SGI |
SPECompG_base2012 = 80.1 |
SGI UV 300 (Intel Xeon E7-8867 v4, 2.40 GHz) |
SPECompG_peak2012 = 84.2 |
OMP2012 license: | 14 | Test date: | May-2016 |
---|---|---|---|
Test sponsor: | SGI | Hardware Availability: | Jun-2016 |
Tested by: | SGI | Software Availability: | Apr-2016 |
Hardware | |
---|---|
CPU Name: | Intel Xeon E7-8867 v4 |
CPU Characteristics: | Intel Turbo Boost Technology up to 3.30 GHz |
CPU MHz: | 2400 |
CPU MHz Maximum: | 3300 |
FPU: | Integrated |
CPU(s) enabled: | 576 cores, 32 chips, 18 cores/chip |
CPU(s) orderable: | 4-32 chips |
Primary Cache: | 32 KB I + 32 KB D on chip per core |
Secondary Cache: | 256 KB I+D on chip per core |
L3 Cache: | 45 MB I+D on chip per chip |
Other Cache: | None |
Memory: | 8 TB (512 x 16 GB 2Rx4 PC4-2133P-R, running at 1600 MHz) |
Disk Subsystem: | 1 x 400 GB SSD (Intel SSD 3500 Series, SATA II) |
Other Hardware: | None |
Base Threads Run: | 513 |
Minimum Peak Threads: | 512 |
Maximum Peak Threads: | 576 |
Software | |
---|---|
Operating System: | SUSE Linux Enterprise Server 12 (x86_64) SP1 Kernel 3.12.57-60.35-default |
Compiler: | C/C++/Fortran: Version 16.0.1.150 of Intel Composer XE for Linux, Build 20151021 |
Auto Parallel: | No |
File System: | ext3 |
System State: | Multi-user, run level 3 |
Base Pointers: | 64-bit |
Peak Pointers: | Not Applicable |
Other Software: | SGI Accelerate 1.12 (Build 714r28.sles12sp1-1604201900), SGI Foundation Software 2.14 (Build 714r28.sles12sp1-1604201900) |
Benchmark | Base | Peak | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Threads | Seconds | Ratio | Seconds | Ratio | Seconds | Ratio | Threads | Seconds | Ratio | Seconds | Ratio | Seconds | Ratio | |
Results appear in the order in which they were run. Bold underlined text indicates a median measurement. | ||||||||||||||
350.md | 513 | 32.9 | 141 | 46.9 | 98.8 | 32.5 | 142 | 576 | 42.9 | 108 | 29.5 | 157 | 29.7 | 156 |
351.bwaves | 513 | 50.1 | 90.5 | 50.2 | 90.2 | 49.1 | 92.2 | 517 | 49.8 | 90.9 | 49.3 | 91.9 | 49.4 | 91.7 |
352.nab | 513 | 101 | 38.7 | 101 | 38.7 | 101 | 38.6 | 520 | 101 | 38.7 | 100 | 38.8 | 100 | 38.8 |
357.bt331 | 513 | 51.0 | 93.0 | 50.9 | 93.1 | 50.8 | 93.3 | 531 | 50.9 | 93.1 | 51.0 | 93.0 | 51.3 | 92.4 |
358.botsalgn | 513 | 38.4 | 113 | 38.4 | 113 | 38.5 | 113 | 576 | 35.5 | 123 | 35.5 | 123 | 35.6 | 122 |
359.botsspar | 513 | 97.9 | 53.6 | 99.2 | 52.9 | 105 | 50.2 | 567 | 104 | 50.5 | 100 | 52.3 | 102 | 51.3 |
360.ilbdc | 513 | 85.2 | 41.8 | 85.2 | 41.8 | 85.1 | 41.8 | 576 | 77.2 | 46.1 | 77.3 | 46.1 | 77.2 | 46.1 |
362.fma3d | 513 | 109 | 34.9 | 109 | 34.9 | 109 | 34.9 | 567 | 95.9 | 39.6 | 95.7 | 39.7 | 95.7 | 39.7 |
363.swim | 513 | 39.9 | 113 | 40.1 | 113 | 39.9 | 114 | 567 | 37.5 | 121 | 37.6 | 120 | 37.6 | 121 |
367.imagick | 513 | 93.7 | 75.0 | 93.9 | 74.9 | 95.8 | 73.4 | 576 | 88.8 | 79.1 | 91.0 | 77.3 | 87.5 | 80.3 |
370.mgrid331 | 513 | 50.1 | 88.2 | 50.2 | 88.1 | 50.4 | 87.8 | 512 | 48.5 | 91.2 | 47.7 | 92.6 | 49.2 | 89.8 |
371.applu331 | 513 | 91.5 | 66.2 | 91.3 | 66.4 | 91.4 | 66.3 | 531 | 91.0 | 66.6 | 90.6 | 66.9 | 90.5 | 67.0 |
372.smithwa | 513 | 22.2 | 242 | 22.6 | 237 | 22.4 | 239 | 576 | 20.5 | 261 | 20.7 | 259 | 20.5 | 262 |
376.kdtree | 513 | 48.2 | 93.3 | 48.1 | 93.6 | 48.8 | 92.2 | 576 | 47.2 | 95.4 | 43.9 | 103 | 45.1 | 99.8 |
The config file option 'submit' was used. For all benchmarks threads were bound to cores using the following submit command: dplace -x2 $command This binds threads in order of creation, beginning with the master thread on logical cpu 0, the first slave thread on logical cpu 1, and so on. The -x2 flag instructs dplace to skip placement of the lightweight OpenMP monitor thread, which is created prior to the slave threads.
Transparent Hugepages : Transparent Hugepages are disabled by echo never > /sys/kernel/mm/transparent_hugepage/enabled Software Environment: export KMP_AFFINITY=disabled export KMP_STACKSIZE=200M export KMP_SCHEDULE=static,balanced export OMP_DYNAMIC=FALSE ulimit -s unlimited
Intel Hyperthreading set to Disabled
372.smithwa (peak): "redundant" src.alt was used. 372.smithwa (base): "redundant" src.alt was used.
350.md: | -free |
367.imagick: | -std=c99 |
-O2 -xCORE-AVX2 -ipo1 -openmp -ansi-alias -mcmodel=medium -shared-intel |
-O2 -xCORE-AVX2 -ipo1 -openmp -ansi-alias -mcmodel=medium -shared-intel |
-O2 -xCORE-AVX2 -ipo1 -openmp -mcmodel=medium -shared-intel |
350.md: | -free |
367.imagick: | -std=c99 |