SMP K-7 (Score 1)

Check out an ongoing story over at AMD Central about the K7. According to the technical materials I have read, the K7 will not only do SMP, but will be more efficient than comparable Intel systems, which in the PIII peak at four-way processing. This limitation is caused by a shared system bus @100MHz. In other words, all four processors in a PIII four-way motherboard share the same 100 MHz of bandwidth to the system chipset.

In a K7 SMP system, each CPU has its own separate 200MHz bus to the system chipset. Also, this bus is separate from the memory bus running at 100MHz+, so there will not necessarily be a big bottleneck at the memory bus. The separate bus space for each CPU will allow more efficient SMP and theoretically higher clock speeds for the chips.

AMD, as pundits at The Register have pointed out, still has a huge manufacturing disadvantage. Suggestions have been increasing their deal with IBM to manufacture the K7 chip, borrowing another USD 2 billion for another fab, or looking for a suitor to save them. Compaq might have been a good choice until Pfeifer's ouster.

The bottom line is that the K7 will be a great chip, but clock speeds will not crank up much past 500MHz until the Dresden fab gets going with the .18 micron process by the end of the year.

Can AMD survive? I hope so, I've got some stock I'd hate to take a bath on, and they have good technology.

Right now, I've got a Celeron 300A overclocked at 450MHz, which of course Intel has quickly rid the market of because it was eating up their PII profits. I want a K7, but it will be priced to compete with the PIII, to finally give AMD the margins it needs to turn a profit.

Intel talks tough about competing with AMD, and they certainly have the money to back up tought talk, but I have heard talk from inside Intel that the 64 bit chip design is a disaster and is draining resources from other projects. Just a rumour ... ?