Donate never ending jobs

[Cruncher Pete]

Basically I am not a fan of using a lot of free horse power to help people crack passwords. The language on their site is nice about help security experts, but the freely available tables are used to crack passwords freely. RainbowCrack and a few other programs exist to do just that. In my opinion, helping make what were secure password setups less secure isn't security. Granted this could be done with massive horsepower via other means, I just don't care to put my GPUs to the effort. My personnel preference is to crunch projects that I feel provide some benefit (of course the key is I feel, so it depends on what speaks to me).

Donate for instance interests me in 2 ways. I am intrigued by the bitcoin concept and more importantly that the funds are use to fund medical research (gpugrid). For me RAC is secondary.

BTW, On the topic of the thread, the app_info.xml change has stabilized Donate for me, I ran overnight without getting a neverending job. However, the "509 problem" is there. The 509 is the wus run extra long and produce empty result so you only get 590 credit instead of 5,900. 10% for 3 to 4 time the computing time. I did notice that this always occurs in pairs. Both wus that are sharing the same GPU will go out to lunch together. I think they have a state timeing problem in the app. Freeing up more CPU didn't stop it.

My sentiments exactly and therefore I second that motion. Having said that, I also have to admit that I am also a credit whore and as such I tend to crunch projects that gives me the best outcome first and than I will crunch the others as a secondary target to meet a specific need.

[EmSti]

Frinken GPUS. Too much work lately. I think I found a combination of cards in the pcs that makes Donate stable in 2 of the 3 machines (well as stable as it seem you can get Donate). Turns out the one problem is the MSI r7970 lightning card. It been impossible to get it to work along side of any of the other cards without cutting the output of one of the other 7970 cards by half. Runs fine by itself. Maybe a Bios level difference issues (it is different from the rest). 12.3 or 12.4 Catalyst didn't matter. It is currently in Hydra6 with a disbled 6970. It will not crunch a Donate WU all the way through with the 6970 card enabled (non-crossfire of course). I didn't have any issues when a GigaBtye 7970 was in there. <sigh>

Also, I have not confirmed, but I highly suspect some of the issues are related to MSI Afterburner or Setting EnableUlps and EnableUlps_NA to 0, which allows MSI AfterBurner to monitor the GPU temp on the second GPU. When I removed Afterburner, stomped on Catalyst (remove, boot driver sweeper, boot, you know the drill), I finally got the GPUs in the 2 machines to run at their full potential. The reason I suspect those 2 items is I finally got FireDragonto work full out, when I remembered I didn't reboot since I installed Afterburn and made the regex changes. Did the reboot and 2nd GPU started to dog. Did the whole, strip and replace bit without those 2 items and its been good since even with reboots. I have no desire at the moment to figure out if its one or both.

I swear I feel like I have reinstalled Catalyst 100 times today, trying different combos of things. Wasted an entire day working on this.

[DrPop]

Dang man, sorry to hear this! What a headache! I am not suggesting you do this yet (because I know next to nothing about 7970) but with the 5870 you could flash the bios from another brand. It think it is TechPowerup that has the BIOS collection from all the gamers, along with a GUI BIOS flashing and editing utility for the Radeons. Anyway, if you are sure it's a BIOS issue, may *possibly* be worth looking into. I'd probably read up on what the gamers are doing first, though. A lot of times they find the brand that gives the best O/C BIOS and then flash other cards to that one. I am sure there will be some good info out there...

Wish this stuff would just work like CPU crunching does (for the most part). The sad part is, there is no way for a team to remain even remotely competitive unless it chases high credit GPU projects as well. What to do? Because I feel you on this one...

Sent from my MB860 using Tapatalk 2

[EmSti]

Dang man, sorry to hear this! What a headache! I am not suggesting you do this yet (because I know next to nothing about 7970) but with the 5870 you could flash the bios from another brand. It think it is TechPowerup that has the BIOS collection from all the gamers, along with a GUI BIOS flashing and editing utility for the Radeons. Anyway, if you are sure it's a BIOS issue, may *possibly* be worth looking into. I'd probably read up on what the gamers are doing first, though. A lot of times they find the brand that gives the best O/C BIOS and then flash other cards to that one. I am sure there will be some good info out there...

Wish this stuff would just work like CPU crunching does (for the most part). The sad part is, there is no way for a team to remain even remotely competitive unless it chases high credit GPU projects as well. What to do? Because I feel you on this one...

