HPC Guru

All things, with one exception, have to end

After years of sharing #HPC news and insights, it's time to log off

It was fun while it lasted and I remain grateful to all of you who followed and interacted

Stay curious, stay kind!

— @hpcguru.bsky.social signing off

#HPC #Farewell

Nvidia now generates roughly $4 million in revenue for each employee they have

That's the highest among all of big tech

#HPC #AI #GPU

ISC 2025 HPSF Community Birds of a Feather (BoF) + marking one year since launch at last year's ISC 🔥 with panelists:
💠 Julien Bigot, CEA
💠 Christian Trott, Sandia National Laboratories
💠 Heidi Poxon, NVIDIA
💠 Todd Gamblin, LLNL
💠 Andy Warner, HPE
#HPSF #HPC #HighPerformanceSoftware #ISC25

Observation from #ISC25: None of the Big 3 cloud service providers have booths this year. Don’t recognize any of the big GPUaaS providers either, not even the ones who debuted on Top500 today. Plenty of quantum though. What’s this say about the state of AI in Europe? And cloud HPC for that matter.

Dynamic Use of Resources for Affordable Exascale and Beyond – BoF at ISC25: Martin Schulz (TUM) Implementing all these ideas means breaking long standing dogmas in HPC One Node = On Job —> On-Node Co-Scheduling —> Requires deep understanding and dynamic tracking of topology ==> Effective use of complementary accelerators. Jobs have static resource usage from start to end —> Dynamic Process Allocation —> Requires new resource management interfaces and management approaches ==> Adapt to external conditions (e.g., contention) Worst case and fixed power distribution —> PowerStack / SEANERGYS —> Requires continuous monitoring and live feedback & interactions with the grid ==> Adaptive power steering on over-provisioned nodes

Dynamic Use of Resources for Affordable Exascale and Beyond – BoF at #ISC25

Take aways for our sysadmin team – we need (rather sooner than later):

👉 Monitoring
👉 Al-Based Data Analytics
👉 Dynamic Scheduling & Resource Mgmt

We‘ll definitely need to break some long standing dogmas as Martin says:

40 years of #ISC – 3 in 1 panel nicely moderated by @addisonsnell to provide a good grasp about the past, present, and future of #ISC

Panelist view from the full-house "ML/#AI for Workload Analysis in #HPC" BoF at #ISC25! Great talks and interaction - we went into "flipped classroom" and started asking questions to the audience. I very much enjoyed the invite from Kadida Konate (LBL) and Sergio Iserte (BSC).

DDN Launches Next-Gen Data Intelligence Platform at ISC 2025 DDN launches AI400X3 and Infinia 2.1 at ISC 2025, boosting AI and HPC performance, scalability, and data intelligence for modern enterprises.

DDN launches next-generation appliance: AI400X3 and enhancements to its end-to-end #AI software, Infinia 2.1

www.ddn.com/press-releas...

#ISC25 #HPC #Storage

The new #TOP500 & #Green500 lists are out:

🔹#EXA_JUPITER is now Europe’s fastest supercomputer, ranked #4 worldwide
🔹LUMI & Leonardo continue to lead in the global top 10.
More details on the rankings of #EuroHPC supercomputers: europa.eu/!pc3bjQ

The new edition of #top500 is out!

🇺🇸 El Capitan remains the most powerful system in the world
🇩🇪 JUPITER is now Europe’s fastest #supercomputer
🇫🇮 LUMI & 🇮🇹 LEONARDO stay in the global top 10
top500.org/lists/top500...
#hpc

image by @eurohpc-ju.bsky.social

❄️ @lumi-supercomputer.eu ranks 9th on the new Top500 list! Over 3,000 scientific projects have already run on LUMI.

✨ Next up: LUMI will be succeeded by an AI-optimised supercomputer called LUMI-AI, as part of one of the EU's first AI Factories.

🔗 Read more: csc.fi/en/media-rel...

Conference Center Hamburg (CCH) where ISC25 is happening this week – inside-outside blend

The Conference Center Hamburg #CCH offers a nice inside-outside blend, definitely helping with #Connecting_the_Dots at #ISC25

At the #ISC25 vendor showdown, I think Bruno from Eviden captured a key difference between HPC and AI businesses. Paraphrasing, HPC people like to fuss with the system. AI customers don’t care about the system. Suggests why NVIDIA has been so successful. Meet the customers where they are.

