Even if the P2 State is corrected for Linux it only saying that the GTX 980 Card are NOT meant to Fold either because they are to Powerful for Folding or they are Bad Cards or Faulty Cards.
This is based on Most Core 21 Projects work fine on Older Cards (That Take 1.5 to 2 Times Longer to Complete) and that I can reduce the Memory to -300MHz on Newer Cards.
Or is it the Code on each of the Projects as we are not really seeing this problem on older Core 18 Project that were written for the New GTX 9xx Cards?
I am back on the Current Driver using -300Mhz on the Memory and running the GPU's at 1500MHz.
Once I update my X99 Rig back to the Current Driver as well I will go down to -300 MHz and See.
Estimated Cost for each Failed "BAD_WORK_UNIT" is about $5.00 of wasted electric energy under Xcel Energy.
 
Only NVIDIA can explain WHY they do what they do as does Standard and we both know they are not going to do that.
GTX 9xx cards are for Gaming Not for Folding in the Future Drivers?
 
My testing show each driver new or old run fine for some Core 21 Projects but not other Core 21 Projects.
We will have to make the call on allowing failed 2 or 3 hour Projects or 10 or 15 hour projects, which cost less waste to your money on?
 
As before all we can really do is Keep Folding or just give it up altogether or move over the BOINC. 
 

I am hanging in there until we hit the 3 Billion for the month, then I am off crunching.... not going to put up with these losses of WU's.
 
I have lost over 50+ WU's now.........

Thank you Grandpa_01 for the input, and Bcav. It's definitely frustrating! :)
 
@ Scott , I am glad your doing better. Hope it stays that way!! And congratulations on becoming a Grandpa, it's a wonderful thing. :)

How are you guys going about lowering the P2 state? I've under-clocked the memory of my cards to -501 ( the lowest precision will allow) and still get errors.
 
As for grandpa's questions, it all seems to be about stable memory, Fermi and Kepler have memory at 6008mhz while Maxwell is at 7010mhz
My best guess is that Nvidia knows that 7010mhz is too high and unstable for compute work and Down-clocks it to 6008 for it to work, something that that more stable memory on Fermi and Kepler don't need as they are running at a lower memory speed to begin with.
 
As for the x21 cores, some seem to over-ride Nvidia's down-clocking as I've seen my GPUs reporting the memory running at 3505mhz (7010 effective) compared to the down-clocked and stable 3004 (6008) that the x18 are running at
 
all in all it seems to be the stability of memory at that speed, and it not downclocking to 6008 memory speed like the x18 units do. (or not needing too in the case of Fermi/Kepler)
 
 
also as mentioned earlier, the P9704 work does seem to down-clock to 6008 as I have one now, anyone with other X21 core units that can double check the memory speed that they are set too?

thanks, I'll drop it way down to start and creep it back up a bit if needed, as it doesn't seem to affect the TPF or it's P0 state that I'd be running in for games, I may just leave it fairly low in fact.

I Spook to Soon: ((((But they are Projects 96xx that are failing and not 72xx))))
 

 I will change the above from -300 to -320 or Max it out.
 
01:29:47:WU00:FS01:0x21:Completed 780000 out of 2000000 steps (39%)
01:31:11:WU00:FS01:0x21:Completed 800000 out of 2000000 steps (40%)
01:31:18:WU00:FS01:0x21:Bad State detected... attempting to resume from last good checkpoint
01:32:42:WU00:FS01:0x21:Completed 720000 out of 2000000 steps (36%)
01:34:06:WU00:FS01:0x21:Completed 740000 out of 2000000 steps (37%)
01:35:30:WU00:FS01:0x21:Completed 760000 out of 2000000 steps (38%)
01:36:53:WU00:FS01:0x21:Completed 780000 out of 2000000 steps (39%)
01:38:17:WU00:FS01:0x21:Completed 800000 out of 2000000 steps (40%)
01:38:25:WU00:FS01:0x21:Bad State detected... attempting to resume from last good checkpoint
01:39:49:WU00:FS01:0x21:Completed 720000 out of 2000000 steps (36%)
01:41:12:WU00:FS01:0x21:Completed 740000 out of 2000000 steps (37%)
01:42:36:WU00:FS01:0x21:Completed 760000 out of 2000000 steps (38%)
01:44:00:WU00:FS01:0x21:Completed 780000 out of 2000000 steps (39%)
01:45:24:WU00:FS01:0x21:Completed 800000 out of 2000000 steps (40%)
01:45:31:WU00:FS01:0x21:Bad State detected... attempting to resume from last good checkpoint
01:45:31:WU00:FS01:0x21:Max number of retries reached. Aborting.
01:45:31:WU00:FS01:0x21:ERROR:Max Retries Reached
01:45:31:WU00:FS01:0x21:Saving result file logfile_01.txt
01:45:31:WU00:FS01:0x21:Saving result file log.txt
01:45:31:WU00:FS01:0x21:Folding@home Core Shutdown: BAD_WORK_UNIT
01:45:32:WARNING:WU00:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
01:45:32:WU00:FS01:Sending unit results: id:00 state:SEND error:FAULTY project:9631 run:1 clone:2 gen:43 core:0x21 unit:0x00000041ab436c9b5609bee2518f85f1
 
Somehow I am stuck at -501 MHz

Grandpa_01
bcavnaugh
 
I do not think Presison X will change the P2 state memory clock you need to Install Nvidia Inspector and look at the P2 state clocks I will bet it says 3004 which is 6008Mhz.

Is it ok to have Precision running ,while using inspector? Didn't know if they might conflict. Thx!

Grandpa_01
bcavnaugh
 
I do not think Presison X will change the P2 state memory clock you need to Install Nvidia Inspector and look at the P2 state clocks I will bet it says 3004 which is 6008Mhz.

Correct and I use both together as well with not issues.

I don't know what happened.. I am on 355.xx (can't say which for sure, sorry.. I'm not close enough to look yet.)

I am on client-type advanced and just sky rocketed from 1.0-1.2m to 1.7m .

I am not using the country to fold, just 4 980's in one system.

 
Questions asked on the Folding Forum:
 
I have GTX 980 Graphics Cards what is the Correct Driver Version we should be using?
I also have some GTX 780 Graphics Cards but have not seen any issues on these cards but what is the correct driver version to use here as well?
Last I have AMD 290X Graphics Cards but have not seen any issues on these cards but what is the correct driver version to use here as well?
Posted here: https://foldingforum.org/viewtopic.php?f=61&t=28284&p=280874#p280874
 
1100 17 Nov 2015 MST Sill No Answer from Stanford.

I am waiting for my Rigs to Complete current PrimeGrid Task and then I will Set two rigs to the Bios Defaults that is No Overclocking on the CPU or Memory. BTY if your computer can complete LLR Tasks for days on end then you have a supper stable Bios Setup.
I will then Down Clock all 6 GTX 980 Cards to the lowest setting I can select in PrecisionX 16 and set the P2 in Nvidia Inspector down to 2000MHz.
Yes this is all a waste of money and waste of electricity to only wait for failed projects from Core 21.
But how else can we show that our hardware IS STABLE and that or GPUs are not BAD.
Yes I am made at Stanford, who is not. It is said that it is NEVER Stanford problem.
OK I am done Venting.
 

I started a new Thread for Testing Here: Testing Issues with the GTX 980 Graphics Cards and Core 21 Projects