Am I stable?

[maniac5999]

It's been really tough for me to say, and I've had annoyingly "wierd" things happen with no rhyme or reason. First of all, I recently replaced my dying P2 940 with an 830, as I talked about here: http://forums.anandtech.com/showthread.php?t=2202875&highlight= Then, in my haste to get up and running again, I didn't wipe the win7 install or anything. it booted up and ran fine at 3.5ghz. It passed a 6 hour OCCT run (medium data set) just fine.

2 Days later win7 wouldn't boot, it would hang every time. OK, I'll go back to stock speeds, reinstall windows, and move up carefully. Maybe I pushed the OC too hard, or maybe the dying 940 left a ticking timebomb of a flipped bit in the OS. Everything goes just fine, and it passes every test I throw at it along the way, and next thing I know, I'm back at 3.5ghz. Now, I figure it might be the memory subsystem. The RAM itself is fine at 1066, but the other 2 CPUs I had on the board wouldn't go much past a base clock of 240, and even with the 830, the board won't POST at 260. (it's at 250 and DDR2 1000 with timings that are known good at 1066) So I decide to run Memtest 86+. Mid way thru the 2nd pass it just hangs, no errors or anything, and the + sign and cursor keep blinking, but the clock stops counting and it stops running tests. I reset memtest, and run it overnight in it's 'safe mode'(supposed to give greater stability to the program), and it completes 10 passes without problem.

I set it to run OCCT while I'm at work today, (large data set to test memory too) and 9 hours in it looks like it failed with "out of memory" and "Generic Error in GDI+" errors, both of which it wanted to report to the Dev team. At the same time windows reports that it authenticated itself this afternoon, which makes me think that it was a resource conflict between the two programs that caused the crash.

Am I being a hopeless optimist and ignoring instability when it shows up, or am I just seeing a lot of random flukes all at once because I'm looking for them?

 

[Plimogz]

Edit: No, you're not stable.

Am I being a hopeless optimist and ignoring instability when it shows up, or am I just seeing a lot of random flukes all at once because I'm looking for them?

What has most probably changed now is that you're not just blaming the CPU, presumably because you -- for want of a better term -- "trust" the new X4 830.

I don't know if it's just me -- can't be bothered to go back to your previous thread and check -- but I got the impression that you mention RAM much more this time around...

So by all means, troubleshoot your RAM, then PSU and Mobo, in whichever order. Really, swap out whatever you can most easily get a replacement for.

And you were right last time the subject was brought up; yours a quite reputable PSU. But that doesn't exclude a bum unit. Although if you say that the OCCT PSU test looks good... meh, what can I say.

Bad motherboard, albeit a bummer of a diagnosis, seems to be the likeliest candidate.

 

[maniac5999]

What has most probably changed now is that you're not just blaming the CPU, presumably because you -- for want of a better term -- "trust" the new X4 830.

I don't know if it's just me -- can't be bothered to go back to your previous thread and check -- but I got the impression that you mention RAM much more this time around...

So by all means, troubleshoot your RAM, then PSU and Mobo, in whichever order. Really, swap out whatever you can most easily get a replacement for.

And you were right last time the subject was brought up; yours a quite reputable PSU. But that doesn't exclude a bum unit. Although if you say that the OCCT PSU test looks good... meh, what can I say.

Bad motherboard, albeit a bummer of a diagnosis, seems to be the likeliest candidate.

Well, I spoke about the RAM just to rule it out. 3.36ghz is known stable territory on the board itself. my Kuma ran with a base clock of 241 for over a year flawlessly. I also ran the ram at 1066 and these 5-5-5-15 timings for over a week on memtest with no problems with the Kuma. (I actually had problems with the initial RAM I bought in 2008, the OCZ Reaper RAM I'm using is an upgrade/replacement for DDR2 1066 OCZ Silver RAM that was out of stock when I RMA'd it--I made EXTRA sure the new RAM was OK)

The thing is, with the 940, I would get "normal" errors. It would fail Prime, or OCCT, etc. and return the standard results "Error detected on core #X" etc. The only time I got one of those with this new setup was when I tried to run the 830 at 3.5ghz on 1.325v. (It failed in the 3rd hour of OCCT) I haven't tested as extensively within 'known good' territory for the board. I.E. base clock of 241 and lower, where the Kuma sat comfortably for a year+, but I never got any errors there either. I'm wondering if anyone has a way to further test this beyond just re-running memtest and OCCT? also, if the answer is just to do such, is there one that I should focus on more than the other?

 

[Cerb]

Sounds to me like RAM, PSU, or mobo, in that order. I don't know what PSU you have. For RAM, either run it a bit slower, or relax the timings, or both. Passing synthetic tests only indicates that those tests cannot verify a fault, not that a fault does not occur. I've had PCs which would run memtest, prime and anything else all day long, but be unstable in real use, then magically become stable with new RAM. Cest la vie.

Since it is a stability issue, do not run your RAM faster than stock, until everything else is fast and stable. If that's not possible (you didn't get a BE), run the CPU at speeds that allow your RAM to run <= its factory ratings.

For that matter, if it is RAM rated at some non-JEDEC weird speed, run at the next lowest standard speed, and don't increase RAM voltage.

If it is stable with the RAM <= stock, then start changing RAM settings and see what happens. If it isn't stable, still, while it could be RAM, I would go :hmm:, maybe the PSU is marginal when the PC is OCed.

Troubleshooting an overclock: another reason to spend more on CPUs with unlocked multipliers.

 

[sm625]

First of all, I recently replaced my dying P2 940 with an 830

It sounds like you need to reevaluate whether the 940 was really "dying". Just because you stuck a kuma in there and it worked doesnt mean the mobo and psu are good.

 

[maniac5999]

Sounds to me like RAM, PSU, or mobo, in that order. I don't know what PSU you have. For RAM, either run it a bit slower, or relax the timings, or both. Passing synthetic tests only indicates that those tests cannot verify a fault, not that a fault does not occur. I've had PCs which would run memtest, prime and anything else all day long, but be unstable in real use, then magically become stable with new RAM. Cest la vie.

Since it is a stability issue, do not run your RAM faster than stock, until everything else is fast and stable. If that's not possible (you didn't get a BE), run the CPU at speeds that allow your RAM to run <= its factory ratings.

The bolded part really helped, thanks, but I think that your conclusion is wrong. The ram is rated as DDR2 1066 5-5-5-15, and I was running it at 1000 with those same timings when I was at 3.5ghz, even though I spent a long time testing it at 1066 with the Kuma 2 years ago. The current most likely culprit in my mind is that the board just isn't quite stable at a base clock of 250, it wouldn't even POST there with any of the other CPUs, and when it would with the 830 I got greedy. I'm going to drop it down to known stable territory for the board at 3.4ghz and see if everything clears up.

 

[Cerb]

If it's not supposed to have a funny voltage at 1066, that should be fine.

 

[birthdaymonkey]

OP, didn't you mention in your previous thread that you'd tried two 940s and they'd both given you problems? If this is the case, now you've used three different quad core CPUs in your system, which would suggest that the problem has nothing to do with the CPU.

Your mobo worked fine with the Kuma, a dual core, but it seems to have issues with anything that requires more power. The 940, being a hot, high-wattage quad, produces the most errors, and the 830 seems a little better, at least until you turn up the voltage and frequency.

This suggests to me that the real problem is with your motherboard. Maybe you should consider taking that 830 back and getting a new board instead... although it's hard to justify spending money on an AM2+ board (940 won't work with DDR3).

 

[sangyup81]

It sounds like you have a CPU-NB problem. Try Prime95 blend and see if it passes 48 hours.