@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20240122

Episode Date: January 22, 2024

- Synopsys to Acquire Ansys - Meta is spending billions on Nvidia chips - OpenAI CEO raising billions to build new chips and new chip factories - Pawesey selects QuEra, US Geological Survey selects Q-CTRL, Quantinuum Raises $300M [audio mp3="https://orionx.net/wp-content/uploads/2024/01/HPCNB_20240122.mp3"][/audio] The post HPC News Bytes – 20240122 appeared first on OrionX.net.

Welcome to HPC NewsBytes, a weekly show about important news in the world of supercomputing, AI, and other advanced technologies. Welcome to HPC NewsBytes. I'm Doug Black. Hi, Shaheen. In a move that consolidates several HPC and AI application areas, Synopsys, a top electronic design automation company, is acquiring ANSYS, which has roots in structural analysis and fluid flow and much else. ANSYS goes back to 1970 and its line of CAE, multi-physics engineering simulation software, is one of the great

workhorses of HPC. You could say ANS Ansys has been a locomotive of HPC. Its software is an engine for HPC adoption and growth. The deal is estimated at $35 billion. Yes. Ansys, with revenues of over $2 billion a year, has been a famed HPC application since 1970 when John Swanson founded Swanson Analysis Systems Inc., SASE. Outside of Pittsburgh, Pennsylvania, they did finite element analysis on mainframes back then. It focused on mechanical computer-aided design, MCAD, which later became MCAE for engineering, competing with other famous codes like Nastran

and then later Abacus, Mark, and others. It grew steadily. It changed hands. It renamed itself to Ansys after the main app. It went public, made series of acquisitions like the app Fluent for fluid dynamics and LS Dyna for crash simulations, L standing for Livermore, all to become a powerhouse for 3D design, materials, thermal, electrical, optics, acoustics, safety engineering, on and on, even autonomous vehicle simulation. Likewise, Synopsys, a Silicon Valley

company founded in 1986 and doing close to $6 billion a year, has been a top EDA company, Electronic Design Automation, along with Cadence and Mentor, which is a US company that is now part of the German company Siemens. The combined company will have a massive portfolio that it can integrate for advanced and multi-physics product design, as you mentioned. Reuters announced findings of their investigative reporting on the effectiveness of banning chips. NVIDIA chips, for example, including the H100 and A100 GPU accelerators, have been banned for export into China, but are turning up in procurement documents at Chinese organizations in violation of US export policy, seemingly so. Regulations are typically effective,

but materials flow is really complicated and making it airtight is just hard. What do you think? Two things should be said about this. One is that an NVIDIA spokesperson said in response to the but materials flow is really complicated and making it airtight is just hard. What do you think? Two things should be said about this. One is that an NVIDIA spokesperson said in response to the Reuters story that if the company learns that a customer has made an unlawful resale to third parties, that NVIDIA will take, quote, immediate action. Not sure what that means. But the second is that it appears very few banned chips are finding their way into China. The amounts are a half dozen or fewer in most cases. Nothing approaching the thousands of GPUs, hyperscalers, and AI development sites need for big AI implementations. That brings us to our second related story in this area.

Mark Zuckerberg said Meta will spend billions on NVIDIA AI chips in 2024 to the tune of some 350,000 H100 class GPUs, an astounding number. That could be a non-trivial fraction of NVIDIA's total output this year. It reinforces the strong momentum of AI infrastructure build out in NVIDIA's existing strengths in AI. Zuckerberg's ultimate goal to put meta in the lead in the pursuit of artificial general intelligence, which is either alarming or exciting, depending on your view of AI. Yeah, related to this, the frenzy to acquire GPUs and build AI factories of the future doesn't seem to be slowing down, though the energy use for these systems, and especially for inference, is going to be a growing obstacle.

If you see it that way and also expect this momentum to continue for the next decade or two, and if you are someone like Sam Altman, the CEO of OpenAI, then you do what he's doing. One, you invest in new low-power CPU architectures for AI, like neuromorphic chips, which we mentioned in a previous episode. And two, as it emerged yesterday, Altman is raising billions to build actual semiconductor

fabs to secure capacity for the mid to long term. So the stakes are massive and becoming even more massive. Imagine setting out in 2024 to compete with NVIDIA and TSMC at the same time. Nice. In quantum computing news, good work by Quera to muscle into Australia, winning a deal at the Posi Center on the west coast of the country, and ahead of indigenous vendors like Silicon Quantum Computing, SQC, which showed a simulation of a small organic molecule on its atomic scale circuit a couple of years ago, and Quantum Brilliance, which is pursuing diamond quantum computers, software, and apps. Yes, lots of news on the quantum front lately. Another Australian company, Q-Control, Q-CTRL, announced a partnership with the U.S. Geological Survey, and as discussed

last week, the RIKEN Research and Supercomputer Center in Japan is now aligned with Quantinuum. There clearly is pairing up going on between major HPC users and supercomputing centers. And speaking of Quantinuum, the company announced last week a venture round totaling $300 million. Okay, that's it for this episode. Thanks so much for being with us. HPC News Bites 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.