My vid card sucks...well you knew that...

over any intake fan. or any opening in general will help with hair and some dust getting into the case. Though we all know that foam filters are prolly the wrose filters you can get, they really don't hamper air flow too at all. Right Iel and Sio? =)

that comment--which reminds me, I need to take the case off and see how things are in there.

Long answer...

Yes, Over Clocking any IC (integrated Circuit) will cause instability. Heres why (massively simplified by the way)…

So say you go out and buy an Intel core i7. You have a ton of choices:

i7-960 4 Cores /8 threads 3.20 GHz up to 3.46 with turbo boost 8 MB 45 nm

i7-950 4 Cores /8 threads 3.06 Ghz up to 3.33 with turbo boost 8 MB 45 nm

i7-940 4 Cores/8 threads 2.93 GHz up to 3.20 with turbo boost  8 MB 45 nm

i7-920 4 Cores /8 threads 2.66 Ghz up to 2.93 with turbo boost 8 MB 45 nm

i7-870 4 Cores /8 threads 2.93 GHz up to 3.60 with turbo boost 8 MB 45 nm

i7-860s 4 Cores /8 threads 2.53 GHz up to 3.46 with turbo boost  8 MB 45 nm

i7-860 4 Cores /8 threads 2.80 GHz up to 3.46 with turbo boost 8 MB 45 nm

Well, looking at all that they look the same, just the clock speed is different. Why is that? When they print the chips at the factory they run them through a quality control process. During this process they visually inspect every chip to look for defects (this is where the AMD 3 core chips come from, they are quad cores with one defective core that is ‘sliced off’ so it won’t mess up the good cores). If there is one, sayonara! To the rubbish bin with you! 25% or more chips per batch are discarded because of this. Some reports put Intel’s core i7 yield (number of good chips per batch) even lower than 75%.

OK, so all the remaining chips are physicaly good, but how are they once powered up? So the chips have current applied to see if the magic smoke gets out. Chip is still good? On to the next step.

So now the chip is put into a test bed and test signals are put through. This is a sort of stress test where every bit fed to the chip is known and specific output is watched for. Still good? Now the testing is ramped up. This is the point where the final clock speed of the chip is figured out. The speed of the input is ramped up until bad bits start flowing out. Ramp the chip back down and voila, a reference clock is known. The chip is badged for whichever model number matches the final good clock speed.

This is highly simplified and not entirely correct, but gives you an idea. Other factors also apply. For one, Intel needs far fewer ‘extreme’ chips than they need lower clocked cheap chips (more people buy the cheaper parts). So for many parts they are rebadged with a lesser part number simply so Intel can sell it. Intel also controls the number of certain parts on the market so they can control the price (basic economics).

So, now you just bought your shiny new core i7 920 and decide that you want to over clock it. Remember, not all chips are born equal. Some have flaws that keep them from running any faster, others might be rebadged 960s that Intel couldn’t (or didn’t want to) unload at the higher price point. If you are lucky it’s the latter, not the former.

So now, why do some chips clock higher? Tiny imperfections in the silicon matrix of the chip. Some of the imperfections and impurities cause resistance in the tiny electrical traces (the wires printed on a chip are called traces). Resistance means heat. Heat causes resistance in a wire. More heat means more resistance which means more heat… eventually you start a fire (this is what happens in electrical house fires)! More likely the resistance in the wire will cause lost data creating an error (lock-up or application crash) or if it is to bad, spontaneous reboots. In the extreme on a printed circuit it just melts that trace or overheats something else and the magic smoke gets released (fried chip).

So You are trying to overclock and you get a little smart; you get yourself a big copper heatsink with a big fan on it (or a nice watercooling setup, or if you have big bucks a phase-change cooler!). This helps because it dissipates the heat from the resistance in the traces giving you more ‘overhead’. The chip can experience more overheating before you get to that critical point where the heat is not getting dissipated fast enough and it reaches a point of failure.

The more you over clock a chip the less time between voltage loads on those hot traces. There is less time for the heat to dissipate so it builds up and gets hotter. Now to compensate for some of the resistance in the traces you can increase the voltage to the chip. If you have enough cooling this may work. But as voltage increases so does heat. In the end this is a losing battle and if you over-volt too much it will just melt the chip, no matter how much cooling you have on it; even liquid nitrogen won’t help after a while as the cooling can’t get all the way into the internals of the chip fast enough (it’s been done and there is an interesting paper about it, the chip melted down in the end).

Hopefully I have shed some light on this and not muddied the waters further. If you have any questions please feel free to ask!

I wrote it to be generic, not just applicable to you, but realy to anyone who was interested. in Tech-Speak I tried to make it 'hardware agnostic.'

And no, most dual core chips were designed to be dual core. And that is curious about your 3ghz chip reading as 3.2ghz. I think I might know whats going on there, but let me do some research before I say for sure. It sounds like a bus-speed or multiplier might be set a bit off.

Intel did do something funny with their orgiginal dual-core P4's and later the Core2Quad chips. They took 2 normal CPU's, for our demonstration here lets say 2 Core2Duo CPU's. They then printed out a small chip that would take all the data going in and out of the 2 Core2Duos and 'glued' it all together. So a Core2Quad is esentialy 2 Core2Duos glued together at the Memory Bus. this means that cores 0 and 1 (computers start counting from 0 so in a quad core you will have cores 0,1,2,3 not 1,2,3,4) can talk to each other directly, and cores 2 and 3 can talk and share data together. But if core 0 wants to share something with core 2 it will have to take a few clock cycles and write the data to RAM then send a little message to core 2 telling it to go pull the sata from RAM and work on it. Very inefficient compared to beeing able to do this on the chip it's self. The core i5 and i7 quad cores are more efficient and can all swap and share data directly on the chip.

AMD has designed each of their multi-core chips from the get-go with an on-chip memory controller (similar to what Intel is using now, years later in the core iX line). Having the memory controller on the chip is what allows 2 cores to be able to adress eachother diretcly to share data.