# Recommendations for using GPU on old hardware
**URL:** https://discourse.julialang.org/t/recommendations-for-using-gpu-on-old-hardware/77257
**Category:** GPU
**Created:** [March 1, 2022, 7:39pm UTC](https://discourse.julialang.org/t/recommendations-for-using-gpu-on-old-hardware/77257 "2022-03-01T19:39:33Z")
**Posts on this page:** 4
**Page:** 1
### Author:  [@quailman1031](https://discourse.julialang.org/u/quailman1031)
#### Post date: [March 1, 2022, 7:39pm UTC](https://discourse.julialang.org/t/recommendations-for-using-gpu-on-old-hardware/77257/1 "2022-03-01T19:39:33Z")
Is there a summary somewhere of what packages or package versions support CUDA capability levels?
I have a Fermi-architecture card (GeForce GT 550M) which is compute capable 2.1 and CUDA driver 9.1 (latest available). I have tried all the way back to CUDA.jl v1.0.0 which issues an error about AddrSpacePtr (under Julia 1.5.2) and CUDA.jl v1.1.0 which issues “ERROR: Device capability v2.1.0 not supported by available toolchain.”
Are there other libraries that would allow me to work with my GPU from julia?
---
### Author:  [@jling](https://discourse.julialang.org/u/jling)
#### Post date: [March 1, 2022, 7:53pm UTC](https://discourse.julialang.org/t/recommendations-for-using-gpu-on-old-hardware/77257/2 "2022-03-01T19:53:08Z")
> [@quailman1031](#):
>
> Fermi-architecture
> **[Support Matrix - NVIDIA Docs](https://docs.nvidia.com/deeplearning/cudnn/support-matrix/index.html)**
>
> These support matrices provide a look into the supported versions of the OS, NVIDIA CUDA, the CUDA driver, and the hardware for the NVIDIA cuDNN 8.9.6 release.
I think you need a new hardware…
> **[CUDA Compatibility :: NVIDIA Data Center GPU Driver Documentation](https://docs.nvidia.com/deploy/cuda-compatibility/index.html#binary-compatibility__table-toolkit-driver)**
>
> CUDA Compatibility document describes the use of new CUDA toolkit components on systems with older base installations.
ctrl-F Fermi shows nothing
---
### Author:  [@jmair](https://discourse.julialang.org/u/jmair)
#### Post date: [March 2, 2022, 8:04am UTC](https://discourse.julialang.org/t/recommendations-for-using-gpu-on-old-hardware/77257/3 "2022-03-02T08:04:22Z")
OpenCL tends to run on pretty much any hardware, even old cards. The Julia bindings are a bit more difficult to learn and use but it could be worth trying.
---
### Author:  [@jmair](https://discourse.julialang.org/u/jmair)
#### Post date: [March 2, 2022, 8:12am UTC](https://discourse.julialang.org/t/recommendations-for-using-gpu-on-old-hardware/77257/4 "2022-03-02T08:12:32Z")
It may also be worth looking at ArrayFire.jl
[https://github.com/JuliaGPU/ArrayFire.jl](https://github.com/JuliaGPU/ArrayFire.jl)
There is a more detailed description of the library on its homepage:
[https://arrayfire.com/](https://arrayfire.com/)
It seems to support for targeting OpenCL, which means it should also run on your hardware, but having easier to use bindings.