Hi mishedlo; Re the AMD LDT bus... Not at all similar.
No one has a patent on 800MHz signaling. Or differential signaling. Or source synchronous clock forwarding. Certainly the rdramreview.com website doesn't claim that Rambus has such. To mom and pop, the two technologies probably look similar, but that is not the case to those who deal with this stuff all the time:
The primary differences between the two technologies, Rambus and LDT, that I noticed just from a very quick glance:
(1) The LDT is unidirectional, while Rambus is bidirectional. This means that the LDT is going to be a lot easier to build, use less power, and not going to have the inter-symbol interference problems that Rambus has been having.
(2) The LDT is point to point, while Rambus channels can have up to 33 active devices. This means that the signals only have to be valid at the ends of the signal lines, a much simpler issue than having to get data to all the intervening points. As an example, they could implement the LDT with simple series resistors at one end. This would mean considerably lower power consumption.
(3) In light of the above, it is natural to assume that the LDT uses something other than the rather difficult RSL for its signalling levels.
I should note that the LDT system is about what I would expect to see between modules in a complicated MCM. The basic technology assumptions that went into its design, as very much distinct from Rambus, is that pins are cheap. So lets make the signaling paths unidirectional. This is going to be what the world of MCM is going to be like, though the technology is also a natural for PCBs with high pinout BGAs.
In short, the LDT is a lot easier than Rambus' stuff. Lower power, easier timing, but more pins.
-- Carl |
