Okay, I've been dancing around it for years--it's time to call it:
There's no economic story around using the cloud for steady-state HPC workloads. It's not even close to cost competitive.
Kīlauea is erupting at Hawaiʻi Volcanoes National Park! 🌋 The eruption is confined to Halema‘uma‘u crater and isnʻt a threat to public safety. There is no reason for travelers to change their plans to Hawaiʻi. For updates, visit https://t.co/8AJ670C2KG.
📸: USGS Volcanoes
DGX GH200: Nvidia ties 256 Grace-Hopper Superchips by 36 NVLink Switches to provide >1 EF FP8 (or ~9PF of FP64)
o 144TB unified memory
o 900 GB/s GPU-to-GPU bandwidth
o 128 TB/s bisection bandwidth
o End-of-year availability
https://t.co/NEIlCG7x5O
#HPC#AI via @HPCwire
#OnThisDay thirty years ago, in 1993, CERN released the World Wide Web software to the public.
Proposed by Sir Tim Berners-Lee, the web was originally created to allow scientists and institutes from all over the globe who were working on CERN data to share information accurately and quickly.
Find out more: https://t.co/6CpbWde774
#CERNImpact #CERNandSociety
To repeat in English, although amazing strides have been made in transformer-based language models from BERT to GPT4, it is anticipated that the rate of improvements will now slow down, as the supercomputers used for training has caught up with the top supercomputers of the world.
the only people I’ve seen dismissing this insightful article are ones who don’t seem to understand compression very well
the only surprise(?) is that it took a skilled writer and thinker, instead of one of us researchers, to make the case crystal clear
https://t.co/Gmr7JxAKuh
Today’s #PhotoOfTheWeek portrays the installation of the @CMSExperiment's silicon strip tracking detectors, successfully completed in 2007. They are part of the system that tracks the particles’ path, as they move through the experiment’s magnetic field.
🔗https://t.co/An3oiTSUyK
This the future, from FLOPS to BYTES. Various research incl ours have shown that ROI for ALUs has diminished for HPC, as majority of high performance apps are bandwidth bound. In fact, even if we had infinite flops, entire workload speed up is mere 7-30% https://t.co/Yiv6sLmKGl
Baidu researchers have developed HelixFold, an efficient protein structure prediction system implementing AlphaFold2 with PaddlePaddle, providing a complete pipeline of training and inference.
https://t.co/atbCmboFmp
Visual Studio Code as a plague.
Myriad has 2184 home directories.
Visual Studio Code creates a remote plug-ins directory approx 1.3GB in size.
That means if every user used VS Code we'd need an additional 2.8TB of storage.
More fun publisher surveillance:
Elsevier embeds a hash in the PDF metadata that is *unique for each time a PDF is downloaded*, this is a diff between metadata from two of the same paper. Combined with access timestamps, they can uniquely identify the source of any shared PDFs.
Very pleased to share our latest preprint: Julia for Biologists
#julialang opens up new possibilities for biologists (computational and beyond) and we discuss how and why.
https://t.co/D3t7OlIaO0
Do you want to attract computational biologists to your project?
Do you want to attract computational biologists to your department?
With a dozen colleagues around the globe, we present "A field guide to cultivating computational biology"!
Read on! https://t.co/mxg6mcQonI
@RonZimmermann on the science cloud project, CLIP, that we built - on the national radio (german only) https://t.co/PUCVis5JVt /cc @timeu_s@ebirn et al.
The HPC system that @ebirn, @PetarForai, Petar Jager and myself built and been managing for the last 3 years. Quite a bit of blood, sweat and tears went into it but I think in the end it payed off.
https://t.co/gZr6RnbA6D