About GPU Driver Management in CKS
Configure and manage GPU driver versions for your Node Pools
GPU driver management is currently a preview feature, and has the following limitations:
- No Cloud Console UI: Configuration must be done through Kubernetes manifests.
- No automatic minor or patch version upgrades: Major version upgrades are supported. Currently, CoreWeave does not automatically upgrade minor or patch versions on Node Pools.
GPU driver version updates require Node reboots. When you set or change a GPU driver version, CoreWeave support will contact you to schedule a maintenance window.
This feature is not available for all clusters. Please contact CoreWeave support if you have questions about availability.
CKS allows you to specify a GPU driver version for your Node Pools. Driver versions can be specified in the Node Pool manifest when deploying a Node Pool. CKS workloads can target Nodes running specific driver versions using the nodeSelector
field in the workload manifest.
Selecting a driver version using the gpu
field is optional. If not specified, the Node Pool uses the latest available driver version.
Supported driver versions
CoreWeave supports all tested release driver (TRD) versions from Nvidia, listed below as major versions.
Although customers are not forced to upgrade from a major version, our recommendation is to upgrade frequently to ensure the best combination of testing coverage, security posture, stability, and performance.
Instance type | Supported ncore image version | Compatible driver versions |
---|---|---|
GB200-4X | ncore-image-2.17.0 | 580 |
GD-1XGH200 | ncore-image-2.17.0 | 580 |
GD-8XA100-I128 | ncore-image-2.17.0 | 535 , 570 |
GD-8XH100-I128 | ncore-image-2.17.0 | 535 , 570 |
GD-8XH100IB-I128 | ncore-image-2.17.0 | 535 , 570 |
GD-8XH200IB-I128 | ncore-image-2.17.0 | 535 , 570 |
GD-8XL40-I128 | ncore-image-2.17.0 | 535 , 570 |
GD-8XL40S-I128 | ncore-image-2.17.0 | 535 , 570 |
RTXP6000-8X | ncore-image-2.17.0 | 535 , 570 |