Hi Tenchusatsu; Re integrated memory controllers...
They do make a lot of sense, and for both the performance and the value markets. They undoubtedly would give lower cost for the value market and higher performance for the performance market (both by the simple and traditional technique of higher integration). The only thing holding delaying the stuff is time to market issues, and risk issues, I suppose. So far, it takes longer to get an integrated design out, and by the time you are shipping, your processor core (and maybe the graphics too) is obsolete. And if you blow either part of the design (ala Timna), you end up having to rev the whole thing. Separate CPUs and chipsets are safer. But I think that the winds of change are coming to this, both in the risk issues and in the time to market issues, and we will see cutting edge integrated solutions before 2 years are out. The risk issue that blew up Timna was its being tied to a peculiar DRAM type, but when it becomes a bit easier to predict the mainstream memory type, that risk goes away. A lot of the other risks are fading too. I have no doubt that all the major players are working hard on integrated solutions. In fact, like you, I'm quite surprised that we don't see more of them already.
The trend towards higher integration is probably one of the reasons that Intel isn't too concerned about losing a good portion of their chipset business to VIA. And of course that is why VIA is working hard on getting into the low end CPU business, they are trying to avoid a soon to be obsolete chipset business.
-- Carl
P.S. Last I looked, the Sri Lanka and Greece market averages were up slightly, so the Intel announcement hasn't completely destroyed the world financial markets. (G) |
