What ATI Must Do

Add Your Comments

Summary: ATI will not be able to seriously challenge nVidia for the video card market until they essentially adapt nVidia’s
approach of using the latest technology across a range of prices. Even if you assume nVidia is inherently better technologically; ATI would still
be much better off if they adapted the nVidia approach.

Before the next ATI barrage, a few quiet observations.

nVidia rules the video card roost, ATI doesn’t. If ATI wants to rule the roost, it must do two things:

  • It has to put out a video card nVidia can’t top in the same generation.
  • It has to have a range of products using the same high-end technology the high-end card uses, and not much less capable. Only the bargain basement stuff can be significantly crippled.

    Remember the Radeon 8500. It, too, came out before nVidia’s new entry. It, too, looked better than the GF3s. It didn’t look better than the GF4 Tis. Not just the top one, all of them.

    In the end, it hardly threatened nVidia’s dominant position.

    The Radeon 9700 will have the same fate unless NV30 comes out and the 9700 does not get flattened by NV30’s product range.

    ATI said the 9700 will not come out at .13 micron, that won’t come until the next generation. NV30 will, and that will almost certainly ensure it will do better.

    (What I would watch for in any previewing is the frequency of the GPU. Provided memory bandwidth is at least equivalent, I suspect that the difference between the NV30 frequency and the R9700 frequency will give you a very rough idea as to where the two stand (really, whether it’s an decent improvement (10-20%) or a blowout (25%-30% or better); I would bet on the first right now).

    Where ATI is really dropping the ball, though, is in the follow-up. Yes, there will be a Radeon 9500. However, the description of Radeon 9700 Lite (which may be the 9500) looks more like 9700 MX than the ATI equivalent of the Ti4400.

    If the Radeon 9500 has only half the rendering pipelines and half the memory channels, this is going to be one pathetic video card compared to its older brother. Maybe you do this to your $99 card, not your $200-250 card.

    (I must admit, this seems so stupid that I have grave doubts this is really the 9500. It does sound more like a card to compete against the MXs. But if this is the 9500, then ATI is clueless about competing with nVidia.)

    Believe me, nVidia will not make that mistake. Their lesser sibling will be about 10-15% less potent than big brother, and youngest brother maybe another 10% less than that.

    This is a winning formula, and it’s crazy for ATI not to copy it. Until ATI copies it, they will not be able to threaten nVidia’s hold on the market.

    That’s because show horses don’t win this kind of competition, work horses win it. nVidia owns the video card market because of its MX and Ti4200 sales, not because of the Ti4600.

    You want to say the Ti4600 makes the success of the MX and Ti4200 possible, yes, that has something to do with it. But if you don’t have the workhorses in place, you don’t get the benefits from having the best show horse.

    If ATI seriously wanted to go toe-to-toe with nVidia, they would have the Radeon 9700. They’d have a card going for $200-250 that was 15% slower. They’d have a card going for $150-200 that was 25% slower. Radeon 9000 Pro would be aimed at the high-end MX crowd, and the 9000 for the low-end crowd.

    There is nothing technically difficult about doing this. All you do is put in slower components in the Radeon 9700 II and III. ATI is certainly doing that now with the Radeon 8500 design in the 9000 series.

    Get While the Getting Is Good

    I don’t think ATI is serious about competing with nVidia head-on-head, every month of the year. Rather, they’re trying to pick their spots and maximize their profits while doing so.

    The Radeon 9700 is basically a rushed dual-memory channel video card. It’s rushed because ATI couldn’t wait for .13 micron and also beat nVidia to the draw. So they went and pushed .15 micron as hard as they could.

    Now dual-memory channel is good for at least some things, no doubt about that. But .13 micron and dual-memory channel (or equivalent) is better, especially when there are signs the video bottleneck moves from memory to the GPU in such a setup.

    So you get a card that’s better for a couple months, then gets thrashed by the more balanced nVidia approach.

    The thrashing that matters (the financial one) doesn’t occur because nVidia’s best being better than ATI’s. The thrashing occurs when nVidia finally shows up and not only its first, but its far cheaper second or even third card does about the same or better, too. Then it’s crash and burn time for the prices on the ATI card.

    Even with this strategy, though, ATI could still give nVidia a lot more fits if they had three Radeon 9700-derived cards that respectively could beat the Ti4600, 4400 and 4200. Now. They don’t. Why not?

    95% of you wouldn’t pay $400 for a card for anything short of your life depending on it. So how much good is any superiority when you don’t have a product that 95% will buy?

    If ATI is at war with nVidia, they seem to be taking the al-Queda approach: one big dramatic attack, and no follow-up. This makes headlines; it doesn’t win wars.

    Let’s presume I’m Joe Sixpack and I see wonderful descriptions of this R9700. I’m not going to lay out $400, but I will lay out $150. Does ATI have something for me? No, but nVidia does.

    Let’s presume I’ll lay out $200-250. Does ATI have something for me? Well, they might, someday, don’t know exactly how much worse than the $400 card it will be. Joe wants it now? Sorry, but nVidia does.

    These price ranges represent far, far more sales than those at the $400 range, and ATI’s current approach essentially forfeits those sales to nVidia for no apparently good (outside of perhaps capacity) reasons (and if it is capacity, then ATI can’t challenge anyway).

    Ruling Vs. Squatting

    You will hear much about ATI ruling or nVidia suddenly becoming the underdog, and it will all be a bunch of nonsense.

    This is not ruling. This is more like plopping your butt on the throne while the king has gone to the bathroom. When the king comes back and you can keep his butt out and yours in the seat, now we’re talking.

    You want to be a fanboy and treat this like a game, fine. The game we’re playing now is called Christmas Sales. Like basketball, it’s the last quarter (or month) that counts, not the first quarter. With the Radeon 9700, ATI will lead in the first quarter against nVidia, by a sizable but not an unsurmountable lead.

    But will you be cheering Christmas Eve?

    All That Matter Is Now

    If you say, “that doesn’t matter,” let me give you a piece of history to ponder.

    Back in 1389, the Serbs faced the Ottoman Turks in a battle. They did well to start with, but eventually the Ottomans kicked their butts, and that began over four hundred years of Turkish rule.

    Sounds like something you’d want to forget, doesn’t it?

    No, Serbs celebrate the Battle of Kosovo. If you ask them why, they’ll essentially tell you, “We were ahead at halftime.”

    This sounds about as insane as Americans celebrating Vietnam or the French Waterloo. Well, it does until you look at their history and find that this was just about the best they’ve ever done against big league competition. (and even that record doesn’t look too shabby compared to a few other European nationalities).

    You do it when you have nothing better to celebrate, and only when you have nothing better to celebrate.

    ATI will have arrived when they end up winning, across the board. Not before.

    Until then, all you’re celebrating is being ahead at halftime (and even if that’s the best you can do, as I’ve pointed out, there’s ways to at least increase the halftime lead).

    You’re Against ATI!

    I’m saying how ATI can do better and increase its marketshare against nVidia, even if it can’t beat them. That’s a strange way to be against them.

    Ed

  • Leave a Reply

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