BSOD only in BOINC

[yosssarian]

Hi,

My System:
Intel i-7 3930K 3.2 GHz (Overclocked to 4.3GHz!)
Cooler Master Glacer Liquid Cooler
Asus X79 Deluxe Motherboard
32 Gb Corsair Dominator DDR3 1600MHz Quad Channel CMP32GX3M4X1600C10
EVGA 850 P2 850W Platinum Power Supply
OCZ Vertex 4 SSD 256Gb
Seagate One Terabyte ST310003
Nvidia GTX 780
Creative SB X-Fi Fatal1ty Champion PCI-e Audio card
Coolermaster Cosmos 2 Case
LG 27EA83 27" IPS Panel
Windows 7 Ultimate 64

My issue, for which I'm seeking assistance, is CLOCK_WATCHDOG_TIMEOUT_c_PROC BSOD appearing in the morning after BOINC SETI has been running overnight, or in the afternoon if it's run while I'm at work. My overclock specifics are in the document attached. It also gives a little history.

I've tested several of the overclock permutations using intel burn, which is mentioned in the doc. I've also tested with 3dmark, and heaven benchmark 4.0 I have not once seen a blue screen during those tests, or have a test not go to completion, ever. In fact, I've never seen a blue screen while using the PC, even while playing a game, like Borderlands, for hours on end. I've never been able to break it. If I let BOINC run overnight it won't go more than a week before crashing.

I thought I had noticed a pattern with the crashes occurring certain days of the week when Acronis was backing up the C drive. I changed the day of the week in which the C drive backup ran, and thought I had fixed the problem as I went almost a month without a crash, but the same old computer **** happened where there's really good evidence that a problem has been resolved, only to see it return.

Many times all I would see in the morning, or when returning from work, would be a black screen, and a completely unresponsive computer. Turning off the power was all I could do. In those cases the event viewer was empty, and sometimes there was no dump file, or it was unusable. Note that prior to recently I was looking only in the windows folder for the RAM dump, and didn't check for the mini dump.

If any more info. is required, please let me know, and I'll be happy to provide. I'm very appreciative of the assistance, but I do make one request. Working in IT, I've had to deal with innumerable PC problems, so I ask that only those who are well informed in this type of problem respond. I really don't want to have to chase down all of the "have you trieds," and "I read somewheres." Sorry for how that sounds, but I did a lot of research on this, and came to this site specifically for the expertise of those who know.

Thanks.

 

[Mr.Scott]

How old is your OS install?
Most Watchdog Timer errors are a result of OS pollution. This is not specific just to BOINC but to any high computation program or benchmark.

 

[yosssarian]

How old is your OS install?
Most Watchdog Timer errors are a result of OS pollution. This is not specific just to BOINC but to any high computation program or benchmark.

Thanks. I did a bare metal install of win7 in June. You wouldn't put 3dmark, and heaven 4.0 in the same category as "high computation program or benchmark?"

 

[Mr.Scott]

Thanks. I did a bare metal install of win7 in June. You wouldn't put 3dmark, and heaven 4.0 in the same category as "high computation program or benchmark?"

No.
By high computation I mean BOINC, F&H, Etc.
The others are, to an extent, but much shorter duration.
Generally, you have a dedicated machine for BOINC and F&H. The more you load/do on those machines that aren't dedicated, the more chance you have of something conflicting with the program operation. Even on a fairly fresh install.
Crunchers are much more sensitive than benchers.

 

[Mandrake4565]

My experience has only been with F@H, though I assume it is similar to Boinc. When stress testing my Pc I use Prime 95 Blend and it works for me. With my overclocked machine I've been able to run F@H for the chimp challenge which I believe was 10 days without issue. That said I do know some people that find even Prime 95 blend isn't enough of a stress test to be able to run F@H 24/7 and usually a bump in Cpu V or Gpu V will take care of the issue.

 

[yosssarian]

My experience has only been with F@H, though I assume it is similar to Boinc. When stress testing my Pc I use Prime 95 Blend and it works for me. With my overclocked machine I've been able to run F@H for the chimp challenge which I believe was 10 days without issue. That said I do know some people that find even Prime 95 blend isn't enough of a stress test to be able to run F@H 24/7 and usually a bump in Cpu V or Gpu V will take care of the issue.

Folding at home is one of the applications you can run under BOINC, just like SETI, which is my choice. The two are very similar.

