Date post: | 05-Apr-2015 |
Category: |
Documents |
Upload: | mitzi-langer |
View: | 105 times |
Download: | 0 times |
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 1 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Part 4
Fundamentals in Computer Technology
Computer Architecture
Slide Sets
WS 2012/2013
Prof. Dr. Uwe BrinkschulteM.Sc. Benjamin Betting
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 2 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Technology trends: things are getting smarter
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 3 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Technology trends: networked systems of the future
NetworkedSystems
pervasive computingdisappearing computer
ambient intelligence
ubiquitous computing
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 4 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Technology trends:heterogeneous hardware-software-systems (HW/SW)
System on Chip Processor-Core, FPGA, RF,Bluetooth ... + Software
Heterogeneity also in the environment (application)
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 5 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Technology trends:cyber physical systems (CPS)
Integration of physical systems and networked computing
In classical embedded systems, the physical environment is controlled by the computer
In cyber physical systems, the physical environement and the computer(s) closely interact and cooperate
Examples: power grids, networks of autonomoues vehicles, air traffic control, …
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 6 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
System-on-Chip
ASICRF
ADC
DataLink
Memory
DSP CPU
AMP
DAC
Clk
Power Management
Communication
Antenna
Sensor
Cable
Control
Video
Audio
Transmitter
Power
RF
ADC
DataLink
Memory
DSP CPU
AMP
DAC
Clk
Power Management
Bus / InterfaceSensors
Control
Video
Audio
Transmitter
Power
Physical Interfaces
BinaryInterfaces
FPGA/FPFAASIC
RF
ADC
DataLink
Memory
DSP CPU
AMP
DAC
Clk
Power Management
Communication
Antenna
Sensor
Cable
Control
Video
Audio
Transmitter
Power
RF
ADC
DataLink
Memory
DSP CPU
AMP
DAC
Clk
Power Management
Bus / InterfaceSensors
Control
Video
Audio
Transmitter
Power
Physical Interfaces
BinaryInterfaces
FPGA/FPFA
System on Chip (SoC)Functional and technology aspects
Technology Aspects: → Process Combinations necessary
Analog/Digital Systems: → Components, Interfaces + Technologies
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 7 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
A
AD/DA
AD/DA
AD
/DA
AD
/DA
D
An analog component with a digital processor coreEmbedded digital processors are mainly used in analog environments. Life science andtechnical applications are mainly analog. Therefore, an analog to digital conversion andvice versa is necessary.
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 8 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Microprocessor
Important component of allmodern analog and digitalapplications.
It became the basic measurefor the technological progressin VLSI.
It became the workhorseof all modern ITapplications
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 9 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Modern processor chips
Intel Pentium Processor
System-on-Chip(Bluetooth SoC: Eynde et al., Alcatel, 2001)
Analog Devices ADSP 21060
Intel Pentium IV IBM Power PC 750
Microphotographs of processor chip layout
Altera FPGA with ARM-core
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 10 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
4004
8080
8086
8028680386
80486
P III
1M
10M
100K
10K
1975 1980 1985 1990 1995 2000
Pentium™
2005 2010
100M
1000 M
P IVCMOSstill state of the art
The Moore curve#
tran
sist
ors
This prediction of Gordon Moore demonstrates the progress of technology in the lastdecades. The complexity of integrated circuits (VLSI) doubles every 18 month.
year
Silicon will be the basic material for the next years.
