Hello, this is my first time on this forum, so please excuse me. I am running a GeForce GT 530 graphics card in my desktop (I know, it’s old but it serves its purpose) and I’m looking to get into CUDA development. According to this [url]https://www.geforce.com/hardware/desktop-gpus/geforce-gt-530-oem/specifications[/url], my card supports CUDA. But according to this [url]https://developer.nvidia.com/cuda-gpus[/url], I don’t see my card listed. But I do remember seeing on some other site (can’t remember where at the moment, so I don’t have a link) that my card does support CUDA 2.1. My first question would be is this information correct? Or would it be an older version?
But the more important question here doesn’t 100% matter on the specific CUDA version. As stated in the title, I am running Ubuntu 18.04 on my computer. Lets just say that the card does support up to 2.1 for the sake of asking the question. According to the download site [url]https://developer.nvidia.com/cuda-toolkit-21-january-2009[/url], the latest version of Ubuntu here is 8.04. This is a 10 year old version of Ubuntu! It is severely outdated and doesn’t receive security updates, so I don’t want to have to install such an old version of Linux. My second question is am I able to download the 8.04 drivers listed here and the CUDA toolkit here, or what do I do? I’m not sure where to start and I’ve searched the deep depths of Google (12+ pages down in fact) and haven’t found a single thing to guide me in the right direction. I do not want to have to buy a newer card just for the sake of exploring new ideas as my current one works fine with my Windows games (it runs Skyrim decently so I’m happy).
I thank anyone who takes the time to help me in advance.
The 2.1 refers to the compute capability of your card, not “CUDA version”. Compute capability is an architectural version. Newer GPUs generally have higher numbered compute capabilities, because they are newer architectures. But a particular CUDA version will support a wide range of compute capabilities.
Compute capability (cc) 2.1 GPUs are supported by CUDA versions up to 8.0. CUDA 9.x do not support cc2.1 GPUs.
If you want to install CUDA 8, your GPU is supported. However, CUDA 8 does not officially support Ubuntu 18.04 (no currently available CUDA toolkit up through 9.2 officially supports Ubuntu 18.04).
You’re welcome to try and get it working on Ubuntu 18.04 if you wish. There are various folks who have tried, and reported their methods, on these forums as well as elsewhere.
A compute capability 2.x GPU is also generally not a good choice for Deep Learning work, as most Deep Learning frameworks get their GPU acceleration via the CUDNN library, and the CUDNN library only supports GPUs of compute capability 3.0 and higher.
Thank you for taking the time to clarify my understanding of this topic. I guess I misunderstood something along the way. I will do some looking around into installing the CUDA 8 toolkit and see about getting it work on my setup.