I was hoping to avoid having to redirect back to what had already been mentioned. You vaguely mentioned a bump in voltage for the CPU, but did you read the attachment in the original post mandrake? That includes the voltages that I have tried. If you have some information regarding the voltages I have used so far, that would be helpful. You also mentioned voltage for the GPU - do you really think that's relevant for the BSOD I said I was having?

 

[Mr.Scott]

You do realize that these are only suggested voltages. None of these are hard fast rules.
All silicon is not created equal. Your particular CPU may require a little more voltage. Same applies to your GPU.

Basically, what Mandrake is saying is that you have an unstable system for what you're doing.
Right now, I tend to agree.

 

[yosssarian]

Thanks. Yes, I know they're suggested voltages, and that no one knows for sure what will, or won't work. No reason we should be at that point after some 20 years of overclocking.

In that same document you'll notice that I documented having used several different CPU voltage + offsets, as well as VCCSA Voltage, and VTT CPU Voltage. Even though I have good water cooling the objective has been in large part to bring down the temps., while maintaining a stable system. Which, by the way, as I'd mentioned, it is. Never once a blue screen while using the PC. I'd call that stable, wouldn't you?

To get back to the matter at hand. Anyone stopping by with well-informed, specific information would be greatly appreciated.

 

[Mandrake4565]

Yes I read what you wrote, sorry if I gave off the impression I did not. Gaming and Benchmarking are no where near as stressful as running Seti and F@H, oh and I did not know one could run F@H in Bonic thought that was Seti only. You can game for hours and that still does not put the same demand on a system as those programs do.

Let me put it a different way and there are people why do not feel the same way as I do. For me Prime Blend is my go to stress test. I have tried numerous others, including Intel Burn, Aida 64, Occt etc. None of which I have found, I can do whatever I want on my Pc when overclocked and not have an issue. With what I do on my Pc it is running up to 10 hours a day normally around 60% and spikes at times to 100% for short periods of time. If my rig were to go down it can be very costly to me. I will normally find a daily overclock by using Prime Blend and then see if I can complete a few Wu with F@H. I have never had an issue running F@H after being able to pass 2+ hours of Prime Blend. I cannot say I've had the same success with the other stressing programs. This may also be because the F@H programs are similar to stressing with Prime.

What I feel you should do, because I do not see any mention of Prime in the stress tests you have run is, run Prime Blend. See if you can pass at least 2 hours of it.

 

[yosssarian]

Thanks Mandrake. But pass or fail, where would that leave me? I already know there is an issue. What I need is specific information to get the overclock working without a blue screen. I need well-informed recommendations on how I've overclocked the computer to guide me towards settings that work. It's always possible that 4.6GHz may be too much for this particular CPU.

I rarely ever post a problem, and would much rather just search for clues. I thought that with thousands of people who do a lot of over clocking I could catch the wisdom of a few gurus.

 

[Mandrake4565]

Well if the Pc passes Prime blend then maybe the issue does not lie in the Overclock settings. Maybe it lies in the program itself which you may find better answers in that section of the forum.

 

[Mr.Scott]

Even though I have good water cooling the objective has been in large part to bring down the temps., while maintaining a stable system. Which, by the way, as I'd mentioned, it is. Never once a blue screen while using the PC. I'd call that stable, wouldn't you?

Not if you can't run BOINC without BSOD, which I believe is why you're here.

 

[yosssarian]

Did I accidentally end up on the forum for conversation?
Mandrake - That's a good thought. However, I've had several versions since the win7 reinstall, and the behavior hasn't changed.
Mr. Scott - (big Star Trek fan here) That depends on your definition of stable. Only one application giving me headaches. I'd call that stable captain, and a wee bit more.

 

[Mr.Scott]

Mr. Scott - (big Star Trek fan here) That depends on your definition of stable. Only one application giving me headaches. I'd call that stable captain, and a wee bit more.

While I agree that stability is in the eye of the beholder pending what you're using the machine for, it would seem you use your machine for BOINC.....and it's not stable.
Not good enough Sonny Jim.

Easy enough to find out. Run BOINC at stock. If it doesn't BSOD.......your OC was unstable. No point in arguing stability.

 

[caddi daddi]

one of the problems i have run into with f@h is vcore sag when it finishes a wu causing me a bsod or a freeze, does seti have a log like folding that you can refer to that might give a clue as to when it crashes?

