Welcome to Runtime! Today: how the most powerful supercomputers in the world are evolving as AI workloads take center stage, Microsoft unveils a new approach to platform services, and the latest funding rounds in enterprise tech.
(Was this email forwarded to you? Sign up here to get Runtime each week.)
Keeping it 500
The biannual release of the Top500 list of the world's most powerful supercomputers does little to change the day to day reality of enterprise tech, but it does give a sense of how hardware trends change over time. The November 2023 edition unveiled Monday — the 62nd such list generated since 1993 — shows how supercomputing designs are changing along with the changing needs of their buyers.
AMD remains at the top of the heap with the only exascale supercomputer on the planet, the massive Frontier system built for the Department of Energy at Oak Ridge National Laboratory.
- That system uses more than 8.6 million CPU and GPU cores combined, and is being used to "model the entire lifespan of a nuclear reactor, uncover disease genetics, and build on recent developments in science and technology to further integrate artificial intelligence with data analytics and modeling and simulation," according to ORNL.
- Intel came in second, but assembly of its Aurora system at the Argonne Leadership Computing Facility in Illinois is only halfway done, and when that's finished (after years of delays) it should break the two-exaflop barrier.
- Intel still powers the most systems on the list given its historical position in the server market, but it is losing ground to AMD, which improved its share of the list by 16% compared to the list released in May.
Microsoft's investment in AI this year really paid off if the goal was to rocket up the Top500 list.
- Eagle, a Microsoft supercomputer built inside Azure, debuted at #3 on the newest list, the only cloud supercomputer in the top ten.
- Supercomputers have been traditionally built onsite for their customers in order to minimize latency, but Microsoft has now placed two systems in the top ten over the last two years.
- Last year Jack Dongarra, the Turing Award winner and professor who has worked closely with ORNL, told Protocol Enterprise's Kate Kaye that he believed future supercomputers would likely be built by cloud providers given how much more efficiently they can add computing power.
- But as The Next Platform pointed out, in recent years the pace at which supercomputing performance has improved has slowed.
- Still, the hybrid CPU/GPU systems that dominate the top of the current list are clearly the future of high-performance computing, and that computing power will start to trickle down to average enterprise users over the rest of the decade.
- AMD's Frontier system also points to the hopeful future of data center design in the AI era: Even though it was the most powerful system overall, it still ranked eighth on the Green500 list of energy-efficient supercomputers.
- "Considering this system was the first machine to achieve exascale, Frontier is proof that power does not need to be sacrificed to achieve an impressive energy efficiency rating," the Top500 organization said in a press release.
As you make your plans for 2024, please consider sponsoring Runtime and getting your message in front of the more than 20,000 enterprise tech industry leaders and decision makers that receive this newsletter each week. We also plan to roll out several new products next year, including special reports, sponsored content, and events, both virtual and live. If you're interested in learning more, contact us here.
That's just like, your opinion, man
Microsoft's .NET is one of the older software development platform technologies out there, and probably not the first choice for startups building cloud apps. But it remains a steady presence in enterprise development shops, and Tuesday Microsoft introduced a new tool within .NET called Aspire designed to help developers struggling to build and deploy cloud apps.
".NET Aspire is an opinionated stack for building resilient, observable, and configurable cloud-native applications with .NET," Microsoft's Glenn Condron wrote in a blog post. The emphasis was in the original sentence and it's an important part of a platform service's product strategy, as we covered in our look at the resurgence in PaaS startups and tools earlier this year.
Platform services that are too opinionated run the risk of slowing down developers as they try to learn how to follow the rules, but ones that leave too many things up to the developer aren't actually that helpful when it comes to simplifying the process. Given that .NET users are already familiar enough with the platform, Aspire should have an easier learning curve.
Deepinfra raised $8 million in seed funding and formally launched the company, which will provide a cloud infrastructure service for running AI inference workloads.
Flip AI landed $6.5 million in seed funding and also made its debut as an observability service built on top of a large-language model.
The Runtime roundup
Nvidia introduced the H200, the next generation of its most powerful line of AI GPUs that is expected to arrive in the second quarter of next year.
There are now more than 1 million public Kubernetes clusters up and running around the world, and that's making them more of a target for hackers, according to The Stack.
Gartner thinks worldwide cloud spending will rise 20.4% to $678.8 billion next year, thanks mostly to the generative AI boom.
Thanks for reading — see you Thursday!