What I’m Waiting For
I believe in eating my own dog food. If I tell people to plan their future computer moves, then I need to do the same. If I’m going to recommend buying something, then I have to be willing to lay out money to buy it (or would be under the proper circumstances).
So, on this day in mid-April, what’s going on in my head?
To Hell With Intel
Sorry, but Prescott is technologically broken, and all the Intel “There’s nothing wrong with it but we’re fixing it” comments in the world do not change that. I don’t expect Intel to really get the heat issue under control (as opposed to patching it up a bit) until Tejas a year from now. So they’re off the list, probably for a long time.
Granted, Intel’s priorities are a bit different than mine. They may well be able to fix Prescott enough to sell them to the OEMs in the world, and if that isn’t good enough for me to overclock the daylights out of them, well, I doubt that’s going to lose Intel execs any sleep. Prescott was always going to be a transitional chip, anyway, so they really don’t need the chip to be capable of hitting 5GHz to make 1% of its buyers happy.
They will need to do that with Tejas, so expect fundamental changes like different electrochemistries then, but not until then (if even then).
That leaves AMD.
Not What, But When
Unless Intel pulls off a miracle, the next system I’ll put together will be a socket 939 system.
Some of you may say, “Well, if you’re going to do that, why do you scream so much more about AMD than Intel?
The answer’s quite simple. Intel can’t deliver what this audience wants, while AMD won’t. Prescott has a fundamental technological problem Intel probably can’t solve anytime soon. With AMD, on the other hand, the technology is fine, it’s the positioning and pricing that’s the problem.
It’s like being the coach of a sports team. One player just doesn’t have the ability to do what you want. Another player does, but he does whatever he feels like, and screws himself up. The first player tells you “I’m doing the best I can”, and tries to improve, but it’s just not working. The second player says, “To hell with you, don’t tell me what to do.”
Whom do you get madder at?
You get madder at the person who won’t do it and is messing up things rather than the one who can’t. You may well toss the one who can’t off the team, but you don’t get mad at him.
Mind you, this isn’t personal. AMD probably doesn’t know I exist, so it’s not a matter of how dare AMD not listen to the great Ed Stroligo. Rather, it’s a matter of following a short-sighted policy which doesn’t seem to be working, and alienating potential allies by telling them, “It’s none of your business what we’re doing.”
Intel is hardly going to win any honesty awards about Prescott, but indirectly, they’ve indicated they know they have a problem. AMD’s attitude, on the other hand, is “Our business is none of your business,” which is a bad thing to say to people like stock analysts whose business is their business.
The only certainty at this point is that I don’t want an AGP only motherboard. The next platform will have to have native PCI Express video, simply because the Radeon 9700 Pro is going to look a bit long in the tooth with these next generation video cards and I see no point in laying out a lot of money for a new AGP card I won’t be able to use for a while. This is strictly a compatibility, not a relative performance issue.
I expect to see such boards come out roughly around September.
The motherboard will also have to allow for a PCI/AGP lock, and be compatible with at least the first generation of 90nm Hammer chips. I don’t expect this to become a problem.
Outside of that, what I buy and when I buy it isn’t really up to me. It’s up to AMD. If they offer a socket 939 processor for a reasonable price (say around $150-$200), I’ll buy sooner rather than later. If they don’t, I’ll wait and see what AMD does with pricing 90nm processors.
I suspect AMD pricing/positioning is going to go something like this:
1. They’ll try to get a premium price on socket 939 chips to start, then introduce a slower model or maybe two at lower prices, maybe one in June/July, another just before 90nm. Prices will probably not go below the $200 level. Socket 754 processors will be cheaper, but not much cheaper than they are today.
2. 90nm will come in more or less on time, and there will be a brief period of premium pricing, followed by a quick replacement of 130nm with 90nm models at the mid- to high-level. Again, we’re probably looking at $200 or more.
3. This is not going to sell a lot of processors, so the dam will break sometime in 2005, perhaps April or so, and we’ll see $100 90nm Hammers around that time. I doubt socket 939 processors will ever become cheap, just cheaper.
So, depending on what AMD actually does in positioning and pricing, I’ll buy somewhere in that September-April timeframe, except that if it starts looking more like April, I might wait a bit more for DDR2 support. I’ll toss in a next-generation video card and a new Raptor, and that will be the core system.
Given what most of the audience has said about the subject, this is roughly what they’re going to do, too, though on the whole, I’ll probably jump sooner rather than later, while they’ll jump later rather than sooner.
However, there’s another, very different scenario lurking inside my head, one that completely trashes this one.
The Dark Horse
There’s an underlying assumption in all this: that socket 939 will become the mainstream overclocking AMD platform.
It’s by no means certain that assumption is going to pan out. It’s quite possible there will be a quite different world in 2005, and it goes like this:
AMD does as described above for socket 939. Come September/October, they introduce Paris. Paris will be effectively a cut-down socket 754 Newcastle with only 256K of cache and no functioning x86-64 circuitry.
AMD will sell these for perhaps $120 tops and down, sub-$100 pricing will come sooner rather than later. They’ll start at 130nm, and go 90nm perhaps middle to later 2005.
Depending on what AMD does with socket 939 pricing, many overclockers could well overlook the performance differences, focus solely on the price differences, and make Paris the AMD overclocking chip of choice.
If a socket 939 core platform ends up costing $200 more than a Paris system, I’d go so far as to say that will probably happen.
This will put AMD in quite a bind, especially if OEMs end up doing the same thing. Rather than pushing AMD fans to the high-end, they may end up shoving them instead into the low-end.
AMD fans in the past have shown again and again that they’ll take less performance for less cash, even when the dollar amounts at stake are much less than, say, $200.
And if it comes down to $200 more in cold cash, frankly, x86-64 fanatics, I don’t think they’re going to give a damn about 64-bits.
What does AMD do if 80 or 90% of the sales end up being Paris sales?
Maybe I ought to keep my ear to the ground before putting hand on wallet.