Advanced search

Message boards : Graphics cards (GPUs) : Changing driver from 314.07 to 314.22

Author Message
Rantanplan
Send message
Joined: 22 Jul 11
Posts: 166
Credit: 138,629,987
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30817 - Posted: 13 Jun 2013 | 13:07:30 UTC

Hello, i had read somewhere there is an performance increasement by 2-2.5 percent by changing drivers to the newer version.

Could that be true, is there a example out there, GPGPU performance tests ?

Anything ?

Greetings

Rantanplan
Send message
Joined: 22 Jul 11
Posts: 166
Credit: 138,629,987
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30823 - Posted: 13 Jun 2013 | 18:09:44 UTC - in response to Message 30817.
Last modified: 13 Jun 2013 | 18:15:35 UTC

ok, no information, but isn'd bad.

I installed the new drivers, looks very good, but gpu usage on nathan is only 87%.

I calculatet the ending time , and ist quiet similar to my old results time, but
i saved the extra power.

Can i configure boinc to use the rest of 13% for a second wu ?

app_config.xml i know but i want use 87% 1st wu and 13% 2nd wu , is that possible ?

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30828 - Posted: 13 Jun 2013 | 19:32:42 UTC - in response to Message 30823.

app_config.xml i know but i want use 87% 1st wu and 13% 2nd wu , is that possible ?

No way. That's CUDA and nVidias driver deciding what to run at which point in time. BOINC just launches the executables but can't regulate anything further.

MrS
____________
Scanning for our furry friends since Jan 2002

Rantanplan
Send message
Joined: 22 Jul 11
Posts: 166
Credit: 138,629,987
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30830 - Posted: 13 Jun 2013 | 19:37:01 UTC

thx for the "furry ones" for that wise advice. keep crunshing

Rantanplan
Send message
Joined: 22 Jul 11
Posts: 166
Credit: 138,629,987
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30841 - Posted: 14 Jun 2013 | 8:48:51 UTC
Last modified: 14 Jun 2013 | 8:49:42 UTC

Changeing drivers was a bad idea, now i am back to 314.07, and whats up ?

Again got my full load of gpu usage .

cheers

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30842 - Posted: 14 Jun 2013 | 9:38:45 UTC

I guess the saying goes with graphics drivers, as well as anything else:

If it ain't broken, don't fix it!
____________

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30846 - Posted: 14 Jun 2013 | 15:22:11 UTC

What about WU times on that host? Any improvements?

MrS
____________
Scanning for our furry friends since Jan 2002

Rantanplan
Send message
Joined: 22 Jul 11
Posts: 166
Credit: 138,629,987
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30850 - Posted: 14 Jun 2013 | 16:41:21 UTC

I calculated that, any wu will end with the same timeamout like the other wus, just calculated, there is no enhancement so far.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30851 - Posted: 14 Jun 2013 | 18:18:28 UTC - in response to Message 30850.
Last modified: 14 Jun 2013 | 18:20:50 UTC

If you have a good setup, driver updates won't improve the app performance at GPUGrid. When new apps are used (such as those recompiled and debugged to move from cuda 3.1 to 4.2) then people using old drivers (3.1) needed to update for the new app to work. Those already using a 4.2 driver got a speed bump from the app.
So, for here driver updates may bring stability, but no speed gain. It's been that way since I started crunching here (and I've tested dozens of driver updates).
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30853 - Posted: 15 Jun 2013 | 11:26:48 UTC - in response to Message 30851.

