THERE HAVE BEEN back channel stories floating around for weeks about tri-core AMD CPUs, but nothing solid. The rumours have been picking up more steam, and now they are finally solid: AMD is doing a three-core chip. The first thing that comes to mind is bad yields, but I am told this is not the case for the three-core product lines. There may be some salvage done on bad quads, but that is not the overriding reason to do this.

The main reason is marketing, it seems AMD is learning from ATI. Most non-top SKU GPUs are simply top SKU die with features turned off, and if you look at the success of people unlocking that, you will see that it is far more than salvage.

AMD is probably doing this for two reasons; the lesser being salvage, the more important one being that Intel can't do it. Intel would have a far harder time making a tri-core part until Nehalem next September – it is easy to fuse off a core, far harder to MCM disparate cores.

This will allow AMD to come out with a lot of mid-range SKUs, having a complete 1-4 core range servicing every market. It also allows for complete market differentiation with a year or so's window into a place where Intel is not.

On the technical side, this is pretty trivial to do: three to core four is just a fuse to blow. What it gets you is a whole lot of choices. Remember the smooth run of SKUs, that was the beginning. If your clocks are thermally constrained, having three instead of four cores gives you a bin or two of speed. Given how few games use a second core fully, this might be a real win.

As far as money goes, assuming there is no salvage, three cores could still be a profit win, but it could be a loss. The selling price of a three core is greater than the price of a dual, and if that difference is greater than the manufacturing cost difference between a dual and a quad, AMD wins. If it is not, or people who would buy a quad buy a tri, then they lose.

Overall, it ends up with greater flexibility for AMD. How the firm uses it will determine whether or not this is a win, loss or draw. In any case, look for it on the consumer side, not the server first, and possibly moving over if it works out. µ