1 distributed loss compensation for low-latency on-chip interconnects class presentation for...
TRANSCRIPT
![Page 1: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/1.jpg)
1
Distributed Loss Compensation for
Low-latency On-chip Interconnects Distributed Loss Compensation for
Low-latency On-chip Interconnects Class Presentation For
Advanced VLSI Design Course
Instructor: Dr.Fakhraie
Presented By: Fahimeh Alsadat Hoseini
Winter 2006University of Tehran
Dept. Electrical and Computer Enginerring
Major Reference: ISSCC 2006 / SESSION 21 / ADVANCED CLOCKING, LOGIC AND SIGNALING TECHNIQUES / 21.7
![Page 2: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/2.jpg)
2
Outline
Scaling trends and motivation Prior work on low-latency repeater-less links by
exploiting transmission-line behavior Negative-impedance converter (NIC) System architecture
Transmitter design Receiver design
Measurement results Summary and conclusions References
![Page 3: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/3.jpg)
3
Scaling Trends - ITRS Roadmap
On-chip wire delay scaling more slowly than gate delay.
Impact of scaling is worst on global wiring.Jose et al., ISSCC, 2006.
![Page 4: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/4.jpg)
4
Motivation Wire delays (D) grow quadratically with wire length,
- Unacceptably great for long wires.
- Wire bandwidths,which are inversely proportional to D, degrade.
Latency is controlled with repeater insertion which allows linear scaling of delay with length.
- Break long wires into N shorter segments
- Drive each one with an inverter or buffer
- Optimal number of repeaters can be determined to minimize delay
- Repeaters, often inserted with CAD tools, consume a significant fraction of on-chip power and area in microprocessors.
2LCRDwirewire
![Page 5: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/5.jpg)
5
Outline
Scaling trends and motivation Prior work on low-latency repeater-less links by
exploiting transmission-line behavior Negative-impedance converter (NIC) System architecture
Transmitter design Receiver design
Measurement results Summary and conclusions References
![Page 6: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/6.jpg)
6
Nearly ‘speed-of-light’ wires
The time-domain solution is:
• In this case, A is a constant. Gamma, the propagation constant, provides information about the characteristics of this line. The imaginary part of gamma, denoted by beta, is inversely related to the phase velocity, and the real part of gamma, denoted by alpha, is the attenuation constant.
R. Chang ,Thesis, 2002.
![Page 7: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/7.jpg)
7
Nearly ‘speed-of-light’ wires
LC dominatedregion
RC region
R. Chang ,Thesis, 2002.
![Page 8: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/8.jpg)
8
Nearly ‘speed-of-light’ wires
Take advantage of the inductance-dominated high-frequency regime of on-chip interconnects
Peak phase velocity is speed of light in SiO2
Reduce low-frequency spectral components of the signal which introduce ISI and lag LC dominated response
Chang et al., JSSC, 2003.
![Page 9: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/9.jpg)
9
Outline
Scaling trends and motivation Prior work on low-latency repeater-less links by
exploiting transmission-line behavior Negative-impedance converter (NIC) System architecture
Transmitter design Receiver design
Measurement results Summary and conclusions References
![Page 10: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/10.jpg)
10
Distributed Loss Compensation
Used in long-distance telephone network before the introduction of optics for long-haul communications
Clock distribution networks- Standing wave oscillators, [O’ Mahony et al., JSSC 2003]
- Rotary traveling-wave oscillator arrays, [Wood et al., JSSC 2001]
Distributed amplifiers use similar ideas to extend the unity-gain bandwidth
![Page 11: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/11.jpg)
11
Negative Impedance Converter
Pole ≈ -gm/(2C) ; zero ≈ 1/(2RC)
Match frequency-dependent loss characteristicsJose et al., ISSCC, 2006.
![Page 12: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/12.jpg)
12
NIC Attenuation Compensation -
With NICs
Without NICs
Increasing R,C=50 fF
Increasing R,C=600 fF
A larger cap C increases the amount of loss compensation at higher frequencies
Negative leads to instability which can lead to excessive ringing or oscillations
Jose et al., ISSCC, 2006.
![Page 13: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/13.jpg)
13
Latency Comparison
NIC links have lower latencies at higher widths due to transmission-line behavior of interconnects for widths > 2 µm
For very small widths (large R), the interconnect is predominantly in the RC domain
Lat
ency
(p
s)
Width (µm)
Length=14mm
NIC Link
Optimally buffered link
Jose et al., ISSCC, 2006.
![Page 14: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/14.jpg)
14
Power Comparison
Power consumed increases rapidly for widths < 4µm due to the large number of NIC elements required
Increasing bit energy in the optimally repeated case is due to large number of repeaters needed to drive the additional C
En
erg
y (p
J/b
it)
Width (µm)
Length=14mm
NIC Link
Optimally buffered link
NIC Attenuation Compensation -
Jose et al., ISSCC, 2006.
