A special memory of Piz Daint - once Europe's fastest and first large #GPU #supercomputer After 13 years of service at #ETH #CSCS, it retired from service. Those two blades shall live on as a milestone in Swiss and European history! Better than spatulas #nvidia? Enjoy the festive season 🎄 #AI #HPC

Switzerland, where you go with your hybrid car by electric train to the mountains 🏔️. Always cool. Happy holidays! ❄️

With the year slowly winding down, we took a bit of time to get together, celebrate and reflect on the amazing 2025 journey and all the learnings and achievements during our Xmas dinner last Friday.🎄 Happy holidays & 2026, here we come!🚀🥳 #HPC #AI @thoefler.bsky.social @csateth.bsky.social

NVIDIA explains why they used FP8 as a scaling factor in NVFP and not MXFP-style E8M0. The idea is to preserve the largest value in each block. buff.ly/SjwVr6w (16:00 ff) We found this to be very true in a study and managed to recover some MXFP accuracy with micro-rotations: buff.ly/gXaCAhF

Programming modern accelerators uses more and more complex forms of parallelism for example pipelining loading and computing at all levels. buff.ly/K85kokX Check out Tiziano's FBLAS and SMI for some fundamentals behind pipelining in applications and networks! buff.ly/wPnHfaJ buff.ly/65mQo7q

Jeff Hammond from NVIDIA gives some historical context on NCCL's lineage going back to MPI, similarly for NVSHMEM. #AI is rooted in a strong #HPC networking history! buff.ly/kMPnbBf (04:00) Learn more about NCCL details in our paper: buff.ly/doin5bt

Celebrating 15 years of the ETH Singapore Center! An impressive journey with one of the quickest growing economies featuring a fabulous R&D story. The SEC is happy to be part of that history. SPCL and I are proud join the future of this journey. #AI #HPC for Good.

Nice overview of PCIe 6.0 key metrics and development - a key goal was to not increase the latency by more than 10ns. Total FEC latency goal <2ns! Watch: buff.ly/c9hBJcV

Ilya Sutskever: "Compression is prediction and vice versa." (15:10). A good compressor will exploit the shared structure - now this forms a strong mathematical reason for exploiting all information in the unlabeled data - independently of the later tasks. Watch: buff.ly/Bnvazym

I talked about a path towards an "AI Climate Scientist" at the ADIA Lab Symposium in Abu Dhabi and the Spanish Royal Academy in Madrid. Some ways to combine #AI and #HPC to address the data and computational challenge in #Climate sciences and accelerate insights. Watch now: buff.ly/7qeGuqf

Bill Dally on NVIDIA networking: "Jensen said: 'Absolutely not, we don't do networking, we're a GPU company'". buff.ly/iUHv62C (4:30) Then he talked the DOE into paying 100%, launching NVLINK Example how a research lab changed the course of its organization. Another example is "all of AI" :-)

Congrats to Saleh Ashkboos, SPCL's 13th PhD graduate (another prime number). A new expert in quantization and LLM optimization is born. Thanks also to the great collaboration with Dan Alistarh from ISTA and James Hensman at Microsoft!

I am very honored to be part of to the two brilliant teams winning the 2025 Gordon Bell Prize for Climate Modeling and the 2025 Gordon Bell Prize (Honorable Mention). Both among the highest honors in #HPC. 🍾 buff.ly/gCkjDRN and buff.ly/4ewhdZ9 Congrats teams - Switzerland 🇨🇭 going strong - onward!

What a week - #SC25 is a wrap! Thanks to all friends and the SPCL team. Right from the plane to the trail to fight the back pain and sleepless night with a traditional post-SC 10k run. Looking forward to teaching #HPC on Monday.

Former member of SPCL, collaborator, and friend Daniele De Sensi speaks at both Broadcom's booth and at the main track of the Supercomputing conference. Onward! We're all proud of you 🎉

🌍 A 26-member team has been awarded the 2025 Gordon Bell Prize for Climate Modelling for their project, “Computing the Full Earth System at 1 km Resolution.” Congratulations to this year’s outstanding team! buff.ly/GKDYnH3

Rajeev Thakur is kicking off our Advanced MPI tutorial at #SC25. It's always an honor to teach this long standing tutorial with esteemed colleagues including Bill Gropp and Pavan Balaji. Great attendance 👌, still some seats in 122. We're looking forward to a productive session.

