ESA cuts the ribbon on 34,000-core Space HPC center tailored for space workloads
- Reference: 1741869009
- News link: https://www.theregister.co.uk/2025/03/13/esa_space_hpc_center_inaugurated/
- Source link:
[1]
The Space HPC facility in Italy – click to enlarge
The aptly named "SpaceHPC" facility is billed as being "demonstrator infrastructure" designed to help Europe's space industry "mitigate risks associated with data processing, modelling, and simulations."
Located in the Italian town of Frascati, 20km outside Rome, Space HPC houses a machine packing 34,000 cores' worth of the "latest generation of AMD & Intel processors." 108 Nvidia H100 GPUs are also present, giving the machine 5 petaflops of raw performance potential.
That power would see Space HPC ranked in around 210th place on the current [2]Top 500 List of Earth's mightiest supercomputers.
The machine uses InfiniBand networking, packs 156 TB of RAM, and includes 3.6 PB of solid state disk storage.
[3]
Direct liquid cooling allowed it to bag a power usage effectiveness score of "below 1.09." The machine is also plumbed into the heating system of the campus where it resides.
[4]
[5]
As is usually the case with supercomputers, Space HPC can be configured to run different workloads. The machine therefore offers partitions dedicated to general compute tasks, and two other partitions that take advantage of the H100s to run AI/ML workloads or other software that needs accelerators.
[6]UK must give more to ESA to get benefits of space industry boom, says Brian Cox
[7]ESA's Integral gamma-ray gazer gasps its last
[8]Einstein Probe finds two stars that have spent 40 million years taking turns eating each other
[9]European Space Agency picks Thales Alenia Space to build lunar lander
ESA's Space Safety Programme has already tested Space HPC to improve its ability to – you guessed it – model space weather. Among other things, it can improve warnings of future solar activity that could pose a danger to infrastructure in orbit or on the ground.
Despite Wednesday's inauguration ceremony the facility remains in its pilot phase until at least April.
"Broader access is planned thereafter, in a ramp-up mode to ensure optimal performance and user experience," the ESA [10]said .
[11]
The org is, however, already considering expressions of interest for time on the machine at a form you can find [12]here . ®
Get our [13]Tech Resources
[1] https://regmedia.co.uk/2025/03/13/supplied_esa_space_hpc.jpg
[2] https://top500.org/lists/top500/list/2024/11/
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Z9MPMvUkJZjo34YU3DrWzgAAAUo&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Z9MPMvUkJZjo34YU3DrWzgAAAUo&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Z9MPMvUkJZjo34YU3DrWzgAAAUo&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2025/03/06/uk_needs_to_contribute_more/
[7] https://www.theregister.com/2025/02/28/esa_integral_end/
[8] https://www.theregister.com/2025/02/19/einstein_probe_twin_stars_revealed/
[9] https://www.theregister.com/2025/01/31/european_space_agency_taps_thales/
[10] https://commercialisation.esa.int/space-hpc/
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Z9MPMvUkJZjo34YU3DrWzgAAAUo&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[12] https://forms.office.com/pages/responsepage.aspx?id=0Kxcmu8r102sXH6-H1T0lVkmHRqFN6BHsPcof67u9bxUOFNRNTlQSkhDTDBEQk03NlRBOTFPQ1AwOS4u&route=shorturl
[13] https://whitepapers.theregister.com/
Re: Theory then practice
At its simplest you can just run a bit of Python/whatever by submitting a shell script to the scheduler (eg Slurm).
What gets more complicated is when you need to spread your code over more than one node. For this, most people use something like [1]MPI to pass "messages" about. At its simplest you could use MPI to do something like "read file X and send back the results"; what's more complicated is doing things like simulating different bits of a protein on separate nodes and then coordinating the calculations.
[1] https://en.wikipedia.org/wiki/Message_Passing_Interface
I wonder if it will run a spillchecker!
The link to send "expressions of interest to has "The ESA Space HPC will allow you to boost your projects by performing complex computations and processing massive datasets more efficiently and rapidely"
It also says "We will come back to you shortly to 'exchange' on your needs." what’s wrong with the word discuss?
Re: I wonder if it will run a spillchecker!
Nothing. I use the word discuss all the time.
good stats
Yeah....good stats...but can it run Crysis?
Theory then practice
A good follow up article would be how you prepare a workload to be run on this or similar beasts. I’m guessing you don’t just remote login and hack away like it’s 1985. How, for example, do you choose when a workload will no longer run efficiently on your Xeon workstation and you need more horsepower?