Experimental core to reduce overheating for large protein WU

Moderators: slegrand, Site Moderators, PandeGroup

Re: Experimental core to reduce overheating for large protein WU

Postby bruce » Mon Jun 22, 2009 5:12 am

Yes, my language was a bit loose when I said ". . .operate exactly like. . . ." (V1.19 needs the fix for checkpointing to make them operate the same way.) I was talking in terms of actual analysis results, and I believe that part is identical except for the heat/speed of the calculation.
bruce
 
Posts: 21067
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Experimental core to reduce overheating for large protein WU

Postby 7im » Mon Jun 22, 2009 1:43 pm

Someone should check to make sure that a WU altered in such a way midstream returns good data though.


PG typically does that BEFORE releasing a client or fahcore to beta testing. That's why updates are not released as immediately as some would expect.

There was a possibility to put the throttle into each client's config settings, but I think that for expediency they used the global version we have now.


Yes, I think I mentioned that a few posts back. ;)
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Experimental core to reduce overheating for large protein WU

Postby John Naylor » Mon Jun 22, 2009 4:45 pm

bruce wrote:One of the things that needs to be confirmed is that if you make no setting of an environment variable, v1.27 works exactly like v1.19.

Unless something different is happening under the surface, I'm using 1.27 with no variables and its folding away at the same PPD and temps as 1.19, so I guess its business as usual in this scenario.
Folding whatever I'm sent since March 2006 :) Beta testing since October 2006. www.FAH-Addict.net Administrator since August 2009.
User avatar
John Naylor
 
Posts: 950
Joined: Mon Dec 03, 2007 4:36 pm
Location: University of Birmingham, UK

Re: Experimental core to reduce overheating for large protein WU

Postby shdbcamping » Mon Jun 22, 2009 5:22 pm

John Naylor wrote:
bruce wrote:One of the things that needs to be confirmed is that if you make no setting of an environment variable, v1.27 works exactly like v1.19.

Unless something different is happening under the surface, I'm using 1.27 with no variables and its folding away at the same PPD and temps as 1.19, so I guess its business as usual in this scenario.

That's great news as well. I just want this to be "win-win" all the way around :D .
shdbcamping
 
Posts: 519
Joined: Mon Nov 10, 2008 7:57 am

Re: Experimental core to reduce overheating for large protein WU

Postby shdbcamping » Tue Jun 23, 2009 3:19 am

I just got some info from Mark. The only core that will be affected by the throttle settings at this point is the core_11 version 1.27. All others will not even know it's there :wink: . Also short term plan is for both the 1.19 and 1.27 versions to stay available. It's up in the air on where the 1.27 will finally end up before going mainstream. Both versions can run on the same system.
shdbcamping
 
Posts: 519
Joined: Mon Nov 10, 2008 7:57 am

Re: Experimental core to reduce overheating for large protein WU

Postby noprob » Wed Jun 24, 2009 7:30 pm

this is in reference to the FahCore_11.exe core; Project: 5758 (Run 10, Clone 663, Gen 27).

Though I haven't figured out how to properly "set" the variable, I am just running this newer version 1.27 core in place of the older version 1.19 as one of the Folding@Home beta testers state
Unless something different is happening under the surface, I'm using 1.27 with no variables and its folding away at the same PPD and temps as 1.19, so I guess its business as usual in this scenario.


My WU did not finish at 4% and returned w/error: EARLY_UNIT_END, when using the older version 1.19 I was already at 6% on this particular WU, so I will run this on one more WU and see what happens and if another sends an error I will just go back to the original 1.19 FahCore_11.exe core as most do at least finish or complete better than 85%.

What is a good temperature range for a Nvidia 8600GT?

Athlon 64 3200+ 1024MB DDR Ram is all stock no over clocks on anything,running XP Pro SP3 w/driver 186 for the Nvidia 8600GT
Image
User avatar
noprob
 
Posts: 78
Joined: Sun Mar 09, 2008 2:48 am
Location: mountains of West Virginia U.S.of A.

Re: Experimental core to reduce overheating for large protein WU

Postby v00d00 » Wed Jun 24, 2009 11:38 pm

Ideally nothing higher than 70C should be fine. Thats all my 8600GTS cooks at.
Image
User avatar
v00d00
 
Posts: 420
Joined: Sun Dec 02, 2007 4:53 am
Location: In the UK

Re: Experimental core to reduce overheating for large protein WU

Postby bruce » Thu Jun 25, 2009 2:59 am

v00d00 wrote:Ideally nothing higher than 70C should be fine. Thats all my 8600GTS cooks at.


NVidia says that they're good up to 105C but a lot of people choose much lower numbers. I think you'll have to make your own decision.
bruce
 
Posts: 21067
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Experimental core to reduce overheating for large protein WU

Postby db597 » Thu Jun 25, 2009 4:21 am

I can't see how NVIDIA can quote 105 degrees figures. I start to get EUEs once my GTS250 exceeds around ~86 degrees. Perhaps 105 is for gaming use where you don't mind some pixels being the wrong colour.
MSI GTX460 336SP 1GB | Intel Q6600 | 2GB DDR2
Palit GTX260 216SP 55nm | Intel E5200 | 2GB DDR2
db597
 
Posts: 123
Joined: Mon Dec 24, 2007 12:26 am

Re: Experimental core to reduce overheating for large protein WU

Postby 7im » Thu Jun 25, 2009 5:20 am

Like CPUs that overclock to different speeds, even from the same model, GPUs are similar in nature. Some have quoted EUEs at stock speeds at low temps, much below your 86. While others routinely run fah without error at around 100. I think 80-90 is what most have said is the typical comfort zone for top temps at top performance.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Experimental core to reduce overheating for large protein WU

