LogMeIn might have a lot of registry entries that are not removed following an uninstall, and security changes that aren't reset. Some of these programs create their own accounts and take ownership of some files.
Were you ever using the BAM! account manager on that system?

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

LogMeIn might have a lot of registry entries that are not removed following an uninstall, and security changes that aren't reset. Some of these programs create their own accounts and take ownership of some files.

I know. For the LogMeIn test, I'll uninstall it, restart the PC, run a registry cleaner, then restart again. The software does not create/use its own Windows user account, but it does install a tray program that controls a Windows Service, that uses the "LogMeIn Mirror Driver" display adapter.

Were you ever using the BAM! account manager on that system?

Yes, I usually use the BAM! account manager. But for these GPUGrid specific tests, I had been attaching to the project directly, without using the account manager. Who do you ask? Seems irrelevant to me.

So, for this test... Now that I still have fresh nVidia 314.22 WHQL drivers (from scratch), I wanted to test if cleaning the BOINC registry settings would solve the problem.

So I:
- Made sure the GTX 460 was removed
- Uninstalled BOINC
- Ensured the following 2 directories were deleted:
* C:\Program Files\BOINC
* C:\ProgramData\BOINC
- Deleted the following 3 Windows Local User Groups:
* boinc_admins
* boinc_projects
* boinc_users
- Ran CCleaner to clean up temporary file directories, and also I manually cleaned them too.
- Restarted the PC
- Did a registry search for anything "BOINC" and anything "Berkeley". For the following entries, which I deemed safe to remove, I exported the registry branch, then deleted the branch/keys:
* Several BOINC installation .exe entries within [HKEY_CURRENT_USER\Software\Microsoft\Windows NT\CurrentVersion\AppCompatFlags\Compatibility Assistant\Store]
* [HKEY_CURRENT_USER\Software\Space Sciences Laboratory, U.C. Berkeley]
* Some startup entries for boincmgr and boinctray within [HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Explorer\StartupApproved\Run]
* [HKEY_LOCAL_MACHINE\SOFTWARE\Space Sciences Laboratory, U.C. Berkeley]
* [HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\Space Sciences Laboratory, U.C. Berkeley]
- Restarted the PC
- Installed BOINC (using the default settings, which were: allow screensaver, don't do a service install, and allow all users to configure)
- When prompted to restart Windows, clicked Yes to restart
- BOINC manager/client started
- Attached to GPUGrid.net project
- Changed BOINC settings to the following (which did not require the creation of a global_prefs_override.xml file):
* Activity -> Run always
* Activity -> Use GPU always
* Activity -> Network activity always available
- BOINC got a Nathan dhfr36 task, assigned instance 0 slot 0
- It completed immediately, exhibiting the behavior.

Alright, so this means that the problem cannot be related to:
- BOINC users/groups/permissions from prior installs (since I had deleted the users/groups)
- BOINC directories/files from prior installs (since I had deleted them)
- BOINC registry keys from prior installs (since I had deleted them)

Dangit. The next test will be the LogMeOn test. I think it might be my last test, before an OS reinstall, so... I hope the problem is a conflict between NATHAN apps and LogMeOn (either the software or the "LogMeIn Mirror Driver" Display adapter)

Regards,
Jacob

So, for this test... Now that I still have fresh nVidia 314.22 WHQL drivers (from scratch), and I verified the issue should not be caused by files/users/regkeys from prior BOINC installations... I wanted to test if uninstalling LogMeIn would solve the problem.

