Skip to content

Commit ee27dda

Browse files
bene2k1bene2k1
committed
feat(gpu): update docs
1 parent d715295 commit ee27dda

File tree

4 files changed

+22
-25
lines changed

4 files changed

+22
-25
lines changed

‎pages/gpu/how-to/use-scratch-storage-h100-instances.mdx‎

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -41,7 +41,6 @@ This enhancement allows us to provide the GPU with a substantial amount of scrat
4141
* for L40S-8-48G Instances: 12.8 TB
4242
* for H100-1-80G Instances: 3 TB
4343
* for H100-2-80G Instances: 6 TB
44-
* for H100-SXM-1-80G Instances: ~1.5 TB
4544
* for H100-SXM-2-80G Instances: ~3 TB
4645
* for H100-SXM-4-80G Instances: ~6 TB
4746
* for H100-SXM-1-80G Instances: ~12 TB

‎pages/gpu/reference-content/choosing-gpu-instance-type.mdx‎

Lines changed: 19 additions & 19 deletions
Original file line numberDiff line numberDiff line change
@@ -22,7 +22,7 @@ It empowers European AI startups, giving them the tools (without the need for a
2222

2323
## How to choose the right GPU Instance type
2424

25-
Scaleway provides a range of GPU Instance offers, from [GPU RENDER Instances](https://www.scaleway.com/en/gpu-render-instances/) and [H100 PCIe Instances](https://www.scaleway.com/en/h100-pcie-try-it-now/) to [custom build clusters](https://www.scaleway.com/en/ai-supercomputers/). There are several factors to consider when choosing the right GPU Instance type to ensure that it meets your performance, budget, and scalability requirements.
25+
Scaleway provides a range of GPU Instance offers, from [GPU RENDER Instances](https://www.scaleway.com/en/gpu-render-instances/) and [H100 SXM Instances](https://www.scaleway.com/en/gpu-instances/) to [custom build clusters](https://www.scaleway.com/en/ai-supercomputers/). There are several factors to consider when choosing the right GPU Instance type to ensure that it meets your performance, budget, and scalability requirements.
2626
Below, you will find a guide to help you make an informed decision:
2727

2828
* **Workload requirements:** Identify the nature of your workload. Are you running machine learning, deep learning, high-performance computing (HPC), data analytics, or graphics-intensive applications? Different Instance types are optimized for different types of workloads. For example, the H100 is not designed for graphics rendering. However, other models are. As [stated by Tim Dettmers](https://timdettmers.com/2023/01/30/which-gpu-for-deep-learning/), “Tensor Cores are most important, followed by the memory bandwidth of a GPU, the cache hierarchy, and only then FLOPS of a GPU.”. For more information, refer to the [NVIDIA GPU portfolio](https://docs.nvidia.com/data-center-gpu/line-card.pdf).
@@ -34,7 +34,7 @@ Below, you will find a guide to help you make an informed decision:
3434
* **Scaling:** Consider the scalability requirements of your workload. The most efficient way to scale up your workload is by using:
3535
* Bigger GPU
3636
* Up to 2 PCIe GPU with [H100 Instances](https://www.scaleway.com/en/h100-pcie-try-it-now/) or 8 PCIe GPU with [L4](https://www.scaleway.com/en/l4-gpu-instance/) or [L4OS](https://www.scaleway.com/en/contact-l40s/) Instances.
37-
* An HGX-based server setup with up to 8x NVlink GPUs with [H100-SXM Instances](<ADD LINK>)
37+
* Or better, an HGX-based server setup with up to 8x NVlink GPUs with [H100-SXM Instances](https://www.scaleway.com/en/gpu-instances/)
3838
* A [supercomputer architecture](https://www.scaleway.com/en/ai-supercomputers/) for a larger setup for workload-intensive tasks
3939
* Another way to scale your workload is to use [Kubernetes and MIG](/gpu/how-to/use-nvidia-mig-technology/): You can divide a single H100 or H100-SXM GPU into as many as 7 MIG partitions. This means that instead of employing seven P100 GPUs to set up seven K8S pods, you could opt for a single H100 GPU with MIG to effectively deploy all seven K8S pods.
4040
* **Online resources:** Check for online resources, forums, and community discussions related to the specific GPU type you are considering. This can provide insights into common issues, best practices, and optimizations.
@@ -62,23 +62,23 @@ Remember that there is no one-size-fits-all answer, and the right GPU Instance t
6262
| Better used for | Image / Video encoding (4K) | 7B LLM Fine-Tuning / Inference | 70B LLM Fine-Tuning / Inference |
6363
| What they are not made for | Large models (especially LLM) | Graphic or video encoding use cases | Graphic or video encoding use cases |
6464

