not sure if I've read the answer to this anywhere, but I think if I had I would remember. I've been assuming that a single memory unit can service multiple requests in parallel, e.g. sdram[read, $blah, blah, 0, 1] sdram[write, $blahh, blahh, 1, 1], sig_done ; ... ctx_arb[sdram] those two commands will take closer to the time it takes one command to complete than the time it takes a two commands to complete serially, right? I would hope thats the case, but I don't know if I've ever seen that explicitely confirmed or repudiated. thanks, Najati