Yea, I'll wait and see what EOC stats says later tonight. Can we use HMF and V7 at same time?

Right, but I know with v7 it has some funky thing where everything is considered a remote call to it. So wasn't sure if it looked at the data folder or if it needed to make calls into the v7 client itself.

jinihammerer

Keep in mind my PPD is average from completed work units not the guestimated PPD from the F@H client. the F@H client PPD will swing wildly back and forth, the only accurate way is by averaging the completed work units.

amen to that.  i'm experiencing this first-hand now.  thx, Jini.

sticks, i've been running my 780s not yet 24 hrs so far.  but i've been monitoring them very closely.  for the most part, i am able to run anywhere from a + 105 gpu offset to a +144mhz gpu offset, w/o throttling nor with failed WUs.
i keep a very close eye on the temps.  if i see temps creeping into the 60C range, i back off my o/c.  unlike w/ my previous 660ti ftw/ 560sc gpu foldingcombo, passive folding activity is not in the cards w/ the 780s.  in the day i could set precision x for a min. offset followed at night by an aggresive offset.
 
to get my numbers, w/ both 780s, i have a midsize vornado fan at the screened side cover of my corsair 650D case.  it moves tons of air.  moreover, today i took the two unused lower 5.25" drive bay covers out.  w/ lots of internal positive pressure, i can feel a lot of hot air moving out of the front of the case now.
i feel the 780s w/ acx are very efficient at shedding heat.  the fans are so quiet, it's misleading that so much heat is coming from the gpu.  afterburner reported temps 5 deg. C higher on my hotter card than presision x.  it's quite possible temps are higher than a user may suspect.  this leads to throttling and as far as i can tell, failed work units.  i was surprised when i opened my case.  from precision x, temps being in the upper 60s C on one card and high 50s C on the other, i expected about the same relatvie temps as when i was running my 660 and 560 24/7, at 65C and 66C for many hours on end.  i opened my case too many times to count, endlessly measuring for when my two 780s came in.  i want to also run my 660 w/ them when the pci-e riser cable comes in.  anyway, the volume of hot air was much more w/ the 780 combo..  i submit to you that the rate at which hot air is shed by my two 780s is less than the rate at which it is being expelled from the case.  so the internal temp. keeps slowly rising and accumulating.
so i'd advise ppl to take a close look at the "actual" heat in their boxes housing 780s as a factor in their 780 maladies.  from managing my heat better today, i took care of my of my problems.  i started off the day before getting a handle on things with 3 falied WUs from the same card.  since addressing the cooling issue, i have had no failed WUs nor any major throttling issues to speak of.
 
i'm probably going to move to a coolermaster haf x case, especially since i want to get my 660ti ftw back in action.  it's good for a solid 75K to 125K ppd.
in addition, if i decide to go w/ a 4930K on sept. 10th, i'd like to go w/ three 780s in that box.  it's something the gf and i have tossed around for a long time.  now if i can only come up with the right pitch.  lol.
 
btw, i grease-up all my gpus w/ arctic silver 5.  they say it takes 200 hours b4 it reaches it's optimum operating temperature.  and the more heat cycles you put through it, the better and quicker you hit it.  supposedly 5 deg. C lower potentially once it's optimized. 
 
well, i hope my 780/heat issue discussion can help you guys.  oh i forgot to mention, i run stock voltage.  sure max. volts, 1200mv can propel me up to 1345mhz, but it's not sustainable for the most part on air in the summer w/o a/c.  i would recommend using stock voltage too.  it adds heat more than it benefits your clock on air in the summer i feel. 
 
to sum up, the classy is super capable, it's to me a matter if you can take away the heat it sheds fast enough.  otherwise you run into the afformentioned issues.  i think evga should let us team members beta test upcoming video cards for them.  some of us live in some real folding sweat boxes, otherwise known as houses.  lol.  good folding.

Yes, go sticks!

To clarify, the 780 is a powerful card. After all, it's a mini titan. Big power = big heat. If you don't get rid of the large "volume" of heat from it, temps rise and we don't want that.

