Disclaimer: This is only reasoned speculation. I have no inside information of the doings of any of the companies mentioned and am just making some semi-educated guesses as to what may happen in the future. No guarantees, so if Intel does something different, get mad at Intel, not me.
Probably by late this summer, Intel will come up with the first Celeron processors that “officially” will operate at 100Mhz FSB. Is this just a ho-hum, where have you been, product announcement? There is reason to think not.
Intel is now committed to a two-tier processor line, the PIIIs for the well-heeled user, and the Celerons for the rest of us. What will the low end of the market look like six months to a year from now?
AMD will eventually get the K6-3 out. The K6-3 is destined/doomed to
become the low-end successor to the K6-2. Once the K7 is out, the K6-3 will drop in price and become the darling of last-gasp socket 7 users and those interested in quick, cheap machines. The 500Mhz K6-3 will be a little much for the 66Mhz Celerons (can’t increase those multipliers forever). Looking a bit further down the road, the AMD low-end will include .18 micron K6IIIs, so that’s 600-700Mhz competition. I don’t think our beloved 300A looks so hot in that company.
So what’s a dominant CPU manufacturer to do? Well, the company may be coy as to when Celeron goes to 100Mhz, but a chipset is a chipset, and they are supposed to have some integrated chipsets coming out, with a definite 100Mhz one coming out in September. I doubt it’s meant for cheap PII diehards.
Second, Intel looks like it’s going to milk 66Mhz for all it’s worth with the current Celerons, going up to a 7X and maybe a 7.5X multiplier. Not unreasonable to presume they might want to do the same with Celery II. They certainly are not going to rush to 133Mhz where it would compete directly with Coppermines.
Third, the maximum cache speed of the current Celerons is about 500Mhz. Until a couple days ago, that was also considered the max for Deschutes chips, though we now know a 550Mhz Katmai is coming out and Intel’s gotten as high as 650Mhz with a .25 micron design. Here, though, we’re talking about 700-800 Mhz. Looks like two things are going to have to happen: 1) A redesign of cache memory on the Celeron and 2) A little shrinkage in design.
If the Coppermines follow Intel history, they should have on-die cache running as fast as the last member of that family. Now that Intel executives are talking about 800Mhz by the end of 1999, it is probably very safe to say any Coppermine will have cache fast enough to handle at least that. I doubt Intel would reinvent the wheel for a new Celeron cache, so it probably can do the same. I still think Intel is going to plop in a bus lock, or maybe something sneakier.** If they don’t, well, you know what to do, though you’ll probably need exotic RAM and be able to run at least a 180Mhz FSB to do it and none of it will come cheap anytime soon. It doesn’t seem to make a whole lot of sense to buslock Coppermines and not Celerons, but then Intel has a history of doing odd things with this chip.
Insofar as shrinkage, Intel will have five fab plants on line making .18 micron designs by fairly early 2000 and no doubt will phase out .25 micron production elsewhere as soon as feasibly possible afterwards. So after seeing one or two .25 100Mhz FSB Celerons first, we could see .18 Celerons by late 1999 or early 2000, probably depending on what AMD does. If the early K7s do head towards the low end, sooner; If not, later. If the K7s do well in FPU, while they’re redesigning cache and shrinking design, they might as well add in KNI/SIMD/SSE or whatever they’ll call it then.
So we are likely to have a .18 micron chip designed for a 100Mhz bus, likely to have KNI, sometime in the next twelve months that should be able to run at 750 or 800Mhz when deemed needed. Of course, what Intel wants and what we want are two different things.
What does the average Intel overclocker have? He is likely to have an Abit BH6, or Asus P2B, or perhaps an AOpen AX6BC or BXR2. All of these handle a 133Mhz FSB with an PCI/4 divider. How handy. What kind of memory is the overclocker likely to have? Many will have Samsung GH or other memory that can be pushed to 133Mhz, and if not, PC133 RAM is right down the pike. Not likely Intel is going to demand any fancy RAMBUS from Celeron users this go-round.
So if Intel comes out with a 550 or 600Mhz Celeron2; I think a lot of people are going to get to 800Mhz sooner and cheaper than they think with more or less their current equipment. Maybe not quite as good with a Coppermine and Camino chipset and some super memory and AGP 4X and maybe PCI66, but not bad at all.
A few things that could stop this – I’ve already mentioned the possibility of a bus lock – there are two other possibilities:
1) These chips will almost certainly not be available in Slot 1 but PPGA. There is the possibility that a Slotket or equivalent may not work at the increased speed. I suspect it will, but it’s a possibility.
2) What may be a greater concern is BIOS support. Intel may not look too kindly on motherboard manufacturers supporting chips on the wrong type board and might do something to mess up that party. The motherboard manufacturers might not be too eager to do that either; after all, they make more money selling new motherboards, especially PPGA boards with 133Mhz FSB. So one may have to buy a new motherboard after all, but that shouldn’t be too bad.
Something to think about for next fall/winter.
**Even motherboard manufacturers inclined to thumb their nose at Intel seem to have a problem with PCI/AGP dividers. If Intel could effectively curtail those dividers in the Camino chipset, they wouldn’t have to bother locking the chip to thwart extreme overclocking. Just a thought.