> 512K of SRAM????!!!!!! Just how much is that going to cost?! > I seem to remember communication on the mailing list last term > that deduced that DRAM would be feasible (and, of course, cheaper) > though it would slower...
Wooooah... hang on... no need to worry that much - most of the cost of this thing has been borne by Martin & me... (ouch... 50 quid for 256k of SRAM...) The finished board will have a whole host of options - either 1Mb of DRAM, fast SRAM, a combination... you set the jumpers, and hey presto - more speed depending on which you plump for. > When you say slow down for /some/ writes, how do you propose doing > this? The only way I can think of is: > > i) using cacheing techniques That's something we're looking into at the moment (possibly necessary to work with the ASIC & standard Megs) > ii) Say that the screen can only reside in certain sections of > memory That's something we'll leave to the programmers - selectable on 4 ports - choose the ports you want to write through to the internal memory. > iii) similar to ii) but copy from fast-ram to sam-ram whenever > the VMPR port is altered (not very elegant). That'll be an optional way of doing it - for things like BASIC for example -- if you're paging an area into the VMPR for the first time, its "copy-thru" bit will be set, and the data will be blatted across. > Another nice way of doing things would be to make your own video > card using VRAMs - just modes 3 and 4. And to think that you were moaning about the costs of SRAM :) :) :) Simon
