Skip to content

Commit

Permalink
Merge pull request #478 from ttakayuki/202406-nvidia-update
Browse files Browse the repository at this point in the history
2024-06 NVIDIA Library Update
  • Loading branch information
ttakayuki authored Jun 20, 2024
2 parents ecedf90 + db2fe5d commit 3dfa7fe
Show file tree
Hide file tree
Showing 6 changed files with 126 additions and 99 deletions.
98 changes: 52 additions & 46 deletions en/docs/gpu.md
Original file line number Diff line number Diff line change
Expand Up @@ -29,72 +29,78 @@ The following is a list of CUDA Toolkit, cuDNN, and NCCL that can be used with t
| cuda/12.2 | 12.2.0 | Yes | Yes | Yes |
| cuda/12.3 | 12.3.2 | Yes | Yes | Yes |
| cuda/12.4 | 12.4.0 | Yes | Yes | Yes |
| cuda/12.4 | 12.4.1 | Yes | Yes | Yes |
| cuda/12.5 | 12.5.0 | Yes | Yes | Yes |

[^1]: Provided only for experimental use. Rocky Linux 8.6 is supported with CUDA 11.7.1 or later.

## cuDNN

Compute Node (V):

| Version | cuda/11.2[^1] | cuda/11.6[^1] | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 |
| ------- | ------------- | ------------- | --------- | --------- | --------- | --------- | --------- | --------- |
| 8.1.1 | Yes | - | - | - | - | - | - | - |
| 8.3.3 | Yes | Yes | - | - | - | - | - | - |
| 8.4.1 | Yes | Yes | Yes | Yes | - | - | - | - |
| 8.6.0 | Yes | Yes | Yes | Yes | - | - | - | - |
| 8.7.0 | Yes | Yes | Yes | Yes | - | - | - | - |
| 8.8.1 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 8.9.7 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 9.0.0[^2] | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Version | cuda/11.2[^1] | cuda/11.6[^1] | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 | cuda/12.5 |
| ------- | ------------- | ------------- | --------- | --------- | --------- | --------- | --------- | --------- | --------- |
| 8.1.1 | Yes | - | - | - | - | - | - | - | - |
| 8.3.3 | Yes | Yes | - | - | - | - | - | - | - |
| 8.4.1 | Yes | Yes | Yes | Yes | - | - | - | - | - |
| 8.6.0 | Yes | Yes | Yes | Yes | - | - | - | - | - |
| 8.7.0 | Yes | Yes | Yes | Yes | - | - | - | - | - |
| 8.8.1 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 8.9.7 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 9.0.0[^2] | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 9.1.1 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |

Compute Node (A):

| Version | cuda/11.2 | cuda/11.6 | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 |
| :------ | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- |
| 8.1.1 | Yes | - | - | - | - | - | - | - |
| 8.3.3 | Yes | Yes | - | - | - | - | - | - |
| 8.4.1 | Yes | Yes | Yes | Yes | - | - | - | - |
| 8.6.0 | Yes | Yes | Yes | Yes | - | - | - | - |
| 8.7.0 | Yes | Yes | Yes | Yes | - | - | - | - |
| 8.8.1 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 8.9.7 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 9.0.0[^2] | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Version | cuda/11.2 | cuda/11.6 | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 | cuda/12.5 |
| :------ | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- |
| 8.1.1 | Yes | - | - | - | - | - | - | - | - |
| 8.3.3 | Yes | Yes | - | - | - | - | - | - | - |
| 8.4.1 | Yes | Yes | Yes | Yes | - | - | - | - | - |
| 8.6.0 | Yes | Yes | Yes | Yes | - | - | - | - | - |
| 8.7.0 | Yes | Yes | Yes | Yes | - | - | - | - | - |
| 8.8.1 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 8.9.7 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 9.0.0[^2] | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 9.1.1 | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |

[^2]: We have confirmed that when cuDNN 9.0.0 is used with CUDA 11.0 to CUDA 11.3, an error occurs when calling the `cudnnRNNBackwardWeights_v8` function.

