Celeron 2: Early Thoughts

First, a URL: ZD Net.

Second, some comments on the article:

Intel is supposed to announce the product March 29. Unfortunately, Intel has also taught us that vaporware is now their way of life, so don’t expect to see it until you start seeing a bunch of reviews of it. When that happens, then you’ll know they’re on their way.

Don’t take the performance differences seriously. Intel is always playing games in its benchmarking to stretch the performance difference between Celerons and their “serious” chips. When the Mendocino Celerons first came out, Intel published benchmarks showing about a 10% difference between the performance of the Celeron vs. PII.

However, when you looked at the equipment the benchmarks were run on, the Celeron got a slow IDE drive while the PII got a high-speed SCSI Cheetah drive. The PII got a much better video card and CD-ROM.

I would lay money these “comparisons” are being done much the same way (matter of fact, the current comparisons are much the same). Also note the infamous phrase “on certain benchmarks” which means in English “we spent weeks looking for that one.”

An overclocked Celeron will be a bit slower than an overclocked PIII running at the same clock speed, due to the difference in the bus speed and the smaller cache on the Celeron, but the difference for most things will be about 7-10% (even less if you are using a VIA board for a Coppermine and a BX for the Celeron). However, there might be ways to compensate for this (see below).

Finally, what you should be looking for (and looking out for) in any Celeron reviews:

1) How fast can it run reliably? Celeron2 overclocking will be quite similar to Celeron1 overclocking, but overclockers have a couple extra weapons this time around.

Something we need to determine right away is whether or not some of the first Celerons are really Coppermine rejects. Some (pretty few) of the first Celerons were just that way, but then Intel was doing a really good job making processors back then, and we don’t know if Intel’s recent production problems have been due because they can’t make enough chips, or can’t make enough good chips. Unless Coppermine yields are truly horrendous, I don’t think this is a long-term problem. I doubt Intel can supply enough Celerons just out of the Coppermine reject bin. Just something to look out for.

Assuming that is not a major problem, the next item we need to look for is how reliably these chips work at 100 MHz. My best guess right now is that the Celerons will be made from the cB0 process, so I think the vast majority of 566s will hit 850, and most 600s will hit 900. I don’t think we can assume any better than that before actual testing.

When the initial Celerons came out, overclockers were often faced with “100 MHz or bust.” If the chip didn’t hit 100 MHz, the only fallback position was 83 MHz, with a PCI divisor of /2, which meant overclocking the PCI bus to 41.7 MHz, which a lot of PCI devices didn’t like.

Some more modern motherboards have settings like 90 or 95 with a PCI divisor of /3. This will be a very good thing to have if you buy a Celeron and it doesn’t quite make 100 MHz.

Looking at this more optimistically, maybe the Celerons will do better (and probably will down the road), with the 566 the most likely candidate to be able to go over 100 MHz right away. Having a number of speeds between the main ones of 100, 112, and 124 will let you get the most out of the overclock (though you don’t have to have a motherboard that moves up 1 MHz at a time). What is probably more important to have if the Celerons are really, really good is the ability to go to a PCI divisor of /4 starting at a speed of 115 or 117.

2) Does it fit my motherboard and/or slocket? The Celerons are supposed to be in FC-PGA format, so it would seem that it should run on any system that is compatible with that, either natively or through a slotket. Then again, we assumed a Coppermine should work on a socket 370-compatible mobo, too. Don’t assume. In any review, make sure that Intel didn’t switch a couple pins to make slotkets useless.

Those of you with socket 370 motherboards like the BM6 need to pay close attention to this. I’ve seen plenty of FC-PGA to slot 1 slotkets, but I haven’t seen any FC-PGA to socket 370 slotkets yet, which is what you are going to need (unless you want to do some soldering on your new CPU) to run this chip on that board if it is FC-PGA compatible.

3) Does my motherboard support this processor yet? Before you buy one of these processors, you should make sure you have a motherboard that supports it. Don’t assume it will. Especially don’t assume it if you have an older motherboard. If your motherboard will not support Coppermines, it will not support these Celerons, either.

A number of you have astutely pointed out to me that your BX boards only go up to a multiplier of 8X, so how can the motherboard support a 9X or better multiplier? That’s a good question. Up to now, I have guessed that the range of multipliers is a BIOS and not a hardware limitation of the BX like the AGP 2/3 divisor. I could be very wrong on this. Maybe this is how Intel will get you to buy a new motherboard, we’ll just have to see.

Assuming I’m right, though, your mobo manufacturer will still have to update the BIOS to allow for these higher multipliers. You should check towards the end of March, early April, to see if they have done this. You could also contact your mobo company if they sponsor a newsgroup and ask them about it. If you already have, or read a response to the question, send it to me.

Occasionally, a motherboard will run a processor even if it doesn’t officially support it. The reason why it works is that the microcode for that family of processors is already in the BIOS. If the microcode to support that family of processors isn’t there, it won’t run. When the Celerons first came out, if your motherboard’s BIOS version didn’t support Celerons, it just sat there. You had to get a PII, start the machine, flash the BIOS, take the PII out, then install the Celeron. The same thing is very likely to happen if you try to install one of these Celerons in a motherboard that doesn’t support it. I can’t say that for certain, but I wouldn’t bet against it.

Should I buy a Celeron?

If you are looking to upgrade your current BX-based setup, you should definitely wait for a Celeron if you don’t want to spend a lot on an upgrade and want to keep using your current equipment. If you buy a Coppermine, you may well find out you need to upgrade your RAM. After you upgrade your RAM, you may find out that your BX board and/or video card doesn’t like high speed, either, so you’ll end up replacing the mobo, too. If you don’t want those kinds of “surprises,” you should wait.

The only real obstacle I see to this is the “Coppermine rejects sold as Celerons” bogeyman. I bring it up because a couple other people have mentioned the possibility, and while I don’t think it will happen, I can’t preclude it quite yet. Let’s put it this way; if the Celeron2s have the same CPUID and microcode as the current Coppermines, then this is something to really start worrying about. 🙂

Email Ed


Be the first to comment

Leave a Reply