How I understood Ratanplans report is this:
- 314.07 with full GPU utilization
- 314.22 with 85% GPU utilization

This could have been some massive difference (not being reported before by anyone here). But since the WU times didn't change I think we can dismiss it.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30856 - Posted: 15 Jun 2013 | 18:08:34 UTC - in response to Message 30853.
Last modified: 15 Jun 2013 | 18:26:07 UTC


- 314.07 with full GPU utilization
- 314.22 with 85% GPU utilization

MrS

If you formed that opinion based on "Again got my full load of gpu usage", then I would say to me it isn't clear that Rantanplan was getting 99% GPU utilization from 314.07. Perhaps to Rantanplan ~93% is 'full GPU utilization'?
When changing driver it's a good idea to uninstall the likes of MSI Afterburner and then reinstall the latest version.
There have been several types of Nathan WU recently, and CPU usage does influence performance somewhat. I'm presently running two different Nate WU's and one used 85% of the GPU while the other used 88% (while running 6 CPU WU's). This rose to 88% and 90% when just running 3 CPU WU's, and 91 and 93% with all CPU WU's suspended (power usage also rose).
Also, if you are using the GPU for something else, the utilization would change...
I don't think I've seen 99% GPU utilization on W7 for any GPUGrid WU with any driver, but its the RunTime that matters anyway, (GPU utilization is just an indicator). If anyone is seeing 99% utilization with 314.07 perhaps it's a reporting difference - NVidia have made lots of power related changes to their drivers:

    314.22 - This release fixes a bug that allowed applications to override the optimum voltage applied by NVIDIA's GPU Boost technology. The voltage depends on the thermal environment of the system, and provides additional application stability. Because of this bug fix, you may find a difference in voltage offsets compared to previous driver releases

    Support for Hybrid Power, a Hybrid SLI technology, is discontinued and not available with this driver...


It's always good to challenge predefined ideas though - things can change.

BTW, and just looking at a few WU's, Rantanplan's GTX670 performance is about,
9% faster than my GTX660Ti when running the NATHAN_KIDc22_full WU's,
6% faster when running NATHAN_KIDc22_noPhos WU's,
3 to 6% faster running the NATHAN_KIDc22_SODcharge WU's.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30857 - Posted: 15 Jun 2013 | 18:44:35 UTC - in response to Message 30856.

Agreed!

MrS
____________
Scanning for our furry friends since Jan 2002

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30859 - Posted: 15 Jun 2013 | 21:07:32 UTC - in response to Message 30857.
Last modified: 15 Jun 2013 | 21:15:09 UTC

I've been using 314.22 on an XP system with a GTX470, and have some reasonably interesting results.

Effect of temperature control and system usage on GPU run time/performance (XP, 7.0.64, GTX470, no CPU tasks)

5 short SANTI_bax1 WU’s were run on a GTX470 with the GPU fan speed set to 2 different settings (very high, high). GPU temperatures ranged from 71°C to 83°C, GPU clocks remained the same. Reducing the fan rate from 84% to 75% caused GPU temps to rise from71°C to 82°C and PCB temps rose from 50°C to 57°C:


High GPU Fan rate (75%) + occasional light system usage
:
I550-SANTI_bax1-12-62-RND6766_1 4522773 15 Jun 2013 | 15:22:36 UTC 15 Jun 2013 | 19:57:05 UTC Completed and validated 16,193.69 1,181.72 19,200.00

Very High Fan rate (84%) no system usage:
I305-SANTI_bax1-15-62-RND2218_2 4520212 15 Jun 2013 | 7:01:08 UTC 15 Jun 2013 | 11:10:57 UTC Completed and validated 14,773.25 1,139.20 19,200.00

Very High fan rate (84%) + occasional light system usage:
I695-SANTI_bax1-11-62-RND3882_0 4521517 15 Jun 2013 | 2:44:49 UTC 15 Jun 2013 | 7:03:47 UTC Completed and validated 15,366.25 1,217.64 19,200.00
I969-SANTI_bax1-12-62-RND8721_0 4522517 15 Jun 2013 | 11:07:10 UTC 15 Jun 2013 | 15:26:10 UTC Completed and validated 15,309.89 1,173.81 19,200.00

High GPU Fan rate (75%) + no system usage:
I917-SANTI_bax1-11-62-RND1813_0 4521286 14 Jun 2013 | 22:20:36 UTC 15 Jun 2013 | 2:47:59 UTC Completed and validated 15,913.81 1,179.19 19,200.00

All Short runs (2-3 hours on fastest card) v6.52 (cuda42)

