Hot Loads !!!

If you think it might be a driver problem, see viewforum.php?f=79

Moderators: slegrand, Site Moderators, PandeGroup

Hot Loads !!!

Postby BrainEater » Wed Feb 22, 2012 12:58 am

Wow , you guys slipped magnum shells into the guns....

Project 8031 WU's are adding 60 watts or so to each of my 580's.I woke up this morning to the "Overload Beeps" of my ups.....6-10 degree rise per card.......

I DO like the points tho.

Did I miss the warning ?

:D
BrainEater
 
Posts: 17
Joined: Sat Mar 28, 2009 3:16 pm

Re: Hot Loads !!!

Postby bruce » Wed Feb 22, 2012 1:12 am

Nothing new on that front. We've always said that a protein size increases, GPUs will be more fully utilized.

And, yes, the new projects analyze bigger proteins, and it has been a while since the typical protein size has been increased.
bruce
 
Posts: 21336
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Hot Loads !!!

Postby Uniwarking » Wed Feb 22, 2012 1:29 am

Yeah, I'm not a fan of the heat and noise... thats why I don't do -advmethods. If these keep up I'm going to stop folding on my GTX 580...

8031-8033 are too much for standard users.
Uniwarking
 
Posts: 5
Joined: Tue Jan 31, 2012 10:48 pm

Re: Hot Loads !!!

Postby BrainEater » Wed Feb 22, 2012 1:38 am

I actually think I like em....

I'm doing the same points with 2 cards.....That's a big increase.

Only thing that did'nt like it was my UPS.Hawhaw

:mrgreen:
BrainEater
 
Posts: 17
Joined: Sat Mar 28, 2009 3:16 pm

Re: Hot Loads !!!

Postby Uniwarking » Wed Feb 22, 2012 1:48 am

I downclocked from 940 to 900 in order to drop my temps a few degrees and keep the fan at or under 80%....
Last edited by Uniwarking on Wed Feb 22, 2012 1:50 am, edited 1 time in total.
Uniwarking
 
Posts: 5
Joined: Tue Jan 31, 2012 10:48 pm

Re: Hot Loads !!!

Postby bruce » Wed Feb 22, 2012 1:50 am

Uniwarking wrote:Yeah, I'm not a fan of the heat and noise... thats why I don't do -advmethods. If these keep up I'm going to stop folding on my GTX 580...

8031-8033 are too much for standard users.


Whoever suggested that not doing advmethods would reduce the heat an noise gave you bad information. All new projects go through a series of steps designed to weed out unstable WUs. If a project successfully completes internal testing, it is sent to early beta testing. If it successfully completes early beta testing, it is sent to advmethods (late beta testing). If it successfully completes late beta testing, it is distributed to everyone.

Virtually all adv projects are migrated to full FAH distribution.

Are you overclocking your GPU? If so, you've mistakenly assumed that the projects you've been running up until now have been utilizing 100% of your GPU resources, overlooking the possibility of a new project moving you one step closer to 100%.
bruce
 
Posts: 21336
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Hot Loads !!!

Postby Uniwarking » Wed Feb 22, 2012 1:52 am

Perhaps I was somewhat misinformed, however, these WU's run 5 - 10 C higher then the WU's I've been running for weeks.

I have bumped my clock down a bit to help with temps and noise as stated above.
Uniwarking
 
Posts: 5
Joined: Tue Jan 31, 2012 10:48 pm

Re: Hot Loads !!!

Postby WiSK » Wed Feb 22, 2012 1:17 pm

bruce wrote:Whoever suggested that not doing advmethods would reduce the heat an noise gave you bad information.


Sure nobody official said it, but the empirical evidence made it true until now. My own experience was to try advmethods, observe that those projects are running too hot and laggy for my machine, so I switched off the setting. This has been fine so far, but now my only choice is to switch off the folding.

In your replies so far about this, bruce, it seems you are quite dismissive about this. Great that better programming can squeeze out 10% more processing power, but if more than 10% of people switch off their GPU folding because of this then has anything been won? So please, suggest to Pande Group to add a new setting -canrunhotandlaggy for folders with watercooled, non-primary-tasked GPUs and move these projects into that class for the assignment server to distribute :)
Image
WiSK
 
Posts: 26
Joined: Tue Dec 06, 2011 7:04 am

Re: Hot Loads !!!

Postby MtM » Wed Feb 22, 2012 1:25 pm

No WiSK, there been some cases where wu's were kept in advmethods because of similar concerns ( temperature ), at least this I can remember. But in general, all advmethods is, is a buffer between beta team work units and regular f@h work units. All work units go through these steps, this is a rule and is why you'll find the people explaining it trying to be very clear about that.

What you are suggesting is basically that -advmethods is going to be used like a separate group of work units.
MtM
 
Posts: 3054
Joined: Fri Jun 27, 2008 2:20 pm
Location: The Netherlands

Re: Hot Loads !!!

