Aerospace and Defense Technology Product Guides, Tech Briefs, and Technical Content
Issue link: https://read.uberflip.com/i/1273994
Mercury Systems, Inc. www.mrcy.com Mercury Systems • 50 Minuteman Road • Andover, MA 01810 • (978) 256-1300 Mercury Systems International • Regus Center, 26 Avenue Jean Kuntzmann • Montbonnot • 38330 France • +33 608 419949 INNOVATION THAT MATTERS ® Mercury Systems and Innovation That Matters are registered trademarks of Mercury Systems, Inc. Other product and company names mentioned may be trademarks and/or registered trademarks of their respective holders. Mercury Systems, Inc. believes this information is accurate as of its publication date and is not responsible for any inadvertent errors. The information contained herein is subject to change without notice. Copyright © 2019 Mercury Systems, Inc. 6576.02E-0919-DS-ServerNVIDIAGPUoptions NVIDIA GPU Options and Specifications Table 1: NVIDIA Ampere microarchitecture GPUs Release date: June 2020 Table 2: NVIDIA Turing microarchitecture GPU Release date: September 2018 Table 3: NVIDIA Volta microarchitecture GPUs Release date: December 2017 Table 4: NVIDIA Pascal microarchitecture GPUs Release date: April 2016 Quadro RTX8000 Quadro RTX6000 NVIDIA T4 CUDA Cores 4608 4608 2560 Tensor Cores 576 576 320 RT Cores 72 72 40 Memory 48 GB GDDR5 ECC 24 GB GDDR5 ECC 16 GB GDDR5 ECC Bus Width 384 bit 384 bit 256 bit Half-Precision Performance (FP16) 32.6 TFLOPS 32.6 TFLOPS 65 TFLOPS Single-Precision Performance (FP32) 16.3 TFLOPS 16.3 TFLOPS 8.1 TFLOPS Double-Precision Performance (FP64) 0.5 TFLOPS 0.5 TFLOPS 0.3 TFLOPS INT8 Precision 206.1 TOPS 206.1 TOPS 130 TOPS Cooling Active Active Passive Max Power Consumption 295W 295W 70W Form Factor 4.4" x 10.5" dual width 4.4" x 10.5" dual width 2.7" x 6.7" single width Advanced Use Case ray tracing, deep learning, visual computing ray tracing, deep learning, visual computing mainstream servers, inference acceleration NVIDIA A100 CUDA Cores 6914 Tensor Cores 432 RT Cores 0 Memory 40 GB HBM2 Bus Width 4096 bit Half-Precision Performance (FP16) 624 TFLOPS Single-Precision Performance (FP32) 312 TFLOPS Double-Precision Performance (FP64) 19.5 TFLOPS INT8 Precision 1248 TOPS Cooling Passive Max Power Consumption 250W Form Factor 4.4" x 10.5" dual width Advanced Use Case universal data center edge computing NVIDIA P100 NVIDIA P40 CUDA Cores 3584 3840 Tensor Cores 0 0 RT Cores 0 0 Memory 16 GB HBM2 ECC 24 GB HBM2 ECC Bus Width 4906 bit 384 bit Half-Precision Performance (FP16) 18.7 TFLOPS 0.2 TFLOPS Single-Precision Performance (FP32) 9.3 TFLOPS 12 TFLOPS Double-Precision Performance (FP64) 4.7 TFLOPS 0.4 TFLOPS Cooling Passive Passive Max Power Consumption 250W 250W Form Factor 4.4" x 10.5" dual width 4.4" x 10.5" dual width Advanced Use Case universal data center inference-throughput servers NVIDIA V100 Quadro GV100 CUDA Cores 5120 5120 Tensor Cores 640 640 RT Cores 0 0 Memory 32 GB HBM2 ECC 32 GB HBM2 ECC Bus Width 4096 bit 4096 bit Half-Precision Performance (FP16) 112 TFLOPS 29.6 TFLOPS Single-Precision Performance (FP32) 14 TFLOPS 14.8 TFLOPS Double-Precision Performance (FP64) 7 TFLOPS 7.4 TFLOPS Cooling Passive Active Max Power Consumption 250W 250W Form Factor 4.4" x 10.5" dual width 4.4" x 10.5" dual width Advanced Use Case universal data center AI design and visualization on workstations