作者: skywalker167 時間: 2017-11-15 10:28 標題: PEZY dominates the new Green500 list
本帖最後由 skywalker167 於 2017-11-15 10:32 編輯
https://fuse.wikichip.org/news/4 ... -new-green500-list/
About a week ago we reported about PEZY reaching the number one spot on the Green500 list with their latest 2,048-core PEZY-SC2 many-core processor. Today, TOP500.org has formally announced their latest November 2017 list.
Dominating the list with four out of the top five most power-efficient systems are the newly installed systems in Japan, all based on the ZettaScaler-2.2 architecture featuring the PEZY-SC2 processors.
Green500 Rank Top500 Rank System Cores Rmax Power Efficiency
1 259 Shoubu system B 794,400 842.0 TFLOPS 50 kW 17.009 GFLOPS/W
50x (8x PEZY-SC2 + 1x Xeon D-1571)
2 307 Suiren2 762,624 788.2 TFLOPS 47 kW 16.759 GFLOPS/W
48x (8x PEZY-SC2 + 1x Xeon D-1571)
3 276 Sakura 794,400 824.7 TFLOPS 50 kW 16.657 GFLOPS/W
50x (8x PEZY-SC2 + 2x Xeon E5-2618L v3)
5 4 Gyoukou 19,860,000 19,135.8 TFLOPS 1,350 kW 14.173 GFLOPS/W
1250x (8x PEZY-SC2 + 1x Xeon D-1571)
Gyoukou, which ranks fifth on the Green500, now also ranks as the fourth most powerful supercomputer. The system also currently holds highest core-count record.
就系呢粒日本研發的眾核心處理器，Green500頭5名有4個都系用Xeon CPU+呢粒野
能源效率甚至可以比Xeon CPU+Volta GPU更加高
Green500第5名Gyoukou系Top500第4名，能源效率比Top500頭5名其他機高好多
對比返，第一名中國的太湖之光用純眾核心處理器SW26010，能源效率只有6GFLOPS/W
Gyoukou能源效率超過14GFLOPS/W，Green500第一名達到17GFLOPS/W，差不多系太湖之光的3倍
而且系已經用緊Xeon CPU
仲勁過Volta GPU....
https://www.top500.org/green500/list/2017/11/
呢度有呢粒野的更加詳細資料
https://fuse.wikichip.org/news/1 ... -a-green500-record/
作者: P14 時間: 2017-11-15 21:16
提示: 作者被禁止或刪除 內容自動屏蔽
作者: Stiggy930 時間: 2017-11-15 23:26
本帖最後由 Stiggy930 於 2017-11-15 23:35 編輯
真正嘅MIC (Many Integrated Cores)。小日本陣營仲有NEC張Vector accelerator。雖然per card 只有2.45 teraflops 。Fujitsu就諗住用ARMv8 based MIC參戰。
美帝嚟緊有Vega 20同Knight mills。美日繼軍事後係HPC再度圍堵中國。
作者: polarhei 時間: 2017-11-16 09:50
同一時間會再改良. 據知仲有其他方法未攞出黍但秘密做緊. 畢竟, 自己人唔畀足數. 加上野要留底(美國太空總署受訓時噉樣講,同時要求自己人盡快搞好個座唔好畀人知太多), 間接增強左北京.
作者: 沙盒A 時間: 2017-11-17 02:26
真正嘅MIC (Many Integrated Cores)。小日本陣營仲有NEC張Vector accelerator。雖然per card 只有2.45 tera ...
Stiggy930 發表於 2017-11-15 23:26
Knoght milling唔係摺左咩？
via HKEPC IR Pro v2.1.0 - iPhone
作者: Stiggy930 時間: 2017-11-17 09:43
Knoght milling唔係摺左咩？
via HKEPC IR Pro v2.1.0 - iPhone
沙盒A 發表於 2017-11-17 02:26
有D咁嘅樹？八月仲講緊㗎禾。
https://www.top500.org/news/inte ... hts-mill-processor/
作者: 沙盒A 時間: 2017-11-17 12:07
有D咁嘅樹？八月仲講緊㗎禾。
Stiggy930 發表於 2017-11-17 09:43
https://itpeernetwork.intel.com/ ... formance-computing/
One step we’re taking is to replace one of the future Intel® Xeon Phi™ processors (code name Knights Hill) with a new platform and new microarchitecture specifically designed for exascale.
https://www.nextplatform.com/201 ... scale-architecture/
NH: What architectural directions seem appropriate for exascale and how did the existing Knights roadmap not fit with that vision?
Davis: We can talk about the right architecture for exascale in general but of course, I can’t be specific [about the Aurora deal] and I’m not going to say “Knights had X and the new Knights has Y”. The needs of the next generation architecture for exascale are fairly well-documented. There’s power, cooling, space, interconnect—and that is a tremendous issue, creating a low latency technology that communicate at that scale. There also compute and storage. There’s a lot to address here.