Gulftown, Bloomfield, Lynnfield, Clarkdale, SB Overclocking Guides (1155/1156/1366)

[Psycho101]

Just to emphasize my thoughts mate, I think you have done a great job with those guides!

Everyone has their own method and as long as it works for them, it's a good method IMO. I am greatful personally for you writing them as they do give very good explanations for what various voltages do. Also for the beginner, as LSD says, it gives the extreme novice a starting point on how to get going on an OC. The guides are logically laid out and well presented. Sounds like a minor detail, but I also like the dark background. I find it much easier to read than white backgrounds with black writing.

I have used my method for as long as I can remember with immense success, however I am open minded to suggestions. For example I freely admit I've only built nine i7/i5/i3 systems, so the intro and voltage explanations were useful to help me clarify in my head what does what.

I also agree that the best "stability" test is using the system for what ever you do the most, be it 3D gaming, modeling, encoding or simply web browsing and word processing. The only reason I usually emphasize a vigorous testing regime (50 IBT runs and 24 hours total of Prime 95 as a minimum) is because I've had problems before where a slightly unstable system has corrupted my precious data. There were always backups, but unless one runs a RAID array with redundancy, there is usually some data loss as nobody really backs up every day, unfortunately.

Don't know if it's worth mentoining or adding to your guides, but having an image of the OS drive before you start clocking is a God-send, especially if clocking the RAM too.Since getting Acronis, I can now simply re-image a slightly corrupted OS in about 3 minutes... amazing. I'm a bit over cautious, but as I often deal with other people's PC's, you can never be too careful with another person's data.

[miahallen]

Yeah, I hear ya :)

Concerning the stability testing bit. I've had really mixed results....for instance I had a Q6600 that was 24hrs prime95 stable but would crash if I was encoding video and playing a gmae at the same time. Once I got it gaming/encoding sable I was happy. On the flip side, the next rig I built (QX9650) would crash in prime95 within 30 minutes (everythime) yet was 100% stable for my daily uses (including encoding/gaming). Since those experiences I've changed my views on stability testing a bit. But since I know that there are plenty of people who would be quite to disagree with me, I don't push my views in my guides, but leave it up to the individuals. My 3 loops/1hr recommendations in the guide were chosen in the interest of finding decent stability in a short amount of time.

Thanks for the feedback Psycho ;)

[Psycho101]

I agree, I've found the same thing myself with both my old Q8200 and my E7400. I was 24 hour pPrime Blend, 12 hour large AND small FFT stable along with 50 IBT runs, yet I could not extract any large RAR files without WinRAR stating the RAR was corrupt (it was actually perfectly fine). After some minor tweaking I discovered it was CPU Reference voltage causing the problem on the Quad ( I added an extra tick at default VTT then increased VTT back) and MCH Ref voltage on the Dual Core (took away three ticks at stock VTT then increased VTT back).

As you suggest, I do what ever stress tests reassure me that the system won't fall over just sitting on the desktop, then I use the PC for a little while. Again, this is why I always quote a 3 day wait after a build if the customer wants an over clocked system. I don't build many machines so I have the time to use each build as my main machine for a couple of days, simply watching TV progs, encoding/time-shifting TV and if they spec a decent GPU I game on it too.

[gigaboy]

I agree, I've found the same thing myself with both my old Q8200 and my E7400. I was 24 hour pPrime Blend, 12 hour large AND small FFT stable along with 50 IBT runs, yet I could not extract any large RAR files without WinRAR stating the RAR was corrupt (it was actually perfectly fine). After some minor tweaking I discovered it was CPU Reference voltage causing the problem on the Quad ( I added an extra tick at default VTT then increased VTT back) and MCH Ref voltage on the Dual Core (took away three ticks at stock VTT then increased VTT back).

As you suggest, I do what ever stress tests reassure me that the system won't fall over just sitting on the desktop, then I use the PC for a little while. Again, this is why I always quote a 3 day wait after a build if the customer wants an over clocked system. I don't build many machines so I have the time to use each build as my main machine for a couple of days, simply watching TV progs, encoding/time-shifting TV and if they spec a decent GPU I game on it too.

OMG I was having this exact problem with rar files not extracting without crc errors with my overclocked IBT 20 pass stable E8400, but I followed your suggestion about tweaking cpu & mch ref voltage and viola, all errors are gone now!! Thank you so much for sharing this great info mate! I tried all sorts of stuffs to eliminate this error but all in vain(upping voltage didn't help but made things worse). This is probably one of the best piece of info I could ever find on entire internet. Sorry I'm bit too excited to write proper English now but hope you understood my gratitude. Again, THANK YOU!

[Salty]

In Miahalen's guide:

My suggestion with QPI, is to set it to the lowest multiplier available (but not “slow mode”).

Recently saw a post by LSDmeASAP saying "Slow Mode" had been fixed for Bloomfield, but I'm finding nothing enlightening on it by Google search.

What is Slow Mode, and what is it useful for? High bclk? I can see a high bclk in a validation of nearly 252MHz and ram at 1007MHz, but what do you do with other bios settings to get there? Any insight welcome.....

[Lsdmeasap]

Myself, and many others, were interested in this being fixed for maximum Bclks. Without slow mode you run into a QPI wall at 8.8Ghz on X58 using x36 multi for QPI, slow mode uses x24 so you have much more room to play.

It's been fixed for a few months now for Bloomfield CPU's. I'm still waiting on it to be fixed for Gulftown though!!!

I have a killer Bclk 980X, can do 246 on water, but same applies frozen since QPI slow mode does not work for these CPU's.

So max Bclk for me with that CPU is the same under dry ice or water, not cool at all :(

[Salty]

This was a bios fix?

[miahallen]

If anyone is interested in following my work, please do so here, thanks :)

[Lsdmeasap]

This was a bios fix?

What, for QPI Slow Mode you mean? If so yes, in the past, up until a few months ago, when you selected QPI Slow in the BIOS it would actually still use x36 (You can verify on old BIOS and use CPU-Tweaker)

[Salty]

What, for QPI Slow Mode you mean? If so yes, in the past, up until a few months ago, when you selected QPI Slow in the BIOS it would actually still use x36 (You can verify on old BIOS and use CPU-Tweaker)

Yeah, that's what I meant. Thank you LSD for sharing this info.