What You Don’t Know Can Kill You
KT400 boards are coming out. You want to look and compare features, fine.
What you need to know if you plan on running at well over 200MHz (which you probably should be planning on doing with future 166MHz FSB TBred/Barton CPUs) is finding out whether or not the motherboard supports a /6 divisor.
Why is this important? It could well be the difference between your machine working and not working at your desired goal.
But My Motherboard Can Run Up To 250MHz!
Can I sell you a bridge?
No, you have a setting on your motherboard that says “250MHz.” Just because you have a setting doesn’t mean it’s going to work. In fact, it almost certainly won’t work.
Why? Read on.
In many ways, a functioning computer is a matter of multiplying and dividing.
Computers run on a system clock. This is generally known as the front side bus (or FSB for short).
Modern CPUs run at a multiplier of the system clock. That means its clock is set to run X times faster than the FSB.
If you have a 1.67Mhz Athlon 2000+, and you see a reference to 12.5 X 133, all that means is that the CPU clock will click 12.5 times
for every click of the FSB clock. It’s like seconds and minutes on a clock (in this case, a clock with ten seconds for every minute).
Most other devices attached to your motherboard run at a fraction of the FSB clock. The primary setting is that of the PCI clock (which governs your hard drives and devices you put in your PCI slots). In modern computers, the speed of that clock is usually one-third or one-fourth of the FSB speed.
If the FSB clock is like minutes on a clock, consider the PCI clock to be like hours on the clock.
Most devices that follow the PCI clock set 33.33Mhz as their standard speed. To derive the PCI clock speed, a motherboard uses a PCI divisor.
If the computer is running at an FSB of 133Mhz; the PCI divisor is normally /4. 133Mhz/4 = 33.33Mhz.
If the computer is running at an FSB of 166Mhz; the PCI divisor on the the newest motherboards should be /5. 166Mhz/5 = 33.33Mhz.
Why Should I Care?
If you overclock your computer by increasing the FSB, you are effectively overclocking the PCI clock, too.
If the computer is running at an FSB of 150Mhz; the PCI divisor stays at /4, and you get 150Mhz/4 = 37.5Mhz. You are running your hard drives and sound cards and the like 12% faster than they were designed to run.
So What? I Still Don’t Care.
You will care very much if your computer fails to work, or you lose everything on your hard drive because of this.
Devices like hard drives and PCI devices generally cannot be overclocked as much as CPUs. A few cannot be overclocked at all.
So I’ll Just Increase The PCI Divisor To A Bigger Number
You can’t. The divisors are built into the hardware. If the chipset doesn’t have a divisor built in, there’s nothing to change. No new BIOS can change that.
Even if the hardware supports a /6 divisor, the BIOS often will not. There’s been some software here and there that promises to be able to do that for you, but whether or not it actually works has been highly disputed.
Give Me An Example
Let’s say you want to run at 220MHz FSB. If you are using a PCI divisor of /6, that means your PCI devices/hard drive are running at 220/6 = 36.67 MHz. Unless you are pretty unlucky, that much of an overclock shouldn’t cause problems.
If you only have a divisor of /5, that means your PCI devices/hard drives are running at 220/5 = 44MHz. At this speed, it becomes fairly likely that at least one of your PCI devices won’t function. Maybe it just won’t work until you lower the speed. Maybe it won’t work until you figure out which device can’t handle it and remove it. Maybe it scrambles your hard drive in the process (it’s happened to me).
Nor is it something you can really avoid. You can buy two of the same thing; one will work, one won’t. You can even have something not work in System A and work in System B.
You could find yourself in the position (as do many of the extreme overclockers right now) of swapping components in and out trying to find a combo that works.
If you’re not in the position to do that, it would be a lot smarter to look upfront for a mobo that supports a /6 divisor to avoid all this.
I can’t give you a 100% answer. Just probabilities. A 10-12% overclock (up to about 37MHz) is usually no big deal, but once in a while it is. Go above 40MHz, and the odds you’ll have a problem increase expotentially the higher you go above that. At 45MHz or better, you’re moving from “probably having problems” towards “definitely having problems,” and if you’re not, that’s luck.
But Is It There?
Does the KT400 support a /6 divisor? I don’t know yet. I read all the KT400 reviews I could find, and I still don’t know. None of the reviewers addressed the issue. I guess they either don’t know about it, or don’t know it’s important.