## NCCL

Compute Node (V):

| Version | cuda/11.2[^1] | cuda/11.6[^1] | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 |
| --------- | ------------- | ------------- | --------- | --------- | --------- | --------- | --------- | --------- |
| 2.8.4-1 | Yes | - | - | - | - | - | - | - |
| 2.11.4-1 | - | Yes | - | - | - | - | - | - |
| 2.12.12-1 | - | Yes | - | - | - | - | - | - |
| 2.13.4-1 | - | - | Yes | - | - | - | - | - |
| 2.14.3-1 | - | - | Yes | - | - | - | - | - |
| 2.15.5-1 | - | - | - | Yes | - | - | - | - |
| 2.16.2-1 | - | - | - | Yes | - | - | - | - |
| 2.17.1-1 | - | - | - | - | Yes | - | - | - |
| 2.18.5-1 | - | - | - | - | - | Yes | - | - |
| 2.19.3-1 | - | - | - | - | - | Yes | Yes | - |
| 2.20.5-1 | - | - | - | - | - | Yes | - | Yes |
| Version | cuda/11.2[^1] | cuda/11.6[^1] | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 | cuda/12.5 |
| --------- | ------------- | ------------- | --------- | --------- | --------- | --------- | --------- | --------- | --------- |
| 2.8.4-1 | Yes | - | - | - | - | - | - | - | - |
| 2.11.4-1 | - | Yes | - | - | - | - | - | - | - |
| 2.12.12-1 | - | Yes | - | - | - | - | - | - | - |
| 2.13.4-1 | - | - | Yes | - | - | - | - | - | - |
| 2.14.3-1 | - | - | Yes | - | - | - | - | - | - |
| 2.15.5-1 | - | - | - | Yes | - | - | - | - | - |
| 2.16.2-1 | - | - | - | Yes | - | - | - | - | - |
| 2.17.1-1 | - | - | - | - | Yes | - | - | - | - |
| 2.18.5-1 | - | - | - | - | - | Yes | - | - | - |
| 2.19.3-1 | - | - | - | - | - | Yes | Yes | - | - |
| 2.20.5-1 | - | - | - | - | - | Yes | - | Yes | - |
| 2.21.5-1 | - | - | - | - | - | Yes | - | Yes | Yes |

Compute Node (A):

| Version | cuda/11.2 | cuda/11.6 | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 |
| :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- |
| 2.8.4-1 | Yes | - | - | - | - | - | - | - |
| 2.11.4-1 | - | Yes | - | - | - | - | - | - |
| 2.12.12-1 | - | Yes | - | - | - | - | - | - |
| 2.13.4-1 | - | - | Yes | - | - | - | - | - |
| 2.14.3-1 | - | - | Yes | - | - | - | - | - |
| 2.15.5-1 | - | - | - | Yes | - | - | - | - |
| 2.16.2-1 | - | - | - | Yes | - | - | - | - |
| 2.17.1-1 | - | - | - | - | Yes | - | - | - |
| 2.18.5-1 | - | - | - | - | - | Yes | - | - |
| 2.19.3-1 | - | - | - | - | - | Yes | Yes | - |
| 2.20.5-1 | - | - | - | - | - | Yes | - | Yes |
| Version | cuda/11.2 | cuda/11.6 | cuda/11.7 | cuda/11.8 | cuda/12.1 | cuda/12.2 | cuda/12.3 | cuda/12.4 | cuda/12.5 |
| :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- | :-------- |
| 2.8.4-1 | Yes | - | - | - | - | - | - | - | - |
| 2.11.4-1 | - | Yes | - | - | - | - | - | - | - |
| 2.12.12-1 | - | Yes | - | - | - | - | - | - | - |
| 2.13.4-1 | - | - | Yes | - | - | - | - | - | - |
| 2.14.3-1 | - | - | Yes | - | - | - | - | - | - |
| 2.15.5-1 | - | - | - | Yes | - | - | - | - | - |
| 2.16.2-1 | - | - | - | Yes | - | - | - | - | - |
| 2.17.1-1 | - | - | - | - | Yes | - | - | - | - |
| 2.18.5-1 | - | - | - | - | - | Yes | - | - | - |
| 2.19.3-1 | - | - | - | - | - | Yes | Yes | - | - |
| 2.20.5-1 | - | - | - | - | - | Yes | - | Yes | - |
| 2.21.5-1 | - | - | - | - | - | Yes | - | Yes | Yes |