 

[yosssarian]

Yikes. From conversation to debate. That's not why I came here. Since it's my PC, allow me the final word. I did not assemble it for the purpose of running SETI. It's just one of many things I do with it. I'm rated in the top .1 of 1% in the US for completed units. It's just icing on the cake.

 

[yosssarian]

Thanks caddi. It has a log, but it records an hour or so of data. I would need days at least.

I remember reading somewhere that one of the CPU voltage settings on the doc I attached is helpful with V sag. Do you know which one, and do you know of a good setting for i7 3930k, on X79 platform?

 

[caddi daddi]

the rig I'm on will not open your doc so i can't tell.
I'll go have a read of your motherboar manual and find out.
I am sure that we will only be covering some file corruption as mr scott has said.

 

[caddi daddi]

looks like vcore manual voltage is what you need to select to highlight, then tap the + key to increase it, try twice.

 

[yosssarian]

caddi, Can't open a word doc? How about if I save it as a text file? It shows the numerous adjustments I've already made to vcore voltage. Let's try a little pasting.

AI Tweaker
Over Clock Auto
BCLK 100 w/TPU
Turbo ratio 43 (4.3GHz) 10/25 46 (4.6GHz)
Clock Gen Filter Enabled
Mem Freq. 1600MHz
Internal PLL over-voltage Auto

Digi Power Control
CPU Load Line Calibration Medium or High
CPU Current Capability 130%
VCCSA Current Capable 120%
VCCSA Load Line Cal. Regular High
CPU Voltage Frequency Auto
CPU Power Phase Control Optimized 2nd option
CPU Vcore Voltage Offset Mode
CPU Vcore Voltage 0.070 (was 0.050 till 11/7)
CPU VCCSA Voltage Offset Mode 10/25 Manual 1.15
CPU VCCSA Offset NA
CPU PLL Voltage Auto
VTT CPU Voltage 1.10000 10/25 Manual 1.15 (Yellow)
CPU Spread Spectrum Disabled
DRAM AB Current Cap 120%
DRAM CD Current Cap 120%
DRAM AB Power Phase Optimized
DRAM CD Power Phase Optimized
DRAM Frequency Auto

CPU Configuration
Intel Adaptive Thermal Monitor Disabled

CPU Power Management
CPU CIE and All Below Enabled

Monitor
Anti-surge Support Disabled
EZ Mode select right most Optimized

EPU Power save Mode Off
Post Delay Additional Time zero

CSM Auto

Boot / Secure Boot Menu OS type Windows UEFI
Fast Boot Enabled

Intel Burn Tests
2/11/2015 0.090 V= 1.392 @ 4.6GHz temps 60’s and 70’s.
3/9/2015 0.080 V=1.392 – 1.408 @ 4.6GHz, temps low 60’s.
3/14/2015 0.070 V=1.376 – 1.392 @4.6GHz, temps high 60’s low 70’s.
3/22/2015 0.060 V=1.360 – 1.376 @4.6GHz, Temps 60’s, one core 70’s.
4/4/2015 0.060 V=1.360 – 1.376 @4.6GHz, Temps 60’s Burn run 10 times.
7/27/2015 0.050 V=1.321 -- [email protected], Temps 70’s Ambient 90F
8/31/2015 0.050 V=1.316 – 1.321 @4.3GHz Temps 60’s and 70’s.
9/19/2015 0.040 V=1.321 - 1.336 @4.3GHz Temps 3 60’s 3 70’s
9/27/2015 0.050 V=1.356 – 1.372 @4.3GHz, Temps 70’s
10/25/2015 0.060 Intel Burn Standard Hardware Monitor
4.6GHz V=1.346 CPU 78c Core 80c
10/25/2015 0.060 Intel Burn High Hardware Monitor
4.6GHz V=1.346 CPU 82c Core 1 80c, 4 70c, 1 60c
11/8/2015 0.070 Intel Burn High
4.6GHz V=1.351 CPU 87c Core ?

Notes:
10/25/2015 - Went back to 4.6GHz. Changed CPU VCCSA Voltage, and VTT CPU Voltage to manual 1.15. Ran Intel burn at standard, high, and very high. Not one hiccup.
10/26/2015 – BSOD first morning. Down clocked to 4.3GHz. XMP?
11/8/2015 4.6GHz, Vcore offset 0.070