Be smarter than that. Find out first before you buy, and if the reviewer didn’t mention it, ask him and don’t accept “I assume.” Nobody should assume when it’s somebody else’s money on the line.
If we find that the KT400 board doesn’t support a /6 divisor, then there’s no overclocking reason to upgrade from a KT333.
Asynchronous Is AGood With The Athlon
People run memory asynchronously on PIV platforms all the time, and it helps them. This is because the PIV bus contains so much bandwidth (3.2Gb/sec at 100MHz; 4.27Gb/sec at 133MHz; 5.33Gb/sec at 166MHz) that there’s plenty of room for ultrafast DDR (RDRAM is a different story) memory and everything else the computer does to all fit on the bus.
Running DDR synchronously on a PIV platform is like having an eight-passenger bus with one occupant, with four people getting on. No problem. Running asynchronously is like having five people get on instead. Still no problem, and there’s still room to spare.
This is not the case with the Athlon platform. It has a smaller bus. At 133Mhz, the FSB can only handle 2.1Gb/sec; at 166MHz, 2.7Gb/sec; at 200MHz, 3.2Gb/sec.
Those capacities are exactly the same as those of DDR memory, so what happens in the Athlon bus is something like three people trying to get into a minibus meant for three that already has a passenger. You can squeeze them all in, but it’s cramped.
For the Athlon, increasing the FSB is like increasing the size of the minibus. So now you have a four-passenger minibus, with four people trying to get in. Still crowded, but at least one more person goes along for the ride.
Running RAM asynchrously on the Athlon is like trying to get not four, but five people in that already occupied four-man minibus. You might be able to get him in by laying him across people’s laps or by having leaving a door open and telling him to hang on tight, but you can hardly say doing that improves the bus ride.
This has been clearly demonstrated in earlier asych attempts.
So it’s a lot more important to run a high FSB on the Athlon than with the PIV. The question is whether or not KT400 motherboards will let you do that any more effectively than KT333s.
If they don’t support a /6, then any 200MHz+ attempts will carry a significant risk of failure, and by 220MHz or more, a very sizable one.
Here’s the other obstacle to 200MHz+.
The KT400 doesn’t like the idea of multiple high-speed stick AT ALL. For instance, the manual for the Asus A7V8X indicates a maximum two sticks of PC2700 and only one stick of PC3200.
Running well over 200MHz can cause memory problems even with the best of sticks, but as we’ve pointed out, it’s not the only reason why a system won’t work at 2xxMHz.
Generally, if memory really can’t handle it, it will crap out almost immediately after boot. If the machine stops working later on in the boot process, it’s probably a PCI component or (more likely) the hard drive. If you slow down, and Windows actually has to make repairs, it probably is the hard drive.
If a machine freezes when you do something that calls upon one of your PCI devices, that’s the first suspect. Random freezes are more likely a memory (or CPU) problem.
What do you think this is, Disney?
Really, this belief is just so dumb when it comes to this. A positive attitude might be a good thing for you to have with other people, but you aren’t trying to run at 220MHz FSB, the computer is. Not only does it not have a positive attitude, it can’t. It can’t have any attitude at all.
The components in your computers are inanimate. They are not sentient. They are completely unaware of your feelings, and completely incapable of being infected by them. This is like expecting a screwdriver or rock to love you.
They will work or not based on physics, not feelings. Don’t take it personally.
Save the positive attitude for personal interrelationships.
Are there people for whom this has worked? Absolutely. Does that mean it has to work for you? Absolutely not, even if you get exactly the same equipment.
It’s like finding somebody who chose “Heads” in a coin flip and won, then concluding that you’ll win any coin flip if you choose “Heads” too.
Know What You’re Getting Into
If you’re looking for a sure thing, this isn’t. That doesn’t make it impossible, just not easy.
The reason why I wrote this the way I did is that I see what people actually do.
First, they spend a bunch of money with a minimum of thought. (I’ve had plenty of people spend a lot of money, THEN ask me about it. More than half the time, they’ve blundered at least fairly badly.)
Then it doesn’t work.
Then they spend more money, usually on the wrong item.
Then it doesn’t work again. At this point, either they spend money until through dumb luck or chance it does work, or they scale back and complain that they got ripped off and will never buy from those companies again.
Yeah, they got ripped off. They ripped themselves off by not thinking, not doing their homework, or being realistic.
Details count doing this.