Appendix. GPU specifications (Coming soon)
NVIDIA’s A100 model easily handles a wide range of acceleration requirements, from small tasks to multi-node workloads. Thanks to the versatility of the A100, IT administrators can always maximize utilization of all GPUs in the data center.
Content update 예정
Default configuration specifications
| Instance family | Description |
|---|---|
| A100 | Configurable as 7 instances with 1g. 10gb specification - Or configurable as 1 instance with 7g. 80gb specification |
MIG instance exception combinations
| Instance family | Description |
|---|---|
| A100 | Configuration that uses 3g. 40gb and 4g. 40gb simultaneously is not allowed - Disallowed combination due to manufacturer policy |
info
For detailed information about the A100, please refer to the NVIDIA > A100 Tensor Core GPU official documentation.
TFLOPS-based configuration specifications
info
During the trial period, only 20TF is provided to all users.
| TF resource | Number of GPU cards | CPU (EA) | Memory (GiB) | GPU (GiB) | Storage (GiB) |
|---|---|---|---|---|---|
| 20TF | A100 x 1EA | 16 | 128 | 80 | 2048 |
| 80TF | A100 x 4EA | 64 | 512 | 320 | 2048 |
| 160TF | A100 x 8EA | 128 | 1024 | 640 | 10240 |