Tomorrow, I'm going to have only the vornado fan intaking on my side cover, directly onto the gpus. I'll reverse my frt. fan to exhaust so all other airways, other than the side are exhausts. Hopefully I'll see a temp drop.

Good luck and good folding.

sticks435

Wow, thanks for the crazy detailed post . I am currently running mine with both sides of my Switch 810 off and one of the stock fans attached to the upper HHD cage angled up blowing air on it. Tomorrow I will move the fan in the lower front space to the bottom right below the card, so its blowing cold air straight into the back half. Going to try to fold at 1200mhz and see what happens.

Well that didn't work out like I had planned. I guess when I was completing those WU's at 1200mhz yesterday, I was actually running close to 1.35volts, as my card runs .142v higher than the Classified tool shows. Tried to back the actual voltage down to 1.2 and completed 1 WU at those clocks before failing 2 in a row, cause the 2nd one to downclock my card to 5xxmhz. I then completed 1 more WU, but at twice the TPF as normal of course.
 
Didn't have time to restart my computer before work this morning to reset the clocks/driver, so will do it tonight and try 1.25v and see what happens.

My SC with the ACX cooler is stable at 1215 MHz with no voltage increases.  The fan is set to a constant 65% and the temps stay in the low 60's.  I have only the one 780 and I have it in my second PCIe slot so that my sound card doesn't block the airflow. 

Ok, think I finally got a handle on my issues. Guess I had my DMM set to the Diode setting instead of DCV night before last, so my readings were high. Now that I am measuring right, I am running 1267mhz@1.2volts rock steady. Completed 4 WU's before I come into work this morning.
 
So, it seems I'm not experiencing the overclocking issues with the core like most everyone else is. I'm thinking it could be 3 things. 1) It was a driver issue and the newest beta fixes it. 2) The custom PCB on the Classified handles overclocking better than the other cards when folding. 3) A combination of the 2 above lol. I know the beta's had a fix for extremely low FPS on the Classy's due to it's custom VRM, so maybe that had something to do with it.
 
Now that it seems my setup can handle overclocks when running this core, I will see how high I can get it. HFM.net said 190K PPD on the 7810 I was running before I left

Ahhh ok, I had thought it was all of them. Yea, I knew you could loose bonus points, it's something like 10 failed or 10% I think, just not sure out of what. I usually test overnight, that way it's not running all day while I'm at work and failing WU's.

80% of all work unit's you've ever completed? or 80% of QRB eligible ones?

likely 80% of all QRB but either way. the 8900 cores when they start failing its very fast almost as soon as they start folding and they can stack up very fast.
If i remember correctly, when the 8900's first came out in beta I had 22 fails in less then 15 min of tweaking around with my card.
 
I was not the only one either many of us seen similar results. It's why the stern warning about overclocking came to be.
 
Why people just started folding factory clock and why the F@H team themselves will tell people with factory overclocks to underclock their cards to factory spec clocks. 
 
If you want to overclock, thats fine, but i strongly suggest you stick with a minor OC or none at all for folding. If you run into 8900 streak of failing for 8 hours you could run your self so far in the hole you may as well start over.
 

jinihammerer

If you want to overclock, thats fine, but i strongly suggest you stick with a minor OC or none at all for folding. If you run into 8900 streak of failing for 8 hours you could run your self so far in the hole you may as well start over.

That is all sound advice, but the work unit failure and QRB 80% threshold is not as cataclysmic as it seems. As long as you don't keep stopping and restarting the client incessantly, it will put itself to sleep for a day after a preset number of failed returns. Built in protection for you and the project.

TimCohn

Here are two settings to help minimize the impact on your completion rate if you're worried about getting a series of failures while trying to tweak your overclock:

<max-slot-errors v='1'/>
<max-unit-errors v='1'/>

Where is this in the FAH control? I take it this has to be added somewhere? I've just recently started folding so I don't know much about this yet so thanks in advance for your patience.

Wow, very comprehensive instructions. Thank you.

Nvidia has identified the 36-hour TDR bug and the fix will be included in the next WHQL drivers.