Skip to content

Commit

Permalink
A better performance table (#593)
Browse files Browse the repository at this point in the history
  • Loading branch information
@sbryngelson
sbryngelson authored Aug 29, 2024
1 parent 530e1a5 commit 56a71b4
Showing 1 changed file with 32 additions and 32 deletions.
64 changes: 32 additions & 32 deletions docs/documentation/expectedPerformance.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -20,38 +20,38 @@ Note:
These are reported as (X/Y cores), where X is the used cores, and Y is the total on the die.
* GPU results are for a single GPU device. For single-precision (SP) GPUs, we performed computation in double-precision via conversion in compiler/software; these numbers are _not_ for single-precision computation. AMD MI250X and MI300A GPUs have multiple graphics compute dies (GCDs) per device; we report results for one _GCD_*, though one can quickly estimate full device runtime by dividing the grind time number by the number of GCDs on the device (the MI250X has 2 GCDs). We gratefully acknowledge the permission of LLNL, HPE/Cray, and AMD for permission to release MI300A performance numbers.

| Hardware | | Grind Time [ns] | Compiler | Computer |
| ---: | ----: | ----: | :--- | :--- |
| NVIDIA GH200 (GPU only) | 1 GPU | 0.32 | NVHPC 24.1 | GT Rogues Gallery |
| NVIDIA H100 | 1 GPU | 0.45 | NVHPC 24.5 | GT Rogues Gallery |
| AMD MI300A | 1 _GCD_* | 0.60 | CCE 18.0.0 | LLNL Tioga |
| NVIDIA A100 | 1 GPU | 0.62 | NVHPC 22.11 | GT Phoenix |
| NVIDIA V100 | 1 GPU | 0.99 | NVHPC 22.11 | GT Phoenix |
| NVIDIA A30 | 1 GPU | 1.1 | NVHPC 24.1 | GT Rogues Gallery |
| AMD MI250X | 1 _GCD_* | 1.1 | CCE 16.0.1 | OLCF Frontier |
| AMD MI100 | 1 GPU | 1.4 | CCE 16.0.1 | Cray internal system |
| NVIDIA L40S (SP GPU) | 1 GPU | 1.7 | NVHPC 24.5 | GT ICE |
| NVIDIA P100 | 1 GPU | 2.4 | NVHPC 23.5 | GT CSE Internal |
| AMD EPYC 9534 (Genoa) | 64/64 cores | 2.7 | GNU 12.3.0 | GT Phoenix |
| NVIDIA A40 (SP GPU) | 1 GPU | 3.3 | NVHPC 22.11 | NCSA Delta |
| NVIDIA Grace CPU (Arm, Neoverse V2) | 72/72 cores | 3.7 | NVHPC 24.1 | GT Rogues Gallery |
| NVIDIA RTX6000 (SP GPU) | 1 GPU | 3.9 | NVHPC 22.11 | GT Phoenix |
| AMD EPYC 7763 (Milan) | 64/64 cores | 4.1 | GNU 11.4.0 | NCSA Delta |
| AMD EPYC 7713 (Milan) | 64/64 cores | 5.0 | GNU 12.3.0 | GT Phoenix |
| Intel Xeon Platinum 8480CL (Sapphire Rapids) | 56/56 cores | 5.0 | NVHPC 24.5 | GT Phoenix |
| Intel Xeon Gold 6454S (Sapphire Rapids) | 32/32 cores | 5.6 | NVHPC 24.5 | GT Rogues Gallery |
| Intel Xeon Platinum 8462Y+ (Sapphire Rapids) | 32/32 cores | 6.2 | GNU 12.3.0 | GT ICE |
| Intel Xeon Gold 6548Y+ (Emerald Rapids) | 32/32 cores | 6.6 | Intel oneAPI 2021.9 | GT ICE |
| Intel Xeon Platinum 8352Y (Ice Lake) | 32/32 cores | 6.6 | NVHPC 24.5 | GT Rogues Gallery |
| Ampere Altra Max Q80-28 (Arm, Neoverse-N1) | 80/80 cores | 6.8 | GNU 12.2.0 | OLCF Wombat |
| AMD EPYC 7513 (Milan) | 32/32 cores | 7.4 | GNU 12.3.0 | GT ICE |
| AMD EPYC 7452 (Rome) | 32/32 cores | 8.4 | GNU 12.3.0 | GT ICE |
| AMD EPYC 7401 (Naples) | 24/24 cores | 10 | GNU 10.3.1 | LLNL Corona |
