90nm Semprons should start showing up in a few weeks, maybe as soon as next week.
We mentioned a little while back that we thought Semprons would be the way to go for the average person reading this.
The initial response was, to put it mildly, underwhelming.
Perhaps the reason for this is that we felt that this was what people would eventually do, while those reading just looked at the situation right now.
We think a number of things are going to happen over the next three-six months that will make most people change their minds, and we’ll detail them here.
First, 90nm The prerequisite for overclockers adopting Semprons will be the introduction of 90nm chips. This in-and-of-itself won’t have much impact, but when overclockers realize that these chips are basically the same as their bigger brothers, Semprons will slowly become more appealing, especially when . . . .
The Price Goes Down When I wrote the Sempron article, a few wrote back to say “Why should I buy a Sempron now when for just $20-25 more, I can get an Athlon 64 2800+?”
The answer is simply, “No, you don’t buy it now, but the price gap is bound to widen. We most probably will never see a $70-80 Athlon 64; we will almost certainly see a $70-80 Sempron. When prices reach that kind of level, AMD fans will start buying.
Of course, if we do eventually see a $70-80 Athlon 64, people will buy that, but that’s probably a bit much to expect.
Not Much Oomph From 90 Looking through various forums, there seems to be this touching belief that these chips will be popping 3GHz with the greatest of ease in a few months.
While we certainly would be delighted if that happened, we’re just as certain that this just isn’t going to happen, and that 3GHz for anyone with less than a freeze unit won’t happen for close to another year (and just maybe not at all). A lot of people haven’t or won’t realize that the world has changed, and it’s no longer business-as-usual in the CPU world.
As time goes by, and the Tooth Fairy doesn’t show up, people will downgrade their expectations, and want to pay less for getting less than what they wanted.
x86-64? Contrary to what certain hypesters will have you believe, there’s little enthusiasm for x86-64 even among AMD fans (as opposed to fanatics). It’s not that they’re against it; it’s more a matter of “Show me the money.”
Give people plenty of software they actually use that takes a lot of advantage of x86-64, and people will warm up quickly enough, but if the average person wakes up next June to find an Windows OS and two games in x86-64, he’s not too likely to be wound up about that.
We think those who are patient enough will end up with an x86-64 enabled Sempron once Intel puts out an x86-64 enabled Celeron, but Intel is hardly dripping with enthusiasm over the whole project. It’s “Yeah, we’ll do it, but enthusiasm is not part of the package.”
So Intel may take its sweet time about doing this, but it may not matter a whole lot to people if there’s little for x86-64 to run on.
Much Smoke, Little Fire Today, we have sockets 939 and socket 754. Whenever AMD feels like going to DDR2, that will mean a new socket or two. Somewhere in the middle of this will be the first hyper-expensive dual cores which (like the Intel dual-cores) will run rather slower than the fastest single-cores, and probably perform worse in most apps.
The prospect for 2005 is for new systems to carry even bigger price tags, yet offer little more power. DDR2 won’t, can’t help much, no memory can. Dual processors need multi-threaded apps that actually benefit from being multithreaded.
Yes, there will be people who’ll buy such systems, but far more will either continue to sit on their wallets, try to stretch current systems with a new video card, or start looking hard at the cheapest upgraded platform they can find.
Conclusion: The Wheels Are Off The Bus
I just got an email which talked about “your eternal quest to proliferate negativity.” I bet some of you out there are nodding your heads in agreement.
He said it like it were a bad thing. 🙂
The problem isn’t that I’m being negative; it’s that reality is. The era of quick, easy performance gains is over, maybe for good. All the chip architects have been saying that to anyone who’ll listen, but people aren’t listening.
Nor is it a matter of loving/hating AMD/Intel. The whole industry is stuck, not just one company, and it is because they’re essentially stuck that we’re even hearing about concepts like dual-core, just to keep up the illusion of progress, pretending that the wheels are still are on the bus, and that the bus is just rolling along.
It’s not the end of the world, just the end of the world as we knew it. Video still has a little breathing room to go, but for CPUs, advances are going to be tough to come by and a lot more expensive: less bang, more buck.
Under those circumstances, the best we can do for the audience is to point out the best you can do under more adverse conditions, and we think it’s going to end up being Semprons.