Longterm instability, what causes it?

[CoW]8(0)]

Well now I'm working hard to get a 24/7 100% load stable OC. But its been very difficult and I still haven't managed to find that nice spot. What I'm looking to do is find out which specific component is causing the instability and lowering the speed of that component (rather than slowing down the entire system).

Hardware:
PSU: Ultra V-series 500W
Video: BFG 7900GS 256MB
CPU: E4400 3.2Ghz (10x Multiplier * 320Mhz FSB, stock voltages)
Motherboard: EVGA 680i LT (1.4V SPP)
RAM: Crucial Ballistix PC2-8500 (800Mhz 4-4-4-12 2T, stock voltage)

Possible instability causes:
-Power supply? (As you can see from the image below, my Vdroop is fairly large. The setting for Vcore in the bios is 1.3250V)
-Cooling? I'm suspecting the Northbridge/Southbridge because they've been known to have bad TIM application and I haven't switched it.
-RAM? I have them set at EPP settings (and SPD settings for the values EPP doesn't specify like tWTR, tRRD etc)

Also, I'll probably use this thread as a log for myself to track down this stability issue.

 

[CoW]8(0)]

Here's a recent everest log

 

[matrix563]

i don't think it will ever be 100% stable long term. overclocking just isn't

my current pc is stock, and it ran for 26 days before i rebooted to get my webcam installed

 

[CoW]8(0)]

I see your point about that. But I'm really just going for maybe a 15 day uptime before something crashes (I assume I'll have to restart more than every 15 days anyway).

 

[matrix563]

well some webpages crashed even for me. but the overall responsiveness never went down on my system. i guess 4gb of ram helps with that

 

[CoW]8(0)]

Well I just remembered that my motherboard states it's specifically spec'd to handle a 1333Mhz FSB. Right now, it's running at 1280Mhz, so I'm beginning to doubt the NB or SB is overheating.

I'm also beginning to suspect it's the Vdroop on the CPU since my 12V rail seems to fluctuate a little bit.

 

[matrix563]

Well I just remembered that my motherboard states it's specifically spec'd to handle a 1333Mhz FSB. Right now, it's running at 1280Mhz, so I'm beginning to doubt the NB or SB is overheating.

I'm also beginning to suspect it's the Vdroop on the CPU since my 12V rail seems to fluctuate a little bit.

it probably is. i had an opty at 2.8ghz but it wasn't prime stable. the 12v was fluctuating pretty bad. it was a 485 watt psu.

 

[CoW]8(0)]

I decided to do some orthos testing because my monitor was flashing on and off and I suspected the RAM (since it stores non-local video memory). So I ran my TV tuner and orthos (Stress RAM test) simultaneously. And not to my surprise, the monitor would flash much more often and sometimes the system would freeze.

Just to make sure my Vdroop wasn't the issue, I raised my Vcore voltage from 1.3250V in the bios to 1.3500V. I reran the same test scenario and it still had the same flickering and lockups.

I decided to restart and further loosen my RAM timings. But then I discovered I was not able to POST. For some reason POST kept stopping at my floppy drive, so I decided to disable it.

Now I'm back in Windows with the same RAM timings and overclock as before (but with Vcore now at 1.3500V). Other changes I made in the bios include the boot order (CDROM->Hard Drive->Removable) and I also disabled the floppy drive.

I did the same orthos RAM stress test with my TV tuner to find that there was no flickering or freezing.

So If I remain stable at these settings, I think I'll lower the Vcore back to 1.3250 and see if that affects anything.

 

[CoW]8(0)]

Well my computer has crashed once, but I'm not certain what caused it (whether it was instability or even an actual crash). I was using my TV tuner card when it occurred. The screen went blank but the audio was still playing in the background.

I probably should have tried memorizing the keystrokes to restart and see if it was truly a crash.

But since then my uptime has been 2 days, 1 hour, 20 minutes with 100% load via Folding SMP client and normal computer usage. So far I haven't changed anything in the bios. From time to time the monitor still flickers though. Usually it happens when I watch a Flash video (like something on youtube). So it's likely that my RAM is having some issues. But still have not crashed.