So I:
- Made sure the GTX 460 was removed
- Uninstalled BOINC
- Ensured the following 2 directories were deleted:
* C:\Program Files\BOINC
* C:\ProgramData\BOINC
- Deleted the following 3 Windows Local User Groups:
* boinc_admins
* boinc_projects
* boinc_users
- Ran CCleaner to clean up temporary file directories, and also I manually cleaned them too.
- Restarted the PC
- Did a registry search for anything "BOINC" and anything "Berkeley", removing keys/branches that I deemed appropriate and safe.
- Uninstalled LogMeIn v4.1.2600 from 12/14/2012, using Programs and Features
- Restarted the PC
- Installed BOINC (using the default settings, which were: allow screensaver, don't do a service install, and allow all users to configure)
- When prompted to restart Windows, clicked Yes to restart
- BOINC manager/client started
- Attached to GPUGrid.net project
- Changed BOINC settings to the following (which did not require the creation of a global_prefs_override.xml file):
* Activity -> Run always
* Activity -> Use GPU always
* Activity -> Network activity always available
- BOINC got a Nathan dhfr36 task, assigned instance 0 slot 0
- It completed immediately, exhibiting the behavior.

So even without the LogMeIn software, the issue persists. But the uninstallation of LogMeIn curiously did NOT remove the "LogMeIn Mirror Driver" Display adapter. It also left a couple regkeys behind, according to CCleaner. So, the next test will involve removing that display adapter, and removing the leftover LogMeIn regkeys using CCleaner.

This better work....

Regards,
Jacob

So, for this test... Now that I still have fresh nVidia 314.22 WHQL drivers (from scratch), and I verified the issue should not be caused by files/users/regkeys from prior BOINC installations... and I had LogMeIn uninstalled, I wanted to test if removing that "LogMeIn Mirror Driver" display adapter and the leftover LogMeIn regkeys would solve the problem.

So I:
- Made sure the GTX 460 was removed
- Uninstalled BOINC
- Ensured the following 2 directories were deleted:
* C:\Program Files\BOINC
* C:\ProgramData\BOINC
- Deleted the following 3 Windows Local User Groups:
* boinc_admins
* boinc_projects
* boinc_users
- Ran CCleaner to clean up temporary file directories, and also I manually cleaned them too.
- Did a registry search for anything "BOINC" and anything "Berkeley", removing keys/branches that I deemed appropriate and safe.
- Used CCleaner to remove the leftover LogMeIn registry keys
- In Device Manager, right-click uninstalled the "LogMeIn Mirror Driver" display adapter, and Windows prompted to restart.
- Restarted the PC
- Installed BOINC (using the default settings, which were: allow screensaver, don't do a service install, and allow all users to configure)
- When prompted to restart Windows, clicked Yes to restart
- BOINC manager/client started
- Attached to GPUGrid.net project
- Changed BOINC settings to the following (which did not require the creation of a global_prefs_override.xml file):
* Activity -> Run always
* Activity -> Use GPU always
* Activity -> Network activity always available
- BOINC got a Nathan dhfr36 task, assigned instance 0 slot 0
- It completed immediately, exhibiting the behavior.

Dang.

I've only got 2 other ideas/tests.
- I'd like to uninstall a couple more applications, and close all programs, before starting BOINC, to see if that helps.
- Also...
I first noticed this issue on 4/1/2013. If you look at the driver listings and dates:
www.nVidia.com -> Drivers -> All NVIDIA Drivers -> Beta and Older Drivers ->
Product Type: GeForce
Product Series: GeForce 600 Series
Product: GeForce GTX 660 Ti
Operating System: Windows 8 64-bit
Language: English (US)
Recommended/Beta: All

... you'll see that 314.22 WHQL'd on March 25, 2013.

That's awful close to when I started noticing the problem.

So, I'd also like to do a test where I uninstall nVidia completely, then reinstall the prior WHQL drivers, 314.07, to see if that makes a difference.

Regards,
Jacob

So, for this test... Now that I still have fresh nVidia 314.22 WHQL drivers (from scratch), and I verified the issue should not be caused by files/users/regkeys from prior BOINC installations... and I had LogMeIn uninstalled and there were no remains (no "LogMeIn Mirror Driver" display adapter and no leftover LogMeIn regkeys)... I wanted to test if any programs/services (like eVGA PrecisionX, or StarDock DeskScapes) would be the cause.