## GDRCopy

Expand Down
6 changes: 3 additions & 3 deletions en/docs/system-overview.md
Original file line number Diff line number Diff line change
Expand Up @@ -161,7 +161,7 @@ The software available on the ABCI system is shown below.
| OS | Rocky Linux | 8.6 | - |
| OS | Red Hat Enterprise Linux | - | 8.2 |
| Job Scheduler | Altair Grid Engine | 8.6.19_C121_1 | 8.6.19_C121_1 |
| Development Environment | [CUDA Toolkit](gpu.md#cuda-toolkit) | 11.2.2<br>11.6.2<br>11.7.1<br>11.8.0<br>12.1.1<br>12.2.0<br>12.3.2<br>12.4.0 | 11.2.2<br>11.6.2<br>11.7.1<br>11.8.0<br>12.1.1<br>12.2.0<br>12.3.2<br>12.4.0 |
| Development Environment | [CUDA Toolkit](gpu.md#cuda-toolkit) | 11.2.2<br>11.6.2<br>11.7.1<br>11.8.0<br>12.1.1<br>12.2.0<br>12.3.2<br>12.4.0<br>12.4.1<br>12.5.0 | 11.2.2<br>11.6.2<br>11.7.1<br>11.8.0<br>12.1.1<br>12.2.0<br>12.3.2<br>12.4.0<br>12.4.1<br>12.5.0 |
| | Intel oneAPI<br>(compilers and libraries) | 2024.0.2 | 2024.0.2 |
| | Intel VTune | 2024.0.0 | 2024.0.0 |
| | Intel Trace Analyzer and Collector | 2022.0 | 2022.0 |
Expand All @@ -183,8 +183,8 @@ The software available on the ABCI system is shown below.
| Container | [SingularityPRO](containers.md#singularity) | 4.1.2-2 | 4.1.2-2 |
| | Singularity Endpoint | 2.3.0 | 2.3.0 |
| MPI | [Intel MPI](mpi.md#intel-mpi) | 2021.11 | 2021.11 |
| Library | [cuDNN](gpu.md#cudnn) | 8.1.1<br>8.3.3<br>8.4.1<br>8.6.0<br>8.7.0<br>8.8.1<br>8.9.7<br>9.0.0 | 8.1.1<br>8.3.3<br>8.4.1<br>8.6.0<br>8.7.0<br>8.8.1<br>8.9.7<br>9.0.0 |
| | [NCCL](gpu.md#nccl) | 2.8.4-1<br>2.11.4-1<br>2.12.12-1<br>2.13.4-1<br>2.14.3-1<br>2.15.5-1<br>2.16.2-1<br>2.17.1-1<br>2.18.5-1<br>2.19.3-1<br>2.20.5-1 | 2.8.4-1<br>2.11.4-1<br>2.12.12-1<br>2.13.4-1<br>2.14.3-1<br>2.15.5-1<br>2.16.2-1<br>2.17.1-1<br>2.18.5-1<br>2.19.3-1<br>2.20.5-1 |
| Library | [cuDNN](gpu.md#cudnn) | 8.1.1<br>8.3.3<br>8.4.1<br>8.6.0<br>8.7.0<br>8.8.1<br>8.9.7<br>9.0.0<br>9.1.1 | 8.1.1<br>8.3.3<br>8.4.1<br>8.6.0<br>8.7.0<br>8.8.1<br>8.9.7<br>9.0.0<br>9.1.1 |
| | [NCCL](gpu.md#nccl) | 2.8.4-1<br>2.11.4-1<br>2.12.12-1<br>2.13.4-1<br>2.14.3-1<br>2.15.5-1<br>2.16.2-1<br>2.17.1-1<br>2.18.5-1<br>2.19.3-1<br>2.20.5-1<br>2.21.5-1 | 2.8.4-1<br>2.11.4-1<br>2.12.12-1<br>2.13.4-1<br>2.14.3-1<br>2.15.5-1<br>2.16.2-1<br>2.17.1-1<br>2.18.5-1<br>2.19.3-1<br>2.20.5-1<br>2.21.5-1 |
| | gdrcopy | 2.4.1 | 2.4.1 |
| | UCX | 1.10 | 1.11 |
| | libfabric | 1.7.0-1 | 1.9.0rc1-1 |
Expand Down
8 changes: 8 additions & 0 deletions en/docs/system-updates.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,13 @@
# System Updates

## 2024-06-18

| Add / Update / Delete | Software | Version | Previous version |
|:--|:--|:--|:--|
| Add | cuda | 12.4.1<br>12.5.0 | |
| Add | cudnn | 9.1.1 | |
| Add | nccl | 2.21.5-1 | |

## 2024-05-28

* In Spot and On-demand Service on Compute Node (V), the upper limits for the number of running jobs at the same time per system have been changed as follows.
Expand Down
Loading

0 comments on commit 3dfa7fe

Please sign in to comment.