Are you running anything like NVIDIA inspector or Afterburner or Precision X now?

I've spent HOURS tonight trying to get TWO Windows 7 machines to accept the 320.18 drivers. Windows kept interfering with update after update and it took multiple re-installs to clear all the problems.

What I noticed was that in each case Precision X was barfing, by which I mean doing weird and unusual things like synching cards that had not been synched.

And every attempt to upgrade the driver caused Windows to try to install drivers or an NVIDIA Windows updated driver *in the middle of the reboot after a clean installation*.

Yes, of course I know how to prevent that from happening, but I didn't know I needed to prevent it from happening until it was happening.

My only guess for you would be to update Windows *thoroughly* leaving nothing for it to do, and start from scratch.

I was *not* a happy camper. I wanted it to do what *I* wanted it to do, not whatever Microsoft thought it should do to idiot-proof the process.

Yes I am running after burner(AB) 2.31 as I'd stated, no I'm not willing to start over with a clean slate on My hdd, I've had gpu down clocks in the past and the occasional Computation (197,) error. I only have one GTX580 card of course. I've seen that, Microsoft tries to install 311.06 I'd let that install and then I went with a clean install of 301.42, since I don't like Microsoft pushing drivers, but when stopping an install won't work, I'll tolerate the install for as long as I have to and then do what I want. Sorry but upgrading the driver in this case is not required, as your problem and mine are not the same, but then I read that thread.

Yeah, I've been seeing the threads.

I wasn't suggesting a clean-slate HDD, I was talking about the drivers.

There was a thread, I don't remember how long ago but in the last six months, that Jason was talking about some Windows update that looked for all the world like a NVIDIA driver update to me, but he actually suggested letting it update for some tangential reason. I do not argue with Jason. I say, "Yes, sir," and hit "install the update."

I don't remember what thread and it's gotten too wee-hoursly for me to look now.

But that's why I was suggesting that you just let Windows do whatever it wants as far as updating-goes, then do whatever you want.

However... I see that what you want to do is go back several NVIDIA driver versions for whatever reason. I'm not sure you don't "undo" whatever that Windows update-thing was that Jason recommended allowing to happen.

Of course, I don't care what you choose to do, I was just giving a recurring problem you seem to be having my best shot in the dark.

The *FIRST* thing I'd do is give-up Afterburner. That's based on a very frustrating experience *I* once had on ONE computer, ONCE, but your mileage, as they say, may vary.

I don't even remember what log-file I was looking in to find what lead me to Afterburner, but it was in a Windows log-file. Them may be lousy breadcrumbs, but they're all I have to offer.

<edit> By the way -- Windows is getting to be like AOL was in the early years - every time you touch the darn thing it wants to "take control" and do whatever it wants for as long as it wants. If you interfere, it lets you know how much it hates you. I **really** find that irritating.</edit>

I will be quiet now and possibly forevermore.

I think you have you logic backwards.

GPU crashes/driver restarts often lead to downclocks, which in turn lead to 197 time limit exceeded errors.

IOW checking your windows event viewer you probably see a driver restart event before you see the task erroring out in the Boinc eventlog.

Downclocks usually require reboots.

I don't do hardware, but I'm getting the feeling you're either trying to push your cards too hard or there's a reason they ended up on ebay in the first place.

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

see here. Maybe helpful
http://setiathome.berkeley.edu/forum_thread.php?id=71748&postid=1372327

http://setiathome.berkeley.edu/result.php?resultid=3015670106
looks like it ran concurrent and it has

Error on call (cudaMemcpy(PowerSpectrumSumMax, dev_PowerSpectrumSumMax, (cudaAcc_NumDataPoints / fftlen) * sizeof(*dev_PowerSpectrumSumMax), cudaMemcpyDeviceToHost)), file c:/[Projects]/__Sources/sah_v7_opt/Xbranch/client/cuda/cudaAcc_summax.cu, line 252: unknown error
Exiting

memcpy errors are symptomatic of such driver hiccups - that task went into temporary exit and recovered.

without access to anything I'd say the task died within a driver restart and for some reason didn't go into any of the failsafes, but was eventually forcefully terminated by boinc. Those coredumps are more difficult to read than tealeaves, but with no expertise whatsoever I'd wager the app decided to wait for Godot.

anyway hiccups happen on the best of systems. be glad it doesn't downclock, forcing you to reboot.

Unless errors happen more freqyently or can be traced to single events it's not much use to grow gray hair over them.

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

Last night nothing unusual happened w/301.24 and 7.0.64 w/the environmental variable(anon platform), so tonight I'm using BoincTasks 1.47 instead of BoincManager cause I can read the text there, BoincManager's text is almost microscopic in size, I wish the text in BoincManager was adjustable size wise(like ctrl+ or ctrl-).

Jord maintains a boinc wishlist on the boinc message boards though I can't see a manager specific wishlist.
He passes on such suggestions, if they look good. You are probably not the only person that could do with an option of larger fonts (I find I'm increasing zoom in word lately), so good chance of at least having it considered. Whether it's implemented then depends on the boinc devs finding it worthwhile too and ease of coding.

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

Last night nothing unusual happened w/301.24 and 7.0.64 w/the environmental variable(anon platform), so tonight I'm using BoincTasks 1.47 instead of BoincManager cause I can read the text there, BoincManager's text is almost microscopic in size, I wish the text in BoincManager was adjustable size wise(like ctrl+ or ctrl-).

Jord maintains a boinc wishlist on the boinc message boards though I can't see a manager specific wishlist.
He passes on such suggestions, if they look good. You are probably not the only person that could do with an option of larger fonts (I find I'm increasing zoom in word lately), so good chance of at least having it considered. Whether it's implemented then depends on the boinc devs finding it worthwhile too and ease of coding.

There is also a longstanding request (well, a series of linked requests, really) in the trac ticket system. [trac]#857[/trac], for example, and [trac]#147[/trac].