Think this is stable enough?

[Nic Miller]

i5 2500k @3.3ghz overclocked to 4.6ghz, Asrock extreme4 gen 3, 8gb gskill ripjaws

So I did some more overclocking testing and reduced my vcore from 1.34 volts which seemed nice and stable to 1.30 - got a BSOD after a couple of hours of Prime blend so this wasn't good. So I upped the voltage to 1.32 and ran a blend test. I checked it after 8 hours and it was fine, then went to bed and checked it in the morning at about 11 hours - there was just a black screen so something happened between 8 and 11 hours. Maximum temps reached were between 51 and 53 degrees across the 4 cores at the 8 hour point.

For a purely gaming system, who would say this is stable enough? Should I run any other tests or should I just go back to 1.34 volts or should I treat this as a fairly stable system?

Thanks

Nic

 

[LtMatt]

I'd go back to 1.34 if it was me, especially since you know you've had problems at 1.32. FYI i require exactly the same amount of volts for my overclock.

 

[Nic Miller]

you're probably right and temps were only 56 degrees at 1.34 volts so I shouldn't be overly concerned with this should I?

Which tests would you do and for how long? Prime? Intel Burn test and for how long?

Part of me wants to just use the 1.32v as I never use the computer for longer than a few hours normally so the 8+ hour prime would suggest it is stable enough. Do games stress the system more or less than Prime though?

Thanks

Nic

 

[Plec]

Part of me wants to just use the 1.32v as I never use the computer for longer than a few hours normally so the 8+ hour prime would suggest it is stable enough. Do games stress the system more or less than Prime though?

I would do your stress tests as these are good early indicators of an unstable clock - but until it's proved iteslf stable with everyday use over the coming weeks you wont know for sure that 100% stable.

If you're worried about extra volts i would personally drop your clock to 4.5GHz and stick with the voltage your comfortable with - 1.32V. The odds are that this voltage should keep things stable at 4.5GHz - and you won't notice any performance difference in losing 100MHz.

 

[LtMatt]

you're probably right and temps were only 56 degrees at 1.34 volts so I shouldn't be overly concerned with this should I?

Which tests would you do and for how long? Prime? Intel Burn test and for how long?

Part of me wants to just use the 1.32v as I never use the computer for longer than a few hours normally so the 8+ hour prime would suggest it is stable enough. Do games stress the system more or less than Prime though?

Thanks

Nic

Those temps are lower than mine. When i run intel burn test i get up to around 79c. As long as i stay under 80c when stressing im happy. Throttling in the 90's remember. Under gaming i never see my load pass 60c.

I personally only run 10 runs of intel burn test on maximum and an optional 8 hours of prime blend, if it passes that im happy that my system wont crash during gaming, which untill now it never has.

I think the general rule of thumb is a maximum of 20 runs of intel burn test, that always seems a little extreme to me so i stick with 10.

I would do your stress tests as these are good early indicators of an unstable clock - but until it's proved iteslf stable with everyday use over the coming weeks you wont know for sure that 100% stable.

If you're worried about extra volts i would personally drop your clock to 4.5GHz and stick with the voltage your comfortable with - 1.32V. The odds are that this voltage should keep things stable at 4.5GHz - and you won't notice any performance difference in losing 100MHz.

This.

 

[Nic Miller]

Is the stability issue solely down to vcore voltage or could changing other paramters make it more less stable? Don't know much about overclocking so just changing vcore at the moment.

PLL overvoltage is set to enabled, speedstep is set to disabled and max wattage at 225.

Rest set to auto I think.

Not overclocked the RAM, so assume I don't need to do anything to it.

I might just live with it as it is at 1.32v and see how stable it is in the real world - surely that's enough power for a 4.6ghz overclock? and why do my tempos always seem so much lower at this overclock than what others report? My cooling is a Corsair H100 with 4 viper fans on push pull, also have another 12 fans in the case - the air coming out is very cold! ANother fan is directed on the back of the CPU, and there is a 20 cm fan facing the top of the CPU. Is my cooling doing a good job then? Bloody noisey though!

Thanks

 

[Plec]

Is the stability issue solely down to vcore voltage or could changing other paramters make it more less stable?

The vCore is the most crucial setting especially with moderate clocks - but other setting do come into play with higher clocks.

Ideally you should set the memory's timings, frequency and voltage manually and be running it as close to 1.5V as possible.

I might just live with it as it is at 1.32v and see how stable it is in the real world - surely that's enough power for a 4.6ghz overclock?

Not really - there are lots of members who are having to push more volts than 1.32V to reach even 4.5GHz - and that's if their chip will do 4.5Ghz.

Not all chips are equal. For example i have a gem of a chip (will run at 4.8GHz at reasonable volts) but i did a build recently that struggled to hit even 4.4Ghz - it's basically luck - although informed tinkering can increase your luck sometimes by a couple of hundred MHz - but this will usually involve vCore (directly or indirectly), PLL and DRAM etc...

 

[LtMatt]

I might just live with it as it is at 1.32v and see how stable it is in the real world - surely that's enough power for a 4.6ghz overclock?
Thanks

All chips vary, yours may require 1.34 for 4.6, but you sound like you want to run it @ 1.32. Why dont you follow plecs advice?

16 fans in one computer?

 

[Nic Miller]

I'm sure one of the programs (think it's CPU-Z) says the timings are 11-11-11-24 for the ram, but my ram is 9-9-9-?....could this be causing an issue? and if so how do I correct it? Am sure the timings are correct in the BIOS though. The stress testing seems fine on the small FFT but failed on the blend - suggesting something other than the overclock?

