TOP500 List Updated With Frontier Remaining At The Top

[phoronix]

Phoronix: TOP500 List Updated With Frontier Remaining At The Top

The 61st edition of the TOP500 supercomputer list has been published this morning. The Frontier supercomputer continues to hold the top spot and the only true Exascale system...

https://www.phoronix.com/news/TOP500-Summer-2023

[Jabberwocky]

Notably absent from the list is the Aurora supercomputer that is to be powered by Intel Xeon Max CPUs and Intel GPU Max accelerators. Aurora is still in the works for what should be a very competitive Exascale supercomputer.

Aurora's history reads like one of the new South African coal power station, "was meant", "to be finished", "expected", "would be around", "delayed", "delayed again for", "would no longer"... Over budget and way behind schedule. Just like the power stations here, I hope they are able to finish it this year.

In 2013 DOE presented their exascale vision of one exaFLOP at 20 MW by 2020. Aurora was first announced in 2015 and to be finished in 2018. It was expected to have a speed of 180 petaFLOPS which would be around the speed of Summit. Aurora was meant to be the most powerful supercomputer at the time of its launch and to be built by Cray with Intel processors. Later, in 2017, Intel announced that Aurora would be delayed to 2021 but scaled up to 1 exaFLOP. In March 2019, DOE said that it would build the first supercomputer with a performance of one exaFLOP in the United States in 2021. In October 2020, DOE said that Aurora would be delayed again for a further 6 months and would no longer be the first exascale computer in the US. In late October 2021 Intel announced Aurora would now exceed 2 exaFLOPS in peak double-precision compute.

https://en.wikipedia.org/wiki/Aurora_(supercomputer)

[schmidtbag]

Well with Intel's 10nm struggles, AMD's performance value, AMD's higher core density, and AMD's better performance-per-watt, it makes sense how Aurora would continued to be delayed. Why put all this effort into something that is obsoleted before it is even operational?

[Flaburgan]

The GREEN500 list is very interesting. The number 6 there is... the number one of the TOP500. Which means, you can be powerful *and* energy efficient. I would love to see even more improvements in that direction. Climate needs it. Well, if we don't have too much rebound effect obviously.

[andrei_me]

Do we know more about Aurora status?

I read somewhere about the challenges faced during the deployment of Frontier, a missing bit/config and so on until all racks were up and running.

That kind of transparency is what we expect nowadays

[grigi]

Aurora's history reads like one of the new South African coal power station, "was meant", "to be finished", "expected", "would be around", "delayed", "delayed again for", "would no longer"... Over budget and way behind schedule. Just like the power stations here, I hope they are able to finish it this year.

In 2013 DOE presented their exascale vision of one exaFLOP at 20 MW by 2020. Aurora was first announced in 2015 and to be finished in 2018. It was expected to have a speed of 180 petaFLOPS which would be around the speed of Summit. Aurora was meant to be the most powerful supercomputer at the time of its launch and to be built by Cray with Intel processors. Later, in 2017, Intel announced that Aurora would be delayed to 2021 but scaled up to 1 exaFLOP. In March 2019, DOE said that it would build the first supercomputer with a performance of one exaFLOP in the United States in 2021. In October 2020, DOE said that Aurora would be delayed again for a further 6 months and would no longer be the first exascale computer in the US. In late October 2021 Intel announced Aurora would now exceed 2 exaFLOPS in peak double-precision compute.

https://en.wikipedia.org/wiki/Aurora_(supercomputer)

Bad memories of giving up dealing with the unrestrained corruption that is Kusile and Medupi.

But an interesting correlation re the marketing of the failures. Although I think Intel is probably going to deliver long before those power stations ever go online.

[Weasel]

Once again showing how x86 is on a different league than RISC shit like ARM when it comes to performance, and efficiency is very good as well.

Get fucked dumbass ARM trolls.

[cb88]

Once again showing how x86 is on a different league than RISC shit like ARM when it comes to performance, and efficiency is very good as well.

Get fucked dumbass ARM trolls.

Not really, Fugaku uses more power because its almost entirely CPUs... so its performance on many tasks will be much greater if the code requires lots of branches or very large/fast node memory. Most of the flops from Frontier are GPUs... it has basically 200k less cpus than Fugaku.

[scalepaper]

The GREEN500 list is fascinating. I read somewhere about the difficulties encountered during the Frontier deployment, such as a lost bit/config and so on until all racks were operational.

basket random