This version (2023/05/17 14:36) is a draft.
Approvals: 0/1
Approvals: 0/1
GPUs available & how to use it
TOP500 List June 2020
<HTML> <!–slide 1–> <!–for nations flags see https://www.free-country-flags.com–> </HTML>
| Rank | Nation | Machine | Performance | Accelerators |
|---|---|---|---|---|
| 1. |  | Fugaku | 416 PFLOPs/s | |
| 2. |  | Summit | 149 PFLOPs/s | <html><font color=“navy”></html>NVIDIA V100<html></font></html> |
| 3. | | Sierra | 95 PFLOPs/s | <html><font color=“navy”></html>NVIDIA V100<html></font></html> |
| 4. |  | Sunway TaihuLight | 93 PFLOPs/s | |
| 5. | | Tianhe-2A | 62 PFLOPs/s | |
| 6. |  | HPC5 | 36 PFLOPs/s | <html><font color=“navy”></html>NVIDIA V100<html></font></html> |
| 7. | | Selene | 28 PFLOPs/s | <html><font color=“navy”></html>NVIDIA A100<html></font></html> |
| 8. | | Frontera | 24 PFLOPs/s | <html><font color=“navy”></html>NVIDIA RTX5000/V100<html></font></html> |
| 9. | | Marconi-100 | 22 PFLOPs/s | <html><font color=“navy”></html>NVIDIA V100<html></font></html> |
| 10. |  | Piz Daint | 21 PFLOPs/s | <html><font color=“navy”></html>NVIDIA P100<html></font></html> |
<HTML> <!–slide 2–> </HTML>
Components on VSC-5
| Model | #cores | Clock Freq (GHz) | Memory (GB) | Bandwidth (GB/s) | TDP (Watt) | FP32/FP64 (GFLOPs/s) |
|---|---|---|---|---|---|---|
| <html><font color=“navy”></html>19x GeForce RTX-2080Ti n375-[001-019]<html></font></html> | ||||||
 | 4352 | 1.35 | 11 | 616 | 250 | 13450/420 |
| <html><font color=“navy”></html>45×2 nVidia A40 n306[6,7,8]-[001-019,001-019,001-007]<html></font></html> | ||||||
 | 10752 | 1.305 | 48 | 696 | 300 | 37400/1169 |
| <html><font color=“navy”></html>60×2 nVidia A100-40GB n307[1-4]-[001-015]<html></font></html> | ||||||
 | 6912 | 0.765 | 40 | 1555 | 250 | 19500/9700 |
<HTML> <!–slide 3–> </HTML>
Working on GPU nodes Interactively
Interactive mode
1. VSC-5 > salloc -N 1 -p zen2_0256_a40x2 --qos zen2_0256_a40x2 --gres=gpu:2
2. VSC-5 > squeue -u $USER
3. VSC-5 > srun -n 1 hostname (...while still on the login node !)
4. VSC-5 > ssh n3066-012 (...or whatever else node had been assigned)
5. VSC-5 > module load cuda/9.1.85
cd ~/examples/09_special_hardware/gpu_gtx1080/matrixMul
nvcc ./matrixMul.cu
./a.out
cd ~/examples/09_special_hardware/gpu_gtx1080/matrixMulCUBLAS
nvcc matrixMulCUBLAS.cu -lcublas
./a.out
6. VSC-5 > nvidia-smi
7. VSC-5 > /opt/sw/x86_64/glibc-2.17/ivybridge-ep/cuda/9.1.85/NVIDIA_CUDA-9.1_Samples/1_Utilities/deviceQuery/deviceQuery
<HTML> <!–slide 4–> </HTML>
Working on GPU using SLURM
SLURM submission gpu_test.scrpt
#!/bin/bash # # usage: sbatch ./gpu_test.scrpt # #SBATCH -J A40 #SBATCH -N 1 #use -N only if you use both GPUs on the nodes, otherwise leave this line out #SBATCH --partition zen2_0256_a40x2 #SBATCH --qos zen2_0256_a40x2 #SBATCH --gres=gpu:2 #or --gres=gpu:1 if you only want to use half a node module purge module load cuda/9.1.85 nvidia-smi /opt/sw/x86_64/glibc-2.17/ivybridge-ep/cuda/9.1.85/NVIDIA_CUDA-9.1_Samples/1_Utilities/deviceQuery/deviceQuery
<HTML> <!–slide 5–> </HTML>
Real-World Example, AMBER-16
| Performance | Power Efficiency |
|---|---|
 |  |