Finally! No issues I can see..Yet!

[GIXXERGUY6]

First off I want to send out a major thanks to:

Adak
ChasR
deadlysyn
Norcalsteve
EarthDog
* Sorry if I've missed anyone as I cleaned out my PM's
** in no particular order of course!

I have WinSMP2 up and running smp 8 verbosity 9 I've completed 5 jobs in 2.5 days I think
I have my 9800gtx+ running SMP console and it's amazing(running 2048 ram) I've completed 8 WU's since 2 a.m EST I was ripping through them at :40-:46 and completing in less than an hour. This one I'm on now is a P5781 and it says 1:46 tpf eta is 2:17..big file???

according to F@H gadget and the stats I'm averaging 24hrs 5344 today is 3307 week is 37056 total is 84550, but stanford stats are showing me at 86xxx.

oh crap sitting here watching HFM I just had my first GPU failed :(

[18:36:36]
[18:36:36] *------------------------------*
[18:36:36] Folding@Home GPU Core
[18:36:36] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[18:36:36]
[18:36:36] Compiler : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[18:36:36] Build host: amoeba
[18:36:36] Board Type: Nvidia
[18:36:36] Core :
[18:36:36] Preparing to commence simulation
[18:36:36] - Looking at optimizations...
[18:36:36] DeleteFrameFiles: successfully deleted file=work/wudata_09.ckp
[18:36:36] - Created dyn
[18:36:36] - Files status OK
[18:36:36] - Expanded 65002 -> 344387 (decompressed 529.8 percent)
[18:36:36] Called DecompressByteArray: compressed_data_size=65002 data_size=3443
87, decompressed_data_size=344387 diff=0
[18:36:36] - Digital signature verified
[18:36:36]
[18:36:36] Project: 5781 (Run 12, Clone 629, Gen 1)
[18:36:36]
[18:36:36] Assembly optimizations on if available.
[18:36:36] Entering M.D.
[18:36:42] Tpr hash work/wudata_09.tpr: 143931600 4122347632 331493708 21088657
44 2878676641
[18:36:42]
[18:36:42] Calling fah_main args: 14 usage=100
[18:36:42]
[18:36:42] Working on Great Red Owns Many ACres of Sand
[18:36:44] Client config found, loading data.
[18:36:45] Starting GUI Server
[18:38:30] Completed 1%
[18:40:15] Completed 2%
[18:42:01] Completed 3%
[18:43:46] Completed 4%
[18:45:32] Completed 5%
[18:47:17] Completed 6%
[18:49:03] Completed 7%
[18:50:49] Completed 8%
[18:52:34] Completed 9%
[18:54:20] Completed 10%
[18:56:05] Completed 11%
[18:57:51] Completed 12%
[18:59:36] Completed 13%
[19:01:22] Completed 14%
[19:03:07] Completed 15%
[19:04:53] Completed 16%
[19:06:38] Completed 17%
[19:08:24] Completed 18%
[19:10:09] Completed 19%
[19:11:56] Completed 20%
[19:13:41] Completed 21%
[19:15:27] Completed 22%
[19:17:13] Completed 23%
[19:18:22] Run: exception thrown during GuardedRun
[19:18:22] Run: exception thrown in GuardedRun -- Gromacs cannot continue furthe
r.
[19:18:22] Going to send back what have done -- stepsTotalG=20000000
[19:18:22] Work fraction=0.2365 steps=20000000.
[19:18:26] logfile size=9252 infoLength=9252 edr=0 trr=23
[19:18:26] + Opened results file
[19:18:26] - Writing 9788 bytes of core data to disk...
[19:18:26] Done: 9276 -> 3396 (compressed to 36.6 percent)
[19:18:26] ... Done.
[19:18:27] DeleteFrameFiles: successfully deleted file=work/wudata_09.ckp
[19:18:31]
[19:18:31] Folding@home Core Shutdown: UNSTABLE_MACHINE
[19:18:34] CoreStatus = 7A (122)
[19:18:34] Sending work to server
[19:18:34] Project: 5781 (Run 12, Clone 629, Gen 1)
[19:18:34] - Read packet limit of 540015616... Set to 524286976.