So I:
- Made sure the GTX 460 was removed
- Uninstalled the following:
* Stardock DeskScapes 8
* Stardock ModernMix
* eVGA OC Scanner
* eVGA PrecisionX
- Restarted the PC
- Closed BOINC
- Closed the following Task programs:
* Trading Paints
* Skype
* GTalk
* Logitech Gaming Software
* SkyDrive
* Core Temp
- Stopped the following services:
* iRacing Helper Service
- Ended the following tasks using Task Manager:
* Skype Update
* Hallmark Card Studio Planner (PLNRnote.exe)
* ZuneLauncher.exe
- Started BOINC
- Allowed BOINC to get a GPUGrid task
- BOINC got a Nathan dhfr36 task, assigned instance 0 slot 0
- It completed immediately, exhibiting the behavior.

Dang.

So, that leaves... what? Rolling back to 314.07 as the last test?
Here goes...

So, I uninstalled back to "Basic Microsoft Display Adapter", then cleanly installed the 314.07 drivers, to test with them.

So I:
- Made sure the GTX 460 was removed
- Uninstalled BOINC
- Ensured the following 2 directories were deleted:
* C:\Program Files\BOINC
* C:\ProgramData\BOINC
- Deleted the following 3 Windows Local User Groups:
* boinc_admins
* boinc_projects
* boinc_users
- Uninstalled the 314.22 drivers
- Restarted the PC
- Closed the following Task programs:
* Trading Paints
* Skype
* GTalk
* Logitech Gaming Software
* SkyDrive
* Core Temp
- Stopped the following services:
* iRacing Helper Service
- Ended the following tasks using Task Manager:
* Skype Update
* Hallmark Card Studio Planner (PLNRnote.exe)
* ZuneLauncher.exe
- Installed BOINC (using the default settings, which were: allow screensaver, don't do a service install, and allow all users to configure)
- When prompted to restart Windows, clicked Yes to restart
- BOINC manager/client started
- Closed BOINC
- Closed the tray programs, stopped the service, and ended the tasks
- Reopened BOINC
- Attached to GPUGrid.net project
- Changed BOINC settings to the following (which did not require the creation of a global_prefs_override.xml file):
* Activity -> Run always
* Activity -> Use GPU always
* Activity -> Network activity always available
- BOINC got a Nathan dhfr36 task, assigned instance 0 slot 0
- It completed immediately, exhibiting the behavior.

I am now stumped, and frustrated.

For you guys that have a GeForce 660 Ti on 314.07 or later drivers, if you get rid of all your tasks, and restart BOINC, and then allow a NATHAN task such that it executes in slot 0, does it work properly for you?

Sigh.

For completeness, I also tested a clean install of 310.90 (which was tricky, because Windows 8 seems to want to auto-install 311.06 if Windows Update downloads it, probably because of my System -> Hardware -> Device Installation Settings)

Anyway, 310.90 also exhibited the behavior. This issue does not appear to be related to driver versions.

I have exhausted everything I can think of within this OS installation. After I verify all of my data is backed up, my next test will be on a clean reinstallation of Windows 8 (/me not happy).

I again ask, though:
For you guys that have a GeForce 660 Ti on 314.07 or later drivers, if you get rid of all your tasks, and restart BOINC, and then allow a NATHAN task such that it executes in slot 0 (as verified in Task Manager), does the task process properly for you?

Thanks,
Jacob

Well... I noticed that, when viewing the properties of some of the BOINC folders, the Security tab would take 5-8 seconds to evaluate the security descriptors for the 3 BOINC Windows user groups.

So, I set both my hard disks for chkdsk, via:
chkdsk c: /f /v
chkdsk d: /f /v
... telling it to run on restart.

The scan results for C: had a section that said:
Cleaning up 2544 unused index entries from index $SII of file 0x9.
Cleaning up 2544 unused index entries from index $SDH of file 0x9.
Cleaning up 2544 unused security descriptors.
CHKDSK is compacting the security descriptor stream

The scan results for D: had a section that said:
Cleaning up 96 unused index entries from index $SII of file 0x9.
Cleaning up 96 unused index entries from index $SDH of file 0x9.
Cleaning up 96 unused security descriptors.

This all sounded promising, but when I restarted, reinstalled BOINC, etc., the issue still occurred.

