Diary/2020-4-11

RTX2080Tiの評価

e-treesのGPUサーバのGPUをGTX1080からRTX2080Tiにアップデート．
折角なのでGTX1080とRTX2080Tiの性能を https://github.com/ekondis/mixbench で比較してみた．
結果は，

RTX2080Tiはだいぶ速そうだぞ，ってことで仕事させないと．

それぞれのスペックは，こんな感じ．

------------------------ Device specifications ------------------------
Device:              GeForce GTX 1080
CUDA driver version: 10.20
GPU clock rate:      1911 MHz
Memory clock rate:   2752 MHz
Memory bus width:    256 bits
WarpSize:            32
L2 cache size:       2048 KB
Total global mem:    8117 MB
ECC enabled:         No
Compute Capability:  6.1
Total SPs:           2560 (20 MPs x 128 SPs/MP)
Compute throughput:  9784.32 GFlops (theoretical single precision FMAs)
Memory bandwidth:    352.32 GB/sec
-----------------------------------------------------------------------
Total GPU memory 8511881216, free 8374583296
Buffer size:          64MB
Trade-off type:       compute with global memory (block strided)

------------------------ Device specifications ------------------------
Device:              GeForce RTX 2080 Ti
CUDA driver version: 10.20
GPU clock rate:      1650 MHz
Memory clock rate:   3500 MHz
Memory bus width:    352 bits
WarpSize:            32
L2 cache size:       5632 KB
Total global mem:    11019 MB
ECC enabled:         No
Compute Capability:  7.5
Total SPs:           4352 (68 MPs x 64 SPs/MP)
Compute throughput:  14361.60 GFlops (theoretical single precision FMAs)
Memory bandwidth:    616.00 GB/sec
-----------------------------------------------------------------------
Total GPU memory 11554717696, free 11379408896
Buffer size:          64MB
Trade-off type:       compute with global memory (block strided)

生データとプロットスクリプト
gpu-comparison.dat(137)
gpu-comparison.plt(128)