Wednesday, May 28

Overclocking A Q6600

It's fair to say that I bought my various bit and pieces of my new build with the express intention of overclocking it. The main enabler of this is the Intel Q6600, a pretty long in the tooth quad core processor that runs at 2.4GHz out of the box. There are newer processors now which run on faster system buses, but their technical designs make them harder to overclock. And of course they're more expensive too and I was on a budget of sorts.

Overclocking is a bit of a dark art with a number of variables to consider, some in your hands and some out. It really is down to luck how successful you'll be even if you have the exact same set up as someone else. For instance, not all Q6600s are made equally; those of a lower quality require a higher voltage to run at stock speeds, and a higher voltage, although safe, will result in higher operating temperatures, something that needs to be kept as low as possible during the overclocking process. This Voltage ID or VID is set by Intel during their tests of each chip at the factory.

Unfortunately my Q6600 had a pretty poor VID of 1.3250 (other Q6600s have VIDs as low as 1.2250). 1.3250 is toward the upper end of possible VIDs so I was pretty unlucky with mine. Still, you play the hand you're dealt and it's precisely because you can overclock bad chips that makes the Q6600 such a good buy.

It's a point of contention as to what exactly a safe voltage is. Stock is certainly safe, but some people think that 1.35 is still stock, while others say that anything up to 1.5 is safe too (since it falls within Intel's own voltage range on the Q6600). One thing everyone agrees on however is that increasing the voltage above your VID may decrease the life of your processor; whether you need it to last the full 15 years it's been designed to operate for is up to you.

I wanted as much as possible "for free"; that means that I didn't want to adjust any voltages or other parameters which would push any of the system beyond their designed specifications and I would stick to the VID (or thereabouts). Within this criteria I was aiming for a core speed of 3.2GHz using a bus speed of 400MHz (the maximum for my RAM), a 33% increase on the the stock speed. Oh, and I didn't want to break 60C in core temperatures while fully loaded.

The basic method for overclocking is one of trial and error. Since I wasn't messing with any of the more complex settings I just had the FSB, CPU multiplier (which when multiplied by the FSB gives you the core speed) and the CPU voltage (which you may need to increase in order to allow the CPU to run at a high speed). Since everything is controlled by the BIOS, you need a reboot for each iteration making it all a long winded process.

Once I got into Windows (if I got that far - if not then I was pushing too far and needed to up the voltage a bit) I used CoreTemp to measure temperatures, CPU-Z to measure effective voltages (it usually drops by the time it gets to your CPU) and the number cruncher Prime95 to stress test the system.

I recorded most of my results here. I was a bit disappointed with the results, although they are pretty in line with those expected from a VID like mine. The good news was that I could reduce the voltage quite a bit to run the chip at stock speeds, but that was an academic conclusion - I wanted to go up in speed, not down.

At a stock voltage of 1.325v I found that I was able to get a stable 3.0GHz (375x8) with some very decent temperatures. Pushing it further to 1.35v allowed me to hit 3.08Ghz (385x8) and still without getting near 60C. To stem my curiosity I hit 3.2GHz (356x9) but only after ramping the voltage past 1.4v (I was lazy and used the auto setting in the BIOS, not a good idea since I now don't know if I could have gone lower). As a result my temperatures hit 62C under load.

Throughout testing I didn't have to touch any of my other components - even the RAM was using its stock voltage at a CL4 speed. I was especially impressed with the performance of my Akasa 965 CPU cooler - I wasn't worried about temperatures at all throughout, and I reckon I might still get 3.2GHz under 60C if I nail the right voltage.

Other things I found was that a faster FSB is better than a higher multiplier (so 400x8 is preferable to 355x9) and that a faster CPU core speed is better than a faster FSB (so 3.0GHz via 375x8 is preferable to 2.8GHz via 400x7). Another thing to note is that power saving technology like Speedstep may not work as well once you adjust some of these settings (for example since Speedstep works by automatically reducing the multiplier under low loads, going from 8x to 6x will have less of an effect than going from 9x to 6x).

Considering all the above and since I like the idea of sticking to stock (for now), I decided to settle with a speed of 3.0GHz via 375x8. I think with a better VID I could have pushed it a bit higher (1.325v alone may have taken me to 3.2GHz), and over time I envisage taking the system to 3.08GHz at least, and possibly even higher provided I can find a voltage that both keeps me there and my temperatures under 60C. But for now I'll take my free 25% increase and leave it at that.

4 comments:

  1. Anonymous12:28

    wanna build mine?

    ReplyDelete
  2. Anonymous09:46

    ....zzzzzzzzzzzz...oh wait did you say something interesting? nope...still talking about computers....zzzzzz

    ReplyDelete
  3. Anonymous16:04

    so you're a computer person...is it possible to build laptops??

    ReplyDelete
  4. Anonymous15:42

    Show off.

    ReplyDelete