Aurora supercomputer is now fully operational, available to researchers
Another ExaFLOPS supercomputer available for researchers.

Argonne National Laboratory this week said that its Aurora supercomputer is now fully operational and is available to the scientific community. The machine, which was announced in 2015 and faced massive delays, now offers over 1 FP64 ExaFLOPS performance for simulation as well as 11.6 mixed precision ExaFLOPS for artificial intelligence and machine learning.
"We are ecstatic to officially deploy Aurora for open scientific research," said Michael Papka, director of the Argonne Leadership Computing Facility (ALCF), a DOE Office of Science user facility. "Early users have given us a glimpse of Aurora's vast potential. We are eager to see how the broader scientific community will use the system to transform their research."
The availability of the Aurora supercomputer for open scientific research may be considered a formal acceptance of the system by ARNL, which marks an important milestone for the troubled machine. Initially planned for 2018, Aurora missed this target due to Intel's decision to discontinue its Xeon Phi processors. After the machine was re-architected, the project faced further setbacks due to Intel's 7nm process technology delay, pushing the completion date to 2021 and then again to 2023.
Even after the hardware was installed in June 2023, it took several months for the system to be fully operational and achieve exascale performance, which it finally reached in May 2024. Yet, for well over half a year, the system was only available to select researchers.
While Aurora is not the most powerful supercomputer for simulations, as its FP64 performance barely exceeds one ExaFLOPS, it is the most powerful system for AI as it can achieve 11.6 mixed precision ExaFLOPS according to the HPL-MxP benchmark.
"A big target for Aurora is training large language models for science," said Rick Stevens, Argonne associate laboratory director for Computing, Environment and Life Sciences. "With the AuroraGPT project, for example, we are building a science-oriented foundation model that can distill knowledge across many domains, from biology to chemistry. One of the goals with Aurora is to enable researchers to create new AI tools that help them make progress as fast as they can think — not just as fast as their computations."
Some of the first research projects using Aurora are detailed simulations of intricate systems, such as the human circulatory system, nuclear reactors, and supernova explosions. The machine's overwhelming performance is also instrumental in processing data from major research centers, such as Argonne's Advanced Photon Source (APS) and CERN's Large Hadron Collider.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
"The projects running on Aurora represent some of the most ambitious and innovative science happening today," said Katherine Riley, ALCF director of science.
"From modeling extremely complex physical systems to processing huge amounts of data, Aurora will accelerate discoveries that deepen our understanding of the world around us."
On the hardware side, Aurora clearly impresses. The supercomputer comprises 166 racks, each holding 64 blades, for a total of 10,624 blades. Each blade contains two Xeon Max processors with 64 GB of HBM2E memory onboard and six Intel Data Center Max 'Ponte Vecchio' GPUs, all cooled by a specialized liquid-cooling system.
In total, Aurora has 21,248 CPUs with over 1.1 million high-performance x86 cores, 19.9 PB of DDR5 memory, and 1.36 PB of HBM2E memory attached to the CPUs. It also features 63,744 GPUs optimized for AI and HPC equipped with 8.16 PB of HBM2E memory. Aurora uses 1,024 nodes with solid-state drives for storage, offering 220 PB of total capacity and 31 TB/s of bandwidth. The machine relies on HPE's Shasta supercomputer architecture with Slingshot interconnects.
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
Blastomonas Wondering how much faster and cheaper this would have been if they had gone AMD :pReply -
bit_user
Frontier and El Capitan are both AMD-based.Blastomonas said:Wondering how much faster and cheaper this would have been if they had gone AMD :p
https://www.tomshardware.com/news/intel-amd-top500-fastest-supercomputer-frontier-aurora-exaflop(Published Nov. 13, 2023)
https://www.tomshardware.com/desktops/servers/hpe-flaunts-el-capitan-supercomputer-blade-with-amds-instinct-mi300-projected-to-be-worlds-fastest-when-finished-this-year(Published May 15, 2024)
Speed-wise, here's the latest published Top 500 list, which features those two in the top slots*:
https://top500.org/lists/top500/2024/11/
* Note that China has stopped submitting its systems to Top 500. -
bit_user
Not social media. They have datacenters full of servers, connected via high-speed networking, but there are some differences between those and proper supercomputers.rammed5559 said:Super computers are used by governments and social media services mostly anyway
AI training clusters are much more like a supercomputer than the racks of machines someone like Facebook would use for serving social media feeds. -
JarredWaltonGPU "We are incredibly relieved to officially deploy Aurora for open scientific research..."Reply
Fixed that for them. LOL. The 2 exaFLOPS supercomputer plan has ultimately resulted in a 1.2 exaFLOPS system that uses significantly more power than AMD's actual 2 exaFLOPS El Capitan. Oops. 🙃 -
P.Amini
You are here? We are waiting for DLSS4 review 🙃JarredWaltonGPU said:"We are incredibly relieved to officially deploy Aurora for open scientific research..."