Kurt Ferreira opens the 14th addition of our ROSS workshop at #SC25! Supercomputer operating systems and middleware going strong! Packed room as always 😀. Starting with an invited talk by NVIDIA's Jeff Hammond on communication systems. Trivia: he probably traveled furthest ✈️

Arrived at #SC25 and just moved into the SPCL den 🏠 - our homebase for the whole on-site team this week. Readying ourselves for a crazy time - my three talks for tomorrow should mainly be set 😀. Looking forward to seeing all of you in person - for those who I have not already met 🤝.

Very nice overview of the emerging UALink standard with nice features such as splitting packets in switches, in-network computing, high energy efficiency, and lowest silicon overhead: buff.ly/AgLvC1g I'll be joining a panel at SC25 contrasting UALink and UEC next Wed: buff.ly/BeCMFcL Join us there

Ram Velega (Broadcom) at OCP "The things that have confined us today that the scale-up domain is less than 100 are going to be changed very soon" (7:55 in buff.ly/9ioPH73) I see this as a call to action for systems and software research. UE 1.0 techniques such as LLR and CBFC play a crucial role!

Ilya Sutskever "Three lines of math can prove all of supervised learning. That's nice" (4:33) "I have not seen an exposition of unsupervised learning that I found satisfying" (7:50) Its optimization objective has little relation to the actual objective you care about Watch: buff.ly/Bnvazym

Can we build an #AI #Climate Scientist? Asked at the ADIA Lab Symposium in Abu Dhabi last week - now online at buff.ly/6igSeyg :-). Much work to be done - this is outlining some directions of indicative results with a lot of potential to accelerate AI for Science.

Collaborator and friend Dan Alistarh talks at ETH about using the new NvFP4 and MXFP4 block formats for inference. Some going from "terrible" accuracy to acceptable using micro rotations to smoothen outliers in blocks. arxiv.org/abs/2509.23202 Great collaboration and cool stuff

Keren Bergman at the 2nd EFCL Workshop: "Huawei combines 3x less performant GPUs with a photonic scale-up network to build higher performance PODs than Nvidia." Nvidia moving towards CPO 😀. Optics everywhere in scale-out. Nice overview of optical networking at all distances.

I was very honored to meet Carnegie Mellon University's President, Dean of the School of CS, and its famous founder Raj Reddy to present a lecture named after him. I tremendously enjoyed speaking with young students and faculty and the evening with CMU's leadership. Thanks for the invitation Raj!

MIT's Sandy Pentland at ADIA Lab symposium: "Modern companies need to structure their incentives to coordiate teams instead of using strict and siloed hierarchies." Enable team leaders to do "what they think is right" instead of inefficient political discussions with leadership.

🎉 Uno accepted to SC25! Unified congestion control + reliable connectivity for intra- & inter-DC traffic to enable inter-DC AI training. 📄Paper: arxiv.org/abs/2510.15802 💻Code: github.com/spcl/Uno_SC25 🤝Collaboration with Microsoft #SC25 #AI #SPCL @thoefler.bsky.social @csateth.bsky.social

MIT's Sandy Pentland at the ADIA Lab symposium 3rd day on #AI in Finance: "More social traders minimize risk using collective (social?) intelligence." This applies to many fields! Do we need social #AI? Corollary: We need more #HPC compute for such social agents :-).

One highlight of the ADIA Lab Symposium was Nobel Laureate Chu's talk towards Net-Zero emissions. "China follows the US textbook from 100 years ago, when the US took products inventend in Europe, such as cars, industrialized and improved them. Now China takes things invented in the west..."

Just arrived at the ADIA Lab symposium in Abu Dhabi to listen to Horst Simon's introduction and Bjorn Stevens' keynote on how to compute the future climate! Featuring our Gordon Bell finalists 🌍🚀 Looking forward to speculating about how to create an #AI climate scientist 😀.

Microsoft's Ultra Ethernet tutorial is now available on youtube 🎥! Saurabh Dighe gives a brilliant motivation for Microsoft's "AI first datacenters" 🤖 followed by Abdul Kabbani and myself explaining the technical details and innovations of Ultra Ethernet supporting this goal 🫡. buff.ly/IPN46ZR

Next was an intriguing talk by @thoefler.bsky.social on computational architectures for more efficiently training large models at @scsatcmu.bsky.social www.youtube.com/watch?v=LnXp... (4/8)

I'm excited to discuss whether we can build an "AI Climate Scientist" in my talk at the ADIA Lab Symposium 2025 🌎 Join us in Abu Dhabi or online from October 27–29! Register here: buff.ly/gCP2K1z #ADIALabSymposium2025

I was shocked to see the first two people in my "masterclass" 🎓 on #AI networking with Ultra Ethernet at #HLF25: David Patterson and Bob Metcalfe 😅! Both Turing award winners - Bob being one of the inventors of Ethernet 🥹. Was great fun also with many enthusiastic students and great discussions 🚀.

