Dale, Re: "I estimate that at 130nm Sledgehammer will have a die size of about 150mm2. This would make it larger than Northwood, but it also includes 1MB of L2 cache already."
I thought you were comparing 130nm Sledgehammer to Prescott. In that case, I would estimate the Prescott die to be quite a bit smaller than Sledgehammer.
But I do find it interesting that you believe that the die area can grow almost 50%, just with 512KB more L2 and two more Hypertransport links. Other than that, though, I don't think there being any other difference from Clawhammer (the 128-bit memory interface shouldn't add anything to die size - likely, the Clawhammer already has the logic for 128-bit DDR, but it is disabled). Any reason why you expect so much growth?
Dale, Re: "Overall, I would guess that the Sledgehammer die at 90nm will cost about 25% more to manufacture than the Prescott die."
That's another interesting prediction. You might be correct. However, I know you've also proposed your theory about AMD being late to 90nm, just as they are now late to 130nm. If you are also correct about that, and 90nm Hammer CPUs won't be coming along until 2004, then wouldn't it be better to compare Prescott against the 130nm Clawhammer, and Nacona (Prescott based Xeon) against the 130nm Sledgehammer? How would you compare the costs between those?
wbmw |
