ipex-llm icon indicating copy to clipboard operation
ipex-llm copied to clipboard

failed to install drivers for A770 in Ubuntu 22.04.05

Open luningxie opened this issue 6 months ago • 9 comments

failed to install drivers for A770 in Ubuntu 22.04.05

I have two A770 GPUs, and follow the advice you provided(https://github.com/intel/ipex-llm/issues/13183) to install the right version of drivers:

here are the instruction:

wget -qO - https://repositories.intel.com/gpu/intel-graphics.key | \
  sudo gpg --yes --dearmor --output /usr/share/keyrings/intel-graphics.gpg


echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy/lts/2350 unified" | \
  sudo tee /etc/apt/sources.list.d/intel-gpu-jammy.list

sudo apt update
sudo apt install -y intel-i915-dkms=1.23.10.92.231129.101+i141-1

echo "deb [arch=amd64,i386 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy unified" | \
  sudo tee /etc/apt/sources.list.d/intel-gpu-jammy.list

sudo apt update
sudo apt install -y intel-level-zero-gpu=1.3.29735.27-914~22.04
sudo apt install -y level-zero=1.14.0-744~22.04 level-zero-dev=1.14.0-744~22.04
sudo apt install -y xpu-smi=1.2.33-52~22.04

After I take all steps and reboot, the screen is black and can't enter the normal windows.

Have tried to blacklist ast and set WaylandEnable=true, it doesn't work either.

env: intel xeon 8480 384 DDR5 memeories Gigabyte MS73 mainboard 2 A770 GPUs

luningxie avatar Jun 23 '25 15:06 luningxie

Hi @luningxie Could you please try installing the drivers by following the steps provided in the following guide?

Install Intel GPU drivers for Linux Kernel 6.5

Let us know if you run into any issues.

liu-shaojun avatar Jun 24 '25 01:06 liu-shaojun

problems remain.

1.downgrade to 6.5.0.35 2.. Enable GPU driver support through force_probe flag 3.For Intel Xeon-W/SP CPU with multiple A770 Arc GPUs. install out of tree drivers and compute runtimes

after the reboot, there are strange and dark colors on the screen that is hard to see. I think it might be related to ast and wayland settings. then I blacklist ast and set waylandenable=true.

then reboot, the screen is black entirely.

luningxie avatar Jun 24 '25 06:06 luningxie

Thanks for the update.

Just to confirm — is your current kernel version 6.5? We noticed you mentioned using the force_probe flag to enable GPU driver support. We didn’t need this step when setting up our own A770 systems — could you share why you chose to use it in your case?

Would it be possible to first troubleshoot the black screen issue without installing the GPU driver, to help determine whether it's actually caused by the driver installation? We haven’t encountered a black screen issue on our side during driver setup, so it might be helpful to isolate the cause before proceeding further.

liu-shaojun avatar Jun 24 '25 06:06 liu-shaojun

for ubuntu 22.04.05,the default kernel is 6.8. So downgrade to 6.5. if I install GPU driver according to the official tutorial(llama-cpp portable version), the screen was ok but failed to load LLM. So definitely the screen problem is caused by installing the intel fw gpu or level drivers.

I misunderstood(https://github.com/intel/ipex-llm/blob/main/docs/mddocs/Quickstart/install_linux_gpu.md#for-linux-kernel-65 )to use the force_probe flag, will try to avoid it again and see what happens.

luningxie avatar Jun 24 '25 08:06 luningxie

Just to confirm — are you installing the GPU driver in order to run the Ollama portable zip, or are you planning to use something else, such as vLLM?

Understanding this will help us give more targeted suggestions.

liu-shaojun avatar Jun 24 '25 08:06 liu-shaojun

thanx for kindly help.

I want to run moe models like deepseek r1 or qwen3 with llama-cpp portable version which is easier to setup. that's why i bought 384G ddr 5 memories.

vllm is not supposed to run moe model as far as i know. And docker version is hard to set up and use.

luningxie avatar Jun 24 '25 09:06 luningxie

even with 6.5.0-35 and not using force_probe flag, the same result again. before rebooting, screen is fine. after it, screen become darker and the color is strange.

luningxie avatar Jun 25 '25 06:06 luningxie

在 Ubuntu 22.04.05 中安装 A770 的驱动程序失败

我有两个 A770 GPU,并按照您提供的建议 (#13183) 安装正确版本的驱动程序:

这是说明:

wget -qO - https://repositories.intel.com/gpu/intel-graphics.key | \
  sudo gpg --yes --dearmor --output /usr/share/keyrings/intel-graphics.gpg


echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy/lts/2350 unified" | \
  sudo tee /etc/apt/sources.list.d/intel-gpu-jammy.list

sudo apt update
sudo apt install -y intel-i915-dkms=1.23.10.92.231129.101+i141-1

echo "deb [arch=amd64,i386 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy unified" | \
  sudo tee /etc/apt/sources.list.d/intel-gpu-jammy.list

sudo apt update
sudo apt install -y intel-level-zero-gpu=1.3.29735.27-914~22.04
sudo apt install -y level-zero=1.14.0-744~22.04 level-zero-dev=1.14.0-744~22.04
sudo apt install -y xpu-smi=1.2.33-52~22.04

在我采取所有步骤并重新启动后,屏幕是黑色的,无法进入正常窗口。

尝试将 ast 列入黑名单并设置 WaylandEnable=true,它也不起作用。

环境:英特尔至强 8480 384 DDR5 内存技嘉 MS73 主板 2 个 A770 GPU

如果是登录界面为黑色,把Wayland设置为false即可

drew-ye avatar Jun 26 '25 07:06 drew-ye

For the dark screen issue, you may need update mesa related package: apt install libegl-mesa0 libglapi-mesa libgl1-mesa-dri libglu1-mesa libglx-mesa0 mesa-vulkan-drivers

rnwang04 avatar Jun 27 '25 02:06 rnwang04