9300 O/C question

those are the lowest clock speeds I've seen using the utility. I personally think it's safe. But I'm also not going to run my card overclocked unless I actually need it to run a game smoothly. In which case I'll clock it to 380/780, play, and then return it to stock.

mine auto detects to 329/669

Out of interest does anyone know exactly how this utility "auto-detects" settings? I haven't used coolbits before, having had ATIs, but I have used ATI-tool and the auto max feature on that basically increases settings until your machine artifacts itself to death. I am not suggesting that coolbits is unsafe, but just trying to get an idea if these settings are found by testing higher and higher settings until weaknesses are detected or if they are relatively arbitrary?

Cheers

I don't know for sure, but it looks like runs in a loop incrementing the clock speeds and rendering 3d (but not to the screen) and as soon as it detects any malfunction it uses the next lower speed. I'm sure they're not arbitrary.

Explain to me how you all feel OCing you video cards will harm them? Assuming you reach a happy speed (no artifacts) and it keeps running (no CTD's or Reboots) then it's just fine to continue forever. All chips (no matter CPU/GPU/Mem/etc) are heat limited. If you get it run under that heat threshold, it will do that forever. There are no moving parts, nothing to wear....as long as heat is kept under control all is good.

I ran my last video card for almost 3 years at more then double the stock speed. (and it will soon go into my boys PC where it will live happily OC'd for quite a bit longer!) My first homebuilt PC CPU ran OC'd from the first week I built it until I repleced it with a better lot CPU and then OC'd it a lot more! Going on its third year it works fine. (800Mhz OC too!)

As long as you manage the heat there is no reason to back off unless your looking to save power when on battery. Otherwise let it rip, just watch the temps now and then. (though they will tell you when its too hot by acting up, when that happens check the temps. Maybe it's time to clean out all the dust!) In fact that gave me an idea, I run filters on my 10 Fan Desktop (all quiet 80mm's), think I'll rig some up for my I9300 sometime......

Two words: Electrostatic migration.
Two more words: Electromagnetic Interference. (EMI)

By mitigating the heat of your overclocked component, you've avoided immediate "burnout". Failing the smoke test. But any time you put two dissimilar, non-inert elements together, you will encounter corrosion, migration, or solution/particulate. And electronic components carry their own baggage with electrostatic migration, in the form of EMI.

Modern computer chips are frequently made from, at least, silicon, copper, and plastic. In some cases, steel, gold, and other substances are used. You might think that there are no "moving parts" in a chip, but you're wrong. Electrons move. With them, they carry a charge. Over time, two things happen:
1. The transistors develop their own "charge", similar to rubbing a needle across a magnet. Usually, this charge is very small, and takes a long time to develop, but eventually it can affect not only it, but the transisors around it. This is a form of EMI, and can permanently disrupt a chip's operation. Turning off your computer for long periods of time between over-clocked uses can mitigate this very well.
2. Over time, connections run too warm and at too high of a frequency will "migrate" into each other. Literally, they can fuse and become useless. Keeping your chip well within engineering tolerances can forestall this. Unfortunately, surface heat measurements aren't useful gauges of what's happening to each individual transistor. Frequently, transistors are actually running much, much hotter than a surface temperature gauge can detect. We're playing with lightning inside of these things

I say this, but with the caveat that electrostatic migration takes years to do its job, and generally only occurs when running circuits at a higher voltage or frequency than they were designed for. Chip designers understand electrostatic migration and EMI issues, and how to mitigate it. That said, running a chip well out of the spec for which it was designed, even if it "works fine" for long periods of time, is also almost certainly damaging it. Slowly.

However, it's normally the difference between reducing the useful lifetime of a chip from 20-30 years to 2-5 years. (I've worked at a silicon design house. Yes, they design these things to last forever, or at least 30 years. Even if they know they won't be used nearly that long.) And when it starts malfunctioning, reducing the clock speed or voltage can give it a new lease on life for a while longer as it begins operating within engineering tolerances.

Given PC life-cycles, you can put a big "OK, I know about electrostatic migration, but I don't care" sticker on this post and ignore it However, you can't assume you aren't damaging your chip by overclocking it. You almost certainly are. The question is how much this damage is worth to you. Given that PC chip prices continually fall, there's a very good chance this damage is worth it in exchange for increased performance; 3 years from now, that same chip won't be worth enough cash to have ever worried about it in the first place.

Well, I EMI would be a concern if we were going out of engineering tolerances...but most of the overclocking people do is not over engineering tolerances.

The 6800 Go GPU has the exact same engineering tolerances as the 6800 Go Ultra GPU. Nvida doesn't make special silicon for each....They simply part them out based on yield quality. As we have progressed in our fabs. yeild quality keeps getting better and better... that still means companies have to fill out their product line. So you end up getting "lower end" chips that are really just as capable as the "higher end" product.

For example..I got a Pentium 4 1.6 Ghz northwood cpu that I was able to take to 2.4 ghz with no voltage increase. After some voltage modification I got that sucker up to 2.8 ghz with the stock HSF unit...This processor is still going !

I will note that heat should always be watched, as the 6800 go Ultra does ship with a different heat pipe setup...though I doubt if the one that comes with the 6800 go is too bad....it is a laptop after all and laptops can get hot.

Wow! Nice long response! I too worked in a Chip Fab (in the early 80's...Western Digital before they made HD's) and yes I know all about electrons, EMI, etc. Your end paragraph put it all into perspective. One it's abouts yields (thats' why my $47 XP1700 CPU runs fine at XP3000+ because it is the same chip and I got into a good lot..I am not out of Eng tolerance with my 800+Mhz OC) and real world product logivity. I have no doubt my CPU will still work fine (and my GPU's) several years from now, but I wont be running them by then!