Sent from my MB860 using Tapatalk 2

You are correct it is TechPowerup and I checked yesterday for MSI Bios. Since the lightning isn't a reference card, I am hesitant to use one that hasn't been built for the card. No word anywhere that I can find on someone upgrading the lightning.

[EmSti]

Yeah making progress. BlackDragon and FireDragon are still running strong with dual 7970s. Hydra6 has been interesting. I got the MSI R7970 lightning to run overnight by down clocking from the factory settings. The first problem is heat. Hydra6 can't keep it cool enough. The driver aborts at 83c but it doesn't stop the Donate job, it pretends it's still running. I happen to be looking when the message briefly showed on the screen. So down clocking to 7970 ref specs (925/1375) and it runs at 76c. Not the best, but a start and it has proven stable. The GigaByte card handled it better (which is why it was in Hydra6 to begin with).

Next problem, the 6970, when enabled, wasn't doing jack and showed as "Disabled Adapter" in Catalyst 12.4. GPU-Z showed it as running at only 250Mhz. Did some searching and found someone reporting this http://devgurus.amd.com/thread/159062. Same issue, different cards and unanswered. But it got me to thinking, what if it wanted a monitor on the card (that old problem). Connected it over to a TV VGA input and sure enough Catalyst enabled management of it and GPU-Z saw it goto 940 Mhz. Woot!. I have had 7970 and 6970 in this box for awhile without a second monitor, so color me surprised by this one. Guess I will need to find old posts for building a fake VGA dongle.

Of course now heat problem is even worse. I am using TThrottle to control Boinc on the GPUs to keep the temp down until I come up with another plan. But at least for the moment I have all 6 GPUs running. Time will tell if they stay stable.

Now, solving the air flow and heat.... Take a look at the Hydra6 pictures, what else can a man do?

[EmSti]

The saga continues...

I now suspect TIM on the R7970 lightning. Bad Tim bad. I have a floor fan pointed into Hydra6 and the card is running 100% at 82c (its own fans going full). Now I need to get up enough nerve to take the card apart and investigate or send it in for repair.

This is like having my own blog.

[DrPop]

Nice! Thanks for the update. Interesting that the old "need a dummy plug" thing reared its head again. Wonder if I still have one somewhere? There is a real simple way to make them out of the correct resistors (3) and a VGA plug. If I remember right YoDude on here had a bunch of them he made up awhile back.

Yeah, heat is becoming more and more the limiting factor on credits per day. I don't know if there is any other option than going full size tower case or open air on a bench with a house fan pointed at it. Even Cheetah and Puma are running hot and they are big cases with lots of airflow - easily on par with the HAF 932 I had, and the Thor V2 case is definitely larger and with more room than that.
My top card is always way hotter than the bottom one. Have you tried putting the 6970 in the top slot (and backing off the clocks on it if necessary), and then put the better 7970 on the bottom? This would allow you to at least get full blast out of the more powerful 7970 while only having to sacrifice a little from the slower card - if at all.
Just some ideas, but I think we're all in about the same boat...

Sent from my MB860 using Tapatalk 2

[EmSti]

Have you tried putting the 6970 in the top slot (and backing off the clocks on it if necessary), and then put the better 7970 on the bottom? This would allow you to at least get full blast out of the more powerful 7970 while only having to sacrifice a little from the slower card - if at all.


Sent from my MB860 using Tapatalk 2

Good call, I will try that before ripping it apart. Now that you mention it, the last time I tried this card in Hydra6 it was better at the bottom. I don't recall it ever running this hot. I wonder if moving it around has loosened things. After dinner I will pull it inspect and swap spots. Silly that I didn't think to try the swap earlier.

[DrPop]

Good deal, please keep us updated! Also, taking the cooler off the GPU is not too hard - I did that with an older HD5830 several months ago, and do you have any Arctic Silver or similar stuff to use? If you are good at applying the very thin layer, and surfaces are tight - cooler to chip, then you will easily see a drop of 5°C over the stock TIM the companies use...

Sent from my MB860 using Tapatalk 2

[EmSti]

I use PK-1 these days. It seems to work better than Artic Silver and doesn't seem to need break-in time. Which is better for machine like these which are always on, they don't get the cooling cycle needed for break-in. The only worry I have on redoing the card is the warranty. The card isn't even 2 months old yet (purchased 4/3). Ok off to move the cards around, again.