[19:18:34] + Attempting to send results [January 29 19:18:34 UTC]
[19:18:34] - Reading file work/wuresults_09.dat from core
[19:18:34] (Read 3908 bytes from disk)
[19:18:34] Connecting to http://171.67.108.21:8080/
[19:18:34] Posted data.
[19:18:34] Initial: 0000; Conversation time very short, giving reduced weight in
bandwidth avg
[19:18:34] - Uploaded at ~9 kB/s
[19:18:34] - Averaged speed for that direction ~47 kB/s
[19:18:34] + Results successfully sent
[19:18:34] Thank you for your contribution to Folding@Home.
[19:18:38] Trying to send all finished work units
[19:18:38] + No unsent completed units remaining.
[19:18:38] - Preparing to get new work unit...
[19:18:38] + Attempting to get work packet
[19:18:38] - Will indicate memory of 2048 MB
[19:18:38] - Connecting to assignment server
[19:18:38] Connecting to http://assign-GPU.stanford.edu:8080/
[19:18:39] Posted data.
[19:18:39] Initial: 43AB; - Successful: assigned to (171.67.108.21).
[19:18:39] + News From Folding@Home: Welcome to Folding@Home
[19:18:39] Loaded queue successfully.
[19:18:39] Connecting to http://171.67.108.21:8080/
[19:18:40] Posted data.
[19:18:40] Initial: 0000; - Receiving payload (expected size: 65505)
[19:18:40] Conversation time very short, giving reduced weight in bandwidth avg
[19:18:40] - Downloaded at ~127 kB/s
[19:18:40] - Averaged speed for that direction ~73 kB/s
[19:18:40] + Received work.
[19:18:40] Trying to send all finished work units
[19:18:40] + No unsent completed units remaining.
[19:18:40] + Closed connections
[19:18:45]
[19:18:45] + Processing work unit
[19:18:45] Core required: FahCore_11.exe
[19:18:45] Core found.
[19:18:45] Working on queue slot 00 [January 29 19:18:45 UTC]
[19:18:45] + Working ...
[19:18:45] - Calling '.\FahCore_11.exe -dir work/ -suffix 00 -priority 96 -check
point 3 -verbose -lifeline 2504 -version 623'

[19:18:45]
[19:18:45] *------------------------------*
[19:18:45] Folding@Home GPU Core
[19:18:45] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[19:18:45]
[19:18:45] Compiler : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[19:18:45] Build host: amoeba
[19:18:45] Board Type: Nvidia
[19:18:45] Core :
[19:18:45] Preparing to commence simulation
[19:18:45] - Looking at optimizations...
[19:18:45] DeleteFrameFiles: successfully deleted file=work/wudata_00.ckp
[19:18:45] - Created dyn
[19:18:45] - Files status OK
[19:18:45] - Expanded 64993 -> 344387 (decompressed 529.8 percent)
[19:18:45] Called DecompressByteArray: compressed_data_size=64993 data_size=3443
87, decompressed_data_size=344387 diff=0
[19:18:45] - Digital signature verified
[19:18:45]
[19:18:45] Project: 5781 (Run 14, Clone 202, Gen 1)
[19:18:45]
[19:18:45] Assembly optimizations on if available.
[19:18:45] Entering M.D.
[19:18:52] Tpr hash work/wudata_00.tpr: 4148752120 580377260 2913308194 2987685
39 754460252
[19:18:52]
[19:18:52] Calling fah_main args: 14 usage=100
[19:18:52]
[19:18:52] Working on Great Red Owns Many ACres of Sand
[19:18:54] Client config found, loading data.
[19:18:54] Starting GUI Server
[19:20:39] Completed 1%
[19:22:26] Completed 2%


Unstable_machine that's the first one since I retarted them.

other than that hiccup that just showed it's head I'm rocking out with my gromacs out!

[Norcalsteve]

congrats on your success! glad to hear i helped in some way... i am a noob myself and owe my knowledge to the elder folders too! ;-P

Now that you are up and steaming away, just keep an eye on those temps... I had the V10 before my water set up, never pushed past 3.8 on it (i had a C0 step chip that hotter ran due to the higher voltage it needed) and temps were about 78C small FFT's in Prime95 durring 24hr stability test runs... glad i upgraded to a D0 and Water when i discovered folding!

[pik4chu]

@Gixxer, to make sure HFM is showing accurate ppd goto the Tools menu and click 'download projects from stanford' that will ensure its got up to date point and project info for the WUs its running. That should account for any discrepency in ppd within reason.

[GIXXERGUY6]

thanks guys.

