[elrepo] Nvidia Driver Performance: El Repo vs. NVidia's .run file

[Manuel Wolfshant]

On 04/25/2016 05:19 PM, Manuel Wolfshant wrote:
> On 04/25/2016 05:06 PM, EBradley at williams-int.com wrote:
>>
>> Hello,
>>
>> I’ve just recently begun using the Nvidia drivers from el repo 
>> installed via yum as opposed to the manually installed .run packages 
>> downloaded from Nvidia.com for our engineering workstations running 
>> RHEL 6. While I love the fact that the el repo drivers do not need to 
>> be reinstalled after every kernel upgrade, one of my users has just 
>> shown me a valid case in which the performance of the el repo Nvidia 
>> drivers is different from the .run package.
>>
>> First, some basic info regarding the environment:
>>
>> User’s System:  Dell Precision T7810
>>
>> Video Card: Nvidia Quadro K620
>>
>> OS: RHEL 6.6
>>
>> Kernel: 2.6.32-573.22.1.el6.x86_64
>>
>> Driver - el repo (361.42): kmod-nvidia.x86-64 nvidia-x11-drv.x86_64 
>> nvidia-detect.x86_64
>>
>> Driver - Nvidia.com: NVIDIA-Linux-x86_64-361.42.run
>>
>> Nouveau driver - blacklisted in /etc/modprobe.d/blacklist.conf
>>
>> I’ll try to be descriptive of the situation without being long-winded:
>>
>> When viewing the results of an analysis in Ansys Mechanical with the 
>> el repo driver, a part is displayed and can be manipulated at will 
>> when viewing the general results. If, however, the user selects one 
>> of the ‘Total Deformation’ selections within the results tree, the 
>> part disappears from the Mechanical window. Conversely, when the el 
>> repo drivers are uninstalled and the Nvidia.com driver is installed 
>> from the .run package, the Total Deformation part is displayed as 
>> expected, showing the regions of varying color gradients depicting 
>> the areas of more/less deformation over the surface of the part. I 
>> was able to reproduce this behavior on my test machine, a Dell 
>> Precision T5500 with an Nvidia Quadro K600.
>>
>> I certainly don’t want to come off like I’m blaming the el repo 
>> driver or anything. Rather, I’m interested to know if anyone has 
>> experienced anything similar (I’ve checked the previous year of posts 
>> and will keep working backward) and also if there’s something I may 
>> have missed or misconfigured when installing the drivers via el 
>> repo/yum vs. manually installing the .run package. Basically, did I 
>> goof or is there a valid difference between the el repo drivers and 
>> the .run package from Nvidia.com?
>>
>> Thanks in advance, and I am happy to provide more detail/logs/answer 
>> questions/etc. And my apologies for any legal jargon that is appended 
>> to this message, it is automatically-generated and outside of my control.
>>
> Hello
>
>     The first thing that pops to mind is to run a ls -lR / before 
> installing any of the drivers, after installing the elrepo package and 
> ( after restoring the system to the pristine status )  after 
> installing via .run.  Then do a diff3 using all three file lists and 
> pastebin ( or attach ) the diffs

Second thing:  a comparison of the Xorg.0.log files generated when using 
the two installation methods would also be interesting.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.elrepo.org/pipermail/elrepo/attachments/20160425/8c88b51e/attachment.html>