by Ed Stroligo.
Ed points out that the pace of new product introduction will only pick
It’s been a really rough summer for computer hardware sites. Nothing significant has been happening in the high performance area.
845 boards? Please. Those are the antiChrists of high performance.
AMD SDRAM boards? With 256Mb of quality DDR RAM going for $35; it’s really hard to see why you’d buy one.
The various DDR boards out there now are making incremental improvements here and there, improvements so small that factors like platform maturity loom bigger.
The various P4 DDR boards coming out? Frankly, I don’t care how some rushed boards perform with soon-to-be-extinct CPUs and compare to a soon-to-be-extinct RDRAM standard. I will care to see how more mature boards will handle the next up-and-coming significantly modified CPUs, and how they compare with
the next generation of RDRAM boards.
While the fall will be better, the only expected item that has me a bit excited are the possibilities of an overclocked Northwood CPU. That’s the only item that may represent a fairly big jump in performance.
Palomino? It will be an improvement, but only earth-shaking for those using SSE-enabled applications. It just doesn’t look like it’s going to overclock all that much.
It’ll be OK, but the exciting AMD chips won’t come until 2002: whether it be Thoroughbred or Clawhammer.
Preliminary reports indicate nForce is a dud compared to the initial hype. We suspect the reason underlying that is that current Athlons simply can’t make much use
of a dual-DDR structure, much as PIII couldn’t do much with a DDR board. We also suspect Thoroughbred/nForce2 will be a better matchup.
DDR333 is coming, and motherboards will accommodate that, but again, looks like 2002.
Even the Northwood mentioned above is really a 2002 chip for overclockers due to cost.
Video cards? What I think we’ll see this fall is not excitement, but affordability. DX8-compliant technology, whether it be GF3 or Radeon 2, should become reachable to the wallets of most high-performance users, and may become a fairly hot item.
I think XBox is going to end up being the hot Christmas item, or at least look that way and get media attention.
Peaks and Valleys
The rest of 2001 does look kind of duddish. 2002 shapes up to be a much more exciting year.
This is perfectly normal. We’re coming to the end of the .18 micron product cycle in CPU, and we’re just not going to see anything earthshaking from what
are essentially tweaked-up chips. The next big jump will occur going to .13 micron.
This is why timing is important in making a purchase. If you had bought a high-end TBird three or four months ago, you’ve enjoyed it three or four months already, and won’t get significantly outclassed until next spring.
If you buy a Palomino for Christmas, you’re outclassed in three months.
AMD’s six-month delay of Palominos (and word is getting out that AMD did have manufacturing problems with these) combined with the additional tweaking of the TBirds took much of the fizz out its introduction.
True, the SMP people will get more benefit from this, but this is still a preference of a relative few.
On the other hand, if a Palomino ends up costing $100 by late October/early November, popping one into an existing system isn’t exactly an international banking decision, either.
What Will I Have A Year From Now?
That’s a real good question. It could be an Intel machine; it could be an AMD machine.
I’ll tell you the one thing it almost certainly will have, a ton of RAM.
I spoke a few months ago about having a 4Gb RAM machine, and essentially do everything off a RAMdrive. When I said that, the necessary technology wasn’t available.
It is now. There are Corsair 1Gb DDR sticks available today. Might be a pain to find a mobo with four slots, but it’s doable.
It would also cost me $2,200 for just the memory, but by the time I’m looking at a next-generation machine, the price will have probably dropped below my personal pain point.
This is not to suggest you should all go off and do this, too, but I think it illustrates a problem we’re all going to be facing the next couple years.
There’s no point having a zillion GHz processor if you end up spending your time waiting for your machine to read rust. A split-second here, a split-second there, pretty soon you’re talking real time. 🙂
Our whole computing infrastructure is built on reading rust, and I don’t think we’re going to be able to afford that in the long run.
I’ll grant you, sticking megagigs on a desktop machine next year is pre-pioneer, and it’s going to be pretty clunky (Windows is pretty hostile to the notion at the moment), but this isn’t going to sound so crazy three years from now.
There are technologies out there which could let storage media give RAM a run for its money, but they’ll look a lot different than today’s hard drives, and we won’t see them any time soon.
Whichever it is, something needs to be done, otherwise this GHz race will just look sillier and sillier.