Troubleshooting very occasional . . annoying re-boot and GTX 570

[BonzaiDuck]

OK, here's the situation.

i7-2600K @ [Turbo] 4.60 Ghz
[idle EIST VCORE 1.008V; light-&-heavy load VCORE 1.35+ to 1.32+ to 1.33V depending on benchmark SW.]
[ VCCIO = 1.125V and see RAM below. Other settings at stock or fixed at known stock values. PLL Overvoltage Enabled -- awaiting my schedule for simply replacing the PLCC/BIOS-chip with ASUS P8Z68V-Pro with V. 706. Mild power settings: over-current 110%.]
4x4GB (16GB) G.SKILL "-GBRL" DDR3-1600 @ 1866 10-10-10-28 1T
RAM or VDIMM voltage ~ >= 1.53V (spec V = 1.50V)


EVGA GTX 570 ~1.3GB [as-is, all stock settings]
Lucid Virtu deployed; all settings of Intel graphics 3000 "Auto" or
stock/as-is, configured for Lucid.

Note: No SW errors with Lucid in last six months. Minor errors seemed to be very correlated with adjustments to features within HDCP.

Seasonic 750W Modular 80+ "Gold" model.

Patriot Pyro ~64GB w 59GB allocated to ISRT caching/acceleration for:

Western Digital 600GB VelociRaptor. -- Accelerated.
Western Digital 500GB Caviar Black -- Marvell controller. [All media streaming and capture is cached or buffered on this drive, and all DVR activity goes to this drive]

Main desktop monitor: Viewsonic 1680x1050max LCD [VS 2235 ??] connected by VGA analog.

Windows Media Center configured for 42" LED-LCD via HDMI passing through ONkyo AV Receiver

Tests: mid-summer, 2012: HCI Memtest: 1300% coverage, 5-6days.
LinX 30 iterations and minimum variation GFLOPS. Prime95 sFFT 10hrs, lFFT 18hrs and Blend 8 hrs -- 0/0 manually terminated.

Long Term Behavior: Build completed August 2011. Most stable, error-free, least troublesome and by far the fastest, happiest system I ever had.

System is used for TV viewing on the LG HDTV almost 24/7, and system downtime may be 1 hour monthly average. All other work and games currently configured for the VSonic monitor. Never had problems with running Media Center on the LG while playing a car racing game or any of two flight simulators, with IE and several office and/or graphics programs open.

I have begun to notice with a frequency of maybe every two weeks, that the system suddenly reboots by itself. Barring this, I can leave the system running for the good part of a week without reboot for some routine reason.

But this "unscheduled reboot" with the indicators in the System log reminds me of an 8800 GTS card I once had in another machine, similarly configured for dual-monitor with HD. Apparently I had overclocked the nVidia gfx with the most commonly used software known applicable for nVidia at that time (but momentarily forget the name). I apparently forgot to reset the video card on that system when installing new OS w delete/repartition and reformat of the boot HDD. And -- I am confident that my OC settings on the 8800 were not perfect -- it was an early experience in OC/test of video cards, and I since decided it wasn't worth the trouble. Took me a year to figure out the cause. That old system would either BSOD or simply reboot every week or so under the same type of usage.

The GTX 570 seems to run at <= 60C at 75F room ambient.

I'll listen to any suggestions on how to proceed to determine and eliminate the cause of this behavior. I will have to reset the BIOS anyway, for the update via chip replacement.

 

[Attic]

What exactly is event viewer showing leading up to the reboot?

This behavior reminds me of the same issue I had on my rig. I believe it was due to the video card (slight OC) but the behavior went away before I could isolate it.

 

[BonzaiDuck]

What exactly is event viewer showing leading up to the reboot?

This behavior reminds me of the same issue I had on my rig. I believe it was due to the video card (slight OC) but the behavior went away before I could isolate it.

