Skip to main content

Machine Performance Specs

NameGenerationCUDA CoresGPU MemoryMemory BandwidthTFLOPS1Tensor Core2vCPUsRAM
GPU+Maxwell1,6648 GB192 GB/sec2.6-830
P4000Pascal1,7928 GB243 GB/sec5.3-830
P5000Pascal2,56016 GB288 GB/sec9.0-830
P6000Pascal3,84024 GB432 GB/sec12.0-830
V100Volta5,12016 GB900 GB/sec14.0112 TFLOPS (FP16)830
V100x32Volta5,12032 GB900 GB/sec15.7125 TFLOPS (FP16)830
RTX4000Turing2,3048 GB416 GB/s7.157 TFLOPS (FP32)830
RTX5000Turing3,07216 GB448 GB/s11.289 TFLOPS (FP32)830
A4000Ampere6,14416 GB448 GB/s19.2153 TFLOPS (FP16)845
A5000Ampere8,19224 GB768 GB/s27.8222 TFLOPS (FP16)845
A6000Ampere10,75248 GB768 GB/s38.7309 TFLOPS (FP16)845
A100Ampere6,91240 GB HBM21,555 GB/s19.5156 / 312 (FP32/16) TFLOPS1290

  1. TFLOPS = Peak single precision (FP32) TeraFLOPS performance
  2. Tensor Core performance = Peak single or half precision (TF32 or TF16) Tensor Core performance