# Installation issue **URL:** \ **Category:** General\ **Created:** [July 3, 2025, 4:59pm UTC](https://discuss.vllm.ai/t/installation-issue/1002 "2025-07-03T16:59:27Z")\ **Posts on this page:** 9\ **Page:** 1
**Author:** ![passin](https://avatars.discourse-cdn.com/v4/letter/p/c89c15/32.png) [@passin](https://discuss.vllm.ai/u/passin)\ **Post date:** [July 3, 2025, 4:59pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/1 "2025-07-03T16:59:27Z")
Hello everyone, i tried to install vLLM today by using the installation process of this [link](https://docs.vllm.ai/en/latest/getting_started/installation/cpu.html), but i got an error. During the installation, i got this message : – Found Torch: /home/me/.venv/lib/python3.12/site-packages/torch/lib/libtorch.so CMake Error at cmake/cpu\_extension.cmake:164 (message): vLLM CPU backend requires AVX512, AVX2, Power9+ ISA, S390X ISA or ARMv8 support. Call Stack (most recent call first): CMakeLists.txt:97 (include) By the way, i would like to use vLLM with the CPU. Thanks for reading this and have a good day. ---
**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\ **Post date:** [July 3, 2025, 4:59pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/2 "2025-07-03T16:59:46Z")
The error means your CPU does not support any of the required instruction sets: AVX512, AVX2, Power9+, S390X, or ARMv8. vLLM’s CPU backend requires at least one of these ISAs to build and run; older CPUs without them are not supported for vLLM CPU inference or serving, and there is no workaround documented in the official sources. You can check your CPU’s supported instructions with `lscpu | grep -i avx` or similar commands. For more details, see [vLLM CPU installation requirements](https://docs.vllm.ai/en/latest/getting_started/installation/cpu.html) and related [GitHub issues](https://github.com/vllm-project/vllm/issues/4692), which confirm this requirement and the lack of support for CPUs without these instruction sets. Would you like a step-by-step explanation or help checking your CPU’s capabilities? Sources: - [vLLM CPU installation requirements](https://docs.vllm.ai/en/latest/getting_started/installation/cpu.html) - [GitHub issue: Installation with CPU with errors](https://github.com/vllm-project/vllm/issues/4692) - [GitHub issue: Compiling VLLM for cpu only.](https://github.com/vllm-project/vllm/issues/5317) - [https://github.com/vllm-project/vllm/issues/13662](https://github.com/vllm-project/vllm/issues/13662) ---
**Author:** ![passin](https://avatars.discourse-cdn.com/v4/letter/p/c89c15/32.png) [@passin](https://discuss.vllm.ai/u/passin)\ **Post date:** [July 3, 2025, 5:01pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/3 "2025-07-03T17:01:59Z")
> [@RunLLM](#): > > lscpu | grep -i avx Got it, i tried this command, but nothing appeared ---
**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\ **Post date:** [July 3, 2025, 5:02pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/4 "2025-07-03T17:02:20Z")
If `lscpu | grep -i avx` returns nothing, your CPU does not support AVX, AVX2, or AVX512 instructions. vLLM requires at least one of these (or Power9+, S390X, or ARMv8) for CPU inference, so your hardware is not compatible with vLLM’s CPU backend and you cannot run vLLM on this machine’s CPU as per the official documentation and error messages. There is no supported workaround for CPUs lacking these instruction sets. See [vLLM CPU requirements](https://docs.vllm.ai/en/latest/getting_started/installation/cpu.html) and [related GitHub issues](https://github.com/vllm-project/vllm/issues/4692) for confirmation. Would you like advice on alternative hardware or LLM inference solutions for older CPUs? Sources: - [GitHub issue: Installation with CPU with errors](https://github.com/vllm-project/vllm/issues/4692) - [vLLM CPU requirements](https://docs.vllm.ai/en/latest/getting_started/installation/cpu.html) ---
**Author:** ![passin](https://avatars.discourse-cdn.com/v4/letter/p/c89c15/32.png) [@passin](https://discuss.vllm.ai/u/passin)\ **Post date:** [July 3, 2025, 5:08pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/5 "2025-07-03T17:08:53Z")
> [@RunLLM](#): > > with vLLM’s CPU backend and you cannot run vLLM on this machine’s CPU as per the official documentation and error messages. There is no supported workaround for CPUs lacking these instruction sets. Are there any other vLLM usable version for my CPU ? ---
**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\ **Post date:** [July 3, 2025, 5:09pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/6 "2025-07-03T17:09:27Z")
No, there are no versions of vLLM that support CPUs lacking AVX, AVX2, AVX512, Power9+, S390X, or ARMv8 instruction sets. All official vLLM CPU backends require at least one of these ISAs, and there is no documented workaround or legacy version for older CPUs without them, as confirmed in the [official documentation](https://docs.vllm.ai/en/latest/getting_started/installation/cpu.html) and [GitHub issues](https://github.com/vllm-project/vllm/issues/4692). Would you like recommendations for alternative LLM inference libraries that support older CPUs? Sources: - [vLLM CPU requirements](https://docs.vllm.ai/en/latest/getting_started/installation/cpu.html) - [GitHub issue: Installation with CPU with errors](https://github.com/vllm-project/vllm/issues/4692) ---
**Author:** ![passin](https://avatars.discourse-cdn.com/v4/letter/p/c89c15/32.png) [@passin](https://discuss.vllm.ai/u/passin)\ **Post date:** [July 3, 2025, 5:13pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/7 "2025-07-03T17:13:06Z")
Yeah, i would like it ---
**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\ **Post date:** [July 3, 2025, 5:13pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/8 "2025-07-03T17:13:24Z")
For older CPUs without AVX/AVX2/AVX512, you can try [llama.cpp](https://github.com/ggerganov/llama.cpp), which is designed for maximum compatibility and can run on CPUs without advanced vector instructions. It is widely used for LLM inference on legacy hardware and supports various quantization options for efficiency. Would you like step-by-step instructions for installing and running llama.cpp? Sources: - [https://github.com/vllm-project/vllm/issues/627](https://github.com/vllm-project/vllm/issues/627) - [https://github.com/vllm-project/vllm/pull/2244](https://github.com/vllm-project/vllm/pull/2244) ---
**Author:** ![passin](https://avatars.discourse-cdn.com/v4/letter/p/c89c15/32.png) [@passin](https://discuss.vllm.ai/u/passin)\ **Post date:** [July 3, 2025, 5:17pm UTC](https://discuss.vllm.ai/t/installation-issue/1002/9 "2025-07-03T17:17:40Z")
No thanks. Also, thank you for you answers