Is Conroe a killer CPU you’ll have to buy?
That entirely depends on what you have to be killed.
Conroe will likely appeal most to those well behind the curve (i.e. pre-Hammer machines for AMDers, Northwoods or earlier for Intelers).
For anybody with anything more recent than that, well, outside of some early Hammer buyers, they’ll probably not get enough of a boost.
The one big question that hasn’t been answered yet about Conroe is its overclockability. Obviously, there’s no hard data yet, but given their Dothan ancestry, I don’t think these things are going to overclock a ton faster than 2.66GHz using orthodox cooling. 3GHz may end up being a real challenge with the first ones; I would be surprised if we saw anything like 3.5GHz anytime soon (or not so soon), again, freeze units not included.
Another point that hasn’t gotten too much attention is that Conroes come with two cache sizes; there’s the 2Mb and 4Mb version. If you want the 4Mb version, you’re going to lay out at least $300 for one when they first come out. The 2Mb version will be more like $200, but less cache means you’ll lose a chunk of extra performance, too.
Overclockers may well be attracted to the 1.86GHz. I don’t doubt those CPUs will be capable of 50% overclocks, if the motherboards will let them.
These CPUs will be running at 266MHz FSB (quad-pumped), so if you’re looking for a 50% plus overclock, you’re talking about running the mobo at 400Mhz+ FSB. That’s a likely bottleneck, at least initially; there’s not too many mobos capable of such speeds around these days, and may not be for a while.
Running memory at 800MHz ought not be a problem, provided you’re willing to pay for it. Low-latency memory may be a different story for a while.
Like most processors, Conroe will mature in the course of time, and there will be growing pains.
For those who cannot contemplate buying something from Intel, you’re just going to have to wait until AMD comes up with something better.
It almost certainly will not be the initial AM2s. While they probably won’t trail the Conroes by as much as current AMD CPUs, a sizable gap will still remain, and it’s hard to see that gap being closed until AMD comes out with 65nm chips with extra goodies built in.
This is part of the reason why 65nm transition is so important to AMD’s competitiveness (production capacity and cost are the rest).
There’s a bunch of scuttlebutt about future improvements and extra help, but realistically, at best, maybe the end of 2006 if AMD gets desperate (i.e., 65nm gets really delayed), 2007 realistically for those items AMD has mentioned, and probably much later if ever for some of the wilder items.
For instance, there’s been some talk about a coprocessor designed by Clearwater. AMD says they’re “talking to them.” Well, obviously not for very long, since Clearwater’s products aren’t even optimized for AMD processors, which might be a good first step.
Let’s be kind and succinct about this, and point out that designing a coprocessor and building a platform infrastructure around it is not something anybody could rationally expect to be done in 2006 or even 2007.
While most AMDers can rationally wait to see when/if the company trumps Conroe for at least a while, maybe a long while, what one ought not do is simply buy an AM2 in the hope that someday within the socket’s life cycle, they’ll manage the feat.
Really, some of the recent comments by AMD execs seem to boil down to, “If you buy AM2 when it’s ready, someday, we’ll have something really good for it.”
If you can manage that leap of faith, you’re ready for religion, or maybe a Mac. 🙂 There’s just been too many cases in the past where compatibility was promised but not delivered (by both Intel and AMD) to base a buying decision on that hope.
As always, we advise you and your wallet, “When in doubt, don’t whip it out.”