Sign in with
Sign up | Sign in
Your question

"Overclock Failed" -- WTF?

Last response: in Systems
Share
a b B Homebuilt system
April 19, 2010 4:20:12 PM

OK, so this is a new one for me. My fiancee calls me this morning and says that when she turns on the computer, it fails to POST and hangs with an error message that says "Overclock Failed." The machine is not overclocked -- though I do have four sticks of 1.7V DDR3 in there, and I had them set at 1.8V to get them stable. Last night the machine worked fine; today it won't POST. It had worked perfectly for about a year straight, and I hadn't touched any of the sittings since.

I am at work, so I can't see any of the specific error messages and am just going by what I've been told, but it seems to me like the only possibility here is that the RAM voltage adjustment was being recognized fine and now suddenly it's not. I've told her to reset the BIOS to defaults and then go back in and try to set the voltage and timings manually, but there's probably only a slim chance of that working (or of her doing it correctly, for that matter). Specs are:

Q9550 @ stock (2.83Ghz)
ASUS P5E3 Pro
4x2GB Patriot Viper DDR3 1333-7-7-7-20, 1.7V (running at 1.8V)
Visiontek HD 4870 512MB
850W NZXT PSU

Mainly, my question is: What the hell could have caused this to happen all of a sudden? Please don't tell me the motherboard just decided to crap out for no good reason. I hate replacing the motherboard. Perhaps I'll see if I can boot with one stick of RAM at a time when I get home, but in the meantime, I think she's pretty helpless to deal with actually taking components in and out herself. Obviously, I won't be able to run memtest until I can actually get the machine to boot, so I'll probably have to wait until later to try that.

More about : overclock failed wtf

a b B Homebuilt system
a b K Overclocking
April 19, 2010 4:40:11 PM

I think you're on the right track. It's likely RAM related. Your ideas to reset the CMOS, reset the RAM values in the BIOS, and run Memtest86+ are exactly what I'd do. It's possible a stick of RAM is dead/dying.
m
0
l
a b B Homebuilt system
April 19, 2010 5:21:20 PM

Update: By resetting the BIOS to default, she was able to get the system running again and at least get into Windows. Nothing has frozen yet, but I'm expecting it to since the RAM voltage is probably on "auto" now. I've told her to try playing a game (Sims 3 always seemed to be great at exposing RAM issues on our last machine) and see if she can get it to hang.

I'm kind of reluctant to reset the timings/voltage if it doesn't freeze, because I don't want to cause that overclock failed" issue again. Though I suppose I ought to at least try it to see if it happens or not. If I use the exact same RAM settings and it doesn't give me an error this time, then I'll really be wondering WTF.

In either case, yeah, I think memtest is in order regardless just to make sure I haven't got a bad stick of RAM. Though I'd far prefer that to a bad motherboard.
m
0
l
Related resources
a b K Overclocking
April 19, 2010 5:54:30 PM

Check the forums ofr your particular brand/model RAM for tolerance to overvolting. I know my GSkill gets flaky on certain OC settings with an additional .1V past stock.

I liken the RAM "innards" to a lightbulb. If it's rated to 170 watts and you pump 180 watts, sure it'll work and give a brighter glow while at the same time accelerating the end of its lifespan.

http://www.overclock.net/intel-memory/346472-bad-overvo...

m
0
l
a b B Homebuilt system
April 20, 2010 5:33:32 AM

OK, so the system apparently worked all day for my fiancee with the RAM settings at "auto" - however, when I got home she had put it in sleep mode and it wouldn't come back on. Had to turn it off by holding down the power switch, at which point it would fail to POST, and power would cycle on and off every few seconds (another sign of RAM problems, I guessed).

Finally turned it off and unplugged the machine from the wall for a few minutes, and was able to get into the BIOS on restart. Setting the voltage to 1.74V (as opposed to 1.8V previously) seemed to get the system stable. Ran memtest86+ and all sticks checked out OK on several passes.

So basically, the only thing I can think of is that that RAM itself was OK and remains OK ... but somewhere between 1.74V and 1.8V I cross some threshold where a tiny bit of instability presents itself maybe once a year? Not sure how likely that is, but it's the best theory I've got. Anyone have any better ideas?
m
0
l
a b K Overclocking
April 20, 2010 6:39:07 AM

capt_taco said:
OK, so this is a new one for me. My fiancee calls me this morning and says that when she turns on the computer, it fails to POST and hangs with an error message that says "Overclock Failed."

