The Lifetime of A Platform

Add Your Comments

The Intel Canterwood/Springdale motherboards are out. People are starting to buy them.

For those of you wondering whether or not to get on the bandwagon, it might be enlightening to track out the likely lifespan of this platform.

Now

The Canterwood/Springdales are transitional boards, part of the migration from the old PII/Athlon mobo standard of AGP and PCI/parallel ATA/SDRAM to the future PCI-something (Express or 64)/serial ATA/DDR-II.

Fully “transitioned” boards won’t be available until middle-to-late 2004, so if you buy/build a computer (or more likely, buy/build for someone else) once an eon, and that computer will be around for an eon, that’s something to consider.

In the meantime, it’s pretty safe to assume a time before obsolescence of 12-18 months.

The latest processors for this platform are the Intel D-1 stepping processors. Right now, they are only available in the 200MHz FSB versions, but 133MHz versions should officially begin to become available in a few weeks.

However, being officially available is by no means the same as you being able to find one. Those of you who looked high and low for C-1 stepping PIVs at lower people will remember that they didn’t start showing up regularly until about five months after they were “officially” available. That was much longer than the usual 2-3 months, and given the continued slovenly state of PC sales, such a long wait could repeat itself.

Early reports on the D-1s seem to indicate that around 3.3GHz is a reasonable expectation with decent air-cooling, with maybe a couple hundred more MHz a reasonable possibility.

It’s not going to get much better than this at 130nm. Intel might come out with an E-stepping, but it’s unlikely there will be any particular significant improvement in GHz until we see 90nm Prescotts.

Unfortunately, memory doesn’t quite jive with the ideal overclocking ratios for low-end 200MHz CPUs. There are rumors floating around about true 4ns memory chips in testing, but less scrupulous memory sellers have been playing such games with claims that the proper attitude of the buyer ought to be “Guilty until proven innocent.”

If your goal is get an easy, hassle-free 3GHz or 3.2GHz machine, that shouldn’t be a problem. Aim for 3.6GHz, though, and you have both CPU and memory at the tail end of the technology, plus a heavily overclocked chipset. When you have that many factors going against you, if you want easy, that’s not what you’re going to get.

Later

One good thing about the Canterwood/Springdales is that they’re supposed to support Prescotts. Well, at least some Prescotts.

Unless there’s a dramatic change, the initial Prescotts will begin as socket 478 CPUs. Some time in spring 2004, socket 775 CPUs will begin to come out, along with motherboards for them.

These motherboards will be equipped to handle either DDR or DDR=II, probably SATA-II, and will have a PCI-Express inteface for video.

The introduction of SATA-II will probably represent the crossover point for the mainstream user, too.

Since these mobos will incorporate two big changes, adaptation is likely to be relatively slow, and delays due to DDR-II or PCI Express video cards being unavailable/not commonly available/too expensive are quite possible.

Prescotts will be made using a 90nm process. This has always allowed for a significant boost in processor speed in the past, and there’s no reason to think this won’t happen again. Based on prior history, a guesstimate on the overclockability of the initial Prescotts will be in low 4GHz+ range, say 4.3GHz.

In the past, Intel has continued to improve CPUs even after they’ve been supplanted by newer technology, so it’s probable that some time in 2004, you’ll see a socket 478 Prescott capable of 4.5-5GHz. Intel will probably continue to make socket 478 chips throughout 2004.

The initial generation of socket 775 boards will probably not provide any compelling reason to upgrade from a Canterwood/Springdale platform, mostly because there will be little improvement initially from DDR-II and PCI-Express video cards. It is also unclear at this moment whether the 2005(?) Nehamem processor (call it a Pentium V; it’s meant to be a redesign) would need a new platform. Probably will, but you don’t need an answer to that even for planning purposes for at least another year, and by then, the answer should be available.

Cost will be the biggest initial problem with Prescotts. Intel will start off with a 3.4GHz CPU carrying a $600 price tag. Such a processor is unlikely to become somewhat affordable (i.e. less than $300) until around April 2004, or reasonably affordable (a little less than $200) until June.

If Athlon64s provide stiff competition, these dates might be pushed up a couple months.

However, this timeline assumes that the 3.4GHz will be the lowest-rated Prescott processor, and this may well not be the case. If Northwood history repeats itself, we could see lower-rated (and priced) processors meant for OEMs show up a few months after Prescott introduction, say January and February, perhaps as low as a 2.6GHz.

Should that happen, a 2.6 or 2.8 Prescott will end up being the overclocker’s choice.

Socket 775 Prescotts will be just an introduction to the Tejas generation of processors, still 90nm, due out in late 2004.

Buy Early, Stay Late

After a few more weeks, there’s not going to be any big price drops in this kind of equipment that would justify waiting months and months for it. For this platform, it’s either worth it to you now or it’s never going to be worth it to you. If it’s the latter (which is probably the case most of the time), wait a year.

If you do buy, unless you’re extremely picky, you’re probably good for at least a year and a half.

Ed

Leave a Reply

Your email address will not be published. Required fields are marked *