NVIDIa Tesla P100 GPU aCCeleRaTOR

1 Tesla P100 PCle | DaTa sheeT | OCT16 World s most advanced data center aCCeleRaTOR for PCIe-based serversHPC data centers need to support the ever-growing demands of scientists and researchers while staying within a tight budget. The old approach of deploying lots of commodity compute nodes requires huge interconnect overhead that substantially increases costs without proportionally increasing Tesla P100 GPU accelerators are the most advanced ever built, powered by the breakthrough NVIDIa Pascal architecture and designed to boost throughput and save money for HPC and hyperscale data centers. The newest addition to this family, Tesla P100 for PCIe enables a single node to replace half a rack of commodity CPU nodes by delivering lightning-fast performance in a broad range of HPC LEAP IN PERFORMANCENVIDIA Tesla P100 for PCIe PerformanceAppl cat on Speed-upNAMDVASPMIL AMBERHOOMD-Blue affe/AlexNet2X K802X P100 (P Ie)4X P100 (P Ie)

2 30 X25 X10 X5 X0 X20 X15 XDual PU server, Intel E5-2698 v3 2 3 Hz, 256 B System Memory, Pre-Product on Tesla P100 SPECIFICATIONSGPU ArchitectureNVIDIA PascalNVIDIA CUDA Cores3584 Double-Precision TeraFLOPSS ingle-Precision TeraFLOPSHalf-Precision TeraFLOPS GPU Memory16GB CoWoS HBM2 at 732 GB/s or 12GB CoWoS HBM2 at 549 GB/sSystem InterfacePCIe Gen3 Max Power Consumption250 WECCYe sThermal SolutionPassiveForm FactorPCIe Full Height/LengthCompute APIsCUDA, DirectCompute, OpenCL , OpenACCTeraFLOPS measurements with NVIDIa GPU Boost technologyNVIDIa Tesla P100 GPU aCCeleRaTOR 2016 NVIDIa Corporation. All rights reserved. NVIDIa , the NVIDIa logo, Tesla , NVIDIa GPU Boost, CUDA, and NVIDIa Pascal are trademarks and/or registered trademarks of NVIDIa Corporation in the and other countries.

3 OpenCL is a trademark of Apple Inc. used under license to the Khronos Group Inc. All other trademarks and copyrights are the property of their respective owners. OCT16To learn more about the Tesla P100 for PCIe visit GIaNT leaP IN PeRFORMaNCeTesla P100 for PCIe is reimagined from silicon to software, crafted with innovation at every level. Each groundbreaking technology delivers a dramatic jump in performance to substantially boost the data center throughput. PASCAL ARCHITECTUREMore than TeraFLOPS of FP16, TeraFLOPS of double-precision, and TeraFLOPS of single-precision performance powers new possibilities in deep learning and HPC HBM2 Compute and data are integrated on the same package using Chip-on-Wafer-on-Substrate with HBM2 technology for 3X memory performance over the previous-generation architecture.

4 B -d rect onal BW ( B/Sec)8006004002000K40P1003X memory boostM40K40M40P100 (FP32)P100 (FP16)2520151050 Teraflops (FP32/FP16)Exponential HPC and hyperscale performanceUnified MemoryCPUGPUPAGE MIGRATION ENGINES impler programming and computing performance tuning means that applications can now scale beyond the GPU s physical memory size to virtually limitless levels. Addressable Memory ( B)10,0001,000100100K40M40P100 Virtually limitless memory scalability