🤍
😀컴덕534
5개월 전

🗨️ 2023년 11월 슈퍼컴퓨터 성능 순위

카테고리   ➡︎   🖥️ 컴퓨터 · 노트북 · 수리| 커뮤니티 정보/소식

2023년 11월 슈퍼컴퓨터 성능 순위가 발표됐습니다. 

 

1위는 프론티어입니다. AMD 에픽 3세대 64코어 프로세서에 인스팅트 MI250X 조합으로 1.679TFlops의 성능을 냅니다. 

 

HIGHLIGHTS - NOVEMBER 2023

This is the 62nd edition of the TOP500.

The 62nd edition of the TOP500 shows five new or upgraded entries in the top 10 but the Frontier system still remains the only true exascale machine with an HPL score of 1.194 Exaflop/s.

The Frontier system at the Oak Ridge National Laboratory, Tennessee, USA remains the No. 1 system on the TOP500 and is still the only system reported with an HPL performance exceeding one Exaflop/s. Frontier brought the pole position back to the USA one year ago on the June 2022 listing and has since been remeasured with an HPL score of 1.194 Exaflop/s.  

Frontier is based on the latest HPE Cray EX235a architecture and is equipped with AMD EPYC 64C 2GHz processors. The system has 8,699,904 total cores, a power efficiency rating of 52.59 gigaflops/watt, and relies on HPE’s Slingshot 11 network for data transfer.  

The Aurora system at the Argonne Leadership Computing Facility, Illinois, USA is currently being commissioned and will at full scale exceed Frontier with a peak performance of 2 Exaflop/s. It was submitted with a measurement on half of the final system achieving 585 Petaflop/s on the HPL benchmark which secured the No. 2 spot on the TOP500.

Aurora is built by Intel based on the HPE Cray EX - Intel Exascale Compute Blade which uses Intel Xeon CPU Max Series processors and Intel Data Center GPU Max Series accelerators which communicate through HPE’s Slingshot-11 network interconnect.

The Eagle system installed in the Microsoft Azure cloud in the USA is newly listed as No. 3. This Microsoft NDv5 system is based on Intel Xeon Platinum 8480C processors and NVIDIA H100 accelerators and achieved an HPL score of 561 Pflop/s.

The Fugaku system at the RIKEN Center for Computational Science (R-CCS) in Kobe, Japan is at No. 4 the first system listed outside of the USA. It previously held from June 2020 until November 2021 the No. 1 position on the TOP500.  Its HPL benchmark score of 442 Pflop/s, is now only sufficient for the No. 4 spot.   

The LUMI system at EuroHPC/CSC in Finland has been further upgraded and is now listed as No. 5 worldwide. It remains the largest system in Europe and is listed with and HPL score to now 380 Pflop/s.  

Here is a summary of the system in the Top 10:

  • Frontier remains the No. 1 system in the TOP500. This HPE Cray EX system is the first US system with a performance exceeding one Exaflop/s. It is installed at the Oak Ridge National Laboratory (ORNL) in Tennessee, USA, where it is operated for the Department of Energy (DOE). It currently has achieved 1.194 Exaflop/s using 8,699,904 cores. The HPE Cray EX architecture combines 3rd Gen AMD EPYC™ CPUs optimized for HPC and AI, with AMD Instinct™ 250X accelerators, and a Slingshot-11 interconnect.

  • Aurora achieved the No. 2 spot by submitting an HPL score of 585 Pflop/s measured on half of the full system. It is installed at the Argonne Leadership Computing Facility, Illinois, USA, where it is also operated for the Department of Energy (DOE). This new Intel system is based on HPE Cray EX - Intel Exascale Compute Blades. It uses Intel Xeon CPU Max Series processors, Intel Data Center GPU Max Series accelerators, and a Slingshot-11 interconnect.

  • Eagle the new No. 3 system is installed by Microsoft in its Azure cloud. This Microsoft NDv5 system is based on Xeon Platinum 8480C processors and NVIDIA H100 accelerators and achieved an HPL score of 561 Pflop/s.

  • Fugaku, the No. 4 system, is installed at the RIKEN Center for Computational Science (R-CCS) in Kobe, Japan. It has 7,630,848 cores which allowed it to achieve an HPL benchmark score of 442 Pflop/s.

     

  • The (again) upgraded LUMI system, another HPE Cray EX system installed at EuroHPC center at CSC in Finland is now the No. 5 with a performance of 380 Pflop/s. The European High-Performance Computing Joint Undertaking (EuroHPC JU) is pooling European resources to develop top-of-the-range Exascale supercomputers for processing big data. One of the pan-European pre-Exascale supercomputers, LUMI, is located in CSC’s data center in Kajaani, Finland.

