• NBT (No Bus Turn Around) functionality allows zero wait
Read-Write-Read bus utilization; fully pin-compatible with
both pipelined and flow through NtRAM™, NoBL™ and
ZBT™ SRAMs
• FT pin for user-configurable flow through or pipeline operation
• IEEE 1149.1 JTAG-compatible Boundary Scan
• ZQ mode pin for user-selectable high/low output drive
• 2.5 V or 3.3 V +10%/–5% core power supply
• 2.5 V or 3.3 V I/O supply
• LBO pin for Linear or Interleaved Burst mode
• Byte Write (BW) and/or Global Write (GW) operation
• Internal self-timed write cycle
• Automatic power-down for portable applications
• JEDEC-standard 119- and 209-bump BGA package
-250 -225 -200 -166 -150 -133 Unit
Pipeline
3-1-1-1
3.3 V
2.5 V
Flow
Through
2-1-1-1
3.3 V
2.5 V
t
KQ
tCycle
Curr (x18)
Curr (x36)
Curr (x72)
Curr (x18)
Curr (x36)
Curr (x72)
t
KQ
tCycle
Curr (x18)
Curr (x36)
Curr (x72)
Curr (x18)
Curr (x36)
Curr (x72)
2.3
4.0
365
560
660
360
550
640
6.0
7.0
235
300
350
235
300
340
2.5
4.4
335
510
600
330
500
590
6.5
7.5
230
300
350
230
300
340
3.0
3.5
5.0
6.0
305
265
460
400
540
460
305
260
460
390
530
450
7.5
8.51010101115ns
8.5
210
200
270
270
300
300
210
200
270
270
300
300
3.8
6.6
245
370
430
240
360
420
195
270
300
195
270
300
4.0
7.5nsns
215
mA
330
mA
380
mA
215
mA
330
mA
370
mA
150
mA
200
mA
220
mA
145
mA
190
mA
220
mA
250 MHz–133MHz
2.5 V or 3.3 V V
2.5 V or 3.3 V I/O
with either ADSP or ADSC inputs. In Burst mode, subsequent
burst addresses are generated internally and are controlled by
ADV. The burst address counter may be configured to count in
either linear or interleave order with the Linear Burst Order (LBO)
input. The Burst function need not be used. New addresses can be
loaded on every cycle with no degradation of chip performance.
Flow Through/Pipeline Reads
The function of the Data Output register can be controlled by the
user via the FT mode . Holding the FT mode pin low places the
RAM in Flow Through mode, causing output data to bypass the
Data Output Register. Holding FT high places the RAM in
Pipeline mode, activating the rising-edge-triggered Data Output
Register.
Byte Write and Global Write
Byte write operation is performed by using Byte Write enable
(BW) input combined with one or more individual byte write
signals (Bx). In addition, Global Write (GW) is available for
writing all bytes at one time, regardless of the Byte Write control
inputs.
FLXDrive™
The ZQ pin allows selection between high drive strength (ZQ low)
for multi-drop bus applications and normal drive strength (ZQ
floating or high) point-to-point applications. See the Output Driver
Characteristics chart for details.
Sleep Mode
Low power (Sleep mode) is attained through the assertion (High)
of the ZZ signal, or by stopping the clock (CK). Memory data is
ns
retained during Sleep mode.
Core and Interface Voltages
The GS8324Z18/36/72 operates on a 2.5 V or 3.3 V power supply.
All input are 3.3 V and 2.5 V compatible. Separate output power
(V
) pins are used to decouple output noise from the internal
DDQ
circuits and are 3.3 V and 2.5 V compatible.
DD
Functional Description
Applications
The GS8324Z18/36/72 is a 37,748,736-bit high performance 2-die
synchronous SRAM module with a 2-bit burst address counter.
Although of a type originally developed for Level 2 Cache
applications supporting high performance CPUs, the device now
finds application in synchronous SRAM applications, ranging
from DSP main store to networking chip set support.
