By mr_debauch
Tue Sep 30, 2008 3:17 am
guyb wrote:Nope, a parallel bus at the same clock speed and with the same profile connecting wire/track will be faster than a serial one. There may be some problems in synchronization due to bus skew but these are solved at design time, by keeping PCB track lengths equal and removing any sharp 90 degree turns etc. SATA300 is quicker than IDE, but its a much newer technology. USB2 is faster than a parallel printer port, but thats just because of the standard set when it was designed. A printer cable has the bandwidth capability to be much faster than a USB2 one, its just that there was no need for that ridiculous data transfer in 1990 since the biggest hard drives were about 100MB, and the hardware to support a standard like that would be prohibitively expensive. The same goes for IDE.
You can get serial memory, but its a hell of a lot slower than parallel addressed modules. Bear in mind that the i/o speed of a hard drive is incredibly slow when compared to a RAM module. You never get sustained transfer speeds from a HDD near the 133MBps limit of IDE.
What AMD were referring to when they criticised the FSB was having one bus to all peripherals and thus one global address space. This means that only one peripheral can only be used at once (excluding things like DMA and interrupts etc). With a point to point system, you still have parallel busses, but there would be one bus between each peripheral that needs to be connected. (This would be a bit overkill, generally its just the memor, cpu and maybe graphics etc that need such high bandwidth). This sort of scheme massively over complicates the PCB layout and design considerations, but can offer better performance.
Amen on the solar panels!
and if only we would have a need for high bandwidth in the video department on an mpc *hint hint akai*
next step, trying to get solid state drives working on the newer mpc's.
MPC-Live/Remix16