Rank Site System Cores Rmax (TFlop/s) Rpeak (TFlop/s) Power (kW)
1 DOE/SC/Oak Ridge National Laboratory
United States		 Frontier - HPE Cray EX235a, AMD Optimized 3rd Generation EPYC 64C 2GHz, AMD Instinct MI250X, Slingshot-11
HPE		 8,699,904 1,194.00 1,679.82 22,703
2 DOE/SC/Argonne National Laboratory
United States		 Aurora - HPE Cray EX - Intel Exascale Compute Blade, Xeon CPU Max 9470 52C 2.4GHz, Intel Data Center GPU Max, Slingshot-11
Intel		 4,742,808 585.34 1,059.33 24,687
3 Microsoft Azure
United States		 Eagle - Microsoft NDv5, Xeon Platinum 8480C 48C 2GHz, NVIDIA H100, NVIDIA Infiniband NDR
Microsoft		 1,123,200 561.20 846.84  
4 RIKEN Center for Computational Science
Japan		 Supercomputer Fugaku - Supercomputer Fugaku, A64FX 48C 2.2GHz, Tofu interconnect D
Fujitsu		 7,630,848 442.01 537.21 29,899
5 EuroHPC/CSC
Finland		 LUMI - HPE Cray EX235a, AMD Optimized 3rd Generation EPYC 64C 2GHz, AMD Instinct MI250X, Slingshot-11
HPE		 2,752,704 379.70 531.51 7,107
6 EuroHPC/CINECA
Italy		 Leonardo - BullSequana XH2000, Xeon Platinum 8358 32C 2.6GHz, NVIDIA A100 SXM4 64 GB, Quad-rail NVIDIA HDR100 Infiniband
EVIDEN		 1,824,768 238.70 304.47 7,404
7 DOE/SC/Oak Ridge National Laboratory
United States		 Summit - IBM Power System AC922, IBM POWER9 22C 3.07GHz, NVIDIA Volta GV100, Dual-rail Mellanox EDR Infiniband
IBM		 2,414,592 148.60 200.79 10,096
8 EuroHPC/BSC
Spain		 MareNostrum 5 ACC - BullSequana XH3000, Xeon Platinum 8460Y+ 40C 2.3GHz, NVIDIA H100 64GB, Infiniband NDR200
EVIDEN		 680,960 138.20 265.57 2,560
9 NVIDIA Corporation
United States		 Eos NVIDIA DGX SuperPOD - NVIDIA DGX H100, Xeon Platinum 8480C 56C 3.8GHz, NVIDIA H100, Infiniband NDR400
Nvidia		 485,888 121.40 188.65  
10 DOE/NNSA/LLNL
United States		 Sierra - IBM Power System AC922, IBM POWER9 22C 3.1GHz, NVIDIA Volta GV100, Dual-rail Mellanox EDR Infiniband
IBM / NVIDIA / Mellanox		 1,572,480 94.64 125.71 7,438

  • The No. 6 system Leonardo is installed at a different EuroHPC site in CINECA, Italy. It is an Atos BullSequana XH2000 system with Xeon Platinum 8358 32C 2.6GHz as main processors, NVIDIA A100 SXM4 40 GB as accelerators, and Quad-rail NVIDIA HDR100 Infiniband as interconnect. It achieved a Linpack performance of 238.7 Pflop/s.

  • Summit, an IBM-built system at the Oak Ridge National Laboratory (ORNL) in Tennessee, USA, is now listed at the No. 7 spot worldwide with a performance of 148.8 Pflop/s on the HPL benchmark, which is used to rank the TOP500 list. Summit has 4,356 nodes, each one housing two POWER9 CPUs with 22 cores each and six NVIDIA Tesla V100 GPUs each with 80 streaming multiprocessors (SM). The nodes are linked together with a Mellanox dual-rail EDR InfiniBand network.

  • The MareNostrum 5 ACC system is new at No. 8 and installed at the EuroHPC/Barcelona Supercomputing Center in Spain. This BullSequana XH3000 system uses Xeon Platinum 8460Y processors with NVIDIA H100 and Infiniband NDR200. It achieved 183.2 Pflop/s HPL performance.

  • The new Eos system listed at No. 9 is a NVIDIA DGX SuperPOD based system at NVIDIA, USA. It is based on the NVIDIA DGX H100 with Xeon Platinum 8480C processors,N VIDIA H100 accelerators, and Infiniband NDR400 and it achieves 121.4 Pflop/s.