Postby WiSK » Wed Feb 22, 2012 1:56 pm

No, I explicitly suggested a new setting -canrunhotandlaggy so that the current official definition of -advmethods can remain intact.
WiSK
 
Posts: 26
Joined: Tue Dec 06, 2011 7:04 am

Re: Hot Loads !!!

Postby GreyWhiskers » Wed Feb 22, 2012 2:41 pm

The 762x projects were, and still are, kept as Advmethods after a similar hue and cry when they came out. That gave the GPU folders some control over their destiny.
User avatar
GreyWhiskers
 
Posts: 767
Joined: Mon Oct 25, 2010 5:57 am
Location: Saratoga, California USA

Re: Hot Loads !!!

Postby MtM » Wed Feb 22, 2012 3:52 pm

GreyWhiskers wrote:The 762x projects were, and still are, kept as Advmethods after a similar hue and cry when they came out. That gave the GPU folders some control over their destiny.


Right, and that's why I said made the post saying that this was a request to change the advmethods flag into something it was not meant to be.

@WiSK

A new flag can not be added to existing clients.
MtM
 
Posts: 3054
Joined: Fri Jun 27, 2008 2:20 pm
Location: The Netherlands

Re: Hot Loads !!!

Postby bruce » Wed Feb 22, 2012 4:39 pm

FAH has dealt with GPU overheating problems for some time now and at least one of the FahCores was modified so a donor could limit their FAH production to, say, 90% of its capacity by inserting idle time into the FAH jobstream. (With GPUs in adjacent slots, the middle one did run hotter than the outer two). The capability of adjusting for a reduced performance setting doesn't work with all the FahCores. There are a couple of enhancement requests to improve this situation, but they involve multiple development groups so I don't predict that changes will come quickly.

If I seem dismissive, it's because Stanford University really has no control over how you choose to overclock your hardware or what fan profile you decide to use. I'm afraid that it's the only good choice available to you right now. Threatening to quit does not lead to a good outcome for anyone.

For years, we've had overclockers who would complain that their CPU is 100% stable and it can run prime95 for long periods of time. They choose to ignore the fact that prime95 does not adequately assess the heat generated by a heavy floating point load and their machine IS UNSTABLE. I don't have a similar example for GPUs, except maybe these proteins, but the same concept applies.

Creating a new class of proteins that are less able to use all of your resources isn't likely to happen. It would require changes to the client, changes to the assignments sever logic, and a rational definition of how to measure what "hotandlaggy" (utilization) means on a wide variety of GPUs. You won't get the same answer with 80 shaders or with 800 or 8000. Ultimately a lot depends on the drivers, too, and AMD/NV change them rather frequently, making maintainnce of this class of WUs a continual process which is likely to be neglected since ultimately experts in biochemistry/molecular simulation/protein modeling are not the folks that will be benchmarking new GPUs or even reading the articles written by those who do.

FAH will gradually finish the simulation of small proteins and concentrate more and more on new, larger proteins which will increase the productivity (and heat) of your hardware. Science progresses slowly and it's a gradual progress, but every time a study of a new family of projects is started, we may just be facing the same issues again, and I'm not sure there is a long-term solution.
bruce
 
Posts: 21336
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Hot Loads !!!

Postby mdk777 » Wed Feb 22, 2012 5:25 pm

See this thread:

viewtopic.php?f=67&t=20880&p=208759#p208759

Just went back to a 6802 WU and temp. on the card went down 10 C.
560 TI, no OC.

I don't know the answer, but lagginess is also significant.
I don't care, but at the point that it becomes difficult to use the machine and fold at the same time, I anticipate that others will.
Transparency and Accountability, the necessary foundation of any great endeavor!
mdk777
 
Posts: 813
Joined: Fri Dec 21, 2007 4:12 am

Re: Hot Loads !!!

Postby DocJonz » Wed Feb 22, 2012 6:49 pm

mdk777 wrote:......
I don't know the answer, but lagginess is also significant.
I don't care, but at the point that it becomes difficult to use the machine and fold at the same time, I anticipate that others will.

I agree. As I noted in the other thread when the 803x came through yesterday (GPU3 Lag), this will cause an issue for some people - even those who don't overclock their GPU's ... I don't OC mine, and they've all increased in temps significantly, and the screen lag (which is a real issue) has returned.
This will be even more of an issue for multiple GPU Folders - I've got some machines with two GPU's, and the temps are such that I'm contemplating reducing the Folding down to one GPU per machine as a result .....
- Please don't mistake these comments as some sort of threat to withdraw* - they are purely practical observations which I think many others will be drawing for themselves.

Perhaps this issue would be of interest in discussion by the DAB - we've not heard from them for a while .....

(* I've been Folding everyday bar power cuts, internet loss etc since the summer of 2005, and this issue won't fundamentally change that!)
User avatar
DocJonz
 
Posts: 354
Joined: Thu Dec 06, 2007 6:31 pm
Location: United Kingdom

Next

Return to V6 GPU3 beta (including Fermi) OpenMM

Who is online

Users browsing this forum: No registered users and 2 guests

cron