a little overclocking isn't a risk at all, or at least not a significant one according to the scores of people who've done it, including people who've done it on dozens of machines for people (who would tend to be frustrated/angy if the expensive smoke came out a week later...). it's also not technically that insane, processor speed is a worst case specification, determined by testing at the plant, under better than worst case conditions a processor can run faster. rough processor speed does depend on design and how it's made, but subtle differences between how each chip actually turns out result in the different speed grades, the raw chips have to be tested to determine the potential of each one. a well cooled and clock chipped cpu is probably more reliable than an unchipped one running at a higher internal temperature (and coincidentally, a lower temperature means a faster maximum speed for reliable operation).
there are also some design tolerances built in, like how tolerant it is to noise on the clock and other lines that affect the speed rating. in a real design these should be smaller than the design allows, meaning there's some head room in the speed that can be exploited by keeping other things more tightly controlled. temperature is actually a big factor, for two reasons: first, obviously the semiconductor material melts or allows the dopants to flow/diffuse at a certain temperature (the dopants are what make it an orderly device rather than a random lump of silicon), second the individual parts on the die are connected by metal, the resistance of metals (most, including the ones currently used on micro's) increases with temperature, which increases the time it takes for a signal to propagate and overcome capacitance. lowering the temperature of the die and metal increases the speed they can run at, cool it well bellow the maximum rated temperature and it can reasonably be run much faster than the speed that it's rated for at maximum temperature (because of the better conductance of the interconnects and lower leakage current which also increases with temperature and causes problems). one of the things that has enabled faster processors has been using copper for those internal connections rather than aluminum. copper conducts better and has allowed higher speeds but people had to figure out how to make it stick to the semiconductor. figuring out how to use copper was a very, very big deal in the semiconductor industry. if gold could be used the same designs would run even faster, but, gold is expensive (though not terribly in the very, very small amounts that would be used) and more importantly no one has (to my knowledge) figured out how to use it on chips other than as the wires that go from the chip to the leads. the ways gold wires are attached to semiconductors wouldn't work well for the many, many interconnections on the chip which have to be formed in bulk rather than one at a time for cost reasons. there are also bias voltages that, because higher precision power supplies cost more money, are allowed to be somewhat loose. bias and operating voltages also affect processor speed, by controlling them more tightly, the safe operating speed of the processor is increased. further, there are specs for how much load other parts can place on each output line and how strongly each input must be driven, variations in load (in different machine designs and different specific instances) and drive levels (from different designs and specific instances, i.e. the other chips all have a range of "normal" load or drive they provide, yours may be at the high or low end or somewhere in between, usually parts perform at the "better" end of the rating because of allowances made in manufacturing) will also have an effect on how fast things can be run before they get random. it's also the pure quest for speed, and in the case of older machines a willingness to experiment knowing that worst case you'll fry a processor that wasn't that valuable anyway. it's curiosity and the need to play (at least for geeks like me, which is how it started, once we geeks had fun with it and proved it more casual users got interested along with "lower level" geeks). when i get my garage workshop setup i plan on some overclocking experiments myself, starting with a 9500 i have several spare processor boards for. i'll use the older machine to gain experience and learn just what can be tolerated before moving on to somewhat more expensive machines. i won't be "happy" that i've found the edge of the envelope until i fry at least one chip, then i'll have a good idea what the "real" limits might be. on the other hand, i'll be measuring and watching things more closely than some clock chippers so i should hopefully have more warning when i'm pushing too hard (semiconductors provide many tell tale warning signs, if you watch for them). of course if you're going to chip, you also want to chip the memory and other busses, which have even more variables involved, some of which can be easily change and some that can't (i.e. if you have faster ram you may be able to run the ram bus faster, if the ram controller can also handle it or have it's operating environment changed to allow it). it's also well known that when most semiconductor makers get a large order for slower chips (slower relative to what they are offering for a particular type), they usually only test the batch to make sure each will meet that spec., knowing that many of the chips would pass even their stringent test for higher speeds (obviously when a chip is "new" they tend to check them all desperately trying to get enough of the top end chips). after a production line has been run for awhile and the process well tuned the yield of higher speed chips improves and there can easily be more high speed chips produced than the market needs which results in fast chips actually being sold as slower chips, knowingly (i.e. the chips have been tested and the maker knows they can run faster, but they are willing to sell them as slower chips at the lower price because they have a surplus) or just because they have enough of the fast chips that they don't fully test the speed of every chip, so long as they have enough of the fast ones and there is still demand for the slower, cheaper chips which are actually made exactly the same way (as best as the manufacturer can control conditions) but which may or may not test out as well. over all though, it's the knowledge that it's possible and the joy of "upgrading" your machine. it's no more insane than filling all of the hard drive bays in a tower machine, both put stresses on things that may shorten life depending on what you do about cooling and other problems. Zoki wrote: ------ > *** I still don't understand why anybody would risc frying his CPU to make > it run 100MHz faster... You gain what by doing this? > > Zoran. -- -- Philip Stortz --"There are reminders to all Americans that they need to watch what they say, watch what they do, and this is not a time for remarks like that; there never is." White House Press Secretary Ari Fleischer, how much blood will it take to win back the first amendment, which the Bush administration clearly does not support! The thought police are amongst us. -- G-List is sponsored by <http://lowendmac.com/> and... Small Dog Electronics http://www.smalldog.com | Refurbished Drives | -- We have Apple Refurbished Monitors in stock! | & CDRWs on Sale! | Support Low End Mac <http://lowendmac.com/lists/support.html> G-List list info: <http://lowendmac.com/lists/g-list.shtml> --> AOL users, remove "mailto:"; Send list messages to: <mailto:[EMAIL PROTECTED]> To unsubscribe, email: <mailto:[EMAIL PROTECTED]> For digest mode, email: <mailto:[EMAIL PROTECTED]> Subscription questions: <mailto:[EMAIL PROTECTED]> Archive: <http://www.mail-archive.com/g-list%40mail.maclaunch.com/> Using a Mac? Free email & more at Applelinks! http://www.applelinks.com