I noticed, though, that if I left a BOINC folder (like projects) properties window open on the security tab, then opening an additional properties window to the security tab was instantaneous.

So I tried leaving a window open, while getting another Nathan task, the issue still occurred.

I suggest you give it one last try and if that fails, perform a new W8 installation. Try with UAC off, the Firewall off, any AV disabled and adjusting for best performance (system properties, advanced, performance settings).
GL

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

I have found the problem/solution!

So, after getting some really good sleep, I woke up and it occurred to me that I could use Process Monitor to log what the acemd process actually does.

So, while doing that, I noticed in Process Monitor, that while it was reading from:
C:\ProgramData\BOINC\slots\0
it was sometimes writing to
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots\0

I did a little digging about the Virtual Store, and found this article, which basically says that Windows will use the Virtual Store whenever the directories already exist there.
http://social.msdn.microsoft.com/Forums/en-US/windowssecurity/thread/004743cd-9151-40cf-bad2-936de0e8eefb/

So... I unattached the GPUGrid project (so BOINC was now not attached to anything), then closed BOINC, and looked at the VirtualStore directory. I found the following BOINC-related directories/files still existing there:

C:\Users\Jacob\AppData\Local\VirtualStore\Program Files (x86)
C:\Users\Jacob\AppData\Local\VirtualStore\Program Files (x86)\BOINC
C:\Users\Jacob\AppData\Local\VirtualStore\Program Files (x86)\BOINC\stderrgfx.txt
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1351159357_344406832_1_2
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1355671946_933565984_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1355671946_933565984_0_1
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366254_1593119541_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366254_1593119541_0_1
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366254_1593119541_0_2
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366626_328715521_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366626_328715521_0_1
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366627_2091527150_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366627_2091527150_0_1
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\boinc.fzk.de_poem\poempp_gpucrystal_1365366627_2091527150_0_2
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\291px34x3-NOELIA_Klebe_Equ-0-1-RND7876_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\291px34x3-NOELIA_Klebe_Equ-0-1-RND7876_0_11
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\291px34x3-NOELIA_Klebe_Equ-0-1-RND7876_0_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I11R20-NATHAN_dhfr36_3-21-32-RND7160_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I11R20-NATHAN_dhfr36_3-21-32-RND7160_0_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I16R67-NATHAN_dhfr36_3-25-32-RND6257_1_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I1R193-NATHAN_RPS1_respawn3-12-32-RND5365_1_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I1R193-NATHAN_RPS1_respawn3-12-32-RND5365_1_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I1R94-NATHAN_RPS1_respawn3-16-32-RND2805_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I1R94-NATHAN_RPS1_respawn3-16-32-RND2805_0_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I20R52-NATHAN_dhfr36_3-7-32-RND2909_1_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I20R52-NATHAN_dhfr36_3-7-32-RND2909_1_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I28R40-NATHAN_dhfr36_3-24-32-RND6728_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I28R40-NATHAN_dhfr36_3-24-32-RND6728_0_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I3R48-NATHAN_dhfr36_3-15-32-RND6324_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I3R48-NATHAN_dhfr36_3-15-32-RND6324_0_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I3R9-NATHAN_dhfr36_3-0-32-RND6788_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I3R9-NATHAN_dhfr36_3-0-32-RND6788_0_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I4R134-NATHAN_RPS1120528-105-166-RND6992_1_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I4R134-NATHAN_RPS1120528-105-166-RND6992_1_4
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I4R22-NATHAN_dhfr36_3-0-32-RND5427_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I4R22-NATHAN_dhfr36_3-0-32-RND5427_0_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I5R2-NATHAN_dhfr36_3-30-32-RND6269_1_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\I5R2-NATHAN_dhfr36_3-30-32-RND6269_1_9
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_1039_run3-NOELIA_RL4_equ-0-1-RND2306_1_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_1039_run3-NOELIA_RL4_equ-0-1-RND2306_1_11
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_1039_run3-NOELIA_RL4_equ-0-1-RND2306_1_4
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_28_2-NOELIA_RC3_equ-0-1-RND0669_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_28_2-NOELIA_RC3_equ-0-1-RND0669_0_11
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_28_2-NOELIA_RC3_equ-0-1-RND0669_0_4
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_601_run1-NOELIA_RL2_equ-0-1-RND1963_0_0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_601_run1-NOELIA_RL2_equ-0-1-RND1963_0_11
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\projects\www.gpugrid.net\trypsin_lig_601_run1-NOELIA_RL2_equ-0-1-RND1963_0_4
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots\0
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots\0\output.restart.coor
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots\0\output.restart.idx
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots\0\output.restart.vel
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots\0\output.restart.xsc
C:\Users\Jacob\AppData\Local\VirtualStore\ProgramData\BOINC\slots\0\stderr.txt