Postby db597 » Thu Jun 25, 2009 9:55 am

Marking 105C that way is like having 4 bottles of milk:

Bottle 1: expires in 3.5 days time
Bottle 2: expires in 3.0 days time
Bottle 3: expires in 4.0 days time
Bottle 4: expires in 2.5 days time

So therefore put the BEST BEFORE date as 4 days time...
db597
 
Posts: 123
Joined: Mon Dec 24, 2007 12:26 am

Re: Experimental core to reduce overheating for large protein WU

Postby shdbcamping » Thu Jun 25, 2009 12:50 pm

noprob wrote:this is in reference to the FahCore_11.exe core; Project: 5758 (Run 10, Clone 663, Gen 27).

Though I haven't figured out how to properly "set" the variable, I am just running this newer version 1.27 core in place of the older version 1.19 as one of the Folding@Home beta testers state
Unless something different is happening under the surface, I'm using 1.27 with no variables and its folding away at the same PPD and temps as 1.19, so I guess its business as usual in this scenario.


My WU did not finish at 4% and returned w/error: EARLY_UNIT_END, when using the older version 1.19 I was already at 6% on this particular WU, so I will run this on one more WU and see what happens and if another sends an error I will just go back to the original 1.19 FahCore_11.exe core as most do at least finish or complete better than 85%.

What is a good temperature range for a Nvidia 8600GT?

Athlon 64 3200+ 1024MB DDR Ram is all stock no over clocks on anything,running XP Pro SP3 w/driver 186 for the Nvidia 8600GT

Your 8600gt wents inside the case correct? If so it will depend a lot on your case temps and case cooling solution you are employing. The brigdes (north/South) are affected a lot by temp and can cause EUE's as well during folding if they are running hot due to poor airflow. For the card 80-90C should be safe for 24/7 folding if you can keep other MOBO components cool. You are also running only 1 gig RAM so that will be heating up as well.

See if CPUID HWMonitor will pick up your case temps and report them. I have had no EUE problens with 1.27 thus far. But I keep my case temps pretty low because I have some serious CFM air movement :wink: .
shdbcamping
 
Posts: 519
Joined: Mon Nov 10, 2008 7:57 am

Re: Experimental core to reduce overheating for large protein WU

Postby 7im » Thu Jun 25, 2009 1:56 pm

I don't disagree... You're stating the obvious to the hard core GPU folders around here. But it's the GPU hardware companies you should be talking to about this. ;)
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Experimental core to reduce overheating for large protein WU

Postby noprob » Sat Jun 27, 2009 2:11 am

In reference to my earlier post all is well now as I had forgot to reboot for the variables to take hold in system environment.

The 10% reduction has applied according to my FAH log file,which I will change to 15 because I feel comfortable with that setting (I prefer max of 160F gpu temp.) ,but from the input it seems I can go higher and I thank everyone for this information.

I had The CUDA beta driver installed and once I updated the drivers to the stable 186 driver everything concerning this newer version 1.27 core is running smoothly.

again Thank You all for your valuable input.

*[Athlon 64 3200+ XP Pro SP3 1gig ram nVidia 8600GT 512 DDR3 GPU: 540 MHz, Shader: 1188 MHz, Memory: 700 MHz (6.14.11.9107 - nVIDIA ForceWare 191.07) all stock frequencies running between 53C to 72C avg. on video card using GPU-Z 0.3.4 to monitor and Folding@Home GPU Core Version 1.31 setting: "Calling fah_main args: 14 usage=85" air conditioned environment set at 67 to 74 degrees Fahrenheit depending on ambient temperature and WU]

*updated - edited on October 23,2009
Last edited by noprob on Fri Oct 23, 2009 7:13 pm, edited 10 times in total.
User avatar
noprob
 
Posts: 78
Joined: Sun Mar 09, 2008 2:48 am
Location: mountains of West Virginia U.S.of A.

Re: Experimental core to reduce overheating for large protein WU

Postby weedacres » Tue Jul 21, 2009 10:35 pm

shdbcamping wrote:
geokilla wrote:So....does it fix the lag issue where we lag in 2D programs? I know the lag's gone if we have Aero turned on in Windows, but some of us fold in Windows XP and what not.


I tried using it for that issue on an 8800GT in a system that was having a huge issue with that in XP32 bit. I set it up as 10%. The lag is not totally gone but it is much more responsive. I did not want to go above 10% as the Science lost would not be worth the difference in what stutter was left. I would assume that if the % was increased enough, the studder/lag could be eliminated.



I got a late start trying this out so many have probably already figured this out.
For me, the experimental core has solved a lot of usability problems. 2D performance is significantly improved. I was also getting regular BSOD's with certain applications so that I would have to shut down my gpu folding when using those applications. That is no longer a problem except with WMP. I'm set at 90% and while I noticed a small decline in gpu folding performance, it had given me back a much more usable machine.
XP32 Q6600 w/notfreds smp-4 and 2 8800gt's
XP32 X2-6000 w/notfreds smp-2 and 4 8800gt's.
Image
weedacres
 
Posts: 394
Joined: Mon Dec 24, 2007 11:18 pm
Location: Eastern Washington

PreviousNext

Return to NVIDIA specific issues

Who is online

Users browsing this forum: Majestic-12 [Bot] and 1 guest

cron