Those little annoying F@H things ..

"ouch" huh ?

guess as the teams real big time SMP folder "ouch" refers usually to a 300k+ point lost.

I think over the last few years I lost a solid 2 to 3 million points due to upload problems...

The last minor lost for me was today. I lost about 10k because of a folding mistake with the last WU. You will see what I mean when the other P8105 finishes... :rolleyes:

A few years ago I folded -bigadv on some virtual machines and can't tell you how many points were lost on 3 i7 9xx comps. -bigadv was in it's infancy then, similar to GPU QRB today. Look at the thing cross-eyed and WU was gone....

Yeah, it was a 8900.

Looks like it was lost, I have just checked it and from what I can tell, the client is folding the same work unit again, with an estimated PPD of 60000.

Bummer

I have had that happen on SMP WUs 99.99% complete but the client has died, haven't seen that on GPU WUs.

Run HFM.NET monitoring software and it may show where the progress was when the client died.
 
I have had that happen on SMP WUs 99.99% complete but the client has died, haven't seen that on GPU WUs.

Run HFM.NET monitoring software and it may show where the progress was when the client died.

I let the WU run to the end, for the time it would take to do almost 2 8900s I got +/- 18000 points rather than the 50000... but these things need doing.

Once that was finished I checked my clocks and all seems stable. I pushed for 1.2ghz on the core but Unigine Heaven showed up a few graphical glitches. I probably wouldn't have noticed them if it wasn't for my obsession a few years ago trying to get my computer optimized and looking for any small graphics glitch when running 3D Mark 2001.

Tried some benchmarks and five mins in a couple of games to test stability and glitches and all seems ok at 1.15ghz.

One question though, do you know if memory speed helps this current crop of GPU work units. I know it didn't use to back when I was folding on my then fairly new 8800GT and that has been the rule of thumb I have carried forward, not really thinking it may have changed but that was some time ago now.
 
These Core 17 WUs seem to be pretty sensitive to overclocking stability and may require some experimenting to achieve the best overclock.

I leave the memory clocks at the factory setting.
 
These Core 17 WUs seem to be pretty sensitive to overclocking stability and may require some experimenting to achieve the best overclock.

I leave the memory clocks at the factory setting.

I am very careful when it come to overclocking. Unlike some who think if it doesn't crash its ok. I don't want anything producing any errors.

I kinda learnt that the hard way back in 2000/2001 when I didn't think about stability testing. I used winrar to compress a load of stuff, later on when I went to access it the files were damaged. Since then my goal has been to balance between speed, stability and noise.

People that don't think furmark is a very good test these days, I disagree. If you can hit a GPU core that hard and it doesn't throw any errors then that's a good sign.

I wouldn't use Furmark by its self as it doesn't load the VRAM or the memory controller but I do include it (or MSI Kombuster) in my range of testing.

My 7950 is currently eating through the Work Units at 1150 core 1600 Memory and the estimated PPD is 94000.

The memory is harder to check as it has built in error correction, so I found a speed that didn't show up any errors (1700mhz) then pushed it harder untill I got some nice on screen patterns (1750mhz) then chucked in a nice margin or error and settled with 1600mhz)

At the lower end I don't think memory speed increase really helps too much, but I wouldn't be supprised if going from 1250mhz to 1600mhz takes a couple of seconds off the frame time to make it worth while at +/- 4min 25 seconds TFP (90000+ PPD)

With my current settings my estimated TPF is 4 min 18 seconds. That's with no CPU folding.
 
I should probably have said that the only OC check I do is to accomadate F@H. With thr number of machines that I have I have to get a OC that I don't have to baby sit, it has to run all of the currently available WU's without error.
 
I should probably have said that the only OC check I do is to accomadate F@H. With thr number of machines that I have I have to get a OC that I don't have to baby sit, it has to run all of the currently available WU's without error.

I have had the luxury to keep checking in on the PC but really my biggest concern is sending back results with errors in.
 
I have had the luxury to keep checking in on the PC but really my biggest concern is sending back results with errors in.
I think Stanford/PG has a pretty good handle on that now. If a WU completes and is accepted by the receive server things are probably good.

I frequent the Folding Forums and have heard little to no talk about faulty/corrupted WU being accepted or causing problems lately
 
after another failed WU I decided to dismiss the OC.

