What is the common rate of success for laptop GPU folding

A forum just for SPCR's folding team... by request.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Thorsted
Posts: 11
Joined: Sat Jun 13, 2009 12:46 pm
Location: Denmark

What is the common rate of success for laptop GPU folding

Post by Thorsted » Tue Nov 17, 2009 4:20 pm

For starters I found it to be pathetic.

Using my Dell xps m1730 with 8700m gt sli I get like 9 completed units in 2 weeks, I don't have the number but many unstable_machine units unfinished.
32 shaders should give several units per day, depending on size off course.

When beginning temperature of gpu was around 70 C , nothing unusual for a gpu to my knowledge, but not able to finish every unit properly.
Slight underclock to keep temps at 65 C - no noticed improvement.
SLI switched on - no help (didn't expect since its not supposed to help/work).
Underclock to 600 MHz on shaders - no improvement.
Did not have latest driver in the beginnig but last ones did not improve it.

Bottom line it will not work for me, anyone else have some success?

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

interesting

Post by cordis » Wed Nov 18, 2009 11:00 am

Well, it's interesting, I've been thinking about getting a laptop with something from the gtx 260/280 mobile series, but now I wonder. You may be overestimating the speed of 32 shaders a little, I have a couple machines with 9400/ION chips and they do about 600ppd, so 32 should give you around 1200ppd, which will finish a few of the smaller ones a day (~353 points each), but it won't churn through the big 1888 pt WUs all that quickly. But yeah, if you're seeing unstable machine messages, then it sounds like something is up. You might try just cutting the clocks down dramatically to see if that helps, I'm not sure how low rivatuner will let you go, but get it as low as possible and see if that helps. If so, then you can push it back up little by little and see if there's a threshold where it works consistently. If not, then it may just be bad hardware. But it's interesting to hear about laptop folding, keep us posted on your progress.

And hey, welcome to the team! Thanks for helping out!

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Nov 18, 2009 2:29 pm

Can you see GPU temps while you fold?

It would be great if you could get inside the thing, find the gpu, and slap a small copper heatsink on it.

The other question, being a laptop, is are they trying to run voltages lower than stock to save power? Are there tools that can bump voltages?

You should not get EUEs at stock speeds, period.

Going back to basics, can you run Prime 95 successfully?

Is this an AMD CPU/chipset?

Thorsted
Posts: 11
Joined: Sat Jun 13, 2009 12:46 pm
Location: Denmark

Post by Thorsted » Wed Nov 18, 2009 2:42 pm

Yeah obviously a 1888 point unit wont finish in a day, according to FahMon for the same project
A 353 point wu would give 1620ppd on the 8700m gt and 5350ppd on a regular 9800gt.
For a 797 point it would give 1350ppd on the 8700m gt and 4330ppd on a regular 9800gt.
Have not had a 1888 point unit on the laptop.

In the beginning I did set up two GPU instances, first would run like ''normal'' but the second one would fail at like 1,2 or 3 % every time, abandoned that quick. So I am suspecting hardware is not that stable, stock speed of core/shader is 625/1250 lowered to 300/600 with one unit going was not really successful. Games do run nicely. I am beginning to think perhaps OS install has been tinkered too much and need a fresh start. With a sample size of one its hard to draw conclusions.

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

ouch

Post by cordis » Wed Nov 18, 2009 2:57 pm

Yeah, cutting speed in half and still not getting success isn't good. I'd be tempted to try something like pointing a fan at the closest external vent to the thing to see if heat is really the issue, although at half speed I can't imagine it would be. I thought I had read about an nvidia video memory test program, have you tried that?

Thorsted
Posts: 11
Joined: Sat Jun 13, 2009 12:46 pm
Location: Denmark

Post by Thorsted » Wed Nov 18, 2009 3:07 pm

Seems we typed at the same time aristide1 :)

Yes the temps show in speedfan, rivatuner, etc. temp should not be too high as the fan doesn't speed up (fan in steps off-low-high) until 75C and average running is 69-70, one strange thing is the exact same temp for both cores so something is not correct. I found a picture somewhere that showed the whole thing as a unit so i think its like a regular x2 core board. Voltage regulation is non-existant Dell bios is so limited, and Rivatuner does not support mobiles properly, have not found anything to tweak with. Speedfan cannot read voltages of the gpu or motherboard.

Cpu is a core2duo T7700 2,4GHz cipset is PM965

I didn't try Prime95 as its gpu not cpu might give it a go, but I think the OS is needing a refresh as first step at the moment

Thorsted
Posts: 11
Joined: Sat Jun 13, 2009 12:46 pm
Location: Denmark

Post by Thorsted » Wed Nov 18, 2009 3:09 pm

omg i type slow....

I will have a look for that memtest

Thorsted
Posts: 11
Joined: Sat Jun 13, 2009 12:46 pm
Location: Denmark

Post by Thorsted » Thu Nov 19, 2009 2:30 pm

Ok a short update
With a fresh os install and latest nvidia drivers to get cuda, success is not exactly within immediate reach.
First unit at 68% -> unstable_machine
next one at 9% -> Early_unit_end
third one at 60% so far

Downloading latest Dell drivers is version 176.78 - no cuda support
CUDA is implemented in 177.xx
Nvidia latest drivers for mobile is 186.81
Which for me sorta hints the hardware is not up for the task...
A few more wasted units and I will abandon it for sure, does early end units/unstable machine units give any value for the project?

If i get any success i will update for sure.

Post Reply