Worried about the stability of my system. Help

[traxevc]

Hello everyone,

after a lot of research and reading -being noob at OCing- I managed to OC my second pc (C2D E6600 @ 3Ghz) with good temps and very stable: >24 hours of error free Prime95.

After this success, I really want to OC my main pc: i7 950, asus p6x58d-e, geil value plus ddr3 1600mhz 2x4gb sticks, prolimatech megahalems rev b (I just ordered this one). At the moment I am running an intel extreme edition heatsink, but I am waiting for my megahalems to arrive. So, before the arrival of my new cooler, I decided to do some stress testing with different voltages in order to know what is the lowest voltage my cpu can run stable at stock speed and settings (including turbo boost; see the following paragraph). I tried 1.1v in bios. After 6 hours of P95 I got an error. then tried 1.15v in bios which gave an error after 12 hours of P95. Then tried 1.18v in bios which gave an error after 5 hours. At the moment I am stress testing with P95 at 1.21 in bios and waiting to see what happens.

My questions are three:

1. how worrying are these errors? Is anything wrong with my cpu?
2. The turbo boost raises the multiplier to 24 from the original 23. I remember when I first built the pc and I had the voltage on auto, the voltages in cpu-z used to show 1.275 or something at 100% cpu load. So, is the turbo boost in combination with my lowring the cpu voltage to 1.11v, 1.15v and 1.18v, the reason why I get errors after many hours of stress testing? Do I have to turn it off in general?
3. is there any point in me stress testing my pc before I OC it? Or shall I just OC it and then stress test it?

Sorry if these questions sound noob-ish, but they have been occupying my mind lately and I needed to ask somebody.

thanks and looking forward to your suggestions.

 

[lowjack989]

Which test are you running with Prime95....Small FFT's, Large FFT, or Blend?

 

[traxevc]

blend

 

[lowjack989]

Its more than likely your RAM giving you the errors then....Run memtest86 to verify

 

[traxevc]

Ive run the memtest thingy once few weeks ago but it didnt give any error. Is there another way to test my ram?

 

[lowjack989]

When OCing always turn off auto clocking software like turbo boost....Also make sure to turn off speedstep...You also may be getting a big vdroop while running prime causing you to become unstable....I believe setting your load line calibration correctly in BIOS will fix your instability. memtest is what I use for testing RAM I don't know of any others. After reading your OP it looks like vdroop i.e. voltage drop may be your problem

 

[traxevc]

Sorry for the noob question but what causes it? Is it turbo boost or speedstep? How can I try to stop it?

 

[lowjack989]

In BIOS look for the speedstep its there ....just disable it...same for Turbo Boost

 

[traxevc]

Thanks very much friend! In the meantime, if you think of any other suggestion, please do not hesitate to give me

 

[jsc]

lowjack, I don't know about that. I'm sort of fanatical about stability. I have 4 OC'd Core2 (they are still fast enough for what I do) systems:
GA-EP45-UD3P | Q9550 OC'd to 3.6 GHz (425 MHz X 8.5) C3 stepping
GA-EP45-UD3L | Q6600 OC'd to 3.6 GHz (400 MHz X 9)
GA-EP35-DS3P | E7500 OC'd to 4.1 GHz (373 MHz X 11)
GA-G41M-ES2L | E6500 OC'd to 3.87 GHz (352 MHz X 11) limited by the G41 chipset

I turn SpeedStep off and work out my OC settings. Final testing is a 24 hour Prime95 run of small fft's and a 24 hour run of large fft's. Then I turn SpeedStep back on and do a 24 hour blend test. I have never had a system fail the blend test after passing both fft tests.

I figure that if your system is not stable with SpeedStep enabled, your system is not stable.
----------
Overclocking since 1978 - Z80 (TRS-80) from 1.77 MHz to 2.01 MHz

 

[lowjack989]

Hey jsc, I know your a stability fanatic I have read your stuff....I was taught never to use any power saving features while OCing but if you say so..then I would believe you over me as you use a lot more Intel stuff and you have a helluva lot more experience than me

 

[jsc]

Hey, thanks, but you need to remember that my experience with overclocking (or anything else ) is my experience.

I do not know where the whole "turn off SpeedStep and leave it off" thing came from. It makes a lot of sense to turn it off while you are working out your OC settings. But it just didn't make sense to leave it off afterwards. I mean, why generate all that heat when the system is idling or under low load?

 

[lowjack989]

With my AMD OCing CnQ has tended to cause stability issues..it doesn't seem to scale voltages properly after manually setting them in BIOS...So I always cut off speedstep too when OCing Intel Rigs as I assume the same thing will happen...but you do make a valid point.

 

[traxevc]

What you are telling me is that speedstep drops my voltage at some point, i.e. when in a blend test P95 swaps from test to test, when there are gaps of non-100% cpu load. This way the cpu does not get enough juice to carry out the processes for a very short period of time, perhaps less than 0.5 second or something, resulting to an error. And all these because of speedstep. Have I undestood correctly?

 

[lowjack989]

The vdroop that I was talking about actually happens whether speedstep is on or not...When you push the processor the voltage tends to drop due to the mobo's vrm's not suppling enough juice to the CPU...This vdroop is very common. Some boards can have large vdroops which when OCing cause instability due to the fact that even though you may have it overvolted when vdroop is figured in is lower than the value that you set

 

[traxevc]

I finally increased the voltage to 1.21 in BIOS (1.19 in cpu-z) and done >24 Prime95 blend test without errors with turbo boost and speedstep on.

thanks for the suggestions guys

 

[lowjack989]

NP..remember to choose a best answer so the thread can be closed..happy tinkering