|
|
Pages: [1]
|
 |
|
Author
|
Topic: Longer Load Time with New GPU Board (Read 1406 times)
|
Purple Rabbit
Squire
Offline
Posts: 33
|
I don't have a problem, but I am curious. Santa gave me (OK, I specified what I wanted) a new GTX 260 board to replace my 8600GTS in my Q6600 Win Vista computer. The old GPU card took 20 seconds (CPU time) to initialize and then off it went. The GTX 260 takes 30 seconds (CPU time). I'm kind of assuming that there is more "stuff" to be sent to the more capable card before it does its thing. All tasks are successfully completing with the new GPU card so I'm quite happy with the 5x increase in speed. I may have just answered my own stupid question, but inquiring minds want to know  I'm running what the 0.2 installer gave me for SETI (including CUDA).
|
|
|
|
« Last Edit: 02 Jan 2010, 04:42:37 pm by Purple Rabbit »
|
Logged
|
|
|
|
|
Jason G
|
Check you grabbed the Cuda 2.3 DLLs & dropped them over the 2.2 ones the installer would have put in. Hovering the mouse over the dll files (cudart.dll & cufft.dll) in the project folder should show a tooltip containing the version number. An improved installer with these optional DLLs included is in the process of development ( taking time because of other issues.) As an aside, another thing you *might* be seeing with a new card, and so new drivers, *could* be a result of having innocently updated/installed your display drivers to the latest 195 series whilst installing the new card. I don't know about others, but I see a performance drop (and more importantly stability) in Win7x64 with the latest drivers, and have now reverted to 191 series. That's with a 9600GSO.
It is possible the dirvers are doing 'more stuff' ... but exactly whatever that is, it hasn't justified either the compromised system stability or performance drop for me.
|
|
|
|
« Last Edit: 02 Jan 2010, 05:22:09 pm by Jason G »
|
Logged
|
|
|
|
Purple Rabbit
Squire
Offline
Posts: 33
|
Thanks Jason. I loaded the CUDA 2.3 dlls eons ago for the old card after the 0.2 installer did its thing. Sorry I didn't mention it before...sigh. I didn't load any new software for the new card. Potato, http://setiathome.berkeley.edu/show_host_detail.php?hostid=4367681 (I'm a wabbit!) seemed to still have what I needed. I assume nothing has gone South (or North for you  ) since then. I have confirmed that I still have 191 series of drivers. Nothing seems to have changed except the new hardware Like I said,this question is no big deal. Everything is working just fine, but why is it working this way?  Rick
|
|
|
|
|
Logged
|
|
|
|
|
Jason G
|
... Like I said,this question is no big deal. Everything is working just fine, but why is it working this way?  .... Mmmm, with those in the clear I don't really know, other than perhaps the higher compute capability of the new card ( 1.3 I presume). activates some extra Cuda preparation of other library functions/kernels. So then the answer might be yes, the startup could be sending larger/better kernels to the card I guess. There is some limited compute capability 1.3 specific code in there that I never played with...
|
|
|
|
« Last Edit: 02 Jan 2010, 06:17:44 pm by Jason G »
|
Logged
|
|
|
|
Purple Rabbit
Squire
Offline
Posts: 33
|
OK, I'll assume that the answer to my question remains a mystery of the universe, unknowable to all  If I stay up at night wondering about this, it's your fault  Rick
|
|
|
|
« Last Edit: 02 Jan 2010, 06:48:24 pm by Purple Rabbit »
|
Logged
|
|
|
|
Pappa
Alpha Tester
Knight o' The Round Table
 
Offline
Posts: 215
|
You might consider "reinstalling" the 191.xx driver package. As nvidia went to a more universal installer of a larger size, it could be that refreshing and insuring that features of the "newer card" are properly being installed/utilized.
Regards
|
|
|
|
|
Logged
|
|
|
|
Purple Rabbit
Squire
Offline
Posts: 33
|
Thanks Pappa,
I'll try that. I think everything is running like I thought, but a reset to known good values can't hurt.
