![]() |
|
|
#1 |
|
"Svein Johansen"
May 2013
Norway
3·67 Posts |
As most of you know, I have a few titan cards, and I started playing around with Cuda 5.5 tonight, and I saw the installation came with driver called NVIDIA RELEASE 320.59.
I immediately noticed a great 20% speed on the Titans. Iteration times from 4.2-4.4 to 3.1-3.2. I installed accidentally the latest beta GTX driver, and then I was back to the normal level I had before. Just wanted you all to know, that the Kepler GPUs might have a great performance boost with this driver. I dont know about Fermi cards since I dont have one. Last fiddled with by Manpowre on 2013-08-07 at 22:07 Reason: error |
|
|
|
|
|
#2 |
|
"Mr. Meeseeks"
Jan 2012
California, USA
23×271 Posts |
Looks like people are whining and complaining that BF3 and Metro 2033 has lower performance and artifacts. Well, it is beta though is what they don't realize.
|
|
|
|
|
|
#3 |
|
May 2013
East. Always East.
32778 Posts |
It is interesting to see the differences between hard computing vs graphics. New driver shows 20% increase in computation but adds artifacts and decreases performance. Strange and interesting... Good for us, I guess. Hang onto those drivers.
|
|
|
|
|
|
#4 | |
|
"Svein Johansen"
May 2013
Norway
20110 Posts |
Quote:
now this is weird. First I installed Cuda 5.5.. and saw a driver being installed.. and started to playing around with it and code, compiling etc... then I noticed 20% better iteration times.. I thought mabye they did something to drivers and then I installed latest beta. I installed the beta, and measured that, equals the old driver I used. Then I installed cuda 5.5 again and got the driver I got so good results with.. and iteration times were the same as the beta.. what the heck. |
|
|
|
|
|
|
#5 |
|
Sep 2010
So Cal
2·52 Posts |
Well, all I can say from my own personal experience is that the latest released NVIDIA Linux x86_64 driver, which is 325.15, has finally solved the issue of wrong residual outputs on my Asus Titan running Cudalucas 2.03.
Previous drivers including the 325.08 beta apparently did not perform as a fully compatible Titan driver, hence the spurious results either on 64 bit DP mode or 32 bit SP.Edit: Recently installed Cuda 5.5 which seems to run about 5 % faster than Cuda 5.0 :-) Last fiddled with by Robert_JD on 2013-08-08 at 09:58 Reason: Additional info |
|
|
|
|
|
#6 |
|
"Svein Johansen"
May 2013
Norway
C916 Posts |
|
|
|
|
|
|
#7 | |
|
Sep 2010
So Cal
2·52 Posts |
Quote:
After that, then I compiled the Cudalucas 2.03 source code, and testing it out, I got an error message concerning missing "libcuda.so" shared libraries. So as a fix, I simply went into ld.so.conf and typed in "/usr/local/cuda-5.5/bin", then "SAVE", and then typed in "ldconfig" at the terminal and the issue was solved. ![]() P.S. I'm currently using Fedora 18. |
|
|
|
|
|
|
#8 |
|
Mar 2010
19B16 Posts |
|
|
|
|
|
|
#9 | |
|
Sep 2010
So Cal
2·52 Posts |
Quote:
|
|
|
|
|
|
|
#10 |
|
"Carl Darby"
Oct 2012
Spring Mountains, Nevada
32·5·7 Posts |
I'm liking the new Linux driver too.
|
|
|
|
|
|
#11 |
|
"Carl Darby"
Oct 2012
Spring Mountains, Nevada
32×5×7 Posts |
Here's the end of stage 2 and the start of stage 1 with CUDAPm1.
|
|
|
|
![]() |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Nvidia announces Volta based Titan V | tServo | GPU Computing | 8 | 2017-12-10 16:04 |
| Nvidia unlocking Titan X (P) through drivers | Karl M Johnson | GPU Computing | 3 | 2017-08-11 08:40 |
| Nvidia announces Titan Xp card | ixfd64 | GPU Computing | 10 | 2017-05-17 15:19 |
| Nvidia announces Titan X | ixfd64 | GPU Computing | 20 | 2015-04-28 00:27 |
| 2x AMD 7990 or 2x Nvidia Titan ?? | Manpowre | GPU Computing | 27 | 2013-05-12 10:00 |