Fixed that for them. LOL. The 2 exaFLOPS supercomputer plan has ultimately resulted in a 1.2 exaFLOPS system that uses significantly more power than AMD's actual 2 exaFLOPS El Capitan. Oops. 🙃 -
bit_user
Not as relieved as Intel is going to be, once those Ponte Vecchio cards go out of warranty!JarredWaltonGPU said:"We are incredibly relieved to officially deploy Aurora for open scientific research..."
; ) -
bit_user
I'm sure he's just blowing off steam, while some bencmarks finish. Plus, he's an editor, so inspecting the final product (articles) is part of the job.P.Amini said:You are here? We are waiting for DLSS4 review 🙃 -
JarredWaltonGPU
It's taking time. As always. And I basically took yesterday off after the review posted. I'm nearly done with my DLSS4 testing on several competing GPUs so that I can write it up. Want some spoilers? Okay...P.Amini said:You are here? We are waiting for DLSS4 review 🙃
Hogwarts Legacy runs like poop if you turn on all the RT bells and whistles. It's horrible. And I blame Unreal Engine 4, because this has been a common problem. So, on the 9800X3D, I'm CPU limited to around 58 FPS with full RT effects enabled. I can run 4K with DLSS Transformers Quality mode and I get almost the same performance as at 1080p on the 5090. In fact, I get almost the same performance from the 4090 as well, and only the 4080 Super falls a bit off the pace (46 FPS at 4K vs. 57 FPS at 1440p, with quality upscaling).
In all cases, there's a ton of micro-stuttering going on. Doesn't matter if it's 4080 Super or 4090 or 5080 or 5090. The engine and game are just trying to do too much in a poor fashion. So... with a hard CPU bottleneck at these settings, framegen and MFG to the rescue?
Nope. The stuttering still exists. FG/MFG cover it up slightly, but they both depend on relatively predictable frame pacing, so when you go from around 60 FPS and then ever 60 frames or whatever you get a stutter down to 30 FPS for a frame or two, the stuttering screws up framegen and you still end up feeling it. Here's some numbers:
HogwartsLegacyFullRT RTX 5090 DLSSQT (4K) - AVG: 55.32 99pMIN: 32.8
HogwartsLegacyFullRT RTX 5090 DLSSQT MFG2X (4K) - AVG: 113.81 99pMIN: 51.3
HogwartsLegacyFullRT RTX 5090 DLSSQT MFG3X (4K) - AVG: 168.80 99pMIN: 68.7
HogwartsLegacyFullRT RTX 5090 DLSSQT MFG4X (4K) - AVG: 222.03 99pMIN: 81.5
And the "full RT" isn't because it's path tracing, it's just what I named the files to distinguish them from the non-RT testing I've already done. And "99pMIN" is what I called "1% low average FPS" if you're wondering. So on the 5090, MFG scaling is almost perfect. You go from 55 to 114 to 169 to 222. Give or take, margins of error, that's pretty interesting. But even at "222" FPS with 4X MFG, the game feels more like a game running at maybe 70-80 FPS with stuttering.
That's only one of the five games I'm testing for RT, DLSS4, etc. And only three are DLSS4 (because I didn't want to jump through hoops to try to get the preview builds of the other two games). Basically, I'll have Alan Wake 2 (native DLSS4 with full RT), Black Myth Wukong (full RT and DLSS3), Cyberpunk 2077 (native DLSS4 with full RT), Hogwarts Legacy (native DLSS4 and advanced RT), and Indiana Jones and the Great Circle (full RT and DLSS3).
I still need to test the RX 7900 XTX in the five games, for comparison, and I need to retest the 5080 on two of the games (that got public DLSS4 patches today, doh!) But I plan to have this finished up by tomorrow, hopefully sooner than later in the day. 🙃