I'm not bothered about running it at 1.34v - it's not that high, I don't use my PC massively as just too busy, and plan to replace it with IvyBridge over the next 2 years anyway.

Cheers

Nic

 

[Plec]

It wont be the cause of your stability issues as your memory is running under stock.

I'm presuming this is your memory - link

In the CPUz memory tab it should read (the SPD will show its XMP spec):

• 800
• 9
• 9
• 9
• 24
• 2T (command rate)

You need to go into the BIOS and set the memory up manually - and keep the volatge as close to 1.5V as possible but i would stick with 1.55V for now as it's still under the 1.58V threshhold.

 

[Nic Miller]

Why 1.55v and not the stock 1.5v? Any risks here?

I still think 8+ hours prime stable is good enough - am going to run IBT tonight on 10 runs - never used it - can anyone help with settings or is all I need to do is just press go?

Cheers

 

[Plec]

Why 1.55v and not the stock 1.5v? Any risks here?

I didn't know the exact model of your G.Skill but if it was the one i linked to then stick to 1.5V.

I'll try and search out a good IBT thread for you - it goes into a lot of detail.

But keep in mind that even if it passes all your stress tests it will still have to prove itself in everyday usage - i've had system run prime all day and 20 passes of IBT and the system would hang while browsing.

 

[Plec]

IBT thread - link

 

[LtMatt]

i've had system run prime all day and 20 passes of IBT and the system would hang while browsing.

Was that when you were using a negative offset? Because i found that could happen when i had the cpu on a strong overclock coupled with a negative offset so the vcore would drop more at idle than normal.

For example i am now running 4ghz overclock with a negative offset of -0.075, it was -0.095 before but that would cause the problem you mentioned.

 

[Plec]

Was that when you were using a negative offset? Because i found that could happen when i had the cpu on a strong overclock coupled with a negative offset so the vcore would drop more at idle than normal.

For example i am now running 4ghz overclock with a negative offset of -0.075, it was -0.095 before but that would cause the problem you mentioned.

My reply was very much a generic response - as i've built many systems and some have hinted at being stable through stress testing and gone on to prove otherwise through general apps use. Oddly enough my present system was very simple to set up - and all the clocks i experiment with are usually simple to setup (this is due to the luck of my chip rather than any great skills on my part).

My present board doesn't have the offset feature, to my annoyance. I bought this basic board (C45 on release) as it had all the features i required and i was keen to test how well it could clock compared to its bigger brothers. Although it hasn't dissappointed i wish i had more BIOS options to play with - if i had my time again i would have bought the GD65 or the Gigabyte equivalent. If i didn't have such a forgiving chip i think i would have run into problems had i needed to fine tune a desired clock.

As it is my chip pretty much sucks up anything thrown at it (without subtle tweaks) so i haven't felt the need to change MB - especially as i get to play with new builds/BIOS quite often for friends and family.

 

[LtMatt]

My reply was very much a generic response - as i've built many systems and some have hinted at being stable through stress testing and gone on to prove otherwise through general apps use. Oddly enough my present system was very simple to set up - and all the clocks i experiment with are usually simple to setup (this is due to the luck of my chip rather than any great skills on my part).

My present board doesn't have the offset feature, to my annoyance. I bought this basic board (C45 on release) as it had all the features i required and i was keen to test how well it could clock compared to its bigger brothers. Although it hasn't dissappointed i wish i had more BIOS options to play with - if i had my time again i would have bought the GD65 or the Gigabyte equivalent. If i didn't have such a forgiving chip i think i would have run into problems had i needed to fine tune a desired clock.

As it is my chip pretty much sucks up anything thrown at it (without subtle tweaks) so i haven't felt the need to change MB - especially as i get to play with new builds/BIOS quite often for friends and family.

Ah fair enough.

 

[Nic Miller]

cheers for the link - will try an IBT tonight. SOme say that is all you need and to ignore prime.

How about 20 runs on IBT at max plus 8 hours prime blend - if it passes, try it in real world situations for a couple of weeks and monitor - sound okay?

Cheers

Nic

 

[Plec]

How about 20 runs on IBT at max plus 8 hours prime blend - if it passes, try it in real world situations for a couple of weeks and monitor - sound okay?

Sounds a very solid test - IBT if set up properly is a great tool, but i still like to use prime95 as it will sometimes draw out an error that IBT misses. Some people will argue that IBT wasn't set up properly if it doesn't draw out an error that Prime95 subsequently finds - but i agree to disagree on that point. Each to their own is my philosphy...

 

[ThePengu1n]

Personally, I run prime for 30mins, do a cinebench run and if it's fine, just start using it everyday. Worst case, you get a bsod and add a touch more VCore, but as far as I'm concerned, if it's stable in what i do everyday, that's good enough

 

[Nic Miller]

Well I switched on pc this evening and went to sort out a new user account for the mrs and get a bsod out the blue! Not even being stressed! So have now upped vcore to 1.34v again. Have also manually set ram to 1.5v and set the timings manually as it was showing it as 11-11-11-28-1n - which is clearly wrong. Dram was previously set to auto. Could these dram issues be the prob?

Now running ibt on maximum, 4 threads and 20 runs - stable through 11 runs. 133.238s and 123.4391 gflops - core temp max 56,62,63,59.

Sound like the test is set up okay?

Cheers

Nic