JUPITER supercomputer, powered by the NVIDIA Grace Hopper platform, is the fastest in Europe — delivering a more than 2x speedup for #HPC and #AI workloads compared with the next-fastest system

nvidianews.nvidia.com/news/nvidia-...

#ISC25

The Blue Lion Supercomputer Will Run on NVIDIA Vera Rubin — Here’s Why That Matters Two new supercomputers. One new architecture. Vera Rubin is built to change how science gets done — in real time, at massive scale.

LRZ’s #BlueLion #supercomputer will be based on Nvidia’s Vera Rubin architecture

It’s the second large system based on Vera Rubin, following NERSC’s Doudna

blogs.nvidia.com/blog/blue-li...

#HPC #AI #ISC25 #GPU

Half of the #Top10 systems on the latest #Top500 are from Europe

Strategic investments by @eurohpc-ju.bsky.social, aided with China's reluctance to submit to this list

#ISC25 #HPC #AI

Though not the intent of this slide, Erich pointed out that a bunch of GPUaaS/neocloud providers (Nebius, Core42, etc) are now submitting HPL numbers. These systems are still modest in size, but if the bigger ones like CoreWeave ever jumped aboard, the significance of Top500 might skew.

#ISC25

If you define hyperscale to be the size of the largest #HPC datacenters, Top500 systems qualify as hyperscale. I’m not sure what the value of this proclamation is though.

In real hyperscale, the qualifier is how many of these 50-200 MW sites you have, not just the fact that you have one.

#ISC25

HPL-MxP:

1. El Capitan with 16.7 Exaflop/s

2. Aurora with 11.6 Exaflop/s

3. Frontier with 11.4 Exaflop/s

Congratulations to LLNL's #ElCapitan for taking the top position in all 3: Top500, HPCG, and HPL-MxP

#ISC25 #HPC #AI via @profamandarandles.bsky.social

"All the systems on the #Top500 combined have about 1 GW peak power consumption today" E. Strohmeier at #ISC25. When will we see the first single datacenter with 1 GW in #AI or #HPC? Some have been announced recently 😉. Talks about SMRs heating up here.

#HPCG: There's a new #1, after a long reign by Fugaku

1. El Capitan

2. Fugaku

3. Frontier

4. Aurora

top500.org/lists/hpcg/2...

#ISC25 #HPC #AI

#Green500: No changes from Nov 2024!

1. JEDI

2. ROMEO-2025

3. Adastra2

top500.org/lists/green5...

#ISC25 #HPC #AI

El Capitan at LLNL sweeps the Top500, HPCG, and HPL-MxP! It’s been a while since we have seen one machine take first on all three. Congrats!

#ISC2025 #hpc

#Top500: Vendors

o Lenovo edges HPE in # of systems 135 to 132

o HPE has the biggest performance share by a huge margin

o Atos/Eviden is #3 in both

#ISC25 #HPC #AI

Today at 11:30, our director Thomas Schulthess will present at #ISC25 "Alps” – one year on from the first GH200 rollout" in the session "HPC in Europe". Don't miss it!

One thing I love about ISC is that Chinese researchers are given top billing. I was surprised to see focus on applying FPGAs for reverse time migration. I’ve not heard of others leading in this field thinking about that.

I also didn’t realize China cared this much about #HPC for oil & gas.

#ISC25

#Top500: China declines to participate and slowly drops in the list as no new Chinese #supercomputers are added

1. USA 173
2. China 46
3. Germany 43

#ISC25 #HPC #AI

#Top500: Interconnect in Top10 systems

o 7 systems use Cray's Slingshot (El Capitan, Frontier, Aurora, HPC6, Alps, LUMI, JUPITER Booster)

o 2 systems use Infiniband (Eagle, Leonardo)

o Fugaku retains its proprietary Tofu interconnect

#ISC25 #HPC #AI

#Top500: CPU in Top10 systems

o 5 systems use AMD processors (El Capitan, Frontier, HPC6, LUMI, Alps)

o 3 systems use Intel (Aurora, Eagle, Leonardo)

o JUPITER Booster relies on a Grace Hopper Superchip

o Fugaku uses a proprietary ARM-based Fujitsu A64FX

#ISC25 #HPC #AI

See the full #Top500 list at:

top500.org

Press Release: top500.org/news/el-capi...

#HPC #AI #ISC25