I am at work, so I can't see any of the specific error messages and am just going by what I've been told, but it seems to me like the only possibility here is that the RAM voltage adjustment was being recognized fine and now suddenly it's not.


No, that's not what the message "means". The message indicates a boot failure. Any boot failure. The engineers at Asus assumed that a boot failure is usually due to overclocking, so instead of giving you a message "System failed to boot, now using default values" it says "Overclocking Failed".

Good luck, I had problems like that for a long time until one of my Patriot 1.7V modules finally died. I don't know what causes a module to be intermittantly good, but eventually it went completely bad.
m
0
l
a b B Homebuilt system
April 20, 2010 4:22:57 PM

Crashman said:
No, that's not what the message "means". The message indicates a boot failure. Any boot failure. The engineers at Asus assumed that a boot failure is usually due to overclocking, so instead of giving you a message "System failed to boot, now using default values" it says "Overclocking Failed".

Good luck, I had problems like that for a long time until one of my Patriot 1.7V modules finally died. I don't know what causes a module to be intermittantly good, but eventually it went completely bad.


Wow, this sounds fun. Basically waiting around for one of my modules to die, without knowing which one, and experiencing random incidents the whole time.

I mean, if I could find out which stick it was, I'd just remove that pair and be fine with 4GB for now, but like I said, they all passed memtest with flying colors. Any suggestions for diagnosing that? I guess I'll just keep running memtest every so often until the bad stick decides to show its face. That's the only plan I've got.
m
0
l
a b B Homebuilt system
April 20, 2010 5:09:23 PM

Quote:
Just put in 2 sticks and see if they act up. If they do, then try the other 2.

I suspect the power supply though.


I tried various combinations of that last night (as well as 1 stick at a time), and at 1.74V, none of them were acting up. I honestly don't know whether they're actually more stable at that voltage than 1.8V, or whether it's what Crashman said and one of them is gradually dying but deciding to behave normally most of the time. I guess wait and see if the problem comes up again is my only option, unless someone has a better idea.
m
0
l
a b K Overclocking
April 20, 2010 7:57:49 PM

capt_taco said:
I tried various combinations of that last night (as well as 1 stick at a time), and at 1.74V, none of them were acting up. I honestly don't know whether they're actually more stable at that voltage than 1.8V, or whether it's what Crashman said and one of them is gradually dying but deciding to behave normally most of the time. I guess wait and see if the problem comes up again is my only option, unless someone has a better idea.


Well, when I said I have no idea...I'm sorry, that wasn't accurate. I have some ideas but cannot verify any of them. For example, I'm pretty sure this is BGA RAM, and there's always the chance that cracked solder balls are making intermittent contact with the PCB.
m
0
l
a b K Overclocking
April 20, 2010 8:51:29 PM

Why not use the the RAM at their rated volatge 1.7? What's the reasoning for 1.74V?
m
0
l
a c 84 B Homebuilt system
April 20, 2010 8:56:03 PM

@100Island: 1.74V because of this:

capt_taco said:
Finally turned it off and unplugged the machine from the wall for a few minutes, and was able to get into the BIOS on restart. Setting the voltage to 1.74V (as opposed to 1.8V previously) seemed to get the system stable. Ran memtest86+ and all sticks checked out OK on several passes.


It wasn't stable at 1.7V.
m
0
l
a b B Homebuilt system
a b K Overclocking
April 20, 2010 8:59:02 PM

Yup, many motherboards require a slight voltage bump when all of the RAM slots are filled.
m
0
l
a b B Homebuilt system
April 21, 2010 7:45:08 AM

Well, it's better than relaxing the timings, if you ask me. I'm just worried now that going to 1.8V may have been overkill because I was pissed off when it wouldn't stabilize at 1.7V or 1.72V back when I built it.

After revisiting the motherboard specs -- it only supports a max of 8GB to begin with -- I'm now wondering if maxing it out could be stressing the northbridge, and that's exacerbating the issue? I wonder if a slight bump on the NB voltage would help with stability if this happens again. I mean, in theory I can see how it might, but in practice, I really have never messed with the northbridge voltage outside of actual overclocking, so it just seems a little unnatural to me. I swear I remembered this board as having a 16GB capacity, but I obviously remembered wrong.

At any rate, if there's any input on that, it would be much appreciated. Otherwise, I guess I'm probably just going to see how long it stays stable at current settings, test the memory every few weeks, and hope it's not one of the dying-RAM scenarios Crashman illustrated.
m
0
l
a b K Overclocking
April 21, 2010 2:05:00 PM

On seeing how the RAM problem was fixed by lessening the voltage, take of the NB auto votage settings and manually enter one of the lower values. If you're stable, keep lowering until you're not stable.

