I tried removing the AMD catalyst set using "express uninstall manager" but am still having apparent opencl problems on project milkyway. There was a suggestion at their forum

Many people in BOINC-land advise that you remove all traces of all AMD drivers before you change to a different one. AMD seem to have heard them, and produced a driver removal tool. The only official-looking link Google could find for me is http://sites.amd.com/us/game/downloads/Pages/catalyst-uninstall-utility.aspx, but that's asking me for a user/password login, which I don't have. Maybe somebody else here does?

but I had the same problem accessing the utility.

It would appear that there is still some opencl ATI driver somewhere causing a problem. The offending system is a old Tyan S2892 server with onboard ATI video (unused and disabled via jumper) with two gtx460 and win7x64pro. It would appear I will have to use some driver cleaner to get rid of ATI opencl. Alternately, there is some other problem. Maybe a guru can spot another problem here. This system runs 7.0.28. I can try a newer version of BOINC. It seems to me that either the project or BOINC should be handling this issue.

I tried removing the AMD catalyst set using "express uninstall manager" but am still having apparent opencl problems on project milkyway. There was a suggestion at their forum

Many people in BOINC-land advise that you remove all traces of all AMD drivers before you change to a different one. AMD seem to have heard them, and produced a driver removal tool. The only official-looking link Google could find for me is http://sites.amd.com/us/game/downloads/Pages/catalyst-uninstall-utility.aspx, but that's asking me for a user/password login, which I don't have. Maybe somebody else here does?

but I had the same problem accessing the utility.

The url you require is:

http://www2.ati.com/drivers/amd_cleanup_util_1.2.1.0.exe

Claggy

Thanks Claggy. Unfortunately, it didn't help. After rebooting I "resumed" milkyway and got some more tasks and they immediately had a computer error. The project has already been reset. I dont know what else to try. Currently the GPUs work fine with collatz and prime grid. The same type of milkway "fit" work units all process fine on another pair of gtx460 but that system is a new one with an intel cpu and never had any ATI stuff.

I put two links above for working and non working but I dont see any difference like library names, versions, etc, so I am at a loss as to what is happening.

EDIT - Just detached and re-attached. Still getting errors and giving up on MW for this system.

Things to try:

Update to Boinc 6.0.64 or 7.2.26, Boinc 7.0.28 has Nvidia GPU memory reporting Bugs, as well as others.

Try reinstalling the Nvidia Graphics drivers, Uninstalling the AMD drivers will remove OpenCL.dll (at least the uninstall utility will), not so sure if Nvidia drivers install it any longer.

Try installing the APP SDK, that should install OpenCL.dll, you'll want to do it with a recent Boinc, older Boinc might freak out.

Claggy

Try installing the APP SDK, that should install OpenCL.dll, you'll want to do it with a recent Boinc, older Boinc might freak out.

The APP SDK may make your system very unstable as well. In this case it may be better to install the Intel OpenCL ICD.

(I linked to the 64bit ICD for 1st and 2nd Generation Intel Core CPUs. If one wants to use the version for the 3rd and 4th generation, start here.)

---

Umm:

AuthenticAMD
Dual Core AMD Opteron(tm) Processor 285 [Family 15 Model 33 Stepping 2]

Claggy

Uhm, when I look at the system he points to: http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=543305:

CPU type GenuineIntel
Intel(R) Core(TM)2 CPU 6400 @ 2.13GHz [Family 6 Model 15 Stepping 6]

Oh wait, I was looking at the system under his 'fine' link. Sorry. Missed the other link. :-(
In that case the APP SDK is the way to go, yes.

---

Solved - Thanks!

Re-install of nVidia driver fixed it.

Took a while to verify that the driver install worked because of MURPHY'S LAW. After the re-install I happened to look at the event log.

A quick check of the event log showed numerous memory errors that were corrected as well as a huge number of IDE controller (not disk) errors. There was no drive on the controller port generating the error which was suspicious. Opening the case (probably 6 months since last time) showed several of the ERCC modules were lifting out of their socket. The locking clips were clearly moved back from their locked position. I pulled all the chips and cleaned them and blew out the slots. One pad on a stick was partially corroded but I managed to clean it and booted "memtest86" with ercc enabled and all sticks passed. I cloned the boot disk from IDE to SATA and disabled the IDE entirely. Took 10 hours before system was back processing. MW tasks are running just fine. I need to check event logs on headless systems more frequently.