@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20231211

Episode Date: December 11, 2023

- AMD MI300 availability as AI Chip Party heats up - Q2B Silicon Valley quantum conference update - Linux foundation's High-Performance Software Foundation and DAOS Foundation - Int'l consortium for ...trustworthy and reliable generative AI models for science - EU agrees landmark deal on regulation of AI - ISC Submission Deadlines extended [audio mp3="https://orionx.net/wp-content/uploads/2023/12/HPCNB_20231211.mp3"][/audio] The post HPC News Bytes – 20231211 appeared first on OrionX.net.

Transcript
Discussion (0)
Starting point is 00:00:00 Welcome to HPC News Bites, a weekly show about important news in the world of supercomputing, AI, and other advanced technologies. Hi, everyone. Welcome to HPC News Bites. I'm Doug Black. Hi, Shaheen. Biggest news this week was also a non-surprise. AMD announced availability of its new MI300 GPU accelerator, something they began
Starting point is 00:00:27 previewing early this year. I've heard general comments that the chip series in its release form is more impressive than expected. And indeed, AMD's stock price jumped about 8% since the announcement, and AMD shares are up about 60% year to date. The MI300 takes aim at NVIDIA GPU technology, including the H100 and the forthcoming H200. Speaking of stock prices, NVIDIA's is the biggest gainer of any stock this year. It's up a whopping 230%, if you can believe that. Yeah, AI is the party everybody wants to crash. Microsoft, Amazon, AMD, Intel, Qualcomm. Google also updated its TPU. Chips for internal use like Meta and Tesla are doing, not to mention Apple, and a dozen or more smaller chip vendors. Crucially, there's been a lot of effort on the required software and
Starting point is 00:01:18 in open source. NVIDIA is ahead with CUDA and a long list of GPU-enabled tools and apps, but others are increasingly able to provide enough to participate. So everyone's at the party and looks like there is enough demand for all to have a good time. That means market leadership will shift to who can ship how many chips, and there, NVIDIA continues to have a commanding lead for now. My favorite quantum technology conference, Q2B, Quantum to Business, was held last week, organized by QCWare, a noted software and services company. I've attended for five to six years, and this year gave a talk on quantum computing in HPC with a focus on energy efficiency of quantum accelerators. We can now add cold molecules to the list of modalities in addition to trapped ions, trapped atoms, photonics, superconducting, topological, diamond effect, annealers, silicon spin, and perhaps others.
Starting point is 00:02:14 Canada, Netherlands, Denmark, UK, Europe in general, Australia, where they're in force with a growing ecosystem and significant government investments. The U.S US is in the lead with government and private investments and control of key supply chain. And we are seeing some significant advances, especially in silicon spin with photonic interconnects, and the all-important quantum error correction and increasing focus on what I call HVC-centric quantum. IBM introduced the 133-qubit quantum Heron, the first in what IBM said will be a series of utility-scale quantum chips. The company said the processor delivers IBM's highest performance metrics
Starting point is 00:02:56 and lowest error rates. Also, QIRA, with Harvard, MIT, University of Maryland, and others, announced what the company said is a quantum breakthrough published in the scientific journal Nature. They implemented an advanced error-correcting algorithm to demonstrate 48 logical qubits and hundreds of entangling logical operations. Q-ERA utilizes the neutral atom modality, and the room temperature aspect of neutral atoms adds to the significance of this announcement. The Linux Foundation is backing several HPC initiatives with two new foundations. The High Performance Software Foundation aims to advance a portable core software stack for HPC
Starting point is 00:03:38 consisting of 10 different programs, and the Distributed Asynchronous Object Storage Foundation is focused on the governance and development of DOS, the open source storage system that's had significant impact on the performance of HPC and AI workloads. Related to that is the formation of the Trillion Parameter Consortium, TPC, to create trustworthy and reliable generative AI models for science. TPC.dev is their website. There's a global research consortium of about 70 organizations, government, academia, industry, all involved.
Starting point is 00:04:13 On the AI policy front, there was big news out of Europe. The European Union reached a provisional deal on the world's first comprehensive laws to regulate responsible use of artificial intelligence. It's a big deal, but it has to go through the regulatory process before it becomes law, which they said would be 2025 at the earliest. Yeah, under these new provisions, consumers would have the right to launch complaints and fines could be imposed for violations. It all has echoes, Shaheen, of Europe's GDPR, data privacy regulations, which are stronger than what we have in the United States. So Europe is taking the lead again.
Starting point is 00:04:50 Now, the European Supercomputing Conference, ISC, will be in Hamburg from May 12th to the 16th. They have extended some of their deadlines, so be sure to check their website. Yeah, it's always a great show. And if your submission is accepted, you get a free registration. So make sure you do that. All right, that's it for this episode. Thank you all for being with us. HPC Newsbytes is a production of OrionX in association with InsideHPC. Shaheen Khan and Doug Black host the show. Every episode is featured on InsideHPC.com and posted on OrionX.net. Thank you for listening.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.