@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20240101

Episode Date: January 1, 2024

- Moore's Law slows down, Intel, TSMC - TSMC 1nm Chips, Arizona plant - Huawei results, Mate 60 Pro smartphone, Kirin 9000S chip - DARPA US2QC, PsiQuantum, Microsoft [audio mp3="https://orionx.net/wp-content/uploads/2023/12/HPCNB_20240101.mp3"][/audio] The post HPC News Bytes – 20240101 appeared first on OrionX.net.

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. Happy New Year. We'll jump in with Moore's Law, which has, of course, driven computing speed up and costs

down exponentially for more than 50 years. But nearly everyone in the industry believes it hit a wall in the around 2015. And the phrase, the post-Moore's Law era is commonly used. But according to Intel CEO Pat Gelsinger, rumors of the law's demise are greatly exaggerated. Now, he concedes it's slowing down, but it is not dead yet. And of course, Intel, the originator of Moore's Law, is naturally invested in the term, whose definition is clearly evolving. Gelsinger made remarks recently at MIT. He also said Intel plans to roll out a one trillion transistor chip by the end of the

decade, and that's 10 times what we have today. TSMC, by the way, has also said they see more life left in Moore's law, and HPC performance as a whole on the system level has actually progressed faster than Moore's law, using parallelism at every level. Yeah, Moore's law being about transistor density per square millimeter is slowing down. But to keep doubling the number of transistors available to designers per chip, manufacturers are using new lithography, new materials. They're dealing with quantum effects as they shrink things. They're using larger chips. And then they're going 3D, which can give them another 5x or so by itself. So all of that is extending the life of Moore's law.

Of course, Cerebras is using a whole wafer with its dinner plate size wafer scale engine. So it's eclipsed the law. But wafers are not getting larger. The last time that happened was more than 20 years ago to 30 centimeters diameter, and the entire supply chain is geared up for that, and also smaller wafers. Speaking of Moore's Law, TSMC announced that it is marching towards one nanometer chips by the year 2030. It had started a lab to pursue that milestone last July, and seems to be pushing forward at its usual breakneck speed. One nanometer is approaching about as far as things can go with silicon. And this is another

instance, a major instance of the end of things theme we've just talked about regarding foundational advanced computing technology. In any case, by 2030, other forms of computing will have evolved further and may carry the torch forward for advanced computing. And speaking of TSMC, the company said it has settled its labor disputes in Arizona and construction of its factory there is back on track. There was news in Business Insider that Chinese vendor Huawei, following several quarters of revenue decline, has reported strong growth again, approaching $100 billion in sales. This is reviving discussions of the impact of Western trade sanctions with that company. While its revenue needs to grow by another 38% to get to its previous peak in 2020, the company grew by 9% and sounded upbeat

about turning a corner. In the Western world, the growth will be analyzed to see what products contributed to it and how much of it came from outside China. But the big topic has been Huawei's Mate 60 Pro smartphone, which was released last August. It uses Huawei's own homegrown ARM-based chip called Karin 9000S. That's a 7 nanometer chip presumably manufactured by Chinese chip manufacturer SMIC. There's a program at DARPA called Underexplored Systems for Utility Scale Quantum Computing, US2QC. Utility-scale is when quantum computers can beat classical methods. IBM says it's there now, but the industry in general appears to believe it remains in the current noisy intermediate-scale era, or maybe a step up from

that which some people call logical intermediate-scale. DARPA aims to find and accelerate promising approaches similar to what it did decades ago for massively parallel processing and then later for high productivity systems. Yeah, DARPA said it has selected Microsoft with its topological approach and PsiQuantum, which has an ambitious photonic strategy to move US2QC to the next phase. The program is expected to run through March 2025. All right, that's it for this episode. Thank you all for being with us and Happy New Year. Happy 2024, everyone.

Thank you. Take care. HPC Newsbytes is a production of OrionX in association with Inside HPC. Shaheen Khan and Doug Black host the show. Every episode is featured on InsideHPC.com and posted on OrionX.net. Thank you for listening.