| Apple M1 Pro | 8/10 cores | 14 | GNU 13.2.0 | N/A |
| Intel Xeon Gold 6226 (Cascade Lake) | 12/12 cores | 17 | GNU 12.3.0 | GT ICE |
| Apple M1 Max | 8/10 cores | 18 | GNU 14.1.0 | N/A |
| IBM Power9 | 20/21 cores | 21 | GNU 9.1.0 | OLCF Summit |
| Intel Xeon E5-2650V4 (Broadwell) | 12/12 cores | 27 | NVHPC 23.5 | GT CSE Internal |
| Hardware | Details | Type | Usage | Grind Time [ns] | Compiler | Computer |
| ---: | ----: | ----: | ----: | ----: | :--- | :--- |
| NVIDIA GH200 | GPU only | APU | 1 GPU | 0.32 | NVHPC 24.1 | GT Rogues Gallery |
| NVIDIA H100 | | GPU | 1 GPU | 0.45 | NVHPC 24.5 | GT Rogues Gallery |
| AMD MI300A | | APU | 1 _GCD_* | 0.60 | CCE 18.0.0 | LLNL Tioga |
| NVIDIA A100 | | GPU | 1 GPU | 0.62 | NVHPC 22.11 | GT Phoenix |
| NVIDIA V100 | | GPU | 1 GPU | 0.99 | NVHPC 22.11 | GT Phoenix |
| NVIDIA A30 | | GPU | 1 GPU | 1.1 | NVHPC 24.1 | GT Rogues Gallery |
| AMD MI250X | | GPU | 1 _GCD_* | 1.1 | CCE 16.0.1 | OLCF Frontier |
| AMD MI100 | | GPU | 1 GPU | 1.4 | CCE 16.0.1 | Cray internal system |
| NVIDIA L40S | Single-precision GPU | GPU | 1 GPU | 1.7 | NVHPC 24.5 | GT ICE |
| NVIDIA P100 | | GPU | 1 GPU | 2.4 | NVHPC 23.5 | GT CSE Internal |
| AMD EPYC 9534 | Genoa | CPU | 64/64 cores | 2.7 | GNU 12.3.0 | GT Phoenix |
| NVIDIA A40 | Single-precision GPU | GPU | 1 GPU | 3.3 | NVHPC 22.11 | NCSA Delta |
| NVIDIA Grace CPU | Arm, Neoverse V2 | CPU | 72/72 cores | 3.7 | NVHPC 24.1 | GT Rogues Gallery |
| NVIDIA RTX6000 | Single-precision GPU | GPU | 1 GPU | 3.9 | NVHPC 22.11 | GT Phoenix |
| AMD EPYC 7763 | Milan | CPU | 64/64 cores | 4.1 | GNU 11.4.0 | NCSA Delta |
| AMD EPYC 7713 | Milan | CPU | 64/64 cores | 5.0 | GNU 12.3.0 | GT Phoenix |
| Intel Xeon 8480CL | Platinum, Sapphire Rapids | CPU | 56/56 cores | 5.0 | NVHPC 24.5 | GT Phoenix |
| Intel Xeon 6454S | Gold, Sapphire Rapids | CPU | 32/32 cores | 5.6 | NVHPC 24.5 | GT Rogues Gallery |
| Intel Xeon 8462Y+ | Platinum, Sapphire Rapids | CPU | 32/32 cores | 6.2 | GNU 12.3.0 | GT ICE |
| Intel Xeon 6548Y+ | Gold, Emerald Rapids | CPU | 32/32 cores | 6.6 | Intel oneAPI 2021.9 | GT ICE |
| Intel Xeon 8352Y | Platinum, Ice Lake | CPU | 32/32 cores | 6.6 | NVHPC 24.5 | GT Rogues Gallery |
| Ampere Altra Q80-28 | Arm, Neoverse-N1 | CPU | 80/80 cores | 6.8 | GNU 12.2.0 | OLCF Wombat |
| AMD EPYC 7513 | Milan | CPU | 32/32 cores | 7.4 | GNU 12.3.0 | GT ICE |
| AMD EPYC 7452 | Rome | CPU | 32/32 cores | 8.4 | GNU 12.3.0 | GT ICE |
| AMD EPYC 7401 | Naples | CPU | 24/24 cores | 10 | GNU 10.3.1 | LLNL Corona |
| Apple M1 Pro | | CPU | 8/10 cores | 14 | GNU 13.2.0 | N/A |
| Intel Xeon Gold 6226 | Cascade Lake | CPU | 12/12 cores | 17 | GNU 12.3.0 | GT ICE |
| Apple M1 Max | | CPU | 8/10 cores | 18 | GNU 14.1.0 | N/A |
| IBM Power9 | | CPU | 20/21 cores | 21 | GNU 9.1.0 | OLCF Summit |
| Intel Xeon E5-2650V4 | Broadwell | CPU | 12/12 cores | 27 | NVHPC 23.5 | GT CSE Internal |

__All grind times are in nanoseconds (ns) per grid point (gp) per equation (eq) per right-hand side (rhs) evaluation, so X ns/gp/eq/rhs. Lower is better.__

Expand Down

0 comments on commit 56a71b4

Please sign in to comment.