Add H200 to user guide

Author: grvlbit

docs/hardware/gpu.md

@@ -12,6 +12,7 @@ The UBELIX GPU-based hardware partition currently consists of about 20 compute n
 | 8  | 8x NVIDIA RTX4090 | 128 (2x64) | 24 GB  | 100Gb/s |
 | 1  | 6x NVIDIA A100 | 128 (2x64) | 80 GB | 100Gb/s |
 | 5  | 8x NVIDIA H100 | 128 (2x64) | 96 GB  | 100Gb/s |
+| 2  | 8x NVIDIA H200 | 128 (2x64) | 141 GB  | 100Gb/s |
 
 See the [Slurm partitions page][slurm-partitions] for an overview of options
 for allocating these nodes.
@@ -43,6 +44,7 @@ follows:
 | Nvidia RTX 4090 | 16 | 90GB |
 | Nvidia A100 | 20 | 80GB |
 | Nvidia H100 | 16 | 90GB |
+| Nvidia H200 | 16 | 90GB |
 
 ## Network
 

docs/runjobs/partitions.md

@@ -9,7 +9,7 @@ We are currently operating the following partitions:
 | --------- | -------- | ---------- | ---------------- | ------------- |
 |**epyc2 (default)** | single and multi-core |AMD Epyc2 2x64 cores <br> AMD Epyc4 2x96 cores | 1TB <br> 1.5TB | 1TB |
 | bdw | full nodes only (x*20cores) | Intel Broadwell 2x10 cores | 156GB | 1TB |
-| gpu | GPU <br> (8 GPUs per node, <br> varying CPUs) | Nvidia RTX 3090 <br> Nvidia RTX 4090 <br> Nvidia A100 <br> Nvidia H100 | 24GB <br> 24GB <br> 80GB <br> 96GB |  1.92TB <br> 1.92TB <br> 1.92TB <br> 1.92TB |
+| gpu | GPU <br> (8 GPUs per node, <br> varying CPUs) | Nvidia RTX 3090 <br> Nvidia RTX 4090 <br> Nvidia A100 <br> Nvidia H100 <br> Nvidia H200 | 24GB <br> 24GB <br> 80GB <br> 96GB <br> 141GB |  1.92TB <br> 1.92TB <br> 1.92TB <br> 1.92TB <br> 1.92TB|
 | gpu-invest | GPU | *see gpu partition* |  | |
 | icpu-*investor* | single and multi-core | *see epyc2 partition* |  | |
 

docs/runjobs/scheduled-jobs/gpus.md

@@ -19,6 +19,7 @@ architecture and use one of the following `--gres` option to select it.
 | Nvidia Geforce RTX 4090 | `--gres=gpu:rtx4090:<number_of_gpus>` |
 | Nvidia A100 | `--gres=gpu:a100:<number_of_gpus>` |
 | Nvidia H100 | `--gres=gpu:h100:<number_of_gpus>` |
+| Nvidia H200 | `--gres=gpu:h200:<number_of_gpus>` |
 
 Alternatively, you may use the `--gpus`, `--gpus-per-node` and
 `--gpus-per-tasks` otions. Note that the GPU type still needs to be specified
@@ -48,6 +49,7 @@ In the past, we observed that GPU resources were often left unused because some
 | Nvidia RTX 4090 | 16 | 90GB |
 | Nvidia A100 | 20 | 80GB |
 | Nvidia H100 | 16 | 90GB |
+| Nvidia H200 | 16 | 90GB |
 
 If you submit a GPU job that requests more resources than are available per GPU, your job will be rejected. If your job requires more CPU and memory resources, you may choose to allocate additional GPUs even if these additional GPUs remain unused by your application.