Longtime instability (Core i7 920)

[Kamse]

Ive been running with an unstable system for a long time now (1 year+ ) and Im starting to get a little desperate finding a solution (even if it means replacing it).

SETUP:
Core i/ 920 Bloomsfield C-stepping 2.67 Ghz
Asus P6T SE rev 1.02 (latest bios version)
Kingston 3x2 Gb Value ram (PC10667)

The other parts should be irrelevant as Ive (over longer periods) tried replacing gfx card, bought new disc (removed the old ones), removed sound card, dvd-drives, used multiple fresh windows installs and bought new PSU (a Corsair HX650 Pro).

I've used the stock Intel cooling for a very long time (with stock cooling on every part in system + case fan) and 6 months ago i fitted in my old watercooling setup, with no improvement (other than temps now hovering around max 45-50 instead of 60 ish). I really dont suspect heat to be an issue.

The system was overclocked slightly in its early days, going up around 150-160mhz baseclock, 1.3 vcore, 3.2ghz, but after a few months i just went back to default and it ran for a couple of years without problems.

PROBLEM:
I keep getting random reboots. There seem to be a very clear pattern with alot of reboots when the system is just starting up (getting warm) the first 30'ish minutes and then it usually ends up running relatively stable most of the day (little to no reboots at all) and then it might start rebooting a few times or alot during the evening. This could obviously just be related to a pattern of how i use the computer. I suspected the harddiscs to cause the problem just for this very reason, but it didnt help replacing it.

Sometimes the reboots occur before I even pass BIOS screen and sometimes its after a full day of use.

ATTEMPTS:
Besides replacing parts, Ive tried using complete default BIOS values.
Using extra VCORE at 1.300ish, PLL 1.88, abit more Ram core from 1.5 to 1.6, while also downclocking processor multiplier to 12, reducing baseclock to 100mhz.
Tried disabling Speedstep, Turboboost, Clock Spread Spectrum, Load Line calibration (i think).

Some have given abit more stability, but Im still getting reboots no matter what. At the moment Im using 1.2 Vcore, 1.8 PLL, 1.5 ram, default clock speed, disabled Speedstep, turboboost, CSS.

Ive used the default bios version that shipped with the board up untill a few weeks ago, where i updated to the latest. No improvement thou.
When the system seems stable, I can run Prime without crashes or abnormal temperatures.
Ive run Ram checks with no errors.

Im abit hesitant to start messing around further with BIOS voltages, but Im hoping someone can point out other steps that might improve stability (BIOS setting or something completely else). I know theres a chance of faulty hardware, but I just wanna try everything else first.

Any ideas/opponiens are much much appreciated!

 

[DejaWiz]

Have you tested each memory module separately and tried a different PSU?

 

[jojo69]

Is the thing on a UPS?

 

[mzs_biteme]

Have you tested each memory module separately and tried a different PSU?

^^ This...

PSU's degrade over time. Maybe you had a borderline decent PSU when you build the PC, but over the years, it's just not cutting it....

 

[Kamse]

Have you tested each memory module separately and tried a different PSU?

I replaced my 450W Enermax with a Corsair HX650 pro last summer. At that time, I had had troubles for atleast 6 months. The last PSU ran it stable, even with OC for a long time, and buying the new PSU didnt help.

I havent tested them seperately. Any reason why this would make a difference, when theyre passing the test together?

And the system is not running on a UPS. Ive also had the system connected for longer periods (weeks or more) at atleast 4-5 different locations (as in different adresses/cities), so I doubt its a power (house) problem.

 

[Kamse]

Any other ideas?

 

[staticlag]

Sounds like the motherboard is dead

 

[Kamse]

Any reason youre saying motherboard specificly?

 

[unclewebb]

Have you tried running the Prime 95 Blend test? Have you tried running MemTest 86+? If you get errors in either of these tests then run 1 stick of memory at a time to see if you can isolate the problem. 3 sticks don't usually fail all at the same time but 1 stick failing is a common problem on older computers.

Listen to DejaWiz. He didn't become a [H]ardForum junkie over night.

 

[DejaWiz]

Exactly what unclewebb said...individual memory module testing can (and usually will) offer a deeper insight if you have a failed/failing module. I know it's a time consuming pain in the ass, but at this juncture, I'd strongly recommend you proceed in doing so. If they all come back as passed, then you have narrowed it down to 2 major items left: MoBo or CPU.

 

[RamonGTP]

Sounds like a memory issue to me as well. I would certainly run Prime 95 Blend on that thing for at least 8-12 hours or until you get an error, whichever comes first. Not a huge fan of Memtest myself because I've had bad sticks pass memtest for however long I chose to run the test but certainly would not hurt to try than in ADDITION to Prime95 Blend

 

[ccityinstaller]

I too think its a memory issue, but it might not be the ram dimms themselves...I would test each stick on its own (as mentioned above), and if they all pass, then I would remove the board from the case and run it like that for a bit and see what happens..

The X58 boards are VERY sensitive to any flexing in the board..My old Rampage II Extreme (E-ATX) did this when I had it shoved into a HAF932 (non E-ATX case)..The 24pin ATX bundle coming out front under the board was causing it to flex, thus creating the random reboots/memory errors I was getting..

 

[Kamse]

#9 + #10: As I noted above, Ive tested the rams together with no errors. I just tried to test each module alone with more memtest runs and they all passed. I ran about 120-150% coverage on each. I still got random reboots on atleast 2 of them. Used a different socket for each one aswell.

A funny thing happened thou, when I squezed in the 3rd module. After that I had a really hard time just getting into windows. Got maybe 10-15 crashes in total, but eventually (and without touching it physically) it just ended up running stable and passing the memtest with no errors.

Not sure if its related to what #12 says thou? Its locked up with watercooling all around it, so I cant just take it out and test it. (Note motherboard itself isnt locked, but i cant take it out without removing everything.)

#11 Theres no point trying to run a Prime 95 Blend test for 8-12 hours. Even under normal circumstances (not running Prime) it will crash way before that. Even if I get a crash after 30 mins, it could just be another random reboot that I would have gotten anyway.

Edit: I have done Prime Blend test runs of up to 30-40 mins without errors/crashes before thou.

 

[mzs_biteme]

You got more patience that I do man! I would NOT LAST TWO DAYS, knowing that my system may crash/freeze an any given point. That thing would of been out of the case and either doing more testing, or parted out and sold or simply trashed.

One last thing though, I skimmed through the posts again, and didn't see if you've reinstalled OS... Have you?

 

[Kamse]

My bank account says i reeeeally shouldnt buy this stuff and that makes it easy being patient But imagine sitting with this random reboot problem and knowing you have to flash BIOS, where a reboot will brick it. Those seconds were tense

And yes, atleast 3-4 fresh OS installs and on atleast 3 different discs (with the rest disconnected.). But again. Sometimes it reboots before making it past BIOS screen.

 

[LigTasm]

If you can't remove it, loosen the screws on the standoffs a little and pull slack on all the cables. Maybe loosen up the CPU block as well a little and see if any of it is stressing the board.

Have you ever removed the CPU and checked for damaged or bent pins?

 

[Battleneter2]

, Ive tested the rams together with no errors. .

Does sound ram related. Ram testing diagnostic software is by no stretch full-proof and you should not fully rely on the results.

I would run 1 dim for a day or two at a time so you have cold starts testing each dim individually.

It may well end up being a mobo issues, maybe a capacitor is tired and taking time to charge etc, really process of elimination is the only way to solve this. Replace Ram, Replace PSU, Replace Mobo, replace processor until you find the issue, no doubt you don't have all these parts just lying about.

 

[Kamse]

Does sound ram related. Ram testing diagnostic software is by no stretch full-proof and you should not fully rely on the results.

I would run 1 dim for a day or two at a time so you have cold starts testing each dim individually.

It may well end up being a mobo issues, maybe a capacitor is tired and taking time to charge etc, really process of elimination is the only way to solve this. Replace Ram, Replace PSU, Replace Mobo, replace processor until you find the issue, no doubt you don't have all these parts just lying about.

I got reboots using only 1 ram module at a time with atleast 2 of the modules. Cant remember the 3rd, but that diagnosis is pretty trustworthy

I guess I should try to run with the 3rd alone again, just to be sure, but It really doesnt seems like the cause is at the memory.

#LigTasm: I might try to losen them abit, but im usually not using too much force on them in the first place. And I dont really get how a bended motherboard would behave like this. It doesnt make sence that the motherboard would be bend, really unstable, get better during the day and then suddenly be bad again much later in the evening/next morning.

The capacitor thing could make sence (I guess), but if its the case, is there anything to do to help it?

Im not just gonna replace mobo/cpu or ram btw. Its an outdated system now, so its unfortunately all or nothing. Atm its really starting to look like an 'all'-replacement =(

 

[Tsumi]

Have you tried using HCI Memtest? I know that for me, it was sometimes able to catch errors that Memtest86+ didn't.

Also, have you tried using the RAM one at a time in different slots? And have you tried increasing QPI voltage?

 

[Telo]

SETUP:
Core i/ 920 Bloomsfield C-stepping 2.67 Ghz
Asus P6T SE rev 1.02 (latest bios version)
Kingston 3x2 Gb Value ram (PC10667)

Bought new PSU (a Corsair HX650 Pro).

If you are overclocking you should have gone a bit higher than 650.

I've used the stock Intel cooling for a very long time (with stock cooling on every part in system + case fan) and 6 months ago i fitted in my old watercooling setup, with no improvement (other than temps now hovering around max 45-50 instead of 60 ish). I really dont suspect heat to be an issue.

Try reseating the CPU, water should ALWAYS give better results than stock cooler.

The system was overclocked slightly in its early days, going up around 150-160mhz baseclock, 1.3 vcore, 3.2ghz, but after a few months i just went back to default and it ran for a couple of years without problems.

What was the lowest possible vcore that CPU needed to run stock properly? 1.3 vcore sounds bit high; if my memory serves me right my i7 950 uses around 1.30v to hit 4.2Ghz [I know it is comparing apple to orange, but the difference shouldn't be that far off].

I keep getting random reboots. There seem to be a very clear pattern with alot of reboots when the system is just starting up (getting warm) the first 30'ish minutes and then it usually ends up running relatively stable most of the day (little to no reboots at all) and then it might start rebooting a few times or alot during the evening. This could obviously just be related to a pattern of how i use the computer. I suspected the harddiscs to cause the problem just for this very reason, but it didnt help replacing it.

HDD issues cause lock ups and BSOD, not reboots; if by reboots you mean you are seeing the desktop one second, and the next you see the POST process.

Sometimes the reboots occur before I even pass BIOS screen and sometimes its after a full day of use.

I'd start by redoing the BIOS flash. If you keep getting reboots before POST, then try to test with a different RAM stick [value stuff is as the name states, cheap].

Besides replacing parts, Ive tried using complete default BIOS values.
Using extra VCORE at 1.300ish, PLL 1.88, abit more Ram core from 1.5 to 1.6, while also downclocking processor multiplier to 12, reducing baseclock to 100mhz.
Tried disabling Speedstep, Turboboost, Clock Spread Spectrum, Load Line calibration (i think).

First thing that screams at me here is your PLL, I've never had to run mine higher than 1.80; and if your highest OC was just 3.2Ghz, a PLL value of 1.80 is more than enough.

Throwing a higher VCore at a stock CPU is useless, besides it does not sound like it is your CPU messing you up.

Applying more volts to cheap RAM is asking for issues, if it was rated for 1.5v [I'm guessing here], then use 1.5v.

All those CPU features you disabled would have had an impact if you were OC'ing, running it stock it wouldn't come into play.

Ive used the default bios version that shipped with the board up untill a few weeks ago, where i updated to the latest. No improvement thou.

You should always apply an update to the BIOS, the one that ships with it always misses some sort of support, specially for OC'ing related.

When the system seems stable, I can run Prime without crashes or abnormal temperatures. Ive run Ram checks with no errors.

It's is pretty obvious your system is far from stable.

Im abit hesitant to start messing around further with BIOS voltages, but Im hoping someone can point out other steps that might improve stability (BIOS setting or something completely else). I know theres a chance of faulty hardware, but I just wanna try everything else first.
Any ideas/opponiens are much much appreciated!

I would suggest you go back to the basics. Give this guide a try:

http://www.xtremesystems.org/forums...verclocking-Guides-(now-includes-Sandy-Bridge)


I apologize in advanced if I'm not allowed to post links to other forums. I hope what I said here helps you out; Godspeed.

 

[Kamse]

If you are overclocking you should have gone a bit higher than 650.

My old 450w did fine so thought a 650w is more than enough. The rest of my system is pretty forgiving on power(ssd disc and an old HD4890) and I aint gonna do any serious OC anyway.

Try reseating the CPU, water should ALWAYS give better results than stock cooler.

It does give better results already. The 45-50 degrees now are at full load and only one 120mm radiator at 7v (or 5, cant remember). The stock was never really above 60 either, so pretty sure cooling is perfect and has always been.

What was the lowest possible vcore that CPU needed to run stock properly? 1.3 vcore sounds bit high; if my memory serves me right my i7 950 uses around 1.30v to hit 4.2Ghz [I know it is comparing apple to orange, but the difference shouldn't be that far off].

You are probarbly right. Not saying i used 1.3 in the end, but Im pretty sure i never went too much 1.3 . Ram was probarbly holding me back anyway.

HDD issues cause lock ups and BSOD, not reboots; if by reboots you mean you are seeing the desktop one second, and the next you see the POST process.

I never get BSOD, only a short complete powerdown and then back to POST, but ye. It aint HDD for sure.

I'd start by redoing the BIOS flash. If you keep getting reboots before POST, then try to test with a different RAM stick [value stuff is as the name states, cheap].

Again? Last bios worked fine for years at default settings and the current one is the latest. What would a reflash achieve?

First thing that screams at me here is your PLL, I've never had to run mine higher than 1.80; and if your highest OC was just 3.2Ghz, a PLL value of 1.80 is more than enough.

Throwing a higher VCore at a stock CPU is useless, besides it does not sound like it is your CPU messing you up.

Applying more volts to cheap RAM is asking for issues, if it was rated for 1.5v [I'm guessing here], then use 1.5v.

All those CPU features you disabled would have had an impact if you were OC'ing, running it stock it wouldn't come into play.

Im pretty sure my old OC was done at default PLL. Setting it to 1.88 is just another thing I tried after the reboots started (at stock speed). I may have used a higher RAM voltage for the old OC, but Ive been running everything at default bios settings pretty much since I bought it, just except for those first few months with a slight OC.

You should always apply an update to the BIOS, the one that ships with it always misses some sort of support, specially for OC'ing related.

On this area (especially in the older days with much higher risc of bricking boards) Ive always used the "If it aint broke, dont fix it!" regarding BIOS updates. Only reason I felt like updating it now, is because the reboots are an obvious "broke" thing. I actually also thought the old bios was feeding incorrect Vcore. For i7 920 it should be 1.2v but it was always doing around 1.04 (its obviously the speedstep thing, but still had to be sure). I thought it may be feeding other incorrect voltage for other parts, but again.. it didnt help.

It's is pretty obvious your system is far from stable.

Obviously, but my point is it can hit really long periods without reboots, even thou its temporary.

I would suggest you go back to the basics. Give this guide a try:

http://www.xtremesystems.org/forums...verclocking-Guides-(now-includes-Sandy-Bridge)

Thanks, Ive read other guides aswell.

Gonna try to up the QPI/VTT voltage abit and see if it does any wonders.

But its really starting to look like I need to order new stuff on monday. =(

 

[x2ezx]

Got the same problem on me x58 system and I found a faulty memory with memtest.

 

[Kamse]

So i changed QPI/VTT voltage for the L3 cache and memory controller from 1.2v to 1.26 and the system has been MUCH MUCH more stable. I have gone from something like 20-30'ish reboots a day to a total of 2 reboots the last 3 days. So it obviously aint fixed, but this is so far completely acceptable (and else I can try to bump it abit more). Just really hope it lasts this time (when I raised Vcore to 1.3 it did stabilize abit, but only for a few days =/ ).

Looks like a lot of you had a good hunch about it being a memory issue

 

[Tsumi]

I've run my QPI voltage at up to 1.35v, so I think you would be safe pushing it that high. However, 1366 processors are finnicky, and too high of a QPI would lead to instability. You have to find that sweet spot, one of the challenges of using a X58 system

Z77 and X79 are too easy IMO.