I'm honored to present the Raj Reddy Artificial Intelligence lecture at Carnegie Mellon University this Thursday. Join us in person at 5pm in Rashid Auditorium, Gates Hillman 4401! buff.ly/1kj9LtN This lecture series is honoring the Turing award winner's work in AI.

Thanks to The New York Times for featuring the ACM A.M. Turing Prize! Indeed, beyond the Nobel Prizes, "there are plenty of other prizes scientists and mathematicians can compete for." Learn more: buff.ly/NAhXVmy

The first ADIA Lab transactions edited by Horst Simon arrive at my desk 📖. An exciting mix of different science areas under the umbrella of advanced scientific computing and #AI 🎯. Congratulations to Horst and all co-authors 👏. Onward 🚀!

Listen to 2024 ACM Prize in Computing recipient Torsten Hoefler share how he tries to bridge academia and industry for societal impact . Full episode here or wherever you get your podcast: learning.acm.org/bytecast/ep7... #ACMByteCast #computing

Cool and impactful project led by the MPI-M: coupled simulation of the full Earth system in 1.25km resolution. Unprecedented complexity, fidelity, and performance at 82.5 simulated days per day. Accelerated by DaCe on GH200 - next generation #HPC for #climate science is here!

I met many of my heroes at my first Heidelberg Laureate Forum #HLF25! Exciting discussions with 28 other laureates of the five highest awards in CS and Math and young researchers! Watch my Spark talk at: buff.ly/SO6ntUb Thanks to the HLF Foundation and Klaus Tschira Stiftung.

Happy 100th Birthday to Seymour Cray! 🎂 We took your relentless pursuit of speed and parallel processing, and built on your ideas to start the #AI revolution. Every neural network owes a debt to your chips. Thanks for the gigahertz, Seymour! We're putting them to good use - your #LLM friend. 😉 #AI

Listen to 2024 ACM Prize in Computing recipient Torsten Hoefler discuss the power of High Performance Computing. Full episode here or wherever you get your podcast: learning.acm.org/bytecast/ep7... #ACMByteCast #computing

Apertus, the Swiss Fully-Open-Data model downloaded more than 379k times - ~13 downloads per minute of up to 140+ GiB! Trending for some weeks among the world's top LLMs. The techreport is now on arXiv: buff.ly/bvSioQH

Bill Gropp speaks at the Modeling to Learning with #HPC #ICERM workshop about MPI performance, emphasizing the characteristic rendezvous "little blip" where bandwidth is lost in transition. This will completely go away to a smooth transition with Deferrable Sends in Ultra Ethernet! buff.ly/0bEBb6y

Microsoft's first of many (connected!) Fairwater GB200 Supercomputers coming online in Wisconsin. 10x faster than the top of the #top500 list powered by an innovative network after years of planning and design. Proud to have contributed :-). Watch out for more info appearing.

Wonderful ICERM meeting on #HPC and #AI honoring Bill Gropp from NCSA. After crossing five US states by car (NJ, NY, CT, RI, MA) I met many old and new friends. Thanks to David Keyes to take the lead of this in a multi-year planning effort! buff.ly/fu8h4mB

Was great to attend the Intel Academic Security workshop in Hillsboro, OR to talk about Marcin's work on performance aspects of Confidential HPC and AI systems as well as giving a keynote on UE from a security perspective. TSS rules - thanks to Eric Spada's leadership! Read more: buff.ly/A9jczpC

📢REPS accepted at EuroSys 2026!🚀 A per-packet load balancer for out-of-order transports. It caches high-performing paths and reroutes away from failures. It requires no switch changes and uses only ~25B per flow. 👉https://arxiv.org/pdf/2407.21625 #HPC @thoefler.bsky.social @csateth.bsky.social

New #ISC25 podcast with Kevin Jackson on ISC highlights on connecting everything, #AI vs. #HPC? No! HPC is the Engine for AI and AI is HPC. Precision boosting, Ultra Ethernet, Reasoning Language Models, AI applications, and Supercomputing as driver for future societies!