Socket T Cometh, But . . .

Looks Like April, Too

XBitLabs translates an x86-secret story which says that we’ll see socket T right around April Fools’ Day, too.

This is going to be a tough call for those who will only accept an Intel system.

Normally, one would be inclined to tell those who are hurting and upgrade infrequently to wait for the next generation of motherboards.

The problem this time around, though, is that this generation gap is so big and incorporates so many brand-new players that one is reluctant to do so.

First, you have to be a little worried the stuff will work, at least initially.

Second, you have to more than a little worried about what the bill for PCIExpress video cards and especially for DDR-II RAM is going to be, at least initially.

Third, none of these technologies are going to be a whole lot better when they get introduced than what we have now. In 2005, yes. Not early to mid 2004.

Fourth, the lifespan of this platform looks to be fairly short. Tejas, which ought to come at the end of the year, is supposed to use a BTX form factor mobo will require different power connectors, and may have other unknown hardware differences put in by Intel to make its x86-64 somehow different from AMD64.

In short, I’m really leery about how happy somebody who buys a socket T mobo next April or so is going to be with his or her upgrade options a year later. If socket T offered a huge perfomance boost, that would be one thing, but I just don’t see that coming from the Prescott generation.

Take all of these into consideration, and it’s hard to tell those who really need an upgrade soon to keep waiting. Moving up socket T is really Intel’s way of saying that it can’t handle high-speed Prescotts with socket 478. This is hardly a good sign for overclockers.

For those with a Springdale/Canterwood, you can always top up a little with a relatively cheap socket 478 Prescott, but don’t expect much. You might want to let somebody else try it out first.

That also applies to prospective socket T buyers. Overclocking prospects will probably be better than with socket 478, but again, let somebody else be the pioneer first.

Maybe the best way to express all this is that socket T doesn’t pass the crucial “buy” test. What’s that? I look at all the facts and prospects and possibilities and prices for the equipment and I say, “Do I want to buy it?”

The answer is “No” because I don’t see enough of an improvement over current systems for the money, and the platform has a limited lifespan.

If I don’t want to buy it, how can I recommend that you buy it?

At this point in time, a Northwood system looks to be a good buy. The next good buy from Intel will probably come from the Tejas generation. Inbetween the two, a socket 939 system ought to be a (fairly) good buy from around the middle of the year on.

Ed