Intel 80C186XL Computer Hardware User Manual


 
10-21
DIRECT MEMORY ACCESS UNIT
10.3.2 DMA Latency
DMA Latency is the delay between a DMA request being asserted and the DMA cycle being run.
The DMA latency for a channel is controlled by many factors:
Bus HOLD Bus HOLD takes precedence over internal DMA requests. Using bus HOLD
will degrade DMA latency.
LOCKed Instructions — Long LOCKed instructions (e.g., LOCK REP MOVS) will
monopolize the bus, preventing access by the DMA Unit.
Inter-channel Priority Scheme — Setting a channel at low priority will affect its latency.
The minimum latency in all cases is four CLKOUT cycles. This is the amount of time it takes to
synchronize and prioritize a request.
10.3.3 DMA Transfer Rates
The maximum DMA transfer rate is a function of processor operating frequency and synchroni-
zation mode. For unsynchronized and source-synchronized transfers, the 80C186 Modular Core
can transfer two bytes every eight CLKOUT cycles. For destination-synchronized transfers, the
addition of two idle T-states reduces the bandwidth by two clocks per word.
Maximum DMA transfer rates (in Mbytes per second) for the 80C186 Modular Core are calcu-
lated by the following equations, where F
CPU
is the CPU operating frequency (in megahertz).
For unsynchronized and source-synchronized transfers:
For destination-synchronized transfers:
Because of its 8-bit data bus, the 80C188 Modular Core can transfer only one byte per DMA cy-
cle. Therefore, the maximum transfer rates for the 80C188 Modular Core are half those calculated
by the equations for the 80C186 Modular Core.
0.25 F
CPU
×
0.20 F
CPU
×