NorCal - folding my temps are 75-78 folding 100%

Pika - I did it and now I noticed it took away 2 wu's no failed on CPU but

[21:52:30]
[21:52:30] *------------------------------*
[21:52:30] Folding@Home Gromacs SMP Core
[21:52:30] Version 1.74 (March 10, 2007)
[21:52:30]
[21:52:30] Preparing to commence simulation
[21:52:30] - Looking at optimizations...
[21:52:30] - Created dyn
[21:52:30] - Files status OK
[21:52:30] this execution.
[21:52:30] - Previous termination of core was improper.
[21:52:30] - Files status OK
[21:52:32] ial work packet
[21:52:32]
[21:52:32] Project: 2653 (Run 29, Clone 91, Gen 138)
[21:52:32]
[21:52:32] - Starting from initial work packet
[21:52:32]
[21:52:32] Project: 2653 (Run 29, Clone 91, Gen 138)
[21:52:32]
[21:52:32] Entering M.D.
[21:52:37] e
[21:52:37]
[21:52:37] Folding@home Core Shutdown: FILE_IO_ERROR
[21:52:37] Finalizing output
[21:52:40] CoreStatus = 7B (123)
[21:52:40] Sending work to server
[21:52:40] Project: 2653 (Run 29, Clone 91, Gen 138)
[21:52:40] - Error: Could not get length of results file work/wuresults_01.dat
[21:52:40] - Error: Could not read unit 01 file. Removing from queue.
[21:52:40] Trying to send all finished work units
[21:52:40] + No unsent completed units remaining.
[21:52:40] - Preparing to get new work unit...
[21:52:40] Cleaning up work directory
[21:52:41] + Attempting to get work packet
[21:52:41] Passkey found
[21:52:41] - Will indicate memory of 600 MB
[21:52:41] - Connecting to assignment server
[21:52:41] Connecting to http://assign.stanford.edu:8080/
[21:52:42] Posted data.
[21:52:42] Initial: 40AB; - Successful: assigned to (171.64.65.64).
[21:52:42] + News From Folding@Home: Welcome to Folding@Home
[21:52:42] Loaded queue successfully.
[21:52:42] Connecting to http://171.64.65.64:8080/
[21:52:45] Posted data.
[21:52:45] Initial: 0000; - Receiving payload (expected size: 2447810)
[21:52:48] - Downloaded at ~796 kB/s
[21:52:48] - Averaged speed for that direction ~584 kB/s
[21:52:48] + Received work.
[21:52:48] Trying to send all finished work units
[21:52:48] + No unsent completed units remaining.
[21:52:48] + Closed connections
[21:52:53]
[21:52:53] + Processing work unit
[21:52:53] Work type a1 not eligible for variable processors
[21:52:53] Core required: FahCore_a1.exe
[21:52:53] Core found.
[21:52:53] Working on queue slot 02 [January 29 21:52:53 UTC]
[21:52:53] + Working ...
[21:52:53] - Calling 'mpiexec -np 4 -channel auto -host 127.0.0.1 FahCore_a1.exe
-dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 4864 -version 629'

[21:52:53]
[21:52:53] *------------------------------*
[21:52:53] Folding@Home Gromacs SMP Core
[21:52:53] Version 1.74 (March 10, 2007)
[21:52:53]
[21:52:53] Preparing to commence simulation
[21:52:53] - Ensuring status. Please wait.
[21:53:10] - Looking at optimizations...
[21:53:10] - Working with standard loops on this execution.
[21:53:10] - Previous termination of core was improper.
[21:53:10] - Going to use standard loops.
[21:53:10] - Files status OK
[21:53:12] Starting from initial work pa- Starting from initial work packet
[21:53:12]
[21:53:12] Project: 2653 (Run 29, Clone 91, Gen 138)
[21:53:12]
[21:53:12] Entering M.D.
[21:53:12] ne 91, Gen 138)
[21:53:12]
[21:53:12] Entering M.D.
[21:53:18] Rejecting checkpoint
[21:53:18] Protein: Protein in POPC
[21:53:18] Couldn't open for writing
[21:53:18] Writing local files
[21:53:19] Extra SSE boost OK.
[21:53:19] Writing local files
[21:53:19] Completed 0 out of 500000 steps (0 percent)
[21:59:49] Writing local files
[21:59:49] Completed 5000 out of 500000 steps (1 percent)


and now my gpu isn't updating

I just restarted both consoles via cntrl c and opening them again.

any suggestions? I have a job to go to right now but will be back later

[pik4chu]

when you say not updating do you mean according to the logs? like the folding on the gpu is hung, or in HFM where its saying it has no frame times? if its the later its because by defaulty HFM uses the last 3 times to calculate ppd, according to your log you only have 2 so it wont go from yellow status to green until it has enough data to sample :)

If its the former then try stopping the gpu client, closing it, updating video drivers (even just rerunning the driver setup is fine) and then reboot and try agian. had this happen on one rig this week. Dont know what caused it only what fixed it ;)

[ChasR]

GIX,
You're running the 9800 GTX in the GPU consloe client. Not SMP console. 2048 would be shaders not memory wouldn't it? THe EUE, likely means you're clocked to high. Setting shaders to 2048 results in a clock of 2052. You probably should lower it to 1998. Anything more than a 2% failure rate indicates you're clocked too high.

