How to update NVIDIA drivers

Features

Pricing

Install

Knowledge Base

Guides

Update or not update?

Before we start let’s answer some typical questions:

Is it necessary to update?

No! If everything works for you - don’t touch.

Why then update?

New drivers include support for new cards such as 1660ti, 2060, 2070, 2080, 2080ti series Sometimes special optimization in some miners have a positive effect on hashrate but only for some algorithms and Pascal family of top-end cards - 10xx.

Still not changing your mind?

Then we will start.

Updating procedure

We strongly recommend using Hive Shell for proper driver update procedure. Of course, if you can use local console - use it or if your rigs located in the same LAN with your PC/laptop then you can use for example SSH-clients such as standlone app PuTTY or working in browser ShellInABox instead Hive Shell.

The update procedure is quite simple. Step 1 - Start Hive Shell Step 2 - Execute the following commands:

apt update

nvidia-driver-update

The driver update procedure takes 5-10 min if you’re using an SSD/HDD and 30+ min if you’re using a slow drive such as an USB 2.0 stick. If the update procedure will finish successfully you will see the following message:

Now you can reboot your rig. After reboot check driver version and miner functionality.

Update/downgrade to desired version

You can update driver to desired version. In this case command line for nvidia-driver-update tool must contain URL to desired driver version. For example downgrade to 396.54: nvidia-driver-update https://download.hiveos.farm/drivers/NVIDIA-Linux-x86_64-396.54.run

Troubleshooting

Driver installation failed (module “nvidia-drm”)

In some cases the update procedure can sometimes fail with the following error:

In case this happens, please retry the update procedure - in most cases it will help fix the error.

Driver installation failed (module “nvidia-uvm”)

If installation stops with the following error:

in this case, reboot rig and repeat the update procedure.

Error during uninstallation previous driver (module “nvidia”)

During the installation new driver nvidia-driver-update uninstalls the previous version of the driver. If uninstallation stops with the following error:

It can be caused by falling current driver or application which is still using driver. Reboot the rig and retry the installation procedure.

Error during uninstallation previous driver (signal SIGBUS)

During installation new driver nvidia-driver-update uninstalls the previous version of the driver. If uninstallation stops with the following error (Received signal SIGBUS; aborting):

No solution yet… Not reproducible.

Not enough disk space

use disk-expand tool to enlarge disk to use all space
remove unused drivers at /hive-driver-pack (to do this, send the rm /hive-drivers-pack/NV* command to your worker - it will delete all the available NVIDIA driver archives)
uninstall miners by hpkg tool

Installation fail when unpacking driver

By default TMP directory is on RAM disk so you need to enable logs to have some space: logs-on reboot

After the update miner don’t start properly

If your miner doesn't start to work of works incorrectly after the update, you can repeat driver update procedure or downgrade driver version as described above (see install desired driver version).

Nothing helps?

Try:

write to the support [email protected]
call community support on Telegram channel
reflash the image
remote update image via hive-replace tool

NVIDIA drivers update guide and CUDA 10.x