AMD to Release Six on the Desk in 2010

AMD to Release Six on the Desk in 2010

the oracle

A small blurb on Tech Connect gives news of a bigger AMD gun to be brought to the desktop shootout in 2010. With the multiple core products that AMD is bringing to the server space, it will be no big deal to bring a Six Pack to the desk. With its 8- and 12-core Magny-Cours Opterons planned for Q1 2010, AMD should be able to more than keep up with Intel on the server front but that’s ‘only’ one battleground for the two chip titans. On the desktop side Intel claims performance supremacy with the Core i7 quads and is currently planning to introduce a six-core LGA 1336 CPU, codename Gulftown, in Q2 2010. To counter Intel’s six-core moves, AMD is reported to be preparing a hexa-core of its own which would be scheduled to debut in about a year, in Q3 2010. Named ‘Thuban’ (meaning ‘dragon’ in Arabic), the six-core AMD desktop processor will be made on 45nm technology (Gulftown is 32nm), come in an AM3 package, and feature an integrated DDR3 memory controller, and likely 6MB of L3 cache. Once out, Thuban should be the flagship CPU of the upcoming Leo platform which will include the 800 Series chipsets and the DirectX 11-supporting Radeon HD 5000 series graphics cards. I’m a fan, so I hope it, 1] comes true, and 2] performs very well. i do however, see a couple of things (possibly) wrong with the news. First, the amount of on chip cache seems very small. We’ve already seen that the latest Phenom II models could have used more L3 cache on die, and now AMD will be doubling the cores, and keeping the L3 the same? And next, the bringing of this six core out on 45nm, instead of 32 nm, would seem to be starting with an extreme disadvantage, and not seeming to acknowledge the fact. Also, bringing it in an AM3 package means that only 2 memory channels would be available, after we’ve been shown what an advantage 3 channels are for i7. I would have thought we’d see a socket 1207, at the least. I’m not sure why AMD seems to be playing it so safe, when, in reality, it needs to leapfrog Intel, to get back some of the prestige and marketshare it was starting to build in the years 2003-2005. What do the rest of you armchair benchmarkers think about this? Is this a mistake to commit to what will certainly be an inferior product to what will be Intel’s best at time of introduction?