• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Overclocking CPUs

Associate
Joined
19 Apr 2010
Posts
474
Location
Surrey
Hi guys,

I have been out of the system-building/overclocking loop for some years now (coming up on 9). Having just built my first machine in nearly a decade, all has gone well, but I'm wondering if overclocking is as difficult as it used to be?

How does CPU overclocking happen now? Is it just BIOS based or something more still?

I understand graphics cards are simply software-based now, but CPUs are a whole other matter to me.

Thanks!
 
CPU's are bios based. There's a nice tutorial in the root of this thread, one of the stickies. Read through it, should take you an hour or so and get you up to date :D

GPU's, yeah, they're software based but are really simple. Up the core clock untill you get artifacts in a stress program, them drop by 20 mhz.. same for memory clock, if you want more, up the voltage

It's not really that difficult matey, you'll never break anything unless you cook it and if your educated, you wont..

*edit : Tell a lie, here's the links to the guides..:

http://forums.overclockers.co.uk/showthread.php?t=17612922
http://forums.overclockers.co.uk/showthread.php?t=17803239
 
Nice one mate. Glad to hear things have got much easier :)

Is lapping on CPUs still a concern?
 
Last edited:
Its definatly easier and cooling is a lot better!! most CPUs and GFX cards will give you a decent overclock on stock voltages.
 
If anything CPU overclocking has got harder, all this bclk/uncore/VTT/GTL voltage/drive strength/clock skew, back in the good old days we had FSB/Vcore/Vdimm and that was pretty much it!
 
So from what I can see, the bulk of overclocking stems around pushing the FSB up in 5mhz increments, right?

In essence, especially if you're core multiplier is locked, however with FSB you also overclock more than just the CPU, meaning you won't quite know what to blame upon hitting an obstacle. As to increments, start by larger and narrow down, if you jump too high at a time and the rig cant handle it you might need to clear the CMOS memory by removing the battery for a a few seconds.

It might be a wise and prudent move to gauge the 'limits' of all the factor components in relative isolation, both at stock volts and overvolted within reason. For example, to find your mobo's 'wall' underclock the CPU and the RAM and try upping the FSB to find out how high can you go.

Likewise test the other bits, this way you will know what's holding you back when everything comes together, what you should work on. Furthermore, if you followed my advice and tested the 'limits' of each part under stock and non-stock volts, you can figure out what combination of overclocks and voltages for each component gives the best overall overclock with the least voltage increases across the system. Figure in things like the relationship between component overclocks and performance (e.g. CPU will yield 'more' relative performance than RAM), temperatures, fan noise to maintain the clocks, stability, gain-to-lifespan-loss, sustainability, power-draw (if you're concerned like that) etc.


If it helps you can draw a crosstabulation of components with their limit clocks under stock volts and maybe 3 other levels of voltage beyond (say 'slight', 'moderate' and 'the most I feel comfortable with') Stare at it for a while and hopefully start noticing which parts 'give in' more easily then others, which ones handle stress the best and/or the worst, which exhibit the fastest diminishing returns for each level of volt bumps etc. Figure out what could work, perhaps 3 plans of how to configure everything to achieve the best benefit-to-cost ratio for the overclock in general. You can go as crazy/complex as you want with this, adding extra layers to the matrix noting test results at various stages and interactions between 'individual' component overclocks and other sets of tested overclock variations.

Heck if you're really nuts and work up a little sample you can bring in cross-correlations, probability statistics and multiple regressions. But only if you're deviant like that :rolleyes:


Naturally, every component -will- behave differently in an overclock depending on what you do to the others and you -will not- know how your plans work out until you put them to the test. Not only is each component quite unique in its tendencies and tolerances, but also each combination of components that make up a rig. This is the reason why the above paragraph is mostly hit or miss, but also most definitely best reserved for the meticulously hardcore and/or the mathematically curious. ( I'm quite sure fascinating and deliciously generalizable trends eagerly await discovery :D )

Nonetheless, its a much more systematic approach to the whole ordeal rather than just upping the FSB and starting to blindly guess when inevitably the rig becomes unstable. Depends on your personality type really, some prefer to work it out like suggested, yet others fancy jumping right in and 'feeling' their way through the process. Whatever works for you, perhaps even a sort of middle ground. Also largely irrelevant if all you're looking for is a modest overclock. Merely trying to illustrate that it can be as involved as you wish it.

[snip] I understand graphics cards are simply software-based now [snip]

Finally, just wanted to note that yes, basically overclocking the GPU has become a rather simplistic software procedure but one little detail to keep in mind when clocking newer cards (especially the ATI range) is that apparently they implemented a fancy memory error-correcting algorithm that will try and compensate by running the 'thing' (do pardon the crude expression) a second time to resolve the error. Meaning that there will come a point where performance will start to degrade as you clock higher, possibly before you can detect visual artefacts. In short, yeah overclock but do also run a benchmark every once in a while to find out whether you're gaining or losing fps, try finding the 'sweetspot'.

[snip] Is lapping on CPUs still a concern?

Lapping... well lapping is lapping, modern CPU's conventionally have nickel-plated caps which very willingly reveal a deliciously conductive layer of copper underneath, unless you grind off too much there's not much to go wrong apart from the certainty of a warranty void. As far as the necessity and temp gain go though, I suspect that CPUs are generally more 'even' than a decade ago. (pure speculation) Components seem to withstand temperatures well and coolers cool adequately. Note also that most modern CPUs clock very well indeed even 'unlapped', especially the intel offerings. (4Ghz on air seems almost commonplace with the i7 920) Lapping will shave off (pun very much intended) +/- 5 degrees most of the time and possibly more if you're unlucky enough with the heatsink and/or CPU. If you do opt to lap, do both the CPU and the heatsink, otherwise the purpose is often defeated.


THIS is what you're aiming for mate.



- Ordokai
 
Last edited:
Ordokai, that has helped immensely my friend. I have just received the last of my order, so I'm going to get started this weekend, thank you.
 
Slow incremental steps.

It's like free money when you get a good overclock. Good luck and look forward to hear how you get on with it.
 
Back
Top Bottom