65-
| | **[H100-SXM-1-80G](https://www.scaleway.com/en/TBD/)** | **[H100-SXM-2-80G](https://www.scaleway.com/en/TBD/)** | **[H100-SXM-4-80G](https://www.scaleway.com/en/TBD/)** | **[H100-SXM-80G](https://www.scaleway.com/en/TBD/)** |
66-
|--------------------------------------------------------------------|-------------------------------------------------------------------|-------------------------------------------------------------------|-------------------------------------------------------------------|-------------------------------------------------------------------|
67-
| GPU Type | 1x [H100-SXM](https://www.nvidia.com/en-us/data-center/h100/) SXM | 2x [H100-SXM](https://www.nvidia.com/en-us/data-center/h100/) SXM | 4x [H100-SXM](https://www.nvidia.com/en-us/data-center/h100/) SXM | 8x [H100-SXM](https://www.nvidia.com/en-us/data-center/h100/) SXM |
68-
| NVIDIA architecture | Hopper 2022 | Hopper 2022 | Hopper 2022 | Hopper 2022 |
69-
| Tensor Cores | Yes | Yes | Yes | Yes |
70-
| Performance (training in FP16 Tensor Cores) | 1x 1513 TFLOPS | 2x 1513 TFLOPS | 4x 1513 TFLOPS | 8x 1513 TFLOPS |
71-
| VRAM | 1x 80 GB HBM2E (Memory bandwidth: 2TB/s) | 2x 80 GB HBM2E (Memory bandwidth: 2TB/s) | 4x 80 GB HBM2E (Memory bandwidth: 2TB/s) | 8x 80 GB HBM2E (Memory bandwidth: 2TB/s) |
72-
| CPU Type | Xeon Platinum 8452Y (2.0 GHz) | Xeon Platinum 8452Y (2.0 GHz) | Xeon Platinum 8452Y (2.0 GHz) | Xeon Platinum 8452Y (2.0 GHz) |
73-
| vCPUs | 16 | 32 | 64 | 128 |
74-
| RAM | 120 GB DDR5 | 240 GB DDR5 | 480 GB DDR5 | 960 GB DDR5 |
75-
| Storage | Boot on Block 5K | Boot on Block 5K | Boot on Block 5K | Boot on Block 5K |
76-
| [Scratch Storage](/gpu/how-to/use-scratch-storage-h100-instances/) | Yes (~1.5 TB) | Yes (~3 TB) | Yes (~6 TB) | Yes (~12 TB) |
77-
| [MIG compatibility](/gpu/how-to/use-nvidia-mig-technology/) | Yes | Yes | Yes | Yes |
78-
| Bandwidth | 10 Gbps | 20 Gbps | 20 Gbps | 20 Gbps |
79-
| Network technology | [NVLink](/gpu/reference-content/understanding-nvidia-nvlink/) | [NVLink](/gpu/reference-content/understanding-nvidia-nvlink/) | [NVLink](/gpu/reference-content/understanding-nvidia-nvlink/) | [NVLink](/gpu/reference-content/understanding-nvidia-nvlink/) |
80-
| Better used for | *To be defined* | *To be defined* | *To be defined* | *To be defined* |
81-
| What they are not made for | *To be defined* | *To be defined* | *To be defined* | *To be defined* |
65+
| | **[H100-SXM-2-80G](https://www.scaleway.com/en/TBD/)** | **[H100-SXM-4-80G](https://www.scaleway.com/en/TBD/)** | **[H100-SXM-80G](https://www.scaleway.com/en/TBD/)** |
66+
|--------------------------------------------------------------------|-------------------------------------------------------------------|-------------------------------------------------------------------|-------------------------------------------------------------------|
67+
| GPU Type | 2x [H100-SXM](https://www.nvidia.com/en-us/data-center/h100/) SXM | 4x [H100-SXM](https://www.nvidia.com/en-us/data-center/h100/) SXM | 8x [H100-SXM](https://www.nvidia.com/en-us/data-center/h100/) SXM |
68+
| NVIDIA architecture | Hopper 2022 | Hopper 2022 | Hopper 2022 |
69+
| Tensor Cores | Yes | Yes | Yes |
70+
| Performance (training in FP16 Tensor Cores) | 2x 1979 TFLOPS | 4x 1979 TFLOPS | 8x 1979 TFLOPS |
71+
| VRAM | 2x 80 GB HBM3 (Memory bandwidth: 3.35TB/s) | 4x 80 GB HBM3 (Memory bandwidth: 3.35TB/s) | 8x 80 GB HBM3 (Memory bandwidth: 3.35TB/s) |
72+
| CPU Type | Xeon Platinum 8452Y (2.0 GHz) | Xeon Platinum 8452Y (2.0 GHz) | Xeon Platinum 8452Y (2.0 GHz) |
73+
| vCPUs | 32 | 64 | 128 |
74+
| RAM | 240 GB DDR5 | 480 GB DDR5 | 960 GB DDR5 |
75+
| Storage | Boot on Block 5K | Boot on Block 5K | Boot on Block 5K |
76+
| [Scratch Storage](/gpu/how-to/use-scratch-storage-h100-instances/) | Yes (~3 TB) | Yes (~6 TB) | Yes (~12 TB) |
77+
| [MIG compatibility](/gpu/how-to/use-nvidia-mig-technology/) | Yes | Yes | Yes |
78+
| Bandwidth | 20 Gbps | 20 Gbps | 20 Gbps |
79+
| Network technology | [NVLink](/gpu/reference-content/understanding-nvidia-nvlink/) | [NVLink](/gpu/reference-content/understanding-nvidia-nvlink/) | [NVLink](/gpu/reference-content/understanding-nvidia-nvlink/) |
80+
| Better used for | LLM fine-tuning, LLM inference with lower quantization and/or larger parameter counts, fast computer vision training model training | LLM fine-tuning, LLM inference with lower quantization and/or larger parameter counts, fast computer vision training model training | Llama 4 or Deepseek R1 inference |
81+
| What they are not made for | Training of LLM (single node), Graphic or video encoding use cases | Training of LLM (single node), Graphic or video encoding use cases | Training of LLM (single node), Graphic or video encoding use cases |
8282

