Quote:
|
Originally Posted by Lasereth
I realize that it's just luck of the draw, but that's the problem. I don't want luck to determine if my computer is going to overclock or not, and I sure as hell don't want to go through costly RMA processes to get new parts. I've had this happen to me with 3 or 4 different sets of PC parts...it's really annoying. Like Rawb said, you'll barely notice an increase in speed by OC'ing your processor. I believe it's better to pay extra and get a processor that doesn't need OC'ing!
OC'ing is fun, and when it works it's very rewarding. My main reasoning is that OC'ing is a lot of luck and most people won't admit it.
-Lasereth
|
Laser,
Yes, a single purchase of components has some luck but the overclocking process is anything but. I've had success for years, in and out of good times when processors are flexible and others aren't. I teach local newcomers how to do it. As I said, always keeping things simple. No exotic cooling unless they want to delve into that on their own. We tune the systems to get the most out of them. Move forward until problems appear, back off to 60% of the increase. The systems are extremely stable when done carefully. When someone ignores my paranoia and runs at the bleeding edge - and some do - they're either comfortable and getting good at it or they're my source of horror stories.
As for the benefits of overclocking, generalizations don't hold water. (<-except that one
) Sometimes there's a significant improvement, other times it's just a learning process. The gain depends entirely on the user and intended use. A gaming system will be mostly GPU & vidRAM bound and so won't benefit much from processor improvements. (mostly) A FreeBSD build system needs the best processor & disk performance it can get, and video means nothing. Even so, there's a mix as subsystems become dependent on each other. This is common sore thumb when components are mismatched such that one is always waiting on the rest. i.e. The best GPU will benefit from a faster stream of shape & bitmap data.
Case in point: my 6800 running Doom3 overclocks to 84fps at ~80% of it's potential. Further overclocking makes no difference, until I notch up the AMD 64 FSB. Combined with careful DRAM timing, now I can bump up the 6800 and it makes the difference between 84fps and 94fps. If I clock to the limit of CPU stability, RAM timing falls apart and I'm back to 86fps.
Locating bottlenecks is the challenge, but that's the key to performance in any system. "Overclocking" or "tweaking" are this analysis and tuning as applied to PC's. Just as with automobile hotrodding, the basics sound the same but you can find different methods and results. Manufactured boxes don't have the luxury of this individual consideration. They're built to a mass-production spec that has to take into account averages of component quality, warranty costs, etc. The manufacturer has to be worried about the lowest-common-combination of bad parts. We as box building individuals are able to choose our limits up to our knowledge, the components we received, and the effort we're willing to expend.
Man, what a diatribe. Sorry if I put you to sleep. I usually have this debate in person and questions interrupt before I get on a roll.