Core 2
Core i7
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 11 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
The scaling (c)
1999 2001 2003 2005 2008 2011 2014
0,05
0,1
0,15
0,2
year
2000 0,182001 0,182002 0,132003 0,132004 0,132005 0,12008 0,072011 0,032014 0,01
Strukture size[µm]
The area scales with c2
Signal delay of active components scales with c
Signal delay of passivecomponents will be nearly constant
Signal delays are dominatedby the delay of the wires(passive components)
wire-centered design insteadof only logic optimized designor better a combination of both
Structuresize [µm]
c
year
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 12 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
A top down look at computer technology
Architecture Level
Microarchitecture Level
Register Transfer Level
Gate Level
Transistore Level
Charge Level
…
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 13 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
A top down look at computer technology
Computer
Architecture Level
Instruction Set Architecture, Memory Sizes, Clock Frequency, …
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 14 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
A top down look at computer technology
Computer
Microprocessor, CPU
control unitfunctional unit
(datapath)
memory(program, data)
input/output
connection (bus)
Microarchitecture Level
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 15 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
A top down look at computer technology
Register Transfer Level
state register
nextstate
outputfunction
control unit
instruction(program)
data register
functionalunits
data in
data path
datapath
control
CPU
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 16 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
functional unit
A top down look at computer technology
Gate Level
&
&
1
1 e 1
e 2
e 3
a
3
0
0
0
1
10
0
00
00
0
1
1
1
11
1
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 17 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
A top down look at computer technology
Transistor Level
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 18 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
A top down look at computer technology
Charge Level
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 19 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Planar process technology
Passivation (field oxide)
oxide 4
oxide 3
oxide 2
oxide 1
substrate
metal layer 3
metal layer 2metal layer 2
metal layer 1
Polysilicon
transistor
capacitancemetal to
substrate
capacitancemetal to
metal
oxide = SiO2
Delay times
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 20 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Modeling of signal delays in wires for the connection of active components (gates) in chip-layouts
Voltage at node nj for a given input voltage UE
)()(1 1
n
j
j
ii
jjEn Rdt
duCUtU
Delay td at node n:
n
j
j
ijji
t
nE uCRdttUUd
1 10
))((
Solution: Approximative solution of the differential equation
2
)1(00
nnCRtd for Ri = R0 Vi Ci= C0 Vi per 1 µm of wire length
current through thecapacitance at node j
RnR1 R2
UE C1 C2Cn
un
DifferentialEquation:
Wire:
Model:
1µm 1µm
Total resistanceat node j
Delay times
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 21 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Typical values for resistance, capacitance and the resulting delay times for integrated wires
tdD 6,3RD 7CJN 1,8n-diffusion
tdP 1,3RP 17CPF 0,15polysilicon
tdm 0,001RM 0,02CMF 0,1metal
delay time ns
per 1mm 3µm
resistance
/µm
capacitance
fF/µm
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 22 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
300 Ω 300 Ω 300 Ω
R1 R2 R3
measure point 1 measure point 2 measure point 3 measure point 4
UE
C1 C2 C3350fF 350 fF 350 fF
Example of a real wire model
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 23 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Plot of the node voltage at different measure points per wire length
measure point 1
measure point 2
measure point 3
measure point 4
measure point 4
measure point 3
measure point 2
measure point 1
4,0 V
2,0 V
0 V0 ns 2 ns 4 ns 6 ns 8 ns 10 ns
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 24 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
The clock
Example:- 10 GHz clock frequency (0,1 ns clock period) - 30 mm distance between two components means 10 clock cycles on the wire (silicon)
The clock skew in this exampleis 1 ns. It is too high for sequentialsynchronous circuits.
The clock skew has to be considered in the design phaseand /or has to be avoided byarchitectural solutions.(clock tree, wave pipelining, etc.)
The increasing size of modern chips (chip area) means, that also the connections between the components on the chip become longer.
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 25 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Memory gap
wirecapacity memory
capacity
memory transistor
Typical design of a DRAM cell
Access time is dominated by charging and decharching of the memory capacity:
ta = Rtransistor Cmemory capacity
C is limited above 10 fF to avoid data loss by alpha particles
R is limited by the area of the memory transistor
=> ta is limited
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 26 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Memory gap
The memory capacity scales tothe square with the size of thechip
As seen before, access time doesunfortunately not scale this way
Therefore a memory gap exists.
Increasing clock speed generatesan increasing memory gap.
Caches are a good solution but theycan bridge the gap only unsufficientbecause of the limited locality of typical programs.
Because of the memory gap, anincreasing clock speed results notautomatically in a higher execution-speed of programs.
Memory
CPU
memorygap
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 27 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Unbalanced von Neumann
CPUcaches, ...
vN bottleneck
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 28 of 35 – Prof. Dr. Uwe Brinkschulte, Prof. Dr. Klaus Waldschmidt
Reduction of power and energy consumption is a big issue today
On high end systems, heat dissipation has to be reduced
Modern mobile embedded systems need the reduction to increase battery lifetime
=> Tradeoff between high performance and low power/energy consumption
Main ways to reduce power and energy consumption
Reduction of clock frequency Reduction of supply voltage Optimization of microