Can someone clarify as to why when you download an NVIDIA driver, there are multiple INF’s inside the display driver folder whereas if you download it from the Manufacturer Site e.g. dell, there is only one inf file?
Are these infs suppsoed to correspond to different manufacuters?
Hi blackops786187, the main reason is that the infs describe detailed instructions for installation per each GPU, per each variant of a GPU, and even Dell typically have multiple inf files for their different groups of products (like consumer, professional, datacenter, notebok or add-in etc.). The Nvidia-posted generic driver always aims at supporting as many of our GPUs in the field as possible, so we combine all infs for as many customers=GPU variants as possible into that posting, hence the many inf files vs. one or just a few in a single customers posting…
Thanks for clarifying. Explains what I was seeing in the Dell/lenovo driver packs the manufacturer provides :)
I’ve got a subsequent question regarding the display driver.
How does the setup.exe determine if the driver package is applicable to a machine? I ask this because when you run the setup exe on a machine which doesn’t have a compatible gpu it throws an error and gives a non zero return code. Does it parse the infs ?
We sometimes deploy newer nvidia gpu drivers through our MDM tool and try to tie the application down so it runs attempts only machines which have specific pnp HWID present (which correspond to a GPU Model). We do this so our reports for the application deployment arent filled with failures
Currently I’m determining if a HWID Is ok by simply running it on a test machine and making a note of the gpu HWID present in the device manager page. Ideally I would like to extract all of the HWID the display driver package is applicable to
Feel free to disregard this question if the logic behind the driver installer is confidential :)
no problem :-). I’m not the right person to give a conclusive full answer, but here’s what I know:
YES, one element of check for our installer to make sure it fits the system is the check of Plug’n’play Device IDs.
So if a DevID in the system is not found in any of the infs of a driver package, the installer will fail gracefully, rather than trying to either force the driver onto the system, or mainly just run the copy files part of the installation, and hence just fill our storage space…
So you could indeed parse the infs files for a match with the HW in your environment. Though, we try to support as many as possible of our GPUs=DevIDs in our generic driver packages posted on nvidia.com, so the number of GPUs not matching our package should be rather small (definitely smaller than an OEMs package…). That is one benefit we consider our packcages shall bring :-).
feel free to come back with more questions…