If you're motherboard needs addtional voltage to the NB even without overclocking, I would think there is something strange about that.

m
0
l
a b B Homebuilt system
a b K Overclocking
April 21, 2010 2:52:31 PM

You're right capt_taco, slightly bumping up the NB voltage is another trick to getting the RAM stable when using all the RAM slots. I would bump it up .1v from stock voltage. I've had to do that before for absolute stability.
m
0
l
a b B Homebuilt system
April 21, 2010 4:39:43 PM

hundredislandsboy said:
If you're motherboard needs additional voltage to the NB even without overclocking, I would think there is something strange about that.


I don't know. Logically, it makes sense to me that if you're working a component hard, you could need to give it more juice. I mean, that's essentially why you increase voltages (including the NB voltage) in overclocking. I've never had to do it for a non-overclocked machine before ... but then again, I don't think I've ever maxed out the absolute RAM capacity on a board using sticks of high-speed/higher-than-normal-voltage memory either. I agree, it does seem kind of weird to me too, though I can see how it theoretically could happen.

Well, regardless, the machine has seemed to regain stability for the past two days using the 1.74V setting. I will keep my fingers crossed that it holds, and if not, possibly see if a small bump in the NB voltage helps. If the problem STILL resurfaces after that, I guess I'll just be waiting around for one of the sticks to die, and hopefully identifying it before the problems become too much of a persistent nuisance.
m
0
l
a b K Overclocking
April 21, 2010 5:23:37 PM

You could also do what I do. None of my motherboards have the RAM slots fully populated. All are 2X2 Gig for 4 Gigs total.

BTW, I have the voltage on the NB of both my P5N-D and P5Q SE PLUS mobos slightly bumped up but that's because I'm overclocking (1600 Fsb) and auto or stock voltage settings did not work. However, the RAM voltages are all set on their rated or stock voltage.
m
0
l
a b B Homebuilt system
April 21, 2010 9:05:57 PM

hundredislandsboy said:
You could also do what I do. None of my motherboards have the RAM slots fully populated. All are 2X2 Gig for 4 Gigs total.


That would be great, if I knew which one the bad stick was (or if there is, in fact, a bad stick). Until I can figure that out, dropping back to 4GB seems mostly unfavorable because there's a 50-50 chance I'll pull out the wrong pair, and even if it turns out I guessed right, I'll have no way of knowing, so for as long as I own the computer, every time I play a game or do something memory-intensive, I'm going to wonder if I really found the bad stick, or if the machine is going to freeze 10 seconds from now.

It also may or may not be what's wrong ... and in either case, I'd essentially be writing off a $100+ set of RAM whether I fixed the problem or not.

Also, while I get that 4GB is plenty for gaming, I do prefer having 8GB because my work occasionally involves large-scale photo editing, and the fiancee is a pretty avid creator of Sims stuff, which means she'll pretty often have several large programs open at once.

Sigh. So far it seems like it's well into its third day of stability this way, but for now I think it remains wait-and-see.
m
0
l
a b K Overclocking
April 21, 2010 9:18:13 PM

Or it may not be the RAM stick/s. Could be a RAM slot going bad? How wuould you test that without special bench equipment?
m
0
l
a b K Overclocking
April 21, 2010 9:44:04 PM

capt_taco said:
Sigh. So far it seems like it's well into its third day of stability this way, but for now I think it remains wait-and-see.


That's a great plan. Like I said, I think MY module problem (same parts) was with a cracked solder ball causing intermittent contact, similar to the problem NVidia had with its mobile graphics modules. You might not have my problem.
m
0
l
a b B Homebuilt system
April 21, 2010 10:10:42 PM

hundredislandsboy said:
Or it may not be the RAM stick/s. Could be a RAM slot going bad? How wuould you test that without special bench equipment?


That's pretty straightforward: Just take a known good stick of RAM and test it in each slot with memtest. If it comes back good in three slots but bad in the fourth, it's a pretty good indication that's the problem.

But that would present the same issue as gradually dying RAM -- you'd have to catch it at a time when the problem happens to be manifesting itself. Which means that you'd probably have to wait for the component to die a little more to be (un)reliable enough to screw up with regularity. Which is why I keep coming back to wait-and-see.
m
0
l
!