HFM tracks completed and failed WUs from the last restart of the client, so when you restarted the gpu, both counters zeroed.

[GIXXERGUY6]

GIX,
You're running the 9800 GTX in the GPU consloe client. Not SMP console. 2048 would be shaders not memory wouldn't it? THe EUE, likely means you're clocked to high. Setting shaders to 2048 re****s in a clock of 2052. You probably should lower it to 1998. Anything more than a 2% failure rate indicates you're clocked too high.

HFM tracks completed and failed WUs from the last restart of the client, so when you restarted the gpu, both counters zeroed.

No my shader is 2010 it wanted to know the memory to used for it and wanted to use 3062(that was for the gpu)

now I don't know what the issue is but the cpu is only working at 60% and it said tpf was 6:30 wtf is the issue(and no WTF wasn't way to fold) Why is it always something :D it worked sooooooooo great until earlier :(

project for cpu is p2653 tpf 6:34 eta is 6:20:52 credit says 1760
project for gpu is p5768 tpf :42 eta is 56:42 credit says 353


What is going on here?

[onefstsnake]

The tpf for your gpu looks fine. Are all cores doing work?
I know when I used VM Ubuntu I can go into the system monitor and see the load on all 8 cores.
When I had problems sometimes 4 of 8 cores were idle and sometimes all were around 70%. I had to change the VM config and it fixed it.

But as far as the new Windows client I really havent had many issues and Im still running VM Ubuntu for my i7.

Good luck on getting everything running smoothly.

EDIT: the gpu shouldnt need to use your system memory much since it has onboard memory.

[GIXXERGUY6]

The tpf for your gpu looks fine. Are all cores doing work?
I know when I used VM Ubuntu I can go into the system monitor and see the load on all 8 cores.
When I had problems sometimes 4 of 8 cores were idle and sometimes all were around 70%. I had to change the VM config and it fixed it.

But as far as the new Windows client I really havent had many issues and Im still running VM Ubuntu for my i7.

Good luck on getting everything running smoothly.

EDIT: the gpu shouldnt need to use your system memory much since it has onboard memory.

yeah as far as the gpu ram thing I think I was confused(very well could have been)

Not at home right now so I can't look at anything right now :D

[ChasR]

If you get an a1 WU, cpu utilization is going to be much lower than with an a3. That's normal. Shaders set at 2010 is really 1998, due to the 54 MHz steps.

[GIXXERGUY6]

If you get an a1 WU, cpu utilization is going to be much lower than with an a3. That's normal. Shaders set at 2010 is really 1998, due to the 54 MHz steps.

ok before I left I backed it back down to 1998 in RT.

[deadlysyn]

Glad I could help, GIX. You are likely to see a slowdown when you get one of the a1 WU's, as ChasR stated. With lower CPU utilization on an i7, plus the lower point value due to the lack of the bonus, HFM will report a lower PPD. There isn't much you can do, aside from finish the WU and move on to the next one in your queue.

[GIXXERGUY6]

I want to know how many WU's I've completed with my passkey.. How can I figure this out?

and it doesn't seem like I'm cranking the numbers I should be.

2 days ago I was at 102k and now I'm at 127k..Doesn't seem right.

[GIXXERGUY6]

Well guys I was hoping the driver was the culprit, but it wasn't I completed 2 wu's with my oc settings 775/2010/1150, but I went to bed and failed 5, now I gotta wait 24 hours to start back on gpu's again:(???????????? Any way around this?

ok back to 1998 I guess, After the driver update I thought it was at my overclock and was getting happy that I had no fails, but I ran 12 wu's without a failure(I was at stock settings, tpf's :48, overclock dropped them to :44) this sucks

[Norcalsteve]

Try just OC'ing the shaders... thats all that matters for folding, to a point. plus OC'ing the rest just adds heat. I got my shaders set from 1475mhz (stock) to 1620mhz and the rest of the cards are stock settings. I have pushed 1720 on my shaders but i have to have the top GPU fan SCREAMING. cant wait for my Arctic chill GPU cooler to get in.