Over the last week I lost 2 or 3 WUs due to some weird error. Maybe it's too hot, maybe my system is over loaded. I don't know and I don't care at this point anymore. I flashed the right bios on that system and I'll let it fold stock. Which also means that the memory is now running on 9-9-9-24

I have no idea how badly it's gonna hurt my PPD but I can see myself loosing the #1 spot in the daily production quite easily.
 
after another failed WU I decided to dismiss the OC.

Over the last week I lost 2 or 3 WUs due to some weird error. Maybe it's too hot, maybe my system is over loaded. I don't know and I don't care at this point anymore. I flashed the right bios on that system and I'll let it fold stock. Which also means that the memory is now running on 9-9-9-24

I have no idea how badly it's gonna hurt my PPD but I can see myself loosing the #1 spot in the daily production quite easily.

I have heard that the current AMD chips are more sensitive to high temperature than Intels. But I really doubt it's heat related, esp when you are running server class equipment which is suppose to be tolerant of hotter conditions.

Unless you are willing to subject the hardware to a lot of stress testing, which is going to eat into time more than anything else, you probably wont be able to find where the error is being generated. Which is a shame.

Fingers crossed that you don't lose too much PPD and that the errors stop.
 
Honestly: I am just too lazy to trouble shoot. I highly doubt it's a big problem because I folded with that setup set up just like this for basically a 100 million points. So I know it has to be something minor.

But I just don't feel like pushing the system anyways. Right now I have 50W less consumption. I'm at 650W which is still a lot. It says my peak in terms of PPD should be in the 600k PPD range. But I usually am 50k under the peak-range. So I will drop back to the #2 spot in terms of Folding@Home contribution in this team most likely... Still the 550k PPD I ought to produce on daily bases should be just fine.
 
Having to slow down clocks/settings a little after running F@H for a period time is not uncommon. Just slow things down alittle and find a new stable clock. Not all WU's will handle the same overclock, so sometimes that plays into things too.

I'm just finally getting my stuff re-tuned for the 8900 WU's, had to slow things down a bit after running 7810/7811 WU's. PPD should start going up nicely if Stanford keeps handing out 8900 WUs.....
 
Yeah. The difference at these speeds is unfortunately quite big. I'm crunching on a P8101 now and the TPF went up from 11:30 to 12:30. This means now the bad big WU results in about 385k PPD instead of 450k PPD. The bad thing really is that it takes now the over 20h to fold. So I guess my points will be low this week; however I'm hoping the next week will be fine after that...

But I figure that I will somehow remain up there and that me falling back for a 150k PPD over all won't stop the team maintaining the 2 million PPD mark.
 
Last edited:
Arg!!!

The wife stacked up so much crap around the comp in her sewing room that it blocked the wireless signal, so about a 100k PPD loss on the 7950 that resides in her email/sewing computer

Can't live with them, can't live with out them:lol:
 
LOL. At first I thought you were going to say it overheated. I'm thinking it's time to run a hard line to that PC!
 
:lol: actually the comp is on the floor, but the wireless receiver is on the table with the keyboard and printer, the little antenna was folded down and there was about a 12" of crap stacked on top of the printer blocking the wireless signal. All removed for now and hope message was sent!

LOL and we are still talking, marriage is bliss?:lol:
 
My GTX 660 is dropping 8900's like mad today, had one fail at around 50% then several crash 'n' burn within 2-3% all with the same error...

ERROR:exception: Error invoking kernel execFFT: clEnqueueNDRangeKernel (-5)

:cussing:
 
Just a quick heads up guys to why my stats have fallen foul again. Last wednesday i was rushed into hospital and had emergency surgery....turns out my crohns thought it would flare up after nearly 20yrs of being mostly dormant...oh well shit happens, think i'll treat myself to another new viddy card when i get home lol
 
Last edited:
Just a quick heads up guys to why my stats have fallen foul again. Last wednesday i was rushed into hostiptal and had emergency surgery....turns out my crohns thought it would flare up after nearly 20yrs of being mostly dormant...oh well shit happens, think i'll treat myself to another new viddy card when i get home lol

I hope you can recover soon. Health is not a matter to take lightly. I wish you a speedy recovery!
 
Back
Top