DDR Roadmap

Late last year, we figured that 2001 wasn’t going to be a lot of fun for prognosticators, and it’s turning out that way.

We are beginning to transition from one generation of computer to the next. The standards of the computer we’ll see in 2003 will bear little resemblance to that available today.

Transitions to new technologies are painful. Projected completion times slip, which creates bottlenecks for everything else. If you can’t upgrade often, and you have to make your purchases count and last, this can be torture.

DDR is one of those technologies. We have felt, and still do feel, that this is a technology worth waiting for in the long run. But people are getting frustrated by waits and uncertainties. So are we. If you feel like you’ve been hung up and left to dry, so do we.

At this point, though, we think we have enough information to draw some conclusions:

  1. We don’t think the Ali chipset is ready yet.

    We have one of these, and we don’t think it’s up to par yet.

  2. We’re not impressed enough by what we’ve read about the AMD 760 chipset to recommend buying it.

    The FIC AD11 seems to be the only AMD 76X board that’s even trying to be overclocker-friendly, and it doesn’t seem quite ready to do that yet. Asus has stated that it does not plan to support multiplier changing on its AMD board; the A7M266.
    We cannot recommend such a board for the long-term. Might not be too big a deal for some in the short-term; but given uncertainties about Palomino, it seems foolish to give up multiplier adjusting when there are other possible choices out there.
    The cost of the boards (about $200USD) is also a big deterrent.

    Maybe the biggest problem with the AMD boards is its position as a placeholder, something to buy until the real motherboards show up. The earlier AMD boards were like that, too.

    We’ve decided to pass on the AMD boards in favor of looking at more Via DDR boards a little later on. If the Vias turn out to be dogs, then we’ll look back to the AMDs, but we don’t think a lot of you will be buying these boards until DDR gets cheap and Via shows up, anyway. There’s much more interest in the 133As, so we’ll focus on those over the next month or so.

  3. We think you should wait for the Via DDR boards to show up.

    Buying DDR right now is like Prince Charming picking up one of Cinderella’s stepsisters because he couldn’t wait for Cinderella to show up.

    We don’t know if the Via boards will be Cinderella, but unlike Prince Charming, at least you know she’s coming. Might as well check them all out first.

    Initial indications are: a little better, a lot cheaper. Is that a winning combination? Don’t know yet, but since you need the Palomino to show up before you can ride home in style, why rush?

  4. We’re not all that impressed by most of the initial 133A boards, either.

    We’ve had three of these: the IWill KK266, the Abit KT7A and now the Asus A7V133. Based on the testing we’ve done so far, the IWill KK266 is clearly superior to the Abit KT7A or Asus A7V133. We’re going to beat them up in daily use pretty shortly before making a decided recommendation, but we’d be surprised if that changes our minds.

    There’s a pretty simple reason for this. The KT7A and A7V133 are basically warmed-over KT7s and A7Vs with a different Via chip tucked in. The KK266 is a new design.

    As Joe has been demonstrating, when you take a stick of RAM and have to play Dr. Frankenstein to get it to work at X speed in Big Name Board A, but don’t have to when you stick the same memory in Board B, this tells us Board B is put together a lot better. When you can’t reliably take Big Name Board past 150Mhz, but you can take the little guy more than a few notches above that, this tell us the same.

    This doesn’t mean that you, personally, will be able to run an IWill at 165Mhz, your other components will determine that. However, it’s a very strong indicator to us that the little guy is a better animal, no matter what the speed.

    We’re more than a bit concerned at least in the case of the Asus that it might not be quite up to Palominos running at 1.7Ghz+ in six months time.

    Not that either the Abit or Asus are bad; they’re not. But what do you want, the best board, or the best marketing?

    We shake our heads at some of the reviews we’ve seen, “It can’t do this, and it can’t do that, and it doesn’t get too far here, but this is the greatest thing on earth.” These are shepherds looking for sheep.

  5. DDR is rolling out much more slowly than promised, but it’s getting there

    This news report indicates where DDR is at now, and where it will be shortly.

    Basically, PC2100 modules are ready, and are awaiting certification before production cranks up. Several modules using Micron memory chips have just passed; the Micron modules themselves are going through testing now.

    Micron is approaching this one a little bit differently than the others. They’ve basically designed these chips to be smaller than the others, so they’re cheaper to make. It’s taking them a bit longer to get it right, but they feel the cost reductions are worth it.

    So will you when Micron starts making a ton of them in the second quarter. They’ve reiterated their pledge to keep their DDR pricing in line with PC133 pricing. Their approach is, “We can make them cheaper, so we can make money on this while others who didn’t think ahead are losing theirs.”

    If this sounds like an advertisement for Micron, when there’s only one guy out of a group who wants to cut the current price of something I want to buy in half, he’s always going to be my friend. 🙂

  6. The horse has to show up before the race begins

    This is a shy horse, the Palomino. Not too sure exactly when it’s going to show up, or in what colors. Right now, we’re guessing late April/early May.

    We’re pretty comfortable at this point saying that the Palominos should get introduced at a fairly reasonable price (around $275). The first price cut should bring the price down to $225; the second to about $190. After that, cuts will be in the $10-15 range.

    If you put a gun to our heads, we’d say that they’ll only come out in 133/266Mhz versions. We could be wrong on this, but it would be prudent to assume that and make sure the motherboard you buy can adjust the multiplier.

    At this point, we’re 80% sure Palomino/DDR combos will become a good buy and be the hot upgrade during the summer. That’s much later than we’d been led to believe. If you just can’t hold out that long, buy now, my son. 🙂

    Something funny is going on with the high-end TBirds being made the last few weeks. We don’t think they’re full-fledged Palominos in disguise, more like proto-Palominos with some of the manufacturing techniques being incorporated into the chips.

    We’ve said all along that DDR in and of itself is a dubious upgrade until matched with this CPU. Since DDR is and will remain in an unsettled state for a while, you might as well save your money now, and do a bigger upgrade in one shot that you might notice, than two smaller ones you might not.

    Some may say, “I can get 133A now, stick a Palomino in later, and live happily everafter.” If you don’t mind giving up about 12% improvement when all is said and done with DDR, fine. If buying 133A and SDRAM now, and replacing it with DDR six or nine months from seems good to you, fine by me.


  7. There is no one ideal approach for everyone.

    The problem I have writing these things is that people are quite different in their needs and priorities.

    I stress “wait, wait, wait” a lot for those who have to make their purchase count long-term. Those folks need some balance to the general atmosphere of “buy, buy, buy”.

    Not all of you are in that boat, but all of you certainly can benefit from knowing what’s coming down the road. Then you can make an informed decision that is best for you. The best decision for you may not be what I suggest.

Email Ed

Be the first to comment

Leave a Reply