# Upgrade cloud GPU driver version **URL:** \ **Category:** Technical Help\ **Created:** [August 4, 2020, 2:13pm UTC](https://deeptalk.lambda.ai/t/upgrade-cloud-gpu-driver-version/1787 "2020-08-04T14:13:20Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![sharner](https://avatars.discourse-cdn.com/v4/letter/s/b5ac83/32.png) [@sharner](https://deeptalk.lambda.ai/u/sharner)\ **Post date:** [August 4, 2020, 2:13pm UTC](https://deeptalk.lambda.ai/t/upgrade-cloud-gpu-driver-version/1787/1 "2020-08-04T14:13:20Z")
Is it possible to upgrade the driver versions of thee cloud GPU? I’m trying to use the Nvidia ODTK with TensorRT and it seems to require nvidia drivers 450. I’ve tried to upgrade the drivers but the installer complains it can’t unload the nvidia kernel module. I tried to unload it, and it said the module was in use. I also wasn’t able to reboot the instance. Ideas? ---
**Author:** ![jeremy](https://sea1.discourse-cdn.com/flex019/user_avatar/deeptalk.lambda.ai/jeremy/32/235_2.png) [@jeremy](https://deeptalk.lambda.ai/u/jeremy)\ **Post date:** [August 5, 2020, 8:48pm UTC](https://deeptalk.lambda.ai/t/upgrade-cloud-gpu-driver-version/1787/2 "2020-08-05T20:48:18Z")
On our older 4-GPU cloud instances we’re unfortunately not allowed (due to licensing restrictions) to help you upgrade the CUDA Toolkit to anything over version 9.X. You can do so yourself, but our hands are a bit tied. The NVidia Persistence daemon will make use of the NVidia module. You can stop it like so: ```sh sudo systemctl stop nvidia-persistenced ``` ---
**Author:** ![terminalkitten](https://avatars.discourse-cdn.com/v4/letter/t/dbc845/32.png) [@terminalkitten](https://deeptalk.lambda.ai/u/terminalkitten)\ **Post date:** [August 18, 2020, 10:29pm UTC](https://deeptalk.lambda.ai/t/upgrade-cloud-gpu-driver-version/1787/3 "2020-08-18T22:29:41Z")
See: [Update CUDA on Lambda Cloud - Technical Help - DeepTalk - Deep Learning Community](https://deeptalk.lambda.ai/t/update-cuda-on-lambda-cloud/1500/3)