Could someone look at my tasks please, they error out as soon as they try to start. I ran GPU grid fine about 4 months ago but just started back and it will not run anymore.
Thanks

Problem solved, i took out card 2 and it ran good with the one card. I put card 2 back in and it works now ??

Per chance are you using Sli?

When you can pick up new tasks, do a system restart, make sure you are using default clocks and the GPU temps are fine, suspend other tasks and see if you can get a task to pass 10seconds.

The system cannot find the path specified. (0x3) - exit code 3 (0x3)

SWAN: FATAL : swanMemcpyDtoH failed

Assertion failed: 0, file swanlib_nv.c, line 390

This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.

Based on some suggestions by SKGiven , I tried some things.
-
First I backup 6.13.1 then uninstalled and deleted the folder.
-
With a new copy of 6.12.34 installed and only attching to GPUgrid so there could be no interference from any other project, I got an immediate failure. I tried using the current location which is on my B drive. After that I uninstalled that one and installed 6.12.34 to C drive in the default locations, that too failed. 6.12.34 runs GPUgrid on 3 other computers, although it is the x32 version not the x64 versaion. Next I tried to install the x32 version, attach only to GPUgrid, that to produced an error immediately.

It is something else, but I still wonder what the application error is looking for, what path ????

<core_client_version>6.12.34</core_client_version>
<![CDATA[
<message>
The system cannot find the path specified. (0x3) - exit code 3 (0x3)
</message>
<stderr_txt>
# Using device 0
# There are 2 devices supporting CUDA
# Device 0: "GeForce GTX 570"
# Clock rate: 1.59 GHz
# Total amount of global memory: 1309212672 bytes
# Number of multiprocessors: 15
# Number of cores: 120
# Device 1: "GeForce GTX 570"
# Clock rate: 1.59 GHz
# Total amount of global memory: 1309212672 bytes
# Number of multiprocessors: 15
# Number of cores: 120
MDIO: cannot open file "restart.coor"
SWAN: FATAL : swanMemcpyDtoH failed

Assertion failed: 0, file swanlib_nv.c, line 390

This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.

</stderr_txt>
]]>

Krunchin-Keith, Are these the GTX570 + ATI card machines? I know at one point nvidia put something into their drivers to disable the ATI driver.

Can you try them as single GTX570 card machines and see how they go? Maybe remove ATI card, uninstall ATI driver, run driver sweeper (or equivilent) and then do a clean install of the nvidia driver before trying GPUgrid again.

Make sure you didn't install BOINC in PAE (protected application execution also known as service) mode. Win 7 and Vista don't allow the graphics drivers to be referenced by a service.

Lastly you did install the x64 drivers didn't you? The BOINC version doesn't matter so much, but the drivers may.

I'm back !!!

After many trials I think I have found a solution to this.

Since these two systems were new builds, I had not played with any rom settings yet, leaving the factor set. After lots of failures while doing something else unrelated yesterday, I decided to look into rom settings on one unit. Almost all the voltage type settings were set to "auto". I decided to change all these to "Standard" or "Normal" to keep them stable and from using any of the level 1 or level 2 savings levels and be sure none were overclocked. I also found one other thing, the PCIE Frequency on #2 was set at 95MHz and it said 100MHz was the standard, so I changed that too. Later I found #1 was on Auto, so I set all it the same to normal on voltages and the PCIE Frequency to 100MHz. This may explain why one system, in the past, trashed more work than the other, being the frequency might have been more off standard than the other.

After reboots and waiting for boinc to load, I re-enabled GPUgrid on #2, #1 was still active but still had only been completing 1 out of 12, all the others errored within those first 12 seconds.

Both systems have the one ATI HD6970 GPU and 2 GTX570 GPU's.

#2 got 4 tasks and began to run 2 on the GTX570's. A surprise to me, both passed the 12 second mark and kept going. after time 1 finished sucessfully and a 3rd started also getting past 12 seconds.

by bedtime #1 had retrieved work and was running 2 long runs having passed the error point.

So as of this morning BOTH systems have had NO Errors on GPUgrid.

#1 has completed (valid and credtied) 2 long runs and running 2 more.

#2 has completed (valid and credtied) 4 short and 1 long and is running 2 more long.

This represents the longest consecutive number run at once WITHOUT error for either host. Long runs are taking 6.3 to 7.3 hours.

It is not a driver conflict between ati and nvidia, nor memory problem, nor gpu overclocking issue. I'm still at gpu standard speeds for these factory overclocked models, but even when i undeclocked them 10% that did not help. I did rarely have 1 run and complete while at the standrd stock overclock speed.

Different boinc versions had no change during testing. Even fresh install. Trying one project only had no change even on several boinc versions, so it is not a project conflict.

I'm now on ATI catalyst 11.9 and NVIDIA 285.62 but during all this I had many other combinations starting at about 11.4 or 270.XX. Those never made such as difference as the last change to voltages/frequency.

I will keep monitoring this to see if problem returns and it will be a while before I go back into rom and make any changes, so i'm not 100% sure whether it was a particular voltage for cpu,memory,gpu or the PCIE frequency.

At least for now i'm back in business and producing valid work here at one of my favorite valued projects.

PCIe does not tolerate frequency change (It's running at 100MHz). Do not over/underclock the PCIe bus. It will decrease the stability of the PCIe bus, which is crucial for GPUGrid tasks running for hours. I've posted similar advice here.
Unlike CPU or GPU core / shader clocks, the PCIe bus has a strict norm of its speed. Even lowering PCIe clock will make it unstable. It's quite misleading, that some BIOSes allow to change the PCIe clock. It's very dangerous to do that.

PCIe does not tolerate frequency change (It's running at 100MHz). Do not over/underclock the PCIe bus. It will decrease the stability of the PCIe bus, which is crucial for GPUGrid tasks running for hours. I've posted similar advice here.
Unlike CPU or GPU core / shader clocks, the PCIe bus has a strict norm of its speed. Even lowering PCIe clock will make it unstable. It's quite misleading, that some BIOSes allow to change the PCIe clock. It's very dangerous to do that.

I did not originally do any under/overclocking on the PCIe bus or anything in the ROM not even any voltage changes or tweeking, those settings I found yesterday were how it was set when I got the mobo's.

I set them back to the 100MHz that the rom said was the standard. I have no idea why one was set to 95MHz and the other was on Auto, it was not done by me, or anyone at my house as notone else has access to my systems. Now both are set at 100MHz.