集群中各服务器的配置与性能
Last modified: December 10, 2024
MATLAB Benchmark
MATLAB benchmark times 6 different MATLAB tasks and compares the execution speed. We take the mean time of 50 repeats of the test.
Version: MATLAB/R2023b
Args: matlab -nodisplay -nodesktop
Code: bench(1);mean(bench(50),1)
Test description:
| #Test | Name | Full Name | Description |
|---|---|---|---|
| Test 1 | LU | LAPACK. | Floating point, regular memory access. |
| Test 2 | FFT | Fast Fourier Transform. | Floating point, irregular memory access. |
| Test 3 | ODE | Ordinary diff. eqn. | Data structures and functions. |
| Test 4 | Sparse | Solve sparse system. | Sparse linear algebra. |
| Test 5 | 2-D | 2-D Lissajous plot. | Animating line plot. |
| Test 6 | 3-D | 3-D SURF(PEAKS)and HGTransform. | 3-D surface animation. |
Time cost (shorter is better):
| Machine | Test 1 | Test 2 | Test 3 | Test 4 | Test 5 | Test 6 | Comment |
|---|---|---|---|---|---|---|---|
loginNode | 0.3732 | 0.3605 | 0.2424 | 0.7200 | 0.3520 | 0.2220 | |
bigMem0 | 0.2605 | 0.1994 | 0.1661 | 0.5254 | 0.2846 | 0.2059 | |
bigMem1 | 0.2681 | 0.2196 | 0.1649 | 0.5679 | 0.2761 | 0.1972 | |
bigMem2 | 0.8643 | 0.2315 | 0.3027 | 3.3614 | 0.3731 | 0.2523 | |
bigMem3 | 0.2042 | 0.1165 | 0.1392 | 4.8798 | 0.2142 | 0.1321 |
CPU / Memory
评测程序 Source: PassMark Performance Testing Linux
版本: 10.2.1003
参数: (-i 3) Test Iterations: 3, (-d 2) Test Duration: Medium.
测试结果仅供参考, 有一定的波动. (?) 为未经有效确认的结果.
| CPU Info | loginNode | bigMem0 | bigMem1 | bigMem2 | bigMem3 |
|---|---|---|---|---|---|
| CPU Brand | Intel Xeon CPU E5-2670 v3 @ 2.30GHz | Intel Xeon Gold 6226R CPU @ 2.90GHz | Intel Xeon Gold 6226R CPU @ 2.90GHz | Hygon C86 7285 32-core Processor | AMD EPYC 9754 128-Core Processor |
| Num of CPUs on Board | 2 | 2 | 2 | 2 | 2 |
| Total Threads | 48 | 64 | 64 | 128 | 512 |
| Base Clock Speed (GHz) | 2.30 | 2.90 | 2.90 | 2.00 | 2.25 |
| Boost Clock Speed (GHz) | 3.10 | 3.90 | 3.90 | 2.50(?) | 3.10 |
| CPU Cache (MiB/CPU) | 30 | 22 | 22 | 64 | 256 |
| Lithography (Nanometer) | 22 | 14 | 14 | 14(?) | 5 |
| CPU Speed | loginNode | bigMem0 | bigMem1 | bigMem2 | bigMem3 |
| Integer Math (Million Operations/s) | 113,171 | 206,926 | 210,833 | 326,869 | 2,054,471 |
| Floating Point Math (Million Operations/s) | 56,058 | 113,990 | 114,338 | 146,645 | 1,149,133 |
| Prime Numbers (Million Primes/s) | 168 | 224 | 223 | 271 | 1,728 |
| Sorting (Thousand Strings/s) | 73,125 | 104,601 | 116,809 | 164,992 | 783,396 |
| Encryption (MB/s) | 10,892 | 25,528 | 26,793 | 100,666 | 492,442 |
| Compression (MB/s) | 484 | 804 | 840 | 1,358 | 6,933 |
| CPU Single Threaded (Million Operations/s) | 1,845 | 2,334 | 2,312 | 1,479 | 2,444 |
| Physics (Frames/s) | 2,281 | 4,047 | 4,205 | 5,643 | 22,446 |
| Extended Instructions (SSE) (Million Matrices/s) | 26,481 | 46,341 | 48,944 | 45,151 | 428,652 |
| CPU Final Mark | 28,093 | 47,282 | 48,524 | 52,494 | 138,716 |
| Memory Info | loginNode | bigMem0 | bigMem1 | bigMem2 | bigMem3 |
| Total Available RAM (GiB) | 983.1 | 1,006.6 | 1,006.6 | 503.8 | 1,511.5 |
