My system is a Dell XPS15 9560 laptop with GTX1050 which run Ubuntu 16.04.1.
I tried to install CUDA toolkit 8.0 but it failed to install the graphic driver.
error:
The driver installation is unable to locate the kernel source. Please make sure that the kernel source packages are installed and set up correctly.
If you know that the kernel source packages are installed and set up correctly, you may pass the location of the kernel source with the '--kernel-source-path' flag.
I tried,
sudo apt-get install linux-headers-$(uname -r)
but the system cannot locate linux-header-4.4.0-62-generic.
Is this the problem of Ubuntu 16.04.1?
However, I also have checked here https://developer.nvidia.com/cuda-gpus where I could not find GTX1050 listed as one of the CUDA GPU’s. So I wonder if the GTX1050 laptop version chip will be supported later in the near future or not?
GTX1050 is supported by CUDA, no worries, that’s not the issue. Your issue is with downloading the appropriate linux kernel headers, which for whatever reason your ubuntu install is not finding in the repositories.
I should also point out, if you attempt to install the NVIDIA driver supplied with the toolkit on your laptop, it will more than likely produce a black screen upon bootup due to the default driver install not considering NVIDIA’s Optimus configuration on Linux. To avoid this, you’d want to install it from Ubuntu’s package manager, by also including nvidia-prime and the required CUDA libs, like so:
That is the same link that OP already included in their original question. It seems to me the GTX 1050 for laptops is a recently added SKU, and the list is simply lagging behind reality a bit.
CUDA 8.0 should work just fine with the GTX 1050 as long as you have the latest drivers installed, as vacaloca has pointed out. Getting the latest driver for laptops is sometimes a bit challenging, depending on the vendor.
For some reason I thought it had been released very recently. I did a quick internet search, the GTX 1050 for laptops was apparently introduced in January of this year. Based on historical observation, NVIDIA aren’t exactly a shining example for a meticulously updated web presence, so I am still leaning towards documentation error. Feel free to file a bug to get it corrected.
You lost me there. What kind of policy are you referring to?
What about the way NVIDIA supports CUDA developers strikes you as unusual? NVIDIA has a whole bunch of highly-qualified people whose job it is to support CUDA developers. If you attend next month’s GTC in San Jose, you will be able to meet many of them in person. If you are interested in becoming part of that team, it seems they are hiring: [url]Workday
Updating the web presence, operating these forums, and providing download packages is a separate task handled by a different group of people.
I would claim that the strong support NVIDIA extends to CUDA developers (an approach carried over from their graphics business) has been one of the key factors in the success of CUDA.
CUDA supports all NVIDIA desktop and laptop GPUs of compute capability 2.0 or higher, released since about 2009 or 2010.
The GTX 1050 laptop model is supported. The cuda gpu list just hasn’t been properly updated for that model, i.e. it is an oversight. It doesn’t mean that CUDA doesn’t support that model of GPU in that laptop.
Hi, I got the dell xps 15 today and followed the CUDA installation steps. There is a point where it says the driver is not found early on but recommends you keep your driver and install the other parts. It installed ok and the the deviceQuery and bandwidthTest pass, so yes I agree I think it’s just a documentation issue you’ve found.
i have a gtx 1050, so after these comments if i continue the installation of CUDA with this warning, is it work properly???
i am installing cuda 8 on win 10.
I just bought a Dell XPS15-9560 (the call it P56F ??)
I was able to install CUDA 9 packages, regardless of warning, and use VisualStudio 2017 Community Edition to compile the nbody.exe sample.
It reports finding one 1050 GPU and runs; same executable (plus DLLs) fails on other systems, where they cannot find an NVIDIA GPU.
It must be noted that some MS and NVIDIA directories had to be added manually into the nbody project and between double quotes as the path contains blanks.
This is though just the first sample, I’ll see if everything else works.