It's got fairly robust consistency.
========
Event ID 41 level criticial Source Kernel-Power

General: The system has rebooted without shutting down first. This error could be caused if the system stopped responding, crashed or lost power unexpectedly.

Frequency: Three observations on 11/19/12, 12/9/12, and 12/23/12.


========

[It apparently has nothing to do with some operation like DVR Recording. It had happened before off an older APC 1200 VA UPS with about the same frequency, and I had moved the connection to a new CyberPower ~1200 VA. It always happens that on the day the system does this unexpected reboot or shutdown "without Windows," I will note a momentary "on battery" beep-warning from either UPS. And it did that -- last night -- about an hour or two before it unexpectedly rebooted through the menu options for "Safe boot" through "Normal" that you would expect. I thought I had adjusted both batteries to be slightly more tolerant of the switching.]

 

[BD231]

I had to raise my offset to get the low power state voltage up a bit myself, CPU wasn't getting enough juice when in an idle.

 

[BonzaiDuck]

I had to raise my offset to get the low power state voltage up a bit myself, CPU wasn't getting enough juice when in an idle.

I remember that such a mild caution was noted in the Anandtech article of 2007 which was both analysis and guide to tweaking the QX9650 -- which was (I think) the flagship Conroe or Kentsfield prior to release of WolfDale and Yorkfield. It was this article which explained the relationship between Offset and the overshoot between load and idle states for VCORE. More emphasis was given to the high momentary voltage at transition, but there was mention of low-power, low voltage instability as another manifestation in connection to LLC settings.

The point of it: You could get absolute stability under the load state, but the transition to a low power state could cause instability or a system crash.

Do you have anything else to add in sharing your experience here? I am all ears . . .

Certainly, it wouldn't hurt for me to bump up my Offset by another two notches -- from +0.005 to +0.015 . . .

 

[coffeejunkee]

You say idle voltage is 1.008, should be plenty for 1.6GHz right? I expect C3/C6 states to be a more likely culprit with a 4.6 Ghz oc, that is if you have them enabled ofcourse.

First thing I would do myself is set everything back to stock and see if the reboots still happen.

 

[BD231]

I remember that such a mild caution was noted in the Anandtech article of 2007 which was both analysis and guide to tweaking the QX9650 -- which was (I think) the flagship Conroe or Kentsfield prior to release of WolfDale and Yorkfield. It was this article which explained the relationship between Offset and the overshoot between load and idle states for VCORE. More emphasis was given to the high momentary voltage at transition, but there was mention of low-power, low voltage instability as another manifestation in connection to LLC settings.

The point of it: You could get absolute stability under the load state, but the transition to a low power state could cause instability or a system crash.

Do you have anything else to add in sharing your experience here? I am all ears . . .

Certainly, it wouldn't hurt for me to bump up my Offset by another two notches -- from +0.005 to +0.015 . . .

I had no issues with my 2550 in an idle state for quite some time before it actually started causing issue, and modern video cards typically reset themselves not your entire computer. I don't use dual monitors though.

 

[ShintaiDK]

Looks to be related to your CPU OC. Did you try stock settings to either eliminate or confirm it?

 

[BonzaiDuck]

Looks to be related to your CPU OC. Did you try stock settings to either eliminate or confirm it?

As I said -- I have to reset everything to replace the PLCC BIOS chip, anyway. (Or ditto, if I were just planning to flash the existing with the new version.) I may get that done between Xmas an New Year.

I agree about the possible source. But it's not a result of the OC under "load" per se, but more likely what our colleague suggested -- likely a result of using LLC without considering the low-power state. The occurrence is so infrequent it will be hard to confirm in a short time. Even so, I can bump up the offset a notch or two and then . . . wait two weeks under my normal operating conditions . . .

I should've posted this on the CPU/OC forum, but thought it might be related to video because the video card is always "in use" and I had a similar problem with the old Kentsfield system. But I explained that I finally cornered the source of that problem. And it was an occasional BSOD -- not a "reset." So . . . . yeah . . .