IBM NeXtScale System, Planning And Implementation Manual

Page: 113 / 266

Chapter 4. IBM NeXtScale nx360 M4 compute node
95
Table 4-22 Comparison of NVIDIA GPUs
Use Case High Performance Computing Virtual Graphics
Technical
Specifications Tesla K40 Tesla K20X Tesla K20 Tesla K10 Grid K2 Grid K1
Peak
double-precision
FP performance
(board) 1.43
Tflops
1.31
Tflops
1.17
Tflops
0.19
Tflops
N/A N/A
Peak
single-precision FP
performance
(board) 4.29
Tflops
3.95
Tflops
3.52
Tflops
4.58
Tflops
4.30 Tflops
(2 x 2.15
Tflops)
N/A
Number of physical
GPUs 1x GK110B 1x GK110 1x GK110 2x GK104 4x GK107
Number of CUDA
cores 2,880 2,688 2,496 3,072 (2 x 1,536) 768
(4 x 192)
Memory size per
board 12 GB 6 GB 5 GB 8 GB (2 x 4 GB) 16 GB
(4 x 4 GB)
Memory bandwidth
per board 288 Gb/s 250 Gb/s 208 Gb/s 320 Gb/s 116 Gb/s
Memory I/O 384-bit GDDR5 320-bit GDDR5 256-bit GDDR5 128-bit DDR3
Max Power 235W 225W 130W
Aux Power 1 x 8-pin and 1 x 6-pin 1 x 6-pin
PCI Interface Gen 3 x16 Gen 2 x16 Gen3 x16
Physical display
connectors None
Cooling Passive, Active Passive Passive, Active Passive
Memory Clock 3.0 GHz 2.6 GHz 2.5 GHz 891 MHz
NVIDIA GPU Boost Yes No
Base Core Clock 745 MHz 732MHz 706MHz 745 MHz 850 MHz
Boost Clocks 810 MHz
875 MHz Not applicable
Memory Error
Protection Yes Enabled (External & Internal) Yes (External Only) No