Roundup: The Best Overclocking Software

"We’ll end with a recommendation: it’s sometimes better to overclock via the BIOS setup program. This is true both because of effectiveness (some programs don’t display the frequencies well, for example) and because of operating system choice (the majority of the programs we’ve mentioned are only for Windows)."

There are advantages to overclocking from windows too actually. Other than being lazy that is.
When you're booting your system at an overclocked state, chances are your vrm's can't keep up at powerup for some reason I don't know. So your oc will fail. That I assume is the reason that gigabyte boards shutdown and restart when changing such settings - to avoid the problem.
If you oc from windows, the changes being made are significantly less than from 0 to preset.

Regarding the programs - I have some comments on them. They might be brilliant in some situations, but they don't nessecarily work.
For instance on my work pc, a hp dc7800 with an e6750, setfsb can only read the fsb, it can't set it - no matter which of the 3 types I tried (mmt, slp and the other one). I suspect hp deliberately did something to prevent it from working. So on oem systems you may not find the solution you were looking for in software.
And atitool didn't work very well with my 8800gtx. It could set the gpu and mem speed, but the artefact check isn't able to figure out if the shader speed is too high. So even though it'd run 20 minutes scanning for artefacts in atitool, a 3dmark test would crash at the nature scene when the lights were rendered. So don't rely on the button to automaticly find the best frequency if the shader is linked to the gpu speed.

ps. on all ati hd48x0 cards I've seen so far, the max gpu setting available in overdrive (790 or so) seemed stable - no need to run the check :)

OC that has trouble at bootup can't survive prime95 anyway. Software oc in Windows simply bypass the bootup by oc once it's already in Windows. It's bypassing the problem, not solving it. The oc is still unstable. For a solid oc, bios > software.

When I purchased my first Motherboard back in 2006 (for the previous 6 years, I had been using a custom-built Alienware, and before that, a Compaq laptop, and before that, a Pentium 166 my dad put together, and before that, and Mac II Se), I didn't understand the point of software overclocking, especially since I had to redo the overclocking every time my machine started up. To me overclocking was changing the FSB jumper from 1x to 1.5x or, if I dared, 2x.

Besides things like cool'n'quiet and whatever Intels version is (Speedstep?), why would chipset manufacturers even open-up software interfaces for overclocking, when for so many years they were completely against it? Now days anyone can click the "+" in Gigabytes Overdrive and gain a few extra percent. Must make it hard to sell processors in granularities of a couple hundred MHz when you can overcome that through software without any changes to voltage or mupltipliers.

Anyway, good article. Its nice to see the 3rd-party software that can be used, especially if you have no options from the hardware vendor.

one other thing about software overclocks. you can oc just when you need it (playing crysis) and set it back to stock speed (or even lower) when doing something not cpu intensive. kinda like poor mans' speed step. thats how i oc my 2.6 ghz p4 northwood in my (very modded) compaq

randomizerStrangely for me, the opposite was true. Windows would BSOD at 3.6GHz or higher on my E6600 even at 1.75V, yet it would be quite stable with much less voltage (although still alot more than most people would dare to try) if I OCed in Windows. I have no explanation for it. Perhaps somebody can explain this BSOD:"The Cryptographic Subsystem has failed a mandatory algorithm self-test in bootstrap." (Error code 0x000000123)Forget googling the error code, you'll only find me asking what that means and something else in Chinese.
My guess is this--your FSB chipset has different timings/subtimings (like ram does) that are invoked when setting the FSB in the BIOS depending on what frequency is set, and these are referred to as "strappings" (I don't know exactly why, but the reference to "bootstrap" in your error code makes me think this is referring to these "strappings")--some timings/subtimings that work at one frequency (one strapping) are too tight for higher frequencies (in another strapping), so as the FSB is changed in the BIOS, a new strapping is used for the new FSB--it might be that when OC'ing your CPU in the BIOS, you got into a regime in your FSB where the timings just don't mesh well for whatever reason (like there's no good set of timings on the FSB), and some errors are encountered relating to the strapping, spitting out that error code. When you OC in windows, the strapping for the FSB initially set in the BIOS is used, regardless of how the FSB is changed in the software (i.e. the timings of the FSB chipset are not changed at all when increasing the FSB in the software since the FSB was not changed in the BIOS at all)--my guess is that the strapping corresponding to the FSB in the BIOS just happens to work better for your clocks, which is counter-intuitive to me, but then again, nothing works exactly like you'd expect... That's my best guess...

Shadow703793I would NEVER recommend OCing Through Windows+1, but I would change this to "I would NEVER recommend CPU OCing Through Windows". RivaTuner is a Godsend for GPU tweaking.