[SOLVED] Nvidea

Ernst0 · 03-05-2012, 06:11 PM

http://forums.fedoraforum.org/showthread.php?t=204752

F16,F15 & F14 Nvidia driver guides

-----------------

I do not see a guide RHEL6 to installing the video driver for this GTX550TI gfx card.

Will following the same steps for F16 work?

A link would be nice. I'll keep looking.

John VV · 03-05-2012, 08:46 PM

is there a question here

i am guessing that it would be
" how to install the 295 nvidia.run driver on RHEL6 "

if so you MUST ( a 100% need ) to have a paid for license ( NO EXCEPTIONS )
then see the red hat knowledgebase
the FULL instructions are in there , just run a search
https://access.redhat.com/knowledge/search

also if you do opelgl codding with MESA you also need to reinstall all of mesa after installing the .run driver

Ernst0 · 03-05-2012, 09:10 PM

Quote:

Originally Posted by John VV

is there a question here

i am guessing that it would be
" how to install the 295 nvidia.run driver on RHEL6 "

if so you MUST ( a 100% need ) to have a paid for license ( NO EXCEPTIONS )
then see the red hat knowledgebase
the FULL instructions are in there , just run a search
https://access.redhat.com/knowledge/search

also if you do opelgl codding with MESA you also need to reinstall all of mesa after installing the .run driver

Yes Sir, 100% paid license and may I add that of all the Linuxs' the RHEL6.2 is what I expect of RedHat.

I believe I want to have dual gfx cards and learn to use them as GPU processors from C
Why not? I think a matched pair of 550s' should offer a nice base.

And I thank this site as well. So far it hasn't been all that bad once I added a network card to replace the MB one.

Okay then! I'll follow the link. Thanks.

Ernst0 · 03-05-2012, 09:21 PM

Hey John..

May I ask questions?

Using a GPU to program with is something I was surprised to learn about recently.

So I have the GTX550TI and I plan on a second so they can work together I take it.

I figure that these two will provide a fair training experience yes?

The Tesla's look cool but are not deserved at this time. I'll have to earn them with good works.

So I need to find that driver you mention? Then reinstall the mesa..

Cool .. I need to look these things up.

I take it you have use GPU for more than graphics?

Tell me is it like independant processors returning outout all the time? If it is written that way yes.

Care to chat a bit on your take on GPU co-processing? I'm all ears and eyes here.

Thanks

Edit.. is the stuff for Tesla cards the same as for GTX550TI?

Edit2 ...

I have that .run

I don't see a how to but I will look up Mesa next.

Edit3..

Quote:

A variety of device drivers allows Mesa to be used in many different environments ranging from software emulation to complete hardware acceleration for modern GPUs.

Do I need the MESA stuff? I don't plan on graphics programming only number crunching.

As you can tell I am a noob asking for advice.

Source for reading here : http://www.mesa3d.org/

Source for driver GTX550TI http://www.nvidia.com/Download/index.aspx?lang=en-us

I'll look for guidence.

But if you will John can you share a bit on using the GPU's for number crunching?
I admit I need only the basics and on that I also will be learning OpenMP I believe.

Fantastic things these moder day come computers are. Power of a 1970 main frame on the desk.

Again thanks

John VV · 03-05-2012, 10:10 PM

I is an old issue with nvidia
in order for them to install an every Nix platform they use THEIR own gl.so and headers .The *.run replaces the GL/gl.h and a few others
do some research or googling

it is an old problem that only affects about 1% of users

normally one would do this

Code:

su -
yum search mesa 
--- from that list ---
yum install mesa mesa-devel mesaglw mesaglw-devel

there also might be mesaglu and mesaglu-devel

now once these are installed THEN FOR EVERY YES EVERY TIME you reinstall the .run you will need to run

Code:

su -
yum reinstall mesa*

and you MUST REINSTALL the *.run for EVERY kernel, x11,xorg update
EVERY TIME !!!
you must MANUALLY reinstall it .

as for GPU processing
see: CUDA
http://developer.nvidia.com/gpu-computing-sdk

Now getting CUDA built and installed is a bit beyond the scope of this thread .There are forums and threads FULL of information on that

Ernst0 · 03-06-2012, 12:55 PM

That is very helpful. I am interested in the Cuda processing. As I wrote my main focus is numbers and number theory.

Again thanks...

cascade9 · 03-06-2012, 01:52 PM

Quote:

Originally Posted by Ernst0

I believe I want to have dual gfx cards and learn to use them as GPU processors from C Why not? I think a matched pair of 550s' should offer a nice base.

You dont need dual cards to use CUDA. Dual cards, either SLI/Crossfire or independent is more of a pain than its worth IMO.

Its also not worth buying another GTX550 IMO. The GTX 560 is not much more, will be almost as fast as 2 x GTX550s (possibly faster if you have a 2nd PCIe x16 slot limited to x8/x4), and will use less power and output less heat. GTX 560 Ti is even better, but at a slightly higher cost.

GTX 550- 192 CUDA cores, 900MHz core, 192bit, 116 watts TDP. $120-180 US.
GTX 560- 336 CUDA cores, 810MHz core, 256bit, 150 watts TDP. $180-240 US.
GTX 560 Ti ('normal' version)- 384 CUDA cores, 822MHz core, 256bit, 170 watts TDP. $220-280 US.

Quote:

Originally Posted by Ernst0

I take it you have use GPU for more than graphics?

Yes, but its not a CPU. Some things run very well on GPUs. Most things dont.

Ernst0 · 03-06-2012, 02:04 PM

I have found the CUDA install section at RH and I have found the forums as well.

I hit a snag in the install and am asking for help there.

Not bad progress for my second day as a RHEL6 user.

This CUDA processing is rather cool stuff. Makes me want a Tesla card but I will do fine with the GTX550TI to start.

Ernst0 · 03-06-2012, 02:14 PM

Quote:

Originally Posted by cascade9

You dont need dual cards to use CUDA. Dual cards, either SLI/Crossfire or independent is more of a pain than its worth IMO.

Its also not worth buying another GTX550 IMO. The GTX 560 is not much more, will be almost as fast as 2 x GTX550s (possibly faster if you have a 2nd PCIe x16 slot limited to x8/x4), and will use less power and output less heat. GTX 560 Ti is even better, but at a slightly higher cost.

GTX 550- 192 CUDA cores, 900MHz core, 192bit, 116 watts TDP. $120-180 US.
GTX 560- 336 CUDA cores, 810MHz core, 256bit, 150 watts TDP. $180-240 US.
GTX 560 Ti ('normal' version)- 384 CUDA cores, 822MHz core, 256bit, 170 watts TDP. $220-280 US.

Yes, but its not a CPU. Some things run very well on GPUs. Most things dont.

Thanks!!!!!
I kinda figured there are limitations to what it can do and honestly, I don't know what I want it to do yet either. But, it is there so set it up.

What I am looking at is learning to use the OpenMP plus the GPU coding. It is what is ahead to learn. It should be enough for a while.

You have to admit today's desktop computers are "way-cool." How can I not set up the CUDA?

LOL..

So, yeah, if and when I am good at CUDA programming I may well step up to a better model.
If a rich relative leaves me a pile of cash then maybe even a Tesla

Yeah that is a dream.. I would need a good reason to spend that much.