These short tasks take about the same amount of time on my GTX650TiBoost (give or take a few seconds):
I931-SANTI_bax1-8-62-RND4360_0 4521906 15 Jun 2013 | 4:44:10 UTC 15 Jun 2013 | 8:39:12 UTC Completed and validated 13,207.76 13,059.90 19,200.00
I584-SANTI_bax1-8-62-RND2813_0 4521547 15 Jun 2013 | 1:03:47 UTC 15 Jun 2013 | 4:58:50 UTC Completed and validated 13,199.83 13,067.90 19,200.00
I798-SANTI_bax1-7-62-RND6565_0 4521314 14 Jun 2013 | 21:23:21 UTC 15 Jun 2013 | 1:18:39 UTC Completed and validated 13,196.02 13,061.58 19,200.00
I984-SANTI_bax1-15-62-RND2199_0 4520877 14 Jun 2013 | 17:43:21 UTC 14 Jun 2013 | 21:38:32 UTC Completed and validated 13,199.99 13,061.20 19,200.00
I227-SANTI_bax1-10-62-RND4329_0 4520345 14 Jun 2013 | 14:03:04 UTC 14 Jun 2013 | 17:58:17 UTC Completed and validated 13,192.88 13,051.96 19,200.00
I750-SANTI_bax1-10-62-RND8395_0 4519889 14 Jun 2013 | 6:42:29 UTC 14 Jun 2013 | 10:38:06 UTC Completed and validated 13,204.29 13,057.58 19,200.00
I389-SANTI_bax1-6-62-RND7521_0 4519775 14 Jun 2013 | 10:22:45 UTC 14 Jun 2013 | 14:18:11 UTC Completed and validated 13,192.88 13,046.38 19,200.00
I893-SANTI_bax1-9-62-RND2261_0 4519200 13 Jun 2013 | 23:22:44 UTC 14 Jun 2013 | 3:17:35 UTC Completed and validated 13,198.84 13,063.14 19,200.00
I508-SANTI_bax1-11-62-RND0477_0 4518866 14 Jun 2013 | 3:02:15 UTC 14 Jun 2013 | 6:57:50 UTC Completed and validated 13,202.08 13,069.61 19,200.00
I837-SANTI_bax1-12-62-RND1017_0 4518172 13 Jun 2013 | 16:02:10 UTC 13 Jun 2013 | 19:57:16 UTC Completed and validated 13,187.07 13,044.21 19,200.00
I545-SANTI_bax1-11-62-RND6015_0 4517965 13 Jun 2013 | 12:21:50 UTC 13 Jun 2013 | 16:16:44 UTC Completed and validated 13,188.70 13,040.15 19,200.00
I499-SANTI_bax1-12-62-RND7133_0 4517750 13 Jun 2013 | 8:41:24 UTC 13 Jun 2013 | 12:36:41 UTC Completed and validated 13,192.67 13,045.60 19,200.00
I169-SANTI_bax1-10-62-RND7672_2 4514412 13 Jun 2013 | 19:42:30 UTC 13 Jun 2013 | 23:37:22 UTC Completed and validated 13,193.90 13,050.12 19,200.00
I209-SANTI_bax1-6-62-RND5967_0 4517216 13 Jun 2013 | 5:01:51 UTC 13 Jun 2013 | 8:56:35 UTC Completed and validated 13,196.80 13,057.62 19,200.00

Outcome – similar tasks took different lengths of time to complete under different system usage conditions and with different fan settings/GPU temps. The overall performance range was 9.6%

It looks like temperature control has a role to play in RunTime, even for yesteryear cards, and that system usage also makes a difference.
I don’t really have the time to repeat this and I want to have a quick look at XP vs W7 again (with the same rig and drivers)…
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30861 - Posted: 16 Jun 2013 | 7:42:28 UTC - in response to Message 30859.

I'd expect system usage to have some impact. However, measuring this would be almost futile, since we normally generate "random" useage, which can't easily be repeated.

The temperature factor looks interesting, but I'd say we need more statistic to judge this. And most importantly: how big is the regular variance of run times on this rig?

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Graphics cards (GPUs) : Changing driver from 314.07 to 314.22

//