| Memory Frequency (MHz) | 2,133 | 2,933 | 2,933 | 2,666 | 4,800 |
| Memory Speed | loginNode | bigMem0 | bigMem1 | bigMem2 | bigMem3 |
| Memory Latency (Nanoseconds) | 51 | 54 | 57 | 61 | 70 |
| Memory Read Cached (MB/s) | 22,296 | 27,648 | 26,946 | 18,461 | 23,561 |
| Memory Read Uncached (MB/s) | 9,206 | 8,923 | 7,479 | 12,863 | 23,334 |
| Memory Write (MB/s) | 8,766 | 8,109 | 7,177 | 7,647 | 23,305 |
| Memory Threaded (MB/s) | 107,696 | 194,455 | 186,943 | 226,623 | 726,276 |
| Database Operations (Thousand Operations/s) | 14,320 | 20,466 | 21,388 | 15,415 | 29,555 |
| Memory Final Mark | 2,524 | 2,562 | 2,334 | 2,217 | 2,876 |
评测程序: tinymembench v0.4
| tinymembench-v0.4 (result in MB/s) | lnnode | bm0 | bm1 | bm2 | bm3 |
|---|---|---|---|---|---|
| C copy backwards | 6126 | 6508.9 | 5813.7 | 6879.5 | 18378.2 |
| C copy backwards (32 byte blocks) | 6120.3 | 6494.9 | 5780.3 | 6891.5 | 16917.2 |
| C copy backwards (64 byte blocks) | 6124.1 | 6505.7 | 5772.2 | 6890.6 | 16935.1 |
| C copy | 5994.4 | 6671.2 | 5938.8 | 6888.2 | 18383 |
| C copy prefetched (32 bytes step) | 5829.5 | 4753.6 | 4232.2 | 7126.9 | 16105.5 |
| C copy prefetched (64 bytes step) | 5817.1 | 4784.1 | 4252.4 | 7096.3 | 16373.7 |
| C 2-pass copy | 5185.8 | 5646.7 | 5101.5 | 5827.4 | 13985 |
| C 2-pass copy prefetched (32 bytes step) | 5415.8 | 3387.9 | 3090.4 | 6242.2 | 10197 |
| C 2-pass copy prefetched (64 bytes step) | 5413.5 | 3409 | 3106.7 | 6322.3 | 10401.7 |
| C fill | 11580 | 13888 | 12338.1 | 8745.6 | 39944.5 |
| C fill (shuffle within 16 byte blocks) | 11583.2 | 13938.1 | 12391.8 | 8736.9 | 39729.9 |
| C fill (shuffle within 32 byte blocks) | 11585.6 | 13949.6 | 12454.8 | 8740.6 | 39756 |
| C fill (shuffle within 64 byte blocks) | 11575.1 | 13939.8 | 12433.3 | 8751.9 | 35720.2 |
| standard memcpy | 12389.5 | 5708.8 | 5589.3 | 11175.3 | 18968.3 |
| standard memset | 11536.2 | 8114.5 | 8134.6 | 10633.1 | 29393.4 |
| MOVSB copy | 5215.3 | 5519.7 | 5330.9 | 7462.8 | 26351.4 |
| MOVSD copy | 5221.1 | 5510.3 | 5303.5 | 7461.4 | 26375.4 |
| SSE2 copy | 6004 | 6793 | 6107.3 | 7767.1 | 22674.3 |
| SSE2 nontemporal copy | 11791.9 | 5174.2 | 5025 | 12468 | 24251.6 |
| SSE2 copy prefetched (32 bytes step) | 5843 | 5250.8 | 4671.7 | 7804.7 | 22168.2 |
| SSE2 copy prefetched (64 bytes step) | 5839.4 | 5357.1 | 4787.2 | 7787.7 | 22943.9 |
| SSE2 nontemporal copy prefetched (32 bytes step) | 11431.3 | 3295.4 | 3184.5 | 13364.1 | 25204.1 |
| SSE2 nontemporal copy prefetched (64 bytes step) | 11528.3 | 3455.2 | 3347.3 | 13285.6 | 25874.4 |
| SSE2 2-pass copy | 5443.6 | 6119.2 | 5460.1 | 6518.2 | 23069.5 |
| SSE2 2-pass copy prefetched (32 bytes step) | 5270.8 | 3962.9 | 3597.6 | 7044.2 | 15304.2 |
| SSE2 2-pass copy prefetched (64 bytes step) | 5254.1 | 4062.2 | 3734.7 | 7080.1 | 15882.8 |
| SSE2 2-pass nontemporal copy | 3749.2 | 2587.7 | 2592.5 | 4019.2 | 3313.2 |
| SSE2 fill | 11614.5 | 13636.6 | 12031.1 | 10619.9 | 41563.1 |
| SSE2 nontemporal fill | 17945.7 | 7462.3 | 7294.9 | 32769.2 | 28722.6 |
Reference:
Hygon IPO Specs pp.124-126
GPU / DCU 计算加速卡
bigMem3 上无加速卡.
