That's correct. Also, the number of physical channels and banks has an effect on latencies. In fact, it's possible that if trace lengths are long enough, or if there's enough loaded (filled) banks, PC800 might have higher latencies then PC600 on simpler (lower bank count etc.) platforms. This is because of artificial delays introduced to compensate for different delays. Of course, PC800s raw-bandwidth would generally more then make up for the latnecy difference.Quote:
Originally posted by FaTs:
PC600 is higher latency than PC800 isn't it ?
Because of the way the DRDRAM memory controller accesses the mapped memory, the memory controller must introduce delays so that all modules have the same latency. I'm not quite sure about this, but I believe the RDRAM controller doesn't use a conventional strobe on a conventional demultiplexer (decoder) like SDRAM controllers. That point is really not important in answering your question, however.
More specifically, since slower speed RDRAM runs on a slower dedicated clock, any signal latencies (propogation delays, etc.) would be amplified and would have to be accordingly compensated for. The exact algorithim for this I'm not sure of, but it is part of Rambus' tight reference specifications for the DRDRAM platform I'm sure.
As always, hope the helps. It's very late here and incoherency is beginning to set in, forgive me for that. https://www.sharkyforums.com/images/.../2005/06/5.gif
------------------
Count Baron of the OC Crusaders
[This message has been edited by bobgod007 (edited October 09, 2001).]