Notice how slots\0 exists in this VirtualStore, but no other slots exist.

So I moved those 2 BOINC directories out of Virtual Store (and into my personal folder where I've been tracking this issue).

Then I started BOINC, attached GPUGrid, watched it get a Nathan dhfr36 task, and IT STARTED PROCESSING NORMALLY (and I was watching the VirtualStore directory, and BOINC directories were not being recreated). I removed GPUGrid, closed BOINC, reopened BOINC, reattached, got another Nathan dhfr36 tasks (just to be sure), and it too processed normally!

So, I think this issue is fixed.

The cause is:
Somewhere, either BOINC or the acemd app or Windows itself gets confused about properly handling the situation of a VirtualStore directory pre-existing.

The solution is (assuming you don't require the VirtualStore):
Close BOINC, and remove any BOINC directories that are within the VirtualStore directory found here:
C:\Users\{UserName}\AppData\Local\VirtualStore

Thanks for all your help guys!
I'll keep monitoring tasks to verify the fix!

PS: For me, I think the VirtualStore BOINC directories were created late March early April, when I created a non-admin Windows user and tested BOINC with it, to see if non-BOINC-admins could control BOINC. A possible bug might be that we (BOINC) might not be properly cleaning the VirtualStore directory at certain times, not sure.

Thanks again,
Jacob (very very happy)

Note: Despite my finding the cause, I still really think the admins should make 2 corrections:
1) Fix the Nathan tasks so that, when this issue happens, the tasks are not completed immediately as successful.
2) Fix the validator to do additional verification, so that tasks are not erroneously marked as valid, and full+bonus credit is not erroneously granted.

Jacob, glad you found the cause. I looked at my Win7/64 machines and only 1 had a VirtualStore directory, and that was empty. I wonder if this mishandling of VirtualStore is a new feature with Windows 8. There has been grumbling about Windows 8 not working well with certain projects. Haven't paid too much attention to it though as I have little desire to use Win8.

VirtualStore has been around since Vista, I believe; it's not new. If anything is mishandling it, it's likely not Windows.

The issue is related to running BOINC under a non-Administrator account, which I did at some point in March, as part of my BOINC alpha testing. David Anderson and Rom Walton, BOINC developers, believe it is a BOINC problem, and have already created a possible fix for the issue. In the next week, I'll be testing it out.

GPUGrid developers should still close up the loophole so that tasks are not erroneously marked successful, and also they should fix the validator to perform additional checks before marking results as valid and granting credit.

Rom's most recent email describing the possible nature of the problem is:

Subject: RE: GPUGrid.net problem - Completing immediately, status successful, and granting credit
Date: Mon, 13 May 2013 12:46:10 -0400
From: romw@romwnet.org
To: jacob_w_klein@msn.com; davea@ssl.berkeley.edu

Jacob,

Thanks again for the bug report.

After going back over the thread, I suspect the actual trigger in your case was that you removed your current account from all the BOINC groups and from the administrators group. When BOINC launched, all files it opens for output were redirected to the VirtualStore. The next time you installed BOINC, the installer fixed the environment but the damage was done.

From that point on Windows was mixing and matching files between the two directory structures.

----- Rom

Wow Jacob - congratulations to this victory, achieved by hard and diligent work!

I was about to suggest you borrow some junk HDD and do a test reinstall there, so that you don't destroy your current installation just in case the issue persisted under the new OS. But I won't do that now ;)

MrS

---

Scanning for our furry friends since Jan 2002

Thank you! I'm really really happy.

I celebrated by spending some time tricking out my system even further.

It's a Dell XPS 730X, and it had been running quite hot, but I remembered that I could control chassis fan curves via an "XPS Thermal Monitor" program. (The nVidia Control Panel's ESA control Chassis section doesn't work well for me). Also, I found out how easy it was to create a custom GPU fan curve in eVGA Precision X. So, I've customized all those fan curves, and my system now runs a bit more noisy and a bit cooler. (Full-load temps are still 80-89*C for all the overclocked CPUs, and 67-78*C for all the overclocked GPUs... and I consider those temps to be good.)

So, with the new fan profiles, my "GTX 660 Ti 3GB FTW" can now stay clocked at base 1046 Mhz + max boost 195 Mhz = 1241 Mhz (it can because I make sure it stays below 70*C)... meaning more BOINC throughput! Previously, it usually downclocked itself due to temperatures, usually to 1215 Mhz.

AND... I decided to stop using my PCI-Express Creative Sound Blaster X-Fi Titanium sound card, to free up a slot (I'm switching to the onboard audio).... so that.... I can put a spare GPU in (I knew I could get 3 in there!)... a GTS 240, which cannot run GPUGrid/Poem/WCG, but CAN run SETI.

And I'm positive that some combination of
- uninstalling DeskScapes
- uninstalling LogMeIn
- ensuring GTX 660 Ti (connected to my display) is at base + max boost, below 70*C
- removing the Creative sound card and removing the Creative software
- installing the most recent Windows updates
... has resulted in Windows 8 being SUPER SILKY SMOOTH right now. It's a night and day difference, everything is bouncy and fluid again.

And yes, it's completely gorgeously smooth, even when using 0.001 cpu_usage via app_config (for both GPUGrid and SETI, since they both use GPU apps that have elevated process/thread priorities, and both could issue tasks that use very little CPU), and even when telling BOINC to use 100% CPU (since I don't want any bit of idleness left on the table).

8 CPUs, overclocked at 3742 Mhz, running 80-89*C
1 eVGA GTX 660 Ti 3GB FTW, clocked at overclocked base + max boost, so 1241 Mhz, running 67-70*C
1 eVGA GTX 460 OC 1GB, clocked at overclocked 763 Mhz, running 67-72*C
1 Dell GTS 240 1GB, clocked at 675 Mhz, running 70-78*C
12 GB overclocked RAM
BOINC runs 8 CPU tasks + (at least) 1 GPU task per GPU + 1 non-CPU-instensive task, all 24/7, while I also use it
1000 Watt power supply
Velvety smooth OS operation, despite being completely-fully-loaded
... and a huge power bill every month.

So impossibly happy.

And I'm positive that some combination of
- uninstalling DeskScapes
- uninstalling LogMeIn
- ensuring GTX 660 Ti (connected to my display) is at base + max boost, below 70*C
- removing the Creative sound card and removing the Creative software
- installing the most recent Windows updates
... has resulted in Windows 8 being SUPER SILKY SMOOTH right now. It's a night and day difference, everything is bouncy and fluid again.

I have confirmed that the reason the graphics feel so much nicer is because I am keeping the GPU's temperature below 70*C.

- When BOINC is not using the GPU, graphics are pretty smooth, even if the GPU gets downclocked from base (1046), to something like 350.
- When BOINC is using the GPU, my old fan settings would let it get quite hot (up to 86*C), and it would throttle down from base+maxboost (1241) to 1215. The graphics would be acceptable, but the mouse cursor felt heavy, and there was a small but noticeable amount of responsiveness lag, especially when scrolling within a window. Since the clock rate (1215) was still higher than base (1046), I suspect that the GPU's thermal response was doing something else in addition to simply downclocking.
- When BOINC is using the GPU, my new fan settings keep it below 70*C. The GPU runs at base+maxboost (1241), below 70*C. And the graphics have never felt better - they are now liquid velvet smooth, and supremely responsive.