评测程序 Source: Mixbench
| GPU/DCU | loginNode | bigMem0 | bigMem1 | bigMem2 |
|---|---|---|---|---|
| Device | NVIDIA GeForce GTX 1080 Ti | Tesla T4 | NVIDIA A30 | HYGON Z100 |
| CUDA driver version | 11.60 | 11.60 | 11.60 | - |
| GPU clock rate (MHz) | 1721 | 1590 | 1440 | 1319 |
| Memory clock rate | 2752 MHz | 2500 MHz | 607 MHz | - MHz |
| Memory bus width | 352 bits | 256 bits | 3072 bits | - bits |
| WarpSize | 32 | 32 | 32 | 64 |
| L2 cache size | 2816 KB | 4096 KB | 24576 KB | 8192 KB |
| Total global mem | 11178 MB | 14910 MB | 24068 MB | 32752 MB |
| ECC enabled | No | Yes | Yes | Yes |
| Compute Capability | 6.1 | 7.5 | 8.0 | - |
| Total SPs | 3584 (28 MPs x 128 SPs/MP) | 2560 (40 MPs x 64 SPs/MP) | 7168 (56 MPs x 128 SPs/MP) | 3840 (60 CUs x 64 SPs/CU) |
| Compute throughput (theoretical single precision FMAs) (GFlops) | 12336.13 | 8140.80 | 20643.84 | 10129.92 |
| Memory bandwidth (GB/sec) | 484.44 | 320.06 | 933.12 | 1024.00 |
磁盘读写速度
评测程序: fio-3.7
命令 $ fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename={} --bs=4ki --iodepth=64 --size=400Mi --readwrite={method} --rwmixread={rwrate},
其中 method=randrw | rw, rwrate=0 | 75 | 100, filename 指定为对应硬盘下的位置.
| Spec | r/w | loginNode |
|---|---|---|
| 400 MiB file (in-cache) | ||
| randrw, 75% read + 25% write | read | 332 MiB/s |
| write | 111 MiB/s | |
| randrw, 100% read | read | 510 MiB/s |
| randrw, 100% write | write | 494 MiB/s |
| rw, 75% read + 25% write | read | 392 MiB/s |
| write | 130 MiB/s | |
| rw, 100% read | read | 516 MiB/s |
| rw, 100% write | write | 536 MiB/s |
| 16 GiB file (out-of-cache) | ||
| loginNode:/home, /scratch | ||
| (data disk) | ||
| randrw, 75% read + 25% write | read | 11.6 MiB/s |
| 2209 IOPS | ||
| write | 2.9 MiB/s | |
| 732 IOPS | ||
| randrw, 100% read | read | 11.1 MiB/s |
| 3060 IOPS | ||
| randrw, 100% write | write | 61.9 MiB/s |
| 15800 IOPS | ||
| loginNode:/ | ||
| (software disk) | ||
| randrw, 75% read + 25% write | read | 1.6 MiB/s |
| 413 IOPS | ||
| write | 0.5 MiB/s | |
| 132 IOPS | ||
| randrw, 100% read | read | 1.1 MiB/s |
| 274 IOPS | ||
| randrw, 100% write | write | 7.4 MiB/s |
| 1817 IOPS |