[ChasR]

GIX,
As I've previously explained, 2010 MHz set in Riva Tuner or EVGA precision IS 1998 MHz shader clock. The shaders clock in 54 MHz steps. So your lowering the Riva Tuner setting to 1998 does absolutely nothing to change your shader clock.
As Steve says, don't mess with anything but the shaders. I've tested extensively and found running the core clock from stock to it's max oc makes a 1% difference in TPF. However, you'll trash a lot of WUs and make less ppd than you will at stock. Memory is the same. THere is so little to gain overclocking the video memory it isn't worth it. All you produce is heat.

Manually set the fan to 85% to keep temps low. On the 9800 GTX+ going from 85% to 100 % fan only reduces temps by 1 degree and is a lot louder, so there isn't much reason to go above 85%. Some cards, GTX260 comes to mind, the sweet spot is about 75%.

Just a reminder, TPF means nothing without saying which WU it applies to.

[GIXXERGUY6]

GIX,
As I've previously explained, 2010 MHz set in Riva Tuner or EVGA precision IS 1998 MHz shader clock. The shaders clock in 54 MHz steps. So your lowering the Riva Tuner setting to 1998 does absolutely nothing to change your shader clock.
As Steve says, don't mess with anything but the shaders. I've tested extensively and found running the core clock from stock to it's max oc makes a 1% difference in TPF. However, you'll trash a lot of WUs and make less ppd than you will at stock. Memory is the same. THere is so little to gain overclocking the video memory it isn't worth it. All you produce is heat.

Manually set the fan to 85% to keep temps low. On the 9800 GTX+ going from 85% to 100 % fan only reduces temps by 1 degree and is a lot louder, so there isn't much reason to go above 85%. Some cards, GTX260 comes to mind, the sweet spot is about 75%.

Just a reminder, TPF means nothing without saying which WU it applies to.

Yeah I'm back to stock GPU settings right now except the shader( 1990 ) but I have to wait 24 hours now:(

[ChasR]

Restart the client.

[GIXXERGUY6]

You are my hero

[GIXXERGUY6]

coincidence or not? I changed the clock to stock accept the shader @ 1990 and now I have a p5785 giving me a tpf of 1:46(ETA 2:46) when they(don't know the number) were giving me anywhere from :43-:48 tpf's.

I don't get all these points and ETA's and TPF's and projects and GPU's, and SMP's :D

[ChasR]

ETA is meaningless in discussing WU production. If the WU is 50% complete it's ETA will be 1/2 of a WU just started.

You're remembering incorrectly. There is no way p5785 ran at 0:43/frame on a 9800GTX+.

Use the benchmarking in HFM (tools, benchmarks), instead of relying on memory. Here's the benchmark data from HFM on p5785 run on a 9800GTX @ stock core and memory with a 1998 MHz shader clock:

Name: CINDY 9800GTX
Path: \\Cprcc-ws03\FAH\FAH GPU\
Number of Frames Observed: 300

Min. Time / Frame : 00:01:54 - 5,934 PPD
Avg. Time / Frame : 00:01:55 - 5,883 PPD

[GIXXERGUY6]

ETA is meaningless in discussing WU production. If the WU is 50% complete it's ETA will be 1/2 of a WU just started.

You're remembering incorrectly. There is no way p5785 ran at 0:43/frame on a 9800GTX+.

Use the benchmarking in HFM (tools, benchmarks), instead of relying on memory. Here's the benchmark data from HFM on p5785 run on a 9800GTX @ stock core and memory with a 1998 MHz shader clock:

Name: CINDY 9800GTX
Path: \\Cprcc-ws03\FAH\FAH GPU\
Number of Frames Observed: 300

Min. Time / Frame : 00:01:54 - 5,934 PPD
Avg. Time / Frame : 00:01:55 - 5,883 PPD

No Chas I wasn't referring to getting 1:43 tpf from a 5785 I meant I went form :4x's on "I forget what number" to this one. I was just baffled is all(didn't know if me jacking with the shader screwed with something)

I bow to your infinite folding wisdom that's for sure.

btw you didn't answer me

Will you be my hero?:rock:

[GIXXERGUY6]

ah I see what you're talking about in the benchmarks(is that just my times)

yes it is nevermind I see the name :D

Project ID: 5772
Core: GROGPU2
Credit: 353
Frames: 100

Names: 9800GTX+
Path: E:\Folding\GPU\
Number of Frames Observed: 300

Min. Time / Frame : 00:00:33 - 9,242.2 PPD
Avg. Time / Frame : 00:00:44 - 6,931.6 PPD


btw where are the 5915, 6012, 10102 it's listed under GPU projects but I haven't seen them yet

7k ppd to 12k ppd that looks yummy :D

[ChasR]

THe various WUs have different point values and TPFs. You simply got a different type of WU than the one before.

