Skip to main content
  1. Home
  2. Computing
  3. News

HPE’s 145,000-core supercomputer could help manage U.S. nuclear stock

Add as a preferred source on Google

Hewlett Packard Enterprise (HPE) has built the world’s most powerful supercomputer based on ARM processors. The “Astra” system packs thousands of processors with a combined 145,000-cores together with a huge pool of memory into a surprisingly small supercomputer and is designed to help handle the U.S.’ nuclear stockpile.

Just shy of 2,600 dual ARM processor servers are packed into Hewlett Packard Enterprise’s (HPE) new supercomputer, making it the most powerful system ever built using ARM hardware. Known as the Astra, the mammoth processing platform will be used to run simulations for the National Nuclear Security Administration (NNSA).

HPE’s Next Frontier: Memory-Driven Computing

While Nvidia may be busy pioneering new, miniature supercomputers powered by parallel computing graphics cards, HPE has been working on a new design constructed entirely of dual-processor ARM servers. The new Astra system is purportedly made up of more than 145,000 processing cores and has a far more efficient way of handling its memory.

Recommended Videos

At its peak, Astra is said to be able to offer 2.3 petaflops of performance, with memory that performs up to 33 percent faster than typical supercomputers of comparable size. That’s the key to the system, its memory performance. That’s what allows the system to handle vast quantities of data, which is its core focus — rather than some of the high-performance computing tasks given to more typical supercomputers.

This is all part of HPE’s push for more “memory-driven computing,” as the company calls it. First debuted as a concept in 2017, HPE has, with Astra, developed a system where each of its thousands of processors has access to a large pool of memory. That should massively reduce the latency typically associated with processors accessing the memory of their companion chips.

Astra certainly isn’t going to hold a candle to the recently unveiled IBM system, which has more than 9,000 22-core IBM CPUs and over 27,000 Nvidia Tesla V100 GPUs (so that’s where all the graphics cards have been hiding). But the Astra should be perfectly capable of handling the tasks the NNSA has for it. Developed in conjunction with the U.S. Department of Energy, Astra will be used for nuclear research and running simulations for the NNSA’s Sandia National Laboratories.

At Sandia, the Astra supercomputer will be given a trial run to start with, but could ultimately be used to help manage the nuclear stockpile of the U.S. As Engadget highlights, Sandia runs many physics simulations on a daily basis and it’s hoped that Astra will be able to accelerate that. As a new style of supercomputer, its potential lies in redefining the types of high-performance computers used at such institutions and could mean that ARM-based supercomputers become far more common in the future.

Updated 06/19/2018 to clarify HPE as a separate entity from HP.

Jon Martindale
Jon Martindale covers how to guides, best-of lists, and explainers to help everyone understand the hottest new hardware and…
Microsoft Edge is about to get more frequent updates, but don’t expect more features
Starting with Edge 152 on August 27, Microsoft is cutting its release cycle in half, with smaller but more frequent updates for Stable channel users.
Microsoft Edge illustration official

Microsoft is accelerating updates to its Edge browser, switching from a monthly release schedule to a biweekly one. The change takes effect with Edge 152, due on August 27, and puts the browser on the same cadence as Google Chrome.

More updates, not more features

Read more
What makes a laptop good for both work and entertainment?
Computer, Electronics, Laptop

This post is brought to you in paid partnership with HP.

The HP OmniBook X Flip is designed as an all‑day AI PC that adapts seamlessly from productivity to entertainment without switching devices.

Read more
Your Windows 11 PC can now natively run AI workloads, even if it lacks the Copilot+ badge
Windows 11 laptop on a table

For the better part of a year, Microsoft has been telling us that the future of AI on Windows belongs to Copilot+ PCs. If you wanted Microsoft’s most advanced local AI features, you needed a machine with a dedicated Neural Processing Unit (NPU). That was the deal. Now, Microsoft appears to be rewriting the rules.

According to updated documentation, Windows 11’s local Language Model APIs can now run on non-Copilot+ PCs, provided they have an Nvidia GeForce RTX 30-series GPU (or newer) with at least 6GB of VRAM. On the surface, this sounds like a developer-focused update. In reality, it could be one of the most significant shifts in Microsoft’s AI PC strategy since Copilot+ PCs launched last year. More importantly, it raises a question that has been lingering ever since the AI PC era began: Did we really need NPUs for all of this in the first place?

Read more