8383
| | **[L4-1-24G](https://www.scaleway.com/en/l4-gpu-instance/)** | **[L4-2-24G](https://www.scaleway.com/en/l4-gpu-instance/)** | **[L4-4-24G](https://www.scaleway.com/en/l4-gpu-instance/)** | **[L4-8-24G](https://www.scaleway.com/en/l4-gpu-instance/)** |
8484
|---------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------|

‎pages/gpu/reference-content/gpu-instances-bandwidth-overview.mdx‎

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -35,7 +35,6 @@ GPU workloads often involve processing large datasets, requiring high-bandwidth
3535

3636
| Instance Type | Internet Bandwidth | Block Bandwidth |
3737
|-------------------|-------------------------|---------------------|
38-
| H100-SXM-1-80G | 10 Gbit/s | 5 GiB/s |
3938
| H100-SXM-2-80G | 20 Gbit/s | 5 GiB/s |
4039
| H100-SXM-4-80G | 20 Gbit/s | 5 GiB/s |
4140
| H100-SXM-8-80G | 20 Gbit/s | 5 GiB/s |

‎pages/instances/faq.mdx‎

Lines changed: 3 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -151,10 +151,9 @@ You can change the storage type and flexible IP after the Instance creation, whi
151151

152152
| Range | Available in | Price |
153153
|-------------------|------------------------|-------------------|
154-
| H100-SXM-1-80G | PAR2 | €X.XX/hour¹ |
155-
| H100-SXM-2-80G | PAR2 | €X.XX/hour¹ |
156-
| H100-SXM-4-80G | PAR2 | €X.XX/hour¹ |
157-
| H100-SXM-8-80G | PAR2 | €X.XX/hour¹ |
154+
| H100-SXM-2-80G | PAR2 | €6.018/hour¹ |
155+
| H100-SXM-4-80G | PAR2 | €11.61/hour¹ |
156+
| H100-SXM-8-80G | PAR2 | €23.028/hour¹ |
158157
| H100-1-80G | PAR2, WAW2 | €2.52/hour¹ |
159158
| H100-2-80G | PAR2, WAW2 | €5.04/hour¹ |
160159
| L40S-1-48G | PAR2 | €1.40/hour¹ |

0 commit comments

Comments
 (0)