[GIXXERGUY6]

Well that sucks!

It was working fine. I left to go on service call and came back and my desktop is all pixelated/artifacts (740/1990/1100) I also folded 2 wu's just fine, but failed 5 while I was gone. Nothing was running, doesn't sleep, nothing turns off.

What gives?

[Norcalsteve]

Back your shaders off a few... your other settings are back at stock?

[GIXXERGUY6]

Back your shaders off a few... your other settings are back at stock?

yes 740/1998/1100 when I left and when i came back.

so I restarted pc and was testing with ati tool artifact and no matter were I put the shader it was showing "yellow" I guess artifacts.

I have rth... now and I don't know how I should run the test and what to look for etc....

Can I start crying now or later?

[Mark620]

If you get an a1 WU, cpu utilization is going to be much lower than with an a3. That's normal. Shaders set at 2010 is really 1998, due to the 54 MHz steps.

yes 740/1998/1100 when I left and when i came back.

so I restarted pc and was testing with ati tool artifact and no matter were I put the shader it was showing "yellow" I guess artifacts.

I have rth... now and I don't know how I should run the test and what to look for etc....

Can I start crying now or later?

As ChasR said back the shaders off. Some GPUs will not run that high.
I have : 3 Identical GPU cards = 2~1998, 1~1950
and 4 Identical GPU cards = 2~1998 , 1~1950, 1~1850
and my Zotac will only do about 1400 shaders without trashing WU's
and the Zotac has P5781 is only producing about 50% of what it does on most other projects.

[ChasR]

GIX,
Drop the shaders one step to 1944 and drop the ram to 1000 and see if that's stable. EVGA precision and nVidia nTune tools will both give you a quick stability check. RTHDRIBL will as well. OCCT will test gpu memory. You can drop the core clock and memory down quite a bit before you get any significant reduction in performance. Of course the card should fold with the core and memory at stock clocks, but some don't out of the box and some degrade over time. Anyway my real point is, it might not be the shaders that are making it unstable. Try underclocking core and vram and see if the tests show stability. If that works, you still need to decide whether to fold on or RMA the card.

[GIXXERGUY6]

GIX,
Drop the shaders one step to 1944 and drop the ram to 1000 and see if that's stable. EVGA precision and nVidia nTune tools will both give you a quick stability check. RTHDRIBL will as well. OCCT will test gpu memory. You can drop the core clock and memory down quite a bit before you get any significant reduction in performance. Of course the card should fold with the core and memory at stock clocks, but some don't out of the box and some degrade over time. Anyway my real point is, it might not be the shaders that are making it unstable. Try underclocking core and vram and see if the tests show stability. If that works, you still need to decide whether to fold on or RMA the card.

I just found out my oc didn't stick since I added that 2nd card and a couple reboots.

I was running the xfx(original card at stock 740/1836/1125 and no fails completed 29 wu's)
the BFG is oc'd stock and is rocking out at 760/1880/1150 or something,,runs cooler than my xfx and hasn't had a fail yet I have it at 760/1998/1125.

my xfx I recently clocked back to the 740/1998/1100 lastnight, I woke up and had a fail :(. I think I might RMA it I don't know what it's issue is.

[Norcalsteve]

keep in mind that the same model cards can have different stability in shader clock speeds... i got 2 gtx285's and one is stable at 1620 and the other will only be stable 1 step lower (54mhz lower)

[GIXXERGUY6]

yeah I know, but I recently just failed 7 wu's in a row with the bfg. I don't get why sometimes I get this issue. I backed it off to 1944 so we'll see. Restarted the client and set to 1944 failed 4 and now it's fine.

[Outback_Jon]

yeah I know, but I recently just failed 7 wu's in a row with the bfg. I don't get why sometimes I get this issue. I backed it off to 1944 so we'll see. Restarted the client and set to 1944 failed 4 and now it's fine.
Certain WUs work the GPU harder and an overclock that has been stable suddenly won't be.