 

[CoW]8(0)]

Well I decided to give a game stress test a go (while folding at the same time). I played around with BF2142 demo to see how that would run. Unfortunately, it didn't go very well. The good news was the system didn't crash, the bad news was the rendering was extremely buggy and I got to see some nice effects (like rainbow colors and some weird trilinear filtering, at least I think it was that).

Same result when I decided to rerun without folding. So I think the problem is not CPU related (the 2 cores were not at 100% load when the game was running). This narrows it down to most likely the Northbridge or the RAM. But regardless of the issue being with either one, I restarted and used linked RAM. So instead of running it at 400Mhz (DDR2-800Mhz), I'm running it linked and synced at 320Mhz with the same timings.

Can't make any solid conclusions yet. But if all goes well I think I'll have some interesting notes about overclocking (possibly some extra tweaks to maximize system stability).

 

[RangerSVT]

demo to see how that would run. Unfortunately, it didn't go very well. The good news was the system didn't crash, the bad news was the rendering was extremely buggy and I got to see some nice effects (like rainbow colors and some weird trilinear filtering, at least I think it was that).

did it ever occur to you that it might be something besides your cpu/ram/northbridge and possibly your video card overheating?

Also, i don't care what your uptime is, if you've gone 2 days fine, until you hit 2 days and 1 hour then u crashed, you're unstable, and your overclock is unstable.

Compare my below system in my signature, 7 days running and I haven't had any problems. I've done all my stress testing and i am certain this overclock is stable for long uptimes. i've done the usual orthos/prime, 1GB+ video editing, games like call of juarez, CS: Source, WoW, plenty of DVD's burned and plenty of CD"s burned, websites visited to the ungodly amount count.

 

[CoW]8(0)]

Well I seriously doubt my video card would be overheating. Not only is it not overclocked, it's also watercooled. My video card load temp is 39C, my idle temp is 36C.

And a stable system is relevant to the person who uses it. Your definition of stable may not be what mine is or someone elses. For example, a stable server would be regarded as one with an uptime in the months range.

And I have done the usual orthos/prime stability tests for 24hrs and the system has been able to run through them without any issues, but that doesn't mean its definitively stable. It still could crash a second after.

And I find that Folding is the better stability test for C2D. I've seen errors on folding which I never see in Prime or Orthos.

And like I've already mentioned, I've done the proper testing (as far as my resources go) to narrow down where the instability is most likely occurring.

Also, I forgot the mention. The gaming instability I mentioned above only occurs when I've had at least 1-2 day uptime. It never occurs after a fresh restart (even with Folding SMP client running in the background). These signs clearly point to my NB or RAM as the issue since longer uptimes mean higher RAM usage and more stress on the NB/RAM.

 

[Slade]

My previous system ran 24/7 since 03 till Aug 07. It was overclocked and the only time it shutdown was to upgrade parts or clean up some internals. I could go months without a shutdown.

I've been running my current sig system, closed case 24/7 since I built it. The longest I've had it running is 18 days before I rebooted due to some hotfix or indulgence in some new video card driver.

I think before you claim initial stability, run it thru prime95 for >24 hours AND try to use it for normal browsing etc (nothing mission critical) with warm ambients (worst I put it through was 30C ambient room).

If you can keep a system stable like that under those conditions, running 2 weeks + with 95% of the time idling will be a breeze for the system.

 

[CoW]8(0)]

Didn't I already mention that I'm constantly folding? The system is never idle.

 

[matrix563]

well mine stopped being stable. stupid via usb pci card is causing bsods. i'll try it in a different pci slot but i think its gonna go and soon

 

[RangerSVT]

Didn't I already mention that I'm constantly folding? The system is never idle.

Doesn't matter dude. Systems with overclocks that are exceeding it's capabilities do crash on idle, hell I've experienced it!

 

[CoW]8(0)]

?
Aren't these observations fairly obvious?

 

[CoW]8(0)]

4 Days, 20 Hours, 59 Minutes, 44 Seconds

 

[RangerSVT]

4 Days, 20 Hours, 59 Minutes, 44 Seconds

Good deal bro, you got it stable?

 

[BillParrish]

i don't think it will ever be 100% stable long term. overclocking just isn't

BS.