The initial 10 second discrepancy on start up is almost inconsequential, but the engineer in me had to ask why.
|
|
|
|
|
Logged
|
|
|
|
Purple Rabbit
Squire
Offline
Posts: 33
|
Thanks again to Pappa. Reinstalling the drivers (actually I put in the new 195.62) solved the "problem". Being just a Silly Wabbit I thought (dangerous!) that swapping one board for another where both used the same drivers would be a plug-and-play. Clearly it wasn't...sigh
Edit: This turned out to not be a wall clock time problem (they were about the same for multiple tasks), but a CPU efficiency problem. The CPU was being used (so it says) much more than needed. This is now (I think) been remedied with the reinstall of the drivers.
Rick
|
|
|
|
« Last Edit: 06 Jan 2010, 08:19:15 pm by Purple Rabbit »
|
Logged
|
|
|
|
Pappa
Alpha Tester
Knight o' The Round Table
 
Offline
Posts: 215
|
Things that happen at the OS level (users never see) tend to confuse things. More than once I have been forced to swap hardware (due to hardware failure) and on restart seen various problems even though it is the "Same Type of Device." History has taught me that when playing with hardware changes "Boot into the Safe Mode," and then resintall drivers. Reboot and things are normally happy, the OS sees the hardware and then automajakly knows where the drivers are .
|
|
|
|
|
Logged
|
|
|
|
Gizbar
Squire
Offline
Posts: 29
|
See you've gone with the 195.62 drivers. I've just taken them off my 2 rigs, and gone back to version 191.07. The 195's were unstable on my Win7-64HP rig, but seemed ok on the Vista-32HP rig.
Kept failing with 'Nvidia driver has stopped responding but successfully recovered' message. Every time it failed, it would forget the WU it was working on, and start a new one. I had about 60 part finished tasks, even though there is a GTX260 and an 8800GTX in there. When it failed 3 times in an hour earlier, I'd had enough.
Went back to the older version, no fails, and it is now reducing the 'part-finished' WU's by completing them.
regards, Gizbar.
|
|
|
|
|
Logged
|
|
|
|
|
Pages: [1]
|
|
|
|
Quote!
There is a theory which states that if ever anybody discovers exactly what the Universe is for and why it is here, it will instantly disappear and be replaced by something even more bizarre and inexplicable. There is another theory which states that this has already happened.- Douglas Adams
|
 |  |  |
| |
| Site Statistics |
| Total Members: | 123 |
| Total Posts: | 29,786 |
| Total Topics: | 892 | | Downloads |
| Apps |
| Windows R-1.x | 0 |
| Windows R-2.0 | 0 |
| Windows R-2.2 | 0 |
| Linux 32bit 1.x | 0 |
| Linux 32bit 2.2 | 0 |
| Linux 64bit 2.2 | 0 |
| Alpha/IA64 | 1,938 |
| FreeBSD | 0 |
| HPUX | 0 |
| Subtotal: | 0 |
| Source packs: | 5,803 |
| Tool/WU packs: | 10,078 |
| Total: | 22,048 | | GBs dl'd: | 309.53 | | Pages served |
| Today: | 6,715 |
| Total: | 8,668,319 |
| (since 6/26/2006) |
| 173 Donations to S@H |
| U.S. Dollars: | 3,196.59 |
| Euros: | 863.90 |
| Last 24h: | $ 0.00 |
| Avg./24h: | $ 3.32 |
| Estim. total: | $ 4,319.66 |
Latest Member: Miep |
| |
 | |  |
 |  |  |
| |
Online users/last 15m
37 Guests, 5 Users
Raistmer, Ghost, Tye, Claggy, Vyper 43 Members/last 24hRaistmer, Ghost, Tye, Claggy, Vyper, perryjay, Jason G, SciManStev, k6xt, Morten, arkayn, Slawek, cristipurdel, benool, Frizz, Purple Rabbit, sunu, Wild6-NJ, corsair, M_M, Franz, PatrickV2, JohnDK, _heinz, cenit, Josef W. Segur, glennaxl, skildude, msattler, mr.mac52, Geek@Play, Gizbar, Devaster, WHRoeder, kit344, Byron Leigh Hatch @ team Carl Sagan, TouchuvGrey, Metod, S56RKO, Questor, VoidPilot, The Grinch, hiamps, Pepi
| |
 | |  |
|