??? 05/17/06 18:56 Read: times |
#116411 - Ideas... Responding to: ???'s previous message |
Erik Malund said:
Does anyone have an idea what happens it I double the clock on a 20+ MHz x-tal, so the freq should become 40+ MHz
The absolutely complete answer is: NO Erik Well, I have some ideas. :) But Erik does have the gist of this right: we don't *know* what's going on inside. We can speculate. We can reason based on experience. But we don't *know*. And I'm guessing you won't find many manufacturers interested in helping you understand why their product doesn't work when you operate it out of spec. The bottom line is that once you start to operate the device out of spec, all bets are off, from correct execution to proper output levels to device lifespan. The functionality and characterization of the device as presented in its data sheet are only valid when all of the operating conditions are met, including the clock frequency. So, what's going to happen when you overclock? If you're lucky, and try to overclock aggressively, as it sounds like you are, the device will fail immediately upon boot, causing you to immediately realize the futility of your proposal. However, if you overclock less aggressively, the outcome is decidedly less predictable. Your device might well execute correctly, under some conditions. Perhaps it will run properly for operations except external bus operations. Perhaps it will run properly on your lab bench, only to fail when exposed to elevated temperatures out in the field--and those temperatures might well be within spec for the device. Perhaps it will run properly until your product is exposed to an electrically noisy environment. You might also squeak by successfully for years, only to experience a bunch of failures with a particular lot of devices. The bottom line is that you might very well not know the extent of the impact of your decision to overclock until you're faced with a very expensive recall. Why do these kinds of things happen? When you increase clock frequency, you decrease the amount of time available for propagation; the result is that you can end up latching values that are still transitioning. Additionally, you can create more EMI inside the device than it was designed to withstand, causing transients that exceed the noise margins of the technology in question. Increased clock frequency also typically results in a linear increase in power dissipation, which implies higher internal temperature--another condition under which proper operation can be compromised. And that's by no means a conclusive list. In short, overclocking is just plain old ignorant and foolish. Unless, of course, proper functionality and reliability is of absolutely no importance... in which case there probably isn't much of a market for what you're developing. --Sasha Jevtic |