![Page 15: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/15.jpg)
15
Outline
Scaling trends and motivation Prior work on low-latency repeater-less links by
exploiting transmission-line behavior Negative-impedance converter (NIC) System architecture
Transmitter design Receiver design
Measurement results Summary and conclusions References
![Page 16: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/16.jpg)
16
Test-Chip Components
3-Gbps link in 0.18-µm technology with a 1.5-GHz system clock 17-bit LFSR for generating PRBS and an error counter for obtaining
BER Far-end and near-end waveforms obtained by pico-probing
5mm
PLL
1.67mm
SRAM
LFSRDriver
Receiver
Serpentineserial link
SRAM Error Counter
Jose et al., ISSCC, 2006.
![Page 17: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/17.jpg)
17
System Architecture
Data Bandwidth : 2 / clock period [bits/sec]
Clock period : 1.5 GHzJose et al., ISSCC, 2006.
![Page 18: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/18.jpg)
18
Transmitter Design
Driver consists of M1-2 and termination resistor RT
Predrivers use pseudo-nmos logic Id sets the bias point of the NICs Jose et al., ISSCC, 2006.
![Page 19: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/19.jpg)
19
Modified StrongARM latch:• Low-swing differential
receiver• Small aperture time for
high data-rate Line termination:
• N-well resistor with value of 2Zo
• Excessive capacitive loading at receiver inputs can introduce ISI
Receiver Design
CalibrationCaps
CalibrationCaps
70 mV far-end swing
Jose et al., ISSCC, 2006.
![Page 20: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/20.jpg)
20
Receiver Sampling-point Calibration
Calibration is performed at the receiver end Clock skew compensation between transmitter and receiver Link latency compensation
Jose et al., ISSCC, 2006.
![Page 21: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/21.jpg)
21
Outline
Scaling trends and motivation Prior work on low-latency repeater-less links by
exploiting transmission-line behavior Negative-impedance converter (NIC) System architecture
Transmitter design Receiver design
Measurement results Summary and conclusions References
![Page 22: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/22.jpg)
22
Measurement Results -
obtained from measured S-parameters The NICs cause noticeable bandwidth reduction at frequencies
beyond ≈ 9 GHz The NICs contribute towards a significant reduction in from ≈
50 MHz to 7 GHzJose et al., ISSCC, 2006.
![Page 23: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/23.jpg)
23
Measurement Results -
obtained from measured S-parameters Phase velocity decreases ( increases) at high frequencies
Jose et al., ISSCC, 2006.
![Page 24: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/24.jpg)
24
Outline
Prior work on low-latency repeater-less links by exploiting transmission-line behavior
Negative-impedance converter (NIC) System architecture
Transmitter design Receiver design
Measurement results Summary and conclusions References
![Page 25: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/25.jpg)
25
Summary
Throughput 3 Gbps 3 Gbps 3 Gbps
Clock frequency 1.5 GHz 1.5 GHz 1.5 GHz
Width/
Length
8 µm /
14 mm
0.3 µm /
14 mm
8 µm /
14 mm
Link-latency 12.1 ps/mm 55 ps/mm 18.6 ps/mm
Number of NICs/repeaters
7 18 5
Power consumed
0.16 pJ/bit/mm 0.17 pJ/bit/mm0.5
pJ/bit/mm
Distributed losscompensation (DDR)
Optimally repeatedlink (DDR)
Optimally repeatedlink (DDR)
Jose et al., ISSCC, 2006.
![Page 26: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/26.jpg)
26
Conclusions
As technology scales, on-chip latencies are increasingly becoming a bottleneck for on-chip performance
Optimally repeated RC delays represent latencies that are as much as 3 X those determined by the speed of light in SiO2
Repeaters consume a growing fraction of power and silicon area
Using distributed loss compensation with NICs leads to
arbitrarily long links with a significant latency and energy/bit/mm
advantage over optimally repeated RC links
![Page 27: 1 Distributed Loss Compensation for Low-latency On-chip Interconnects Class Presentation For Advanced VLSI Design Course Instructor: Dr.Fakhraie Presented](https://reader036.vdocument.in/reader036/viewer/2022062322/56649f0d5503460f94c21d99/html5/thumbnails/27.jpg)
27
References
• A. P. Jose and K. L. Shepard, “Distributed Loss Compensation for Low-latency On-chip Interconnects,” IEEE International Solid-State Circuits Conference, 2006.
• A. P. Jose, G. Patounakis and K. L. Shepard, “Near Speed-of-light Onchip Interconnects using Pulsed Current-mode Signaling,” Symp. VLSI Circuits, June, 2005.
• R. T. Chang, et al, “Near speed-of-light signaling over on-chip electrical interconnects,” IEEE J. Solid-State Circuits, vol. 38, no. 5, May, 2003.
• R. Chang , “ Near Speed-of-Light On-Chip Electrical Interconnects,” A DISSERTATION SUBMITTED TO THE DEPARTMENT OF ELECTRICAL
ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY, November 2002.