Title: Memory Design Considerations That Affect Price and Performance
1Memory Design Considerations That Affect Price
and Performance
- Bill Gervasi
- Technology Analyst, Transmeta
- Chairman, JEDEC Memory Parametrics
- bilge_at_transmeta.com
2Posed at the Last Conference
- Why will DDR-I at 400 MHz data rate be a
boutique solution? - Why will DDR-II at 400 MHz data rate be a
mainstream solution?
3Agenda
- JEDEC/Industry Roadmap
- Factors for Market Acceptance
- Difficulties in Achieving 400 MHz
- Factors Affecting Cost
- Wild Cards What Can Change?
4RAM Evolution
5400MB/s
DDR667
4300MB/s
MainstreamMemories
DDR533
3200MB/s
DDR II
DDR400
2700MB/s
Simple,incrementalsteps
DDR333
2100MB/s
DDR I
DDR266
5Factors for Market Acceptance
- Industry Focus
- Number of Competing Suppliers
- JEDEC standard
- Laws of Physics
6Industry Focus
- The JEDEC roadmap represents the industry focus
for mainstream products - DDR-I tops out at 333 MHz data rate
- DDR-II starts at 400 MHz data rate
- This DOES NOT mean that DDR-I at 400 MHz data
rate will not ship in volume - It DOES mean that there will be price premiums
for this speed bin
7What do I mean by Focus?
- There is serious work to hit 400 MHz
- Vendor interoperable solutions
- Mix and match module configurations
- Signal integrity analysis
- We are counting picoseconds
- No JEDEC standard yet proposed for DDR-I at 400
MHz data rate
8For Example
- How we are getting more refined in timing
analysis with DDR-II - The Charge Transfer Model for input timing
measurement and derating
9DDR-I Input Timing Model
CLOCK
CLOCK
Setup
Hold
INPUT
Timing derating by input signal slew
rate 1.0V/ns base value 0.5V/ns base value
50ps 0.4V/ns base value 100ps This got us
through DDR333
The Old Way
10However
- This simplified model was good enough for DDR333
data rates, but leaves picoseconds of available
timing lying around needed for 400!!! - DDR266 Data Setup/Hold 750 ps DDR333 Data
Setup/Hold 600 ps DDR400 Data Setup/Hold 400
ps DDR533 Data Setup/Hold 350 ps
Cant waste time!!!
11Focus on Input Timing
The New Way
INPUT
tINT
tEXT
tT
- DDR-II Charge Transfer Timing Model
- All inputs have a slew rate dependent aspect tEXT
and an independent aspect tINT - Summing tEXT tINT gives input transition time
tT - Transition time tT has min and max values
- Differential input transitions inherently
different
12tEXT for Slow Slew Rate, Single Ended
AT Charge to Transition
VIHAC VSAT
VIHDC
VREF
tEXT
tINT
VILDC
VILACVSAT
13tEXT for Fast Slew Rate, Single Ended
VSAT Saturation Voltage
AT ASAT AADD
VIHAC VSAT
ASAT Charge to Saturation
VIHDC
AADD Charge after Saturation
VREF
tSAT
VILDC
tEXT
tINT
VILAC VSAT
14tEXT for Slow Slew Rate, Differential
AT Charge to Transition
VIHAC VSAT
VIHDC
VREF
VILDC
VILAC VSAT
tINT
tEXT
15tEXT for Fast Slew Rate, Differential
AT ASAT AADD
VIHAC VSAT
VIHDC
VREF
VILDC
VILAC VSAT
tEXT
tINT
16Focus on Timing
CLOCK
tTmin
CLOCK
Setup
tTmax
INPUT
- DDR-II Charge Transfer Timing Model
- Setup tTmax of input - tTmin of reference
17Focus on Timing
CLOCK
tTmax
CLOCK
Hold
tTmin
INPUT
- DDR-II Charge Transfer Timing Model
- Hold tTmax of reference - tTmin of input
18How does this help?
- The Charge Transfer Model gives a higher accuracy
for setup and hold relationships - It also provides a way to accurately describe
derating for input slew rate - These models are negotiated with all suppliers to
define an industry standard
19DDR-II Input Derating Tables
Clock (mV/ps avg)
Strobe (mV/ps avg)
SETUP
SETUP
2.0
0.5
1.0
2.0
0.5
1.0
0.5
0.5
Data (mV/ps)
Addr (mV/ps)
¾
¾
0
0
1.0
1.0
¾
¾
¾
¾
¾
¾
2.0
2.0
Strobe (mV/ps avg)
Clock (mV/ps avg)
HOLD
HOLD
2.0
0.5
1.0
2.0
0.5
1.0
0.5
0.5
0
¾
Addr (mV/ps)
0
¾
Data (mV/ps)
1.0
1.0
¾
¾
¾
¾
¾
¾
2.0
2.0
20Derating Using Charge Transfer
- Accuracy from derating both signals and
references - Result is a two dimensional matrix relating
inputs their references - Identified inherent asymmetries in derating of
setup hold when mixing single ended with
differential signals - Memory module mixes impact slew rates
- The Charge Transfer model controls system cost by
enabling more complex timing analysis
21Charge Transfer on DDR-I?
