gpu benchmark 2020







K80V100 SXM2V100 PCIEV100S PCIEP100P40P4Titan VTitan XP2080TiTitan RTX1080TiRTX 5000RTX 6000RTX 8000T4
ArchitectureKepler 2.0Volta GV100Volta GV100Volta GV100Pascal GP100PascalPascalVolta GV100Pascal GP102Turing TU102TU102PascalTuring TU104TU102TU102
CUDA Cores2496 x251205120512035843840512038404352460835843,0724,6084,6082,560
Tensor Cores0640640N/AN/AN/A640N/A576576N/A384576576320
NVIDIA RT Cores72487272
雙精度64效能(TFLOPS):1.877.578.24.77.50.38
單精度32效能(TFLOPS)5.6151416.49.3125.51512.114.216.3111.411.216.316.38.1
半精度16效能(TFLOPS)32.818.70.1932.6265
INT8(TFLOPS)4722130
INT4(TFLOPS)
GPU Memory(GB)12 GB x2 GDDR516/32 HBM216/32 HBM216/32 HBM212/16248121211241116244816
Memory Bandwidth(GB/sec)900 GB/sec900 GB/sec1134 GB/s549/732347192652.8547.7616672484320+
InterconnectBandwidth GB/secNVLINK 300PCIE 32PCIE 32
MaxPower(WATTs)300W250W250W250W250W75W250W250W260W280W250W265 W295 W70W




https://docs.google.com/spreadsheets/d/e/2PACX-1vQh1HDjcn1UBAS8IxtvYG6ZrToW7cenOz8gtni1veLmsyKv80w_KqUvNgEygWQn4LCoBSyOZKMm0WXV/pubhtml

gpu benchmark 2020

我們從事基礎人工智能技術的開發工作。所有技術,從語音識別到計算機視覺、自然語言處理、數據倉庫、用戶理解,我們用人工智能技術支持很多國際業務並孵化新業務方向。

鴻鵠國際股份有限公司

業務窗口   蔡先生
行動電話:0910-218-322      Line ID: tsaicm1011  wechat微信ID: tsaicm                                                 
公司電話:02-2929-9388 #10
公司傳真:02-2929-7579
Email: sales1@honghutech.com


Comments