hpl-2.0_FERMI_v15 can not full of GPUs memory

Accelerated Computing CUDA CUDA Programming and Performance
1

I want to run a HPL test on our 6P40 prodoct.I tried to install the hpl-2.0FERMI_v15 and run this on my 6p40 or 8*v100 platform. The result was just 10%~20% of the system. And the GPU mem usage was just 10% of the GPU have. Do you have some configuration to make the HPL or CUDA to use all of the GPU memory. Or you have a newer version of the HPL-GPU? Thanks,

Wed Jul 17 18:39:58 2019
±----------------------------------------------------------------------------+
| NVIDIA-SMI 418.67 Driver Version: 418.67 CUDA Version: 10.1 |
|-------------------------------±---------------------±---------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla P40 On | 00000000:3B:00.0 Off | 0 |
| N/A 35C P0 49W / 250W | 2285MiB / 22919MiB | 0% Default |
±------------------------------±---------------------±---------------------+
| 1 Tesla P40 On | 00000000:60:00.0 Off | 0 |
| N/A 34C P0 50W / 250W | 2285MiB / 22919MiB | 0% Default |
±------------------------------±---------------------±---------------------+
| 2 Tesla P40 On | 00000000:61:00.0 Off | 0 |
| N/A 36C P0 50W / 250W | 2285MiB / 22919MiB | 0% Default |
±------------------------------±---------------------±---------------------+
| 3 Tesla P40 On | 00000000:86:00.0 Off | 0 |
| N/A 43C P0 52W / 250W | 2285MiB / 22919MiB | 0% Default |
±------------------------------±---------------------±---------------------+
| 4 Tesla P40 On | 00000000:DA:00.0 Off | 0 |
| N/A 34C P0 51W / 250W | 2285MiB / 22919MiB | 0% Default |
±------------------------------±---------------------±---------------------+
| 5 Tesla P40 On | 00000000:DB:00.0 Off | 0 |
| N/A 31C P0 49W / 250W | 2285MiB / 22919MiB | 0% Default |
±------------------------------±---------------------±---------------------+

Wed Jul 17 18:41:55 2019
±----------------------------------------------------------------------------+
| NVIDIA-SMI 418.67 Driver Version: 418.67 CUDA Version: 10.1 |
|-------------------------------±---------------------±---------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla V100-SXM2… Off | 00000000:1A:00.0 Off | 0 |
| N/A 44C P0 89W / 300W | 2454MiB / 32480MiB | 100% Default |
±------------------------------±---------------------±---------------------+
| 1 Tesla V100-SXM2… Off | 00000000:1B:00.0 Off | 0 |
| N/A 46C P0 93W / 300W | 2454MiB / 32480MiB | 100% Default |
±------------------------------±---------------------±---------------------+
| 2 Tesla V100-SXM2… Off | 00000000:3D:00.0 Off | 0 |
| N/A 46C P0 77W / 300W | 2454MiB / 32480MiB | 100% Default |
±------------------------------±---------------------±---------------------+
| 3 Tesla V100-SXM2… Off | 00000000:3E:00.0 Off | 0 |
| N/A 43C P0 72W / 300W | 2454MiB / 32480MiB | 18% Default |
±------------------------------±---------------------±---------------------+
| 4 Tesla V100-SXM2… Off | 00000000:88:00.0 Off | 0 |
| N/A 46C P0 121W / 300W | 2454MiB / 32480MiB | 100% Default |
±------------------------------±---------------------±---------------------+
| 5 Tesla V100-SXM2… Off | 00000000:89:00.0 Off | 0 |
| N/A 46C P0 78W / 300W | 2454MiB / 32480MiB | 100% Default |
±------------------------------±---------------------±---------------------+
| 6 Tesla V100-SXM2… Off | 00000000:B2:00.0 Off | 0 |
| N/A 45C P0 129W / 300W | 2454MiB / 32480MiB | 91% Default |
±------------------------------±---------------------±---------------------+
| 7 Tesla V100-SXM2… Off | 00000000:B3:00.0 Off | 0 |
| N/A 41C P0 80W / 300W | 2454MiB / 32480MiB | 1% Default |
±------------------------------±---------------------±---------------------+

2

The hpl-2.0FERMI_v15 while still available, is considered obsolete and NVIDIA doesn’t provide support for it.

P40 will be uninteresting from an HPL standpoint. P40 is based on the Pascal cc6.1 family processor, which has very low rate double-precision throughput.

In general the memory usage of HPL will be affected/controlled by the settings in the HPL.dat file.

3

Thanks,

4

I want to run a HPL test on our 8v100 prodoct.I tried to install the hpl-2.0FERMI_v15 and run this on my 8v100 platform. The result was just 10%~20% of the system,Thanks,
where can download hpl_cuda_10_ompi-3.1_volta_pascal_kepler_9-27-18_ext_v2.tgz file?

5

It’s not available publicly.