What is the developer driver?

[font=“Courier New”]To develop NVIDIA CUDA programs, the tools needed are:

  1. Developer drivers
  2. CUDA Toolkit
  3. GPU Computing SDK

What is this developer driver? Installing it silently overrides any NVIDIA graphics driver that is already installed. So, how is the developer driver different from the usual NVIDIA graphics driver? What extra components or functionality does it contain?

The developer driver is available in both 32-bit and 64-bit flavours. If I am developing 32-bit Visual C++/CUDA projects on 64-bit Windows 7, which flavour of the developer driver do I choose? Why?[/font]

You’ll notice that when you follow the “developer driver” link, it just takes you to the normal driver page. All normal NVIDIA drivers support CUDA - the developer driver is just a way to point to the oldest driver version that supports that version of CUDA.

Sometimes, when a new CUDA version has just come up, the developer driver links to a beta driver newer than the current stable one.

Hey there,

It makes me think about an issue i got.

I have a computer with a Quadro, a Geforce and a Tesla plugged in. How should i install this machine properly ? Because it seems that the developer driver is only meant to be installed for Geforce products…

The thing is that if i get a bare installation, with no driver, i cannot install the driver if i haven’t the Geforce plugged in.

One more thing, since 263.03 driver (i tried 263.06, 266.58 also), i get crashes if i use the Quadro display, random crashes requiring a reboot. If i use the C2050 display everything’s ok…

Anyway, thanks for any tips.

What about purely CUDA capable devices like Tesla? Those do have different NVIDIA drivers than those for graphics cards? How is the common CUDA developer driver related to these then?

To put it in different terms, this is a quote from the web page: http://developer.nvidia.com/object/cuda_3_2_downloads.html

cU

In other words, to be sure that you are getting the latest CUDA driver that supports your hardware, go to www.nvidia.com/drivers and fill out the form with your hardware and OS details.

In general, I’ve found that the latest version of the Tesla driver is the same as the latest GeForce driver running the same silicon. On windows, there is the TCC driver - but you’ll have to get someone else to answer questions on that. I don’t run CUDA on windows.

Packaging-wise, TCC driver = standard driver as of R260. (that was several months of my life to get that)

Thanks DrAnderson42 and tmurray for your input. I hope this summary is correct:

    [*]Both the NVIDIA graphics driver and the CUDA developer driver are derived from the same source and ship with the same CUDA support. That is, the NVIDIA graphics driver has all the CUDA capabilities that are in the CUDA developer driver. There are no extra developer features in the CUDA developer driver that are missing in the NVIDIA graphics driver.[*]The CUDA developer driver tries to support a wider set of hardware, i.e., the set of all possible CUDA capable devices.[*]The NVIDIA graphics driver (which you download after choosing your hardware in an online form) is typically aimed at a subset of the devices (one of which is your hardware).[*]The CUDA developer driver is based off the NVIDIA graphics driver version available during its release. Since new CUDA versions are released once in 6 months or more, over time the CUDA developer driver will lag behind the latest NVIDIA graphics driver.

Or put another way, if you are a CUDA developer, just get the latest NVIDIA graphics driver for your device, that is right choice. :biggrin: