I don't think so... I think this is a token move from Intel, to try to placate their critics over the next 6 months. Intel want to show their OEMs that there is a plan B, but I don't think they plan to execute it with any urgency. By the time Intel has design data ready for any PC133 chipset, it will be clear that Rambus will fly, and Rambus prices will be coming down due to mass production and competition. In the meantime OEMs will hold off from going with VIA, because "there is an Intel PC133 chipset coming"
In the "Rambus is expensive" argument, something important is being forgotten. A RIMM has far less pins than a DIMM. The sockets will be cheaper when they reach similar volumes, laying out the boards is difficult the first time you do it, but once the techniques are learnt its not that big a deal (I designed an impedance matched memory card back in 1991).
The one big thing that people are forgetting is the pincount on the chipset. Yield in IC manufacturing is not just about die size. The number of bond wires (>= no of pins) to be attached is a factor too. Wire bonding to the die is a risky operation. There is a risk of mispositioning, there is a risk of thermal or mechanical damage, there is a risk of contamination. The fewer bond wires, the better the yield of finally packaged chips. The package itself is also cheaper. I would expect that Camino and Carmel will cost Intel less to manufacture than older chipsets. Sure they will take a bigger margin while they can, but if it comes to a market share crunch, I am sure they will have room to manouver.
So I still have confidence in the story and will hold.
Ian |
