Coppermine Overclocking Issues

What I get asked

I get quite a few questions from readers about Coppermine overclocking,. They pretty much boil down to the same question:

I want to overclock a Coppermine now, but want to do it safely. How do I do it?

What I really want to ask is “How lucky do you feel?” How much risk are you willing to take? Is your attitude “What the hell, I’m going to give it a shot, and if things fail, well, those are the breaks and I’ll just have to buy new stuff?” If it isn’t, then you shouldn’t be overclocking Coppermines right now, because you are taking some uncertain risks that can’t be avoided by flipping the right switch or by some magic setting.

The issue should not be “how should I do it?” The issue should be “should I do it now?” And my answer to that question is “No.” Why?

Every motherboard out there outside of the BX board right now stinks, and the BX board is not meant to run at these speeds. We have no idea if or how much damage is being caused by running at these speeds, especially if you are doing heavy gaming.

CPUs are not the same as video cards. When Intel or AMD make CPUs, they don’t make 500 and 600 and 700MHz CPUs, they make CPUs. Ideally, every CPU they make should run as fast as the fastest speed it is designed for. So if you buy a CPU on the low end of the totem pole, you are buying a CPU which was initially meant to go much faster.

It makes economic sense for Intel or AMD to build CPUs that way; the cost is in the design of the chip and the machinery made to build the chips. Given how the nominal cost of the silicon, how the process yields results and the wide range of speeds manufacturers sell, it saves money to have just one process rather than several. You can still sell most of the chips that are not quite good enough to run at the fastest speeds, and if it turns out that just about all of them meet the grade, it didn’t cost you any more than if you ended up with not-quite-as-good a batch.

It’s not quite the same story with other components. For one thing, hard drives and video cards don’t come out of the oven done like CPUs. They require a lot of separate components put together. For another, they usually aren’t sold in a wide range of speeds, either. The hard drive company that makes 7200rpm hard drives isn’t making 11,000rpm hard drives from the same production line.

Video cards are a little more like CPUs, but video card makers are usually pushing the limit, and usually don’t get the vast majority of processors that can run far faster than the rated speed. Besides, there are things like video memory chips that limit speed. Not entirely, there is some tolerance, otherwise you couldn’t overclock the video card. But there is nowhere near as much as you often can with a CPU. How many of you have ever run your hard drive at 50MHz, which would be the same as running a Coppermine 500 at 750?

When you run your FSB at 133MHz, the video card is transferring information at 33% faster than it was designed for. At 150MHz, it is 50% faster. Is that safe? What do you mean by safe? Can some video cards do it for a while? Certainly, but some can’t. Can all the video cards that can do it now keep doing it for six months or a year? We don’t know . This is not just a matter of “will it work or not now” but rather “will it keep working?” I can put a hard drive in a case at 150 degrees, and it will work fine for a while, but I wouldn’t bet on it working for three or five years under those conditions.

There is no doubt in my mind that you increase the risk of failure by doing this. The question is: by how much? If I could offer you a billion dollars with a one percent chance of you dying within a year, I suspect most of you would take it. If the offer were the same, except that the odds of dying were fifty percent, most of you wouldn’t. Well, in this case, we don’t know what the odds of survival are, because we don’t know what the tolerances are like we do for CPUs.

For instance, I’m using a PII-333. It’s been running at 400MHz for twenty months. Default voltage within the design parameters. People have been overclocking processors since the 8088, and there’s been a very low failure rate (outside of acts of insanity). So I feel pretty safe saying that with a few precautions, overclocking processors is pretty safe.

On the other hand, we have little experience running AGP video cards at far over specifications. We know PCI devices don’t necessarily take too kindly to overclocking, but this is a different animal that isn’t an absolute slave to the FSB; it has its own clock, for starters, and sends all sorts of different signals to the rest of the world. We know we’re overclocking something when we do it, but exactly what? It sure doesn’t show up in benchmarks. .

Now the limited experience we have indicates that this isn’t as risky as sticking your head inside a unknown lion’s mouth, but how risky? I don’t know. Most people try to figure out the rewards that come with the risks, which brings us to the next point:

What do I get out of this?

Most of you are gamers. You are overclocking in the hopes of increasing your fps to give you an edge in your firefights. But does doing this really help? The reality is current video cards, even the latest, maximize output at a certain point. It doesn’t matter if you are running a zillion MHz processor, the card is only going to put out so much, and that is that.

Have you noticed that benchmarks of video cards lately are usually run at low resolutions like 800X600 or even 640X480? Do you know why? It’s because the higher the resolution, the more work the video card has to do just to fill the screen, and at resolutions like 1024X768, the video card does as much as it can, and it’s not going to do anymore, even if you hooked it into Data’s brain.

At 1024X768, there is not a single video card out there that will show any significant improvement with an 800MHz processor. The video card becomes the bottleneck. Video card performance seems to top out at 650MHz at most; anything faster than that is “wasted.” So sure, if you take a 500MHz Coppermine and get it to 750MHz, you’ll get some improvement, but not as much as you’d probably hope for. If you are already running at 550MHz or 600MHz, you may not get any improvement.

To make matters even worse, many of you have reported having to make adjustments to get the card to work at all, like reducing the AGP speed to 1X, or reducing aperture size, or disabling sidebanding. All of these are more likely than not to REDUCE video performance, not increase it. Just a CPU improvement won’t do it, you need to bolster the video card, too, and the next generation of cards isn’t here yet.

Why can’t you wait?

It’s a little frustrating writing articles suggesting that people wait a couple months for new products, then get letters that essentially ignore everything you’ve said and say, “I want to do this now.” There are good times, and not-so-good-times to buy equipment. We are approaching the end of a not-so-good time. The next sixty to ninety days will begin what I see as a good time to buy a new computer or upgrade.

Right now, Coppermines are scarce, and going for premium prices. In about sixty days, you’ll be able to buy a significantly faster and/or cheaper processor than you can now and not be overcharged. You can either buy a higher-speed, improved step Coppermine that is likely to get you 900MHz or better for about what you would pay now, or there’s a good chance you can buy a Celeron-2 and get about the same performance you’d get with current Coppermine without having to possibly upgrade other parts of your system. We should at least begin to see motherboards that are meant to run Coppermines properly. By then, there should be video cards that will actually take advantage of the speed boost.

Nor are the benefits solely on the Intel side. In sixty days, .18 micron Athlons should become affordable, and you can take a look at those, too. The Spitfires should be out, and I’m sure they’ll be a very competitive chip for those on a budget, too. If you can hold off until the second half of the year, I think the Athlons with on-die cache and DDR will be the powerhouses for quite a while. After that, we’ll probably have to wait about a year until the following Willamette/Sledgehammer choices become affordable, so from about April-May on will be a good time to buy.

You’re going to have a ton of choices during the spring. So why buy now when you’ll pay more than you should in the midst of shortages for little benefit and some risks? If any of you think I’m missing something, and there really is a good reason to buy now rather than wait, I’d like you to write me. I know people buying now are going to regret it in a couple months.

Email Ed

Be the first to comment

Leave a Reply