Originally Posted by APOLAUF
(Post 119403)
This is making a large assumption. The first assumption is that you're using a similar address bus on the Mogul as the Touch. While they do use the same chipset, and therefore you'd expect the same bus width, it doesn't mean that all the wires for the address bus are present on the Mogul.
Take for instance the fact that 64MB of RAM doesn't require an address bus larger than 24 bits, assuming that you have a 4-word-addressable memory. Addressing 128MB of RAM consequently requires 25 bits in our 4-word scheme. Why would they add the extra PCB traces to accommodate, say, the full 32-bit address bus, if the memory installed will never be more than what the factory intended? So add either 24 or 25 traces for the address bus, and you've saved realestate on a small embedded device platform.
The above doesn't imply that the address bus doesn't have the full 32 bits available, but it is nonetheless an assumption, and may be problematic.
The second assumption is that memory timing and latency is the same. It may be, and if it is standard SDRAM, a slower timing will work just fine on faster, newer memory. But what if the power requirements are different? It's well-known that more RAM means a bigger drain. This is because of CMOS leakage current within the device. The more gates/transistors you have, the more current leakage you are going to have. (SDRAM uses capacitive memory, but it still must be switched. Also, given that it is capacitive memory, it takes more power to recharge the RAM cells than if you had fewer of them, as in 128MB vs 64MB.) A way around this is to change your operating device characteristics. Build a lower-voltage device circuit, and change your device characteristics (transistor size, gate width, etc.) so that your leakage isn't as bad. If the power supply to the memory is different, then there is essentially no chance that you will see a RAM upgrade, especially not from the Touch.
These are just some points - simply moving a chip because the dimensions match says nothing about their interoperability. It'd be NICE, but keep in mind that embedded devices aren't as standardized as PC memory, where the end-user is actually responsible for their own memory upgrades (usually anyway.) Therefore, it makes it a lot more difficult, in the long run, to make embedded device parts interchangeable, and leaves us SOL because the manufacturer chose one part over another.
|