  • Sierra, a system at the Lawrence Livermore National Laboratory, CA, USA is at No. 10. Its architecture is very similar to the #7 system’s Summit. It is built with 4,320 nodes with two POWER9 CPUs and four NVIDIA Tesla V100 GPUs. Sierra achieved 94.6 Pflop/s.

Highlights from the List

  • A total of 186 systems on the list are using accelerator/co-processor technology, up from 184 six months ago. 78 of these use NVIDIA Ampere chips, 65 use NVIDIA Volta, and 11 systems with 17.

        Count System Share (%) Rmax (TFlops) Rpeak (TFlops) Cores
    1 NVIDIA Tesla V100 50 10 193,697 366,711 3,807,768
    2 NVIDIA A100 27 5.4 329,415 494,544 3,340,224
    3 NVIDIA A100 SXM4 40 GB 17 3.4 183,039 256,548 1,863,032
    4 NVIDIA Tesla A100 80G 11 2.2 125,864 167,923 1,088,032
    5 NVIDIA A100 SXM4 80 GB 11 2.2 64,250 75,189 582,720
    6 AMD Instinct MI250X 10 2 1,741,739 2,445,340 12,660,800
    7 NVIDIA Tesla V100 SXM2 10 2 88,521 177,436 1,997,584
    8 NVIDIA Tesla A100 40G 9 1.8 60,912 98,037 642,884
    9 NVIDIA Tesla P100 5 1 42,903 60,653 839,200
    10 NVIDIA H100 5 1 590,169 891,871 1,194,024
    11 NVIDIA Volta GV100 4 0.8 269,439 362,565 4,408,096
    12 NVIDIA H100 SXM5 80 GB 4 0.8 176,375 290,081 743,888
    13 Intel Data Center GPU Max 1550 2 0.4 23,335 59,843 199,872
    14 Intel Data Center GPU Max 2 0.4 602,531 1,087,283 4,892,568
    15 NVIDIA A100 SXM4 64 GB 2 0.4 242,205 309,448 1,849,824
    16 NVIDIA Tesla K40 2 0.4 7,154 12,264 145,600
    17 NVIDIA Tesla P100 NVLink 1 0.2 8,125 12,127 135,828
    18 NVIDIA H100 64GB 1 0.2 138,200 265,574 680,960
    19 Preferred Networks MN-Core 1 0.2 2,180 3,348 1,664
    20 Nvidia Volta V100 1 0.2 21,640 29,354 347,776
    21 NVIDIA HGX A100 80GB 500W 1 0.2 2,146 3,508 21,168
    22 NVIDIA Tesla K80 1 0.2 2,592 3,799 66,000
    23 Intel Xeon Phi 31S1P 1 0.2 2,071 3,075 174,720
    24 Deep Computing Processor 1 0.2 4,325 6,134 163,840
    25 Intel Xeon Phi 5110P 1 0.2 2,539 3,388 194,616
    26 NVIDIA Tesla K20x 1 0.2 3,188 4,605 72,000
    27 Matrix-2000 1 0.2 61,445 100,679 4,981,760
    28 AMD Instinct MI210 64 GB 1 0.2 9,087 19,344 96,768
    29 NVIDIA 2050 1 0.2 2,566 4,701 186,368
    30 NVIDIA Tesla K40m 1 0.2 2,478 4,947 64,384

     

     

  • Intel continues to provide the processors for the largest share (67.60 percent) of TOP500 systems, down from 72.00 % six months ago. 140 (28.00 %) of the systems in the current list used AMD processors, up from 24.20 % six months ago.

  • The entry level to the list moved up to the 2.02 Pflop/s mark on the Linpack benchmark.

  • The last system on the newest list was listed at position 455 in the previous TOP500.

  • Total combined performance of all 500 exceeded the Exaflop barrier with now 7.03 exaflop/s (Eflop/s) up from 5.24 exaflop/s (Eflop/s) 6 months ago.

  • The entry point for the TOP100 increased to 7.89 Pflop/s.

  • The average concurrency level in the TOP500 is 212,627 cores per system up from 190,919 six months ago.

General Trends

Installations by countries/regions:

    Count System Share (%) Rmax (TFlops) Rpeak (TFlops) Cores
1 United States 161 32.2 3,725,851 5,711,149 40,077,020
2 China 104 20.8 407,240 770,545 22,605,992
3 Germany 36 7.2 256,266 391,133 4,292,044
4 Japan 32 6.4 669,834 866,458 11,374,380
5 France 23 4.6 173,226 249,513 3,967,832
6 United Kingdom 15 3 81,707 142,086 2,048,112
7 Italy 12 2.4 351,758 463,800 4,427,376
8 South Korea 12 2.4 151,323 207,845 2,021,812
9 Netherlands 10 2 86,851 150,917 921,120
10 Canada 10 2 41,208 71,912 845,984
11 Brazil 9 1.8 55,709 106,041 868,096
12 Saudi Arabia 7 1.4 90,911 138,590 2,676,084
13 Russia 7 1.4 73,715 101,737 741,328
14 Australia 6 1.2 44,908 63,185 565,824
15 Sweden 6 1.2 47,601 68,808 408,256
16 Norway 5 1 15,955 24,964 425,920
17 Taiwan 5 1 31,595 43,027 559,728
18 Poland 4 0.8 13,813 20,495 224,064
19 Ireland 4 0.8 11,394 22,450 290,560
20 India 4 0.8 19,453 25,252 325,832
21 Singapore 3 0.6 9,038 15,786 234,112
22 Switzerland 3 0.6 26,667 36,407 584,992
23 Spain 3 0.6 184,773 322,242 1,559,936
24 Finland 3 0.6 391,388 546,193 3,116,992
25 Czechia 2 0.4 9,589 12,914 163,584
26 Austria 2 0.4 5,038 6,809 133,152
27 Slovenia 2 0.4 6,918 10,047 156,480
28 Bulgaria 2 0.4 7,045 9,154 164,224
29 Luxembourg 2 0.4 12,807 18,291 172,544
30 Morocco 1 0.2 3,158 5,015 71,232
31 Argentina 1 0.2 3,876 5,991 43,008
32 United Arab Emirates 1 0.2 7,257 9,492 107,568
33 Belgium 1 0.2 2,717 3,094 23,200
34 Hungary 1 0.2 3,105 4,508 27,776

HPC manufacturer:

    Count System Share (%) Rmax (TFlops) Rpeak (TFlops) Cores
1 Lenovo 169 33.8 607,475 1,102,594 15,455,216
2 HPE 103 20.6 2,450,983 3,433,844 30,023,316
3 EVIDEN 48 9.6 687,601 1,044,501 9,740,576
4 Inspur 34 6.8 77,234 192,784 1,661,600
5 DELL EMC 32 6.4 184,767 314,595 3,588,956
6 Nvidia 18 3.6 334,354 464,638 2,368,288
7 NEC 12 2.4 70,990 99,067 725,072
8 Fujitsu 12 2.4 566,894 717,062 9,803,088
9 Sugon 9 1.8 21,988 43,119 637,056
10 MEGWARE 7 1.4 22,774 32,680 372,840
11 Penguin Computing, Inc. 6 1.2 19,409 29,686 569,368
12 IBM 6 1.2 201,915 273,679 3,292,832
13 Microsoft Azure 6 1.2 150,370 224,747 1,328,640
14 NUDT 3 0.6 66,082 108,454 5,342,848
15 ACTION 3 0.6 7,994 63,814 178,368
16 Huawei Technologies Co., Ltd. 2 0.4 5,872 9,390 101,184
17 Intel 2 0.4 590,953 1,069,119 4,870,328
18 Liqid 2 0.4 5,393 7,518 102,144
19 IBM / NVIDIA / Mellanox 2 0.4 112,840 148,759 1,860,768
20 Quanta Computer / Taiwan Fixed Network / ASUS Cloud 2 0.4 11,298 19,563 220,752
21 YANDEX, NVIDIA 2 0.4 37,550 50,051 328,352
22 Cray Inc./Hitachi 2 0.4 11,461 18,250 271,584
23 NRCPC 1 0.2 93,015 125,436 10,649,600
24 NEC / DELL 1 0.2 6,708 7,627 125,440
25 ASUSTeK / ASUS Cloud / Taiwan Web Service Corporation 1 0.2 3,533 4,000 62,496
26 Preferred Networks 1 0.2 2,180 3,348 1,664
27 Supermicro 1 0.2 3,700 6,024 85,568
28 Amazon Web Services 1 0.2 9,950 15,107 172,692
29 MEGWARE / Supermicro 1 0.2 9,087 19,344 96,768
30 Nebius AI 1 0.2 46,540 86,792 218,880
31 Fujitsu / Lenovo 1 0.2 9,264 15,142 204,032
32 Netweb Technologies 1 0.2 8,500 13,170 81,344
33 NVIDIA, Inspur 1 0.2 12,810 20,029 130,944
34 ClusterVision / Hammer 1 0.2 2,969 4,335 64,512
35 Lenovo/IBM 1 0.2 2,814 3,578 86,016
36 Atipa Technology 1 0.2 2,539 3,388 194,616
37 Self-made 1 0.2 3,307 4,897 60,512
38 Format sp. z o.o. 1 0.2 5,051 7,709 47,616
39 Microsoft 1 0.2 561,200 846,835 1,123,200