To the OP, why are you not running 8 x 400 with a 1:1 (linked) ? Will the board not do it ?

 

[Marc Adams]

Regarding your instability.

I have a friend with a quad core CPU that was trying to overclock it and run Folding at home.
Would you believe that F@H would crash almost immediately if he OC'ed his Q6600 by even 5MHz

Every other program reported perfect stability up to 3.5GHz. F@H refused to work with ANY oc for him.

P.S. For your boot order I find it faster to have it set to HDD first. If you need to boot from your cd-rom drive just press F12 (May be different on your mobo) when your PC posts and you then select which drive to boot to. Takes about 5-10 seconds off your boot time.

 

[CoW]8(0)]

BS.

To the OP, why are you not running 8 x 400 with a 1:1 (linked) ? Will the board not do it ?

I don't think the CPU can handle it, remember the CPU's stock FSB is only 200Mhz.

I took advantage of the 10x multiplier on the E4400 to get 3.2Ghz.

Well, I decided to give 8x400Mhz another try since the last time I tested it was with an older bios. This new bios supposedly has "Improved 1333Mhz CPU overclocking" so I'm assuming they also moved the straps around on the board.

 

[CoW]8(0)]

Regarding your instability.

I have a friend with a quad core CPU that was trying to overclock it and run Folding at home.
Would you believe that F@H would crash almost immediately if he OC'ed his Q6600 by even 5MHz

Every other program reported perfect stability up to 3.5GHz. F@H refused to work with ANY oc for him.

P.S. For your boot order I find it faster to have it set to HDD first. If you need to boot from your cd-rom drive just press F12 (May be different on your mobo) when your PC posts and you then select which drive to boot to. Takes about 5-10 seconds off your boot time.

And yes, in my opinion F@H is probably the best program to test stability with. It seems to find issues with an OC that other programs don't. But it's still not perfect.

Yes I usually do change my boot order to avoid CD drive spinup, but I've always forgotten to whenever I'm in the bios. And I don't restart very often anyway (usually never unless some program I installed absolutely requires it).

 

[BillParrish]

I don't think the CPU can handle it, remember the CPU's stock FSB is only 200Mhz.

hmm my E6300 with a default cpu FSB of 200 has no issues with an FSB of 475MHz. x 7 but every board and cpu is different and if it will not do it, it will not do it.

I would also take a look via the task manager at an idle state of your free/available virtual and physical memory over time and the processes and number of threads running. You are ignoring the software. Many games have small memory "leaks" or fail to release memory and it is not always a game. Slowly over time they hog resources and free memory and then "crash" a reboot clears it all out and everything seems fine.

However the first thing I would do after reviewing your hardware setup is give the cpu more Vcore if stability is indeed your concern. Running an e4400 at 3.2GHz at stock voltage is remarkable but not condusive to stability, try it at 1.350v 400 x 8 1:1 linked and see how long she will stay up. If happy you can unlink and try again for long term stability with a memory mulitplier/setting to run your ram at full spec speed.

Anywho just some suggestions.

 

[CoW]8(0)]

Actually I believe the E6300 has an FSB of 266Mhz.

And I'm fairly certain it isn't software because I haven't used up all my RAM over time, just more and more of it.

I've already raised the Vcore slightly higher than stock to basically keep Vdroop out of the equation. Although it didn't seem to fix the stability issue I still keep it slightly higher than stock just to keep the CPU further out of the equation.

The settings which got me the 4days 20hrs stability posted earlier, I didn't see any stability issues at all. Everything was perfectly stable.
Those settings were:
CPU: 3.2Ghz (10x Multiplier w/ 320Mhz FSB)
RAM: 640Mhz (linked @ 4-4-4-12 2T)

When I tried 3.2Ghz with a 8x multiplier x 400Mhz FSB, my system started showing those instability signs again when I was in a game (although I noticed a drastic increase in my memory bandwidth).

So right now I'm trying:
CPU: 3.2Ghz (10x Multiplier w/ 320Mhz FSB)
RAM: 640Mhz (linked @ 3-3-3-9 2T)

Seeing if I can get away with tighter RAM timings.