AMD and FSB

Going through the survey and a couple forums, I found that many people don’t know the real story about AMD and FSB.

Many people think AMD locking the multiplier is no different than Intel locking it two years ago.

For them, let me tell you that you can’t increase FSB on Athlon boards the way you can with PIII boards. Not even close.

But it seems that most of those who do know that, even people running AMD-related sites, don’t seem to understand why that is. They say it’s because of
the EV-6 bus or come up with some other reason.

It is none of these.

These FSB restrictions were done deliberately.

I first mentioned this about two months ago in No Better Than Intel, but I guess
the story bears repeating.

First, here’s what Anandtech found out from Via about why Athlon boards don’t O/C FSB too much.

“VIA seemed to indicate that the reason current KX133/KT133 motherboards can’t reach the 133MHz FSB is because they physically made it impossible to reach that FSB setting on current designs. According to the representatives we talked to they claim that they “didn’t want people messing around with [the setting].”

As we’ll see later, they don’t want people “messing around with the setting” but not just for the reason you might think.

If that isn’t sufficient proof, all you have to do to make these FSB problems go away is replace a PLL on an Athlon motherboard. Some Japanese overclockers did just that with a clock generator called Turbo-PLL. Want to see a KX with an FSB of 155Mhz (310Mhz in Athlon DDR doubletalk)? Go here (it’s a big page, and it’s towards the end of the page). Obviously no problem with the EV-6 bus there.

(BTW, we’ve looked into doing this, but the Turbo-PLL chip costs more than $100, and requires rather precise soldering and unsoldering).

Why?

You have two possibilities, could be either, probably both.

AMD

You can remark one of two ways; you can either increase the multiplier, or you can increase the FSB. If you are really out to stop it, you have to attack it both ways, and that’s what AMD appears to have
done by pressuring Via to do what they said they did.

Via

Via obviously has no problem making chipsets that run at 133Mhz or better, they make them for the PIII with many of the same chips used in the Athlon boards. However, Via has to reach 133Mhz on PIII boards to accommodate EB chips. There are no EB chips in the Athlon arena, yet. There will be when DDR boards come along. However, once there are, if Via allowed
a 133Mhz bus now, people might want to stick with the old motherboard rather than buy a DDR motherboard, which after all is just a 133Mhz board times two.

What’s important to understand is that the who or the why may be in question, but the what is not. FSB is being restricted now, and that will severely restrict overclocking once you can’t change the multiplier.

What Happens With DDR?

Current AMD chips are rated at 200Mhz (100Mhz X 2). When AMD does its next revision, they are coming out with chips rated at 266Mhz (133Mhz X 2).

When AMD comes out with these new chips, which are supposed to hit rates as high as 1.5Ghz, will they come out with two varieties of chips just like Intel does with the E and EB chips?

If they do, then you have a 33% overclocking potential by taking a 200Mhz chip and running it at 266Mhz.

However, it doesn’t seem too likely that AMD has gone through all this fuss and bother to stop overclocking both on the FSB and multiplier level just to hand you a 33% overclock when DDR comes along.

What seems a lot more likely is that AMD’s new core revision will only come out in 266Mhz versions, or if they do have some 200Mhz versions come out, they’ll somehow block such chips from running at those speeds.

Looking a couple months down the road, it seems likely that AMD won’t be making anything less than 900Mhz-1Ghz chips, and the current version of the chips can handle that without being able to be overclocked 33%. The aluminum chips certainly won’t, and it will be tough getting even a current blue meanie to run at 1.2Ghz.

So they’ll make those chips for a little while, then shift completely over to 266Mhz for TBirds, which can be FSB overclocking limited just as it is now.

Buy Intel? What Intel?

You can say, “Forget it, I’ll buy Intel.” Fine, but buy Intel what?

You’ll have Willamette shortly, but it will be expensive, and Intel isn’t going to make very many of them until they can redo them in .13 micron. They won’t be cheap (can you say dual-channel RDRAM?), and it probably won’t perform as well as the T-Bird at the same speed.

Forget any major improvement from the PIII until the .13 micron shrink, about the middle of next year. Even with the cC0 stepping, you’re only looking at a bit over 1Ghz as the likely result, and it won’t get better for another nine months.

For all intents and purposes, Intel is effectively leaving the high-end (outside of a relatively few Willamettes) to AMD for the next nine months. You want significantly more than a Ghz before next summer, you have one source, and they’re going to make you pay full price for it.

Meet the new boss, worse than the old boss.

Email Ed