an unvarnished perspective on what I've learned after spending three years working on supercomputing in the cloud. It's hastily written and lightly edited, but I hope others find it interesting: blog.glennklockwood.com/2025/07/less...

#HPC