S?bastien Bourdeauducq wrote: > Actually we're not using them (they're still there in the "regular" > design as a remnant of previous experiments and removed entirely in > the high speed design) - it can work without. And they have a bunch > of quirks too, starting with large PVT effects.
I think what you're saying it that it may not be a bad idea to pay attention to trace lengths in M1r4 :-) Micron themselves recommend on page 12 of [1] to keep the lengths of the differential clock traces within +/- 20 mil (0.5 mm) of each other, and clock to DQ within +/- 500 mil (12.7 mm or about 0.05 ns). That doesn't sound excessively demanding, even without zig-zagging. [1] http://download.micron.com/pdf/technotes/DDR/tn4614.pdf - Werner _______________________________________________ http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org IRC: #milkymist@Freenode
