Now we’re going to see three core processors from AMD:
Why? “Market demand.” Yeah, right. Where? Surely not Earth. Maybe Mars?
Is this being done to get some money for not-quite-working-right quad-cores? AMD says no. Yeah, suuuuuuuuuure.
Look at it this way, this tricore would still have the L3 cache and memory controller of a quad, so you’d have 75% of a quad chewing up 85% of a quad’s die space. That’s hardly the best use of die space, unless a large proportion of these tris would otherwise have to be trashed or sold as duallies.
God only knows how current programming written to handle two or four processors will handle three; it may well be no big deal, but I suspect something funny is bound to happen with at least some multithreading code.
Yes, three cores would probably be less absurd than four for the average user, and some might like a three-core bargain, but does AMD really need to come up with new ways to make LESS money from their chips?
Yes, three cores would generate less heat, which might improve the speed grades, and at least for a while, three heads may be just as good as four.
So yes, there are some practical grounds to think tris are a good idea, beyond lemonade-making. However, if you think Intel is going to pound the K10 quads into the ground with pricing, what do you think is going to happen to tri-pricing when that happens?
I think any tri-pricing is going to end up little higher than dually. Keep in mind that it’s in Intel’s competitive interest to push quadcores hard once they get to 45nm, simply because the more quads AMD gets forced to make, the fewer chips AMD can make overall.
People talk a lot about Intel having a cost advantage using 45nm vs. AMD’s 65nm. That actually isn’t too big a deal, less than a handful of dollars per CPU.
The big deal is Intel using the smaller process technology to increase cores, which forces AMD to follow and make fewer CPUs with more cores. When you combine that with relative Intel superiority forcing AMD’s pricing into the bargain basement, AMD gets forced into building fewer and bigger chips for (relatively) less money.
That’s what is killing AMD.
Tags: Systems & Components