Even though RIKEN is known for its CPU-based leadership in HPC, this is one of the most grounded perspectives on AI challenges I’ve seen at #ISC24 so far. None of that “we need to train a ten trillion parameter model using all the data!” nonsense.

👇 Median employee pay at Nvidia rose 17% to $266,939 during 2023

#AI #HPC #GPU via Dan Nystedt

Some you here at #ISC24 have asked whether you can apply for membership if you're from a small organization.

General Membership is open to users from all sizes of organizations! If you're a user of #HPC or #AI technology or both, we invite you to apply at hpcaileadership.org/services/

Thanks to those who participated in this poll

Half of us expect #Frontier to be #2 this Nov

A case could be made for the other 3 options too

For it not to be in the Top3: ElCapitan, Aurora and a 3rd system need to submit and be faster

#HPC #AI #Top500 #ISC24

Very excited about our 6th generation TPU: 'Trillium TPUs achieve an impressive 4.7X increase in peak compute performance per chip compared to TPU v5e. We doubled the High Bandwidth Memory (HBM) capacity and bandwidth, and also doubled the Interchip Interconnect (ICI) bandwidth…

Intel tells partners that Ponte Vecchio is starting its sunsetting process

@Intel is focusing its production capacity on Gaudi2/Gaudi3 and then ramping its converged #HPC and #AI Falcon Shores #GPU for 2025

servethehome.com/intel-ponte-ve…

#ISC24 via Patrick J Kennedy

Researchers Oak Ridge Lab explore #LLM training strategies on #Frontier

“At extreme scale, we achieved 30% efficiency — which means we left about 70% of Frontier’s computing power on the floor. We need much more optimization …”

ornl.gov/news/going-big…

#AI #GenAI #ISC24

Our new list is released here: io500.org Thank you for the submissions and waiting your feedback for our BOF discussion. #ISC24 #HPC

AMD GPUs are much heavier than I expected. Handy carrying handle though.

'GPT-4o is going to be free and it will be able to do all homework' - Ethan Mollick

Democratizing Access: Likely the biggest impact of GPT-4o is not technical, but a business decision - everyone, whether they are paying or not, will get access

oneusefulthing.org/p/what-openai-…

#AI #GenAI

#Frontier & #Aurora : DOE Lands Top Two Spots on List of Fastest #Supercomputers

And 3 in the Top10 - including #Summit Oak Ridge Lab

DOE Office of Science U.S. Department of Energy are understandably proud of their accomplishment

energy.gov/science/articl…

#HPC #AI #Top500 #ISC24

HPC Guru TOP500 Arguably several aspects of the list need modernizing.

Eg outdated assumption that supercomputers are fixed size - cloud HPC can scale capacity as needed - so what size do those customers submit - their max burst size, some annual average, or are those HPC buyers ineligible?

👉 The OpenMP BOF just took place at #ISC24 . Attendees got a peek at what the #OpenMP API Version 6.0 will look like, when it is released in November 2024!

👉 Note that you also can have a look at the preview.

openmp.org/wp-content/upl…

#HPC #Embedded Simon McIntosh-Smith Sunita Chandrasekaran

TOP500 5/

TPM: 'We remain convinced that the #HPC market is looking for an alternative to the Nvidia stack for price/performance reasons as those buying #AI training systems

We are looking at peak Nvidia in both the HPC and the AI realms, as hard as that might be to believe'

#ISC24

Groundbreaking advancements at #ISC24 ! Marshall Choy unveiled Fugaku-LLM on Samba-1, which was trained on Japan's fastest #supercomputer - #Fugaku - RIKEN. Learn more: sambanova.ai/blog/fugaku-ll…. #AI #MachineLearning

Listening in on the Urgent Computing session at #ISC24 . Sounds like a lot of the challenges here are not technical, but more policy and value.