Although Intel had a nasty shock when Prescott flopped so spectacularly, it was well aware that these problems were approaching, but presumably had hoped that they would hit some time in the future.
As far back as 1989 engineers at Intel knew that the day would come when conventional processors would grow too big, complicated and expensive and that there would inevitably be a move to multi-core processors. One of the engineers who worked on the dual core report was Pat Gelsinger who is now close to the top of Intelâ€™s corporate tree. Gelsinger was evangelising the dual core technology at the Intel Developer Forum last month, which we reported on here.
Taken at the most basic level a dual core processor works much like a dual processor computer, except that it only has one processor socket. The big difference is that in a dual Xeon workstation or server the two processors have to communicate with each other, as well as with the system memory, so thereâ€™s inevitably a performance hit as the processors negotiate which data they will each work on. AMD takes a very different approach to SMP (Symmetric Multi-Processing) as Opteron and Athlon 64 have the memory controller on the processor die. In a two-way Opteron workstation or four-way Opteron server the processors can communicate directly with each other over Hyper Transport, with the result that performance increases in a linear fashion as the processor count increases.
A Smithfield dual core processor die.
In a Pentium 4 PC you already have two virtual processors thanks to Hyper-Threading which Intel tends to refer to as HT, but this is not to be confused with Hyper Transport. A dual Xeon workstation that uses HT will have four virtual processors working away on two sockets, so the architecture for a dual core desktop processor isnâ€™t really a big step for Intel to take. In fact there is plenty of evidence that Intel would have been more than happy to wait until 2006 or 2007 to introduce dual core for the desktop as 2005 was clearly intended to be the year when Intel worked on its EM64T extensions to Windows XP Pro 64-bit, along with its XD hardware buffer overflow protection.
Once Prescott forced Intel to move up its dual core plans, the company had to take a decision about exactly which core it would use, and there was very little choice in the matter. The 1MB L2 cache Prescott was the only candidate and there was no time to modify the core so it came up with a piece of silicon which is two Prescotts laid side-by-side on the same die. This is a pain for Intel as it requires both cores to come out of fabrication perfectly, but if one suffers a fault we presume that it can chop the die in half and use the good bit as a Prescott.
The dual core is twice the size of a Prescott but slightly smaller than a Willamette Pentium 4. The staggering figure is that the dual core processor uses 230 million transistors, so itâ€™s no great surprise that the TDP is 130 Watts. It is hard to imagine that anyone in the engineering department at Intel wanted to produce a processor with those sorts of numbers, but it is equally easy to see that Intel had no choice. It had to move forward and this was the only option for 2005.