And to think, I suffered that "heavy cursor" issue ever since I got the card in mid-January.
I am uber-happy to have this issue also fixed!

When the GPU gets too hot for the thermal design the GPU core frequencies get stepped down in steps of 13MHz . I suspect this creates some sort of imbalance and bottleneck that results in the lag.

A while back I posted about the temperature and clock rates, and suggested that people increase the fan speed as it prevents this sort of downclocking. I didn't experience any sluggishness from the system when the temps were too high/the clocks decreased, but this could be GPU, OS or driver specific. Also my PCIE3 controller is on the CPU and I almost always have a free CPU. Anyway, the issue of lag might just have been hidden on my system. I wonder if WU's are more inclined to fail when the clocks creep down like this. Might explain some peoples experiences using Linux without coolbits. Whatever the case, using fans to keep the GPU cooler helps keep the clocks up, the power usage down, prevents lag, and possibly improves stability. More reasons to use fan control/profile software for those who don't.

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

I'm not convinced the un-/smoothness is influenced by the GPU temperature changing from say 60°C to 85°C. Would you mind double-checking by switching back to the default fan profile?

I'd rather suspect a different type of WU is responsible. I've seen types change lately, and the smoothness (i.e. the intervals between individual work packages crunched by the GPU) depends on the complexity of the WU (the more atoms are simulated, the longer each time step takes, i.e. the longer screen refreshes are blocked).

MrS

---

Scanning for our furry friends since Jan 2002

I'm not convinced the un-/smoothness is influenced by the GPU temperature changing from say 60°C to 85°C. Would you mind double-checking by switching back to the default fan profile?

I'd rather suspect a different type of WU is responsible. I've seen types change lately, and the smoothness (i.e. the intervals between individual work packages crunched by the GPU) depends on the complexity of the WU (the more atoms are simulated, the longer each time step takes, i.e. the longer screen refreshes are blocked).

MrS

I just re-tested, per your request. This time, my 660 Ti was running a NOELIA unit (which is rarer than Nathans, and has a higher GPU Load than Nathans)

My test involved opening calculator (Scientific mode, plus History), and moving it around in a perfect circle at a high speed, while paying very close attention to how responsive it was being redrawn and how 'heavy' it felt. I'm sure a better test would be to grab a complicated window, like a browser with many tabs, and drag it around in a perfect circle quickly.

I have, again, confirmed the following behavior:
- I put the fans on standard stock profiling
- As the temps went up, it downclocked from 1241 to 1228, and my trained eye could tell that the paint refreshes were slightly slower
- As the temps went higher yet, it downclocked to 1215, and my trained motor skills could actually feel in the mouse the perception that it was heavier to move, an effect created because of the increased lag in video response.
- I put the fan profiles to maximum to bring the temp down
- Once it got down to 70*C for a couple of seconds, it throttled back up to 1241. Now the calculator motion was velvety smooth with instant responsiveness, and the mouse felt frictionless.

I hope that's the test you wanted. This is very real. I'm even noticing that video response whilst I'm typing here, i.e.: the response time needed to place the character, is definitely improved. And scrolling up/down never felt more responsive. And OS animations (like Start Menu, or fade in/out of new or closed windows, fade in/out of Jump Lists, etc.), is gorgeous.

You might be able to test it yourself, to prove it to yourself.

Note: I have heightened senses, so I'm sure some people may not notice it at all. But to me, it's very very clear. I now fully endorse keeping your GPU temps below any temp that would affect the GPU clock.

Side note: When the new non-sluggish behavior initially presented itself a couple days ago, I was completely baffled as to what was going on. For a little while, I had thought that the feeling of fluidity was because the bottom of my mouse was dirty enough to slide better. But, I cleaned it, and then realized it was all caused by the 70*C GPU threshold. :)

If we want to discuss the any more, we should create a separate thread.

Regards,
Jacob