Controls
Addresses, data I/Os, chip enable (E1), address burst control
inputs (ADSP, ADSC, ADV), and write control inputs (Bx, BW,
GW) are synchronous and are controlled by a positive-edgetriggered clock input (CK). Output enable (G) and power down
control (ZZ) are asynchronous inputs. Burst cycles can be initiated
Specifications cited are subject to change without notice. For latest documentation see http://www.gsitechnology.com.
NoBL is a trademark of Cypress Semiconductor Corp.. NtRAM is a trademark of Samsung Electronics Co.. ZBT is a trademark of Integrated Device Technology, Inc.
IChip Enable; active low
IChip Enable; active low (x72/x36 Versions)
IChip Enable; active high (x72/x36 Versions)
IOutput Enable; active low
IBurst address counter advance enable
Preliminary
Byte Write Enable for DQC, DQD I/Os; active low
(x72/x36 Versions)
Byte Write Enable for DQE, DQF, DQG, DQH I/Os; active low
ISleep Mode control; active high
IFlow Through or Pipeline mode; active low
ILinear Burst Order mode; active low
IMust Connect High
IMust Connect High (x72 and x36 versions)
Must Connect Low
Must Connect Low (x18 version)
IWrite Enable; active low
Parity Bit Enable; active low (High = x16/32 Mode, Low = x18/36
I
FLXDrive Output Impedance Control
I
(Low = Low Impedance [High Drive], High = High Impedance [Low
Specifications cited are subject to change without notice. For latest documentation see http://www.gsitechnology.com.
Die B
x36
32Mb
Preliminary
GS8324Z18(B/C)/GS8324Z36(B/C)/GS8324Z72(C)
Functional Details
Clocking
Deassertion of the Clock Enable (CKE) input blocks the Clock input from reaching the RAM's internal circuits. It may be used to
suspend RAM operations. Failure to observe Clock Enable set-up or hold requirements will result in erratic operation.
Pipeline Mode Read and Write Operations
All inputs (with the exception of Output Enable, Linear Burst Order and Sleep) are synchronized to rising clock edges. Single cycle
read and write operations must be initiated with the Advance/Load pin (ADV) held low, in order to load the new address. Device
activation is accomplished by asserting all three of the Chip Enable inputs (E1, E2, and E3). Deassertion of any one of the Enable
inputs will deactivate the device.
Read operation is initiated when the following conditions are satisfied at the rising edge of clock: CKE is asserted low, all three
chip enables (E1, E2, and E3) are active, the write enable input signals W is deasserted high, and ADV is asserted low. The address
presented to the address inputs is latched into the address register and presented to the memory core and control logic. The control
logic determines that a read access is in progress and allows the requested data to propagate to the input of the output register. At
the next rising edge of clock the read data is allowed to propagate through the output register and onto the output pins.
Write operation occurs when the RAM is selected, CKE is active, and the Write input is sampled low at the rising edge of clock.
The Byte Write Enable inputs (BA, BB, BC, and BD) determine which bytes will be written. All or none may be activated. A write
cycle with no Byte Write inputs active is a no-op cycle. The pipelined NBT SRAM provides double late write functionality,
matching the write command versus data pipeline length (2 cycles) to the read command versus data pipeline length (2 cycles). At
the first rising edge of clock, Enable, Write, Byte Write(s), and Address are registered. The Data In associated with that address is
required at the third rising edge of clock.
Flow Through Mode Read and Write Operations
Operation of the RAM in Flow Through mode is very similar to operations in Pipeline mode. Activation of a Read Cycle and the
use of the Burst Address Counter is identical. In Flow Through mode the device may begin driving out new data immediately after
new address are clocked into the RAM, rather than holding new data until the following (second) clock edge. Therefore, in Flow
Through mode the read pipeline is one cycle shorter than in Pipeline mode.
Write operations are initiated in the same way, but differ in that the write pipeline is one cycle shorter as well, preserving the ability
to turn the bus from reads to writes without inserting any dead cycles. While the pipelined NBT RAMs implement a double late
write protocol in Flow Through mode a single late write protocol mode is observed. Therefore, in Flow Through mode, address
and control are registered on the first rising edge of clock and data in is required at the data input pins at the second rising edge of
clock.