- This model would also help design high speed
DDR-I systems - However, the work to retrofit this to DDR-I needs
to be done to benefit from it
22DDR-II Improvements
- DDR-II introduces technical improvements that
reduce the cost of achieving high speeds - Prefetch 4
- Differential data strobe
- I/O Calibration
- Lower I/O Voltage
- On-Die Termination
23Prefetch 4
24Moving to the Next Level
- Todays SDRAM architectures assume an inexpensive
DRAM core timing - DDR I (DDR200, DDR266, and DDR333) prefetches 2
data bits increase performance without
increasing timing costs - DDR II (DDR400, DDR533, DDR667) prefetches 4 bits
internally, but keeps DDR double pumped I/O
25Prefetch 2 Versus 4
CK
READ
data
Core access time
Prefetch 2
Prefetch 4
Costs
Column cycle time
Essentially free
Costs
26Prefetch Impact on Cost
- By doubling the prefetch depth, cycle time for
column reads writes relaxed, improving DRAM
yields
Pre-fetch
DDR Family
Data Rate
Cycle Time
Starts to get REAL EXPENSIVE!
2
266
7.5 ns
DDR-I
2
333
6 ns
Comparable to DDR266 in cost
2
400
5 ns
4
400
10 ns
DDR-II
4
533
7.5 ns
4
667
6 ns
27Why Not Prefetch 8?
- DIMM width 64 bits
- PCs use 64b, servers use 128b (2 DIMMs)
- 64 byte prefetch okay for PC, but
- 128 byte prefetch for servers wastes bandwidth
- DDR-II must service all applications well to
insure maximum volume ? minimum cost
28Differential Data Strobe
29Differential Data Strobe
- Just as DDR added differential clock to SDR
- DDR II adds differential data strobe to DDR I
- Transition at the crosspoint of DQS and DQS
30Differential Data Strobe
VREF
DQS
DQShigh time
DQSlow time
Normal balanced signal
VREF
DQS
DQShigh time
DQSlow time
Mismatched Rise Fall signal
Error!
31Differential Data Strobe
DQS
VREF
DQS
DQShigh time
DQSlow time
Normal balanced signal
DQS
VREF
DQS
DQShigh time
DQSlow time
Mismatched Rise Fall signal
Significantly reduced symmetry error
32I/O Calibration
33I/O Calibration
- Balance pull-up and pull-down driver strength
- Reduces timing errors from signal asymmetry
- Insures signal rise and fall times are similar
Data
DRAM
Controller
Data
Reference
341.8V I/O Voltage
351.8V Signaling
2.5V
VDDQ
1.8V
1.60V
VDDQ
VIHac
1.43V
1.15V
VIHdc
1.25V
VIHac
VREF
1.03V
VILdc
VIHdc
1.07V
0.90V
VREF
VILac
VILdc
0.77V
0.90V
VILac
0.65V
VSS
VSS
0V
SSTL_18
SSTL_2
36I/O Voltage Impact on Timing
- Assume 1mV/ps edge slew rate
- DDR-I 700 mV (VIL?VIH) 700 ps
- DDR-II 500 mV (VIL?VIH) 500 ps
- Helps meet the need for speed
- Signal integrity is a serious challenge at DDR-I
and 400 MHz data rate
37On-Die Termination
38On-Die Termination
VTT VDDQ 2
Data
Controller
DDR-I
DRAM
Data
Controller
DDR-II
DRAM
VDDQ 2
VDDQ 2
- Reduces system cost while improving signal
integrity
39What Can Change?
40Wild Cards
- 100 yield of 5 ns cycle time cores (magic?)
- Industry gets excited about engineering DDR-I at
400 MHz - DDR-II slow transition from schedule or price
- Feature creep
- Die penalties
- DRAM guys trying to make money for once
41Conclusions
- DDR-I at 400 will ship in volume but
- not likely to cross over /bit
- Industry focus is on transition to DDR-II for
400 MHz data rates
42Thank You