June 2024 Top 500 (supercomputers)

[Yoused]

Top 500 has updated the standings for the first half of this year.

Still at the top is the Frontier machine at Oakridge National Laboratories, boast 1.2EFlops of performance.

In second place is Aurora, at Argonne NL. Curious thing, Aurora has more cores and uses 70% more juice but Frontier is still 20% faster. Naturally, Aurora is based of AMD EPYC + AMD Instinct GPU, while Aurora is Xeon + Intel Data Center GPU.

The power draw on these machines is kind of breath-taking. Aurora uses 39 GW, which sounds to me like rather a lot. I am not sure whether that number includes the cooling system.

The ARM-based Fugaku that once held the top spot has dropped to 4th. It looks like most of the other installations get their push from GPU, but Fugaku is entirely ARM cores, no GPUs. And it is certainly not more energy efficient than Frontier.

There is a Grace installation in Switzerland in 6th place, which looks like it could possible contend for the number one spot if it could be scaled up to a similar size.

 

[throAU]

The power draw on these machines is kind of breath-taking. Aurora uses 39 GW, which sounds to me like rather a lot. I am not sure whether that number includes the cooling system.

Excuse me what?

Surely that's a typing error and should be 39MW because the entire gold/copper mine site I used to work at only used a 100 MW gas turbine power station for their open pit operations including all the admin buildings, the processing plant, other smaller surrounding mines, the local village accomodation, etc.


edit:
Must be a typo, its more power than the worlds largest power station puts out:

Three Gorges Dam - Wikipedia

en.wikipedia.org

 

[dada_dave]

Excuse me what?

Surely that's a typing error and should be 39MW because the entire gold/copper mine site I used to work at only used a 100 MW gas turbine power station for their open pit operations including all the admin buildings, the processing plant, other smaller surrounding mines, the local village accomodation, etc.


edit:
Must be a typo, its more power than the worlds largest power station puts out:

Three Gorges Dam - Wikipedia

en.wikipedia.org

Indeed @Yoused just made a typo. It’s 38MW in the chart - it’s listed as >38,000KW.

 

[throAU]

Indeed @Yoused just made a typo. It’s 38MW in the chart - it’s listed as >38,000KW.

Cool, cool carry on. Figured it was maybe typo-ed in the original source perhaps as well. Either way that’s more like it and still a metric fuckload of electricity

 

[cbum]

Fuckload is definitely NOT metric... could certainly be imperial...

 

[cbum]

.

 

[Yoused]

Fuckload is definitely NOT metric... could certainly be imperial...

fuckton is metric

 

[Jimmyjames]

The power draw on these machines is kind of breath-taking. Aurora uses 39 GW, which sounds to me like rather a lot. I am not sure whether that number includes the cooling system.

Crazy that they’re running TWO i9-14900Ks! Maniacs

 

[dada_dave]

Nvidia Grace Hopper Superchips help three new supercomputers top the Green500 list — GH200 is more efficient than rival supercomputers with AMD's Instinct MI250X

The Green500 introduces new supercomputers powered by Nvidia in the top three, while the Top500 list stays the same.

www.tomshardware.com

On the other hand, JEDI has a power rating of 67.31 kW (73 GFlops/watt), Isambard-AI Phase 1 requires 117.08 kW (68 GFlops/watt), and Helios needs 316.88 kW (67 GFlops/watt). These devices aren’t in the top 10 regarding raw computing performance, but they aren’t that far behind. JEDI is ranked at 189, while Isambard reached 128. Helios GPU is the most powerful of the bunch, sitting in the 55th position. However, their combined power requirements are still dwarfed by the Aurora, which requires 38,698 kW (26.15 GFlops/watt) and is the second most powerful supercomputer.

All three new entrants to the Green500 are powered by Nvidia’s Grace Hopper super chips, which use the Arm architecture and are inherently more efficient than traditional x86 systems. Aside from the top three systems, other Grace Hopper-powered supercomputers also made the top 10 in the Green500 list—preAlps (5) and Venado (8) also made it, making Nvidia the preferred supplier of 5 out of the ten most efficient supercomputers. This demand likely helped the company reach a record $26 billion 24Q1 revenue.

 

[throAU]

fuckton is metric

That’s fucktonne to you sir

 

[Cmaier]

When i was at AMD, we got *really* excited when we made it on the list. If I remember correctly, we did a deal with cray, and ended up at th top for awhile. I could be misremembering. Opteron was really a great chip. Still a miracle that it got done.

 

[dada_dave]

Intel removing the last of Xeon Phi support from compilers. Why post that here? Read on …

Intel throws in the towel on the processors that killed the first Aurora supercomputer — Knights Mill and Knights Landing support removed from LLVM

Larrabee is gone forever?

www.tomshardware.com

The Knights Mill processors faced numerous delays and reportedly didn't meet performance targets, ultimately resulting in the cancellation of the first revision of the Aurora supercomputer after several years of delays. The DoE later changed the design of Aurora to employ Intel's Sapphire Rapids and Ponte Vecchio compute GPUs and faced yet more delays and performance issues primarily due to problems with Intel's Sapphire Rapids CPUs and Ponte Vecchiocompute GPUs.

These hardware issues, along with cooling system malfunctions and other stability problems, still prevent Aurora from achieving its full potential. However, the system is now on track for full deployment sometime this year, a whopping nine years since Aurora was first announced.

 

[Yoused]

Seems kind of disconcerting that the most power-hungry SCs are owned by the DoE.

 

[Citysnaps]

Seems kind of disconcerting that the most power-hungry SCs are owned by the DoE.

Guessing nearby's Stanford Linear Accelerator Center (SLAC) operated for the DoE, and also Lawrence Livermore National Lab.

No doubt NSA has the latest and greatest.