AM486DX5-133W16BHC AMD (ADVANCED MICRO DEVICES), AM486DX5-133W16BHC Datasheet - Page 27

• Manufacturer Part Number: AM486DX5-133W16BHC
• Manufacturer: AMD (ADVANCED MICRO DEVICES)

Specifications of AM486DX5-133W16BHC

• Family Name: Am486
• Device Core Size: 32b
• Frequency (max): 133MHz
• Instruction Set Architecture: RISC
• Supply Voltage 1 (typ): 3.45V
• Operating Supply Voltage (max): 3.6V
• Operating Supply Voltage (min): 3.3V
• Operating Temp Range: 0C to 85C
• Operating Temperature Classification: Commercial
• Mounting: Surface Mount
• Pin Count: 208
• Package Type: SQFP
• Lead Free Status / Rohs Status: Not Compliant

Step 3 If RDY is returned instead of BRDY during a
3.8.5.2
The use of AHOLD as the control mechanism is often
found in systems where an external second-level cache
is closely coupled to the microprocessor. This tight cou-
pling allows the microprocessor to operate with the least
amount of stalling from external snooping of the on-chip
cache. Additionally, snooping of the cache can be per-
formed concurrently with an access by the microproces-
sor. This feature further improves the performance of
the total system (see Figure 11).
Note: To maintain proper system timing, the AHOLD
signal must remain active for one clock cycle after HITM
transitions active. Deassertion of AHOLD in the same
clock cycle as HITM assertion may lead to unpredictable
processor behavior.
The following sections describe the snooping scenarios
for the AHOLD implementation.
3.8.5.3
Scenario : This scenario assumes that a processor-ini-
tiated access has already started and that the external
logic can finish that access even without the address
being applied after the first clock cycle. Therefore, a
snooping access with AHOLD can be done in parallel.
In this case, the processor-initiated access is finished
first, then the write-back is executed (see Figure 12).
The sequence is as follows:
Figure 11. Closely Coupled Cache Block Diagram
Address Bus
Data Bus
write-back, the HOLD signal can be reasserted
at any time starting one clock after ADS goes
active in the first transfer up to the final transfer
when RDY is asserted. Asserting RDY instead
of BRDY will not break the write-back cycle if
HOLD is asserted. The processor ignores
HOLD until the final write cycle of the write-back.
AHOLD Bus Arbitration Implementation
Normal Write-Back
L2 Cache
DRAM
CPU
Enhanced Am486DX Microprocessor Family
Peripheral
Interface
I/O Bus
Slow
P R E L I M I N A R Y
Address Bus
Data Bus
Address Bus
Data Bus
Step 1 The processor initiates an external, simple,
Step 2 In the same cycle, AHOLD is asserted to indi-
Step 3 During the next clock cycles, the BRDY or RDY
Step 4 Two clock cycles after AHOLD is asserted, the
Step 5 Two clock cycles after EADS is asserted, the
Step 6 In this cycle, the processor-initiated access is
Step 7 Two clock cycles after the end of the processor-
Step 8 As an example, AHOLD is now removed. In the
Step 9 The write-back access is finished when BLAST
Step 10 In the clock cycle after the final write-back
The status of the snooped and written-back line is now
either shared (INV = 0) or is changed to invalid (INV = 1).
non-cacheable read access, strobing ADS = 0
and W/R = 0. The address is driven from the
CPU.
cate the start of snooping. The address bus
floats and becomes an input in the next clock
cycle.
signal is not strobed Low. Therefore, the pro-
cessor-initiated access is not finished.
EADS signal is activated to start an actual
snooping cycle, and INV is valid. If INV is 0, a
read access caused the snooping cycle. If INV
is 1, a write access caused the snooping cycle.
Additional EADS are ignored due to the hit of a
modified line. It is detected after HITM goes in-
active.
snooping signal HITM becomes valid. The line
is modified; therefore, HITM is 0.
finished.
initiated access, the cache immediately starts
writing back the modified line. This is indicated
by ADS = 0 and W/R = 1. Note that AHOLD is
still active and the address bus is still an input.
However, the write-back access can be execut-
ed without any address. This is because the
corresponding address must have been on the
bus when EADS was strobed. Therefore, in the
case of the core system logic, the address for
the write-back must be latched with EADS to
be available later. This is required only if
AHOLD is not removed if HITM becomes 0.
Otherwise, the address of the write-back is put
onto the address bus by the microprocessor.
next clock cycle, the current address of the
write-back access is driven onto the address
bus.
and BRDY both transition to 0.
access, the snooping cache drives HITM back
to 1.
27