Interconnect Technologies:

    Count System Share (%) Rmax (TFlops) Rpeak (TFlops) Cores
1 Infiniband 219 43.8 2,907,744 4,352,058 33,362,748
2 Gigabit Ethernet 209 41.8 3,092,269 4,917,199 38,035,068
3 Omnipath 33 6.6 166,863 248,630 3,748,420
4 Custom Interconnect 30 6 331,560 496,015 21,875,448
5 Proprietary Network 9 1.8 533,405 645,722 9,291,776

Processor Technologies:

    Count System Share (%) Rmax (TFlops) Rpeak (TFlops) Cores
1 Intel Cascade lake 137 27.4 477,852 932,186 11,253,276
2 Intel Skylake 95 19 344,621 655,060 7,839,400
3 AMD Zen-2 (Rome) 69 13.8 719,594 1,006,710 11,863,896
4 AMD Zen-3 (Milan) 66 13.2 2,090,713 2,957,925 20,296,256
5 Intel Ice Lake 36 7.2 405,639 591,413 4,899,532
6 Intel Sapphire Rapids 25 5 1,663,979 2,752,425 10,358,728
7 Intel Broadwell 18 3.6 69,439 86,885 2,048,008
8 Intel Haswell 11 2.2 55,575 75,864 1,653,660
9 Fujitsu ARM 8 1.6 530,839 641,021 9,105,408
10 Intel IvyBridge 7 1.4 82,953 132,610 5,841,172
11 Intel Xeon Phi 7 1.4 80,727 153,898 3,502,580
12 Power 7 1.4 311,567 417,833 5,081,600
13 NEC Vector Engine 5 1 47,143 60,692 208,064
14 AMD Zen-4 (Genoa) 5 1 48,755 55,442 1,167,456
15 ShenWei 1 0.2 93,015 125,436 10,649,600
16 Intel Nehalem 1 0.2 2,566 4,701 186,368
17 X86_64 1 0.2 4,325 6,134 163,840

Green500

HPCG Results

HPL-MxP Results

On the HPL-MxP (formally HPL-AI) benchmark, which measures performance for mixed-precision calculation, Frontier already demonstrated 9.95 Exaflops!

The HPL-MxP benchmark seeks to highlight the use of mixed precision computations. Traditional HPC uses 64-bit floating point computations. Today we see hardware with various levels of floating point precisions, 32-bit, 16-bit, and even 8-bit. The HPL-MxP benchmark demonstrates that by using mixed precision during the computation much higher performance is possible (see the Top 5 from the HPL-MxP benchmark), and using mathematical techniques, the same accuracy can be computed with the mixed precision technique when compared with straight 64-bit precision.

Rank (HPL-MxP) Site Computer Cores HPL-MxP (Eflop/s) TOP500 Rank HPL Rmax (Eflop/s) Speedup of HPL-MxP over HPL

1

DOE/SC/ORNL, USA

Frontier, HPE Cray EX235a

8,699,904

9.950

1

1.1940

8.3

2

EuroHPC/CSC, Finland

LUMI, HPE Cray EX235a

2,752,704

2.350

5

0.3797

6.18

3

RIKEN, Japan

Fugaku, Fujitsu A64FX

7,630,848

2.000

4

0.4420

4.5

4

EuroHPC/CINECA, Italy

Leonardo, Bull Sequana XH2000

1,824,768

1.842

6

0.2387

7.7

5

DOE/SC/ORNL, USA

Summit, IBM AC922 POWER9

2,414,592

1.411

7

0.1486

9.5

About the TOP500 List

The first version of what became today’s TOP500 list started as an exercise for a small conference in Germany in June 1993. Out of curiosity, the authors decided to revisit the list in November 1993 to see how things had changed. About that time they realized they might be onto something and decided to continue compiling the list, which is now a much-anticipated, much-watched and much-debated twice-yearly event.

 

qEnBWgjA4ALkhMhUaT9AAB.jpg

 

 

https://www.top500.org/lists/top500/2023/11/highs/

37
0
댓글
0
덕질 비회원은 댓글을 볼 수 없습니다. 로그인을 해주세요.
Deokjil non-members cannot see comments. Please log in.
1 2 3 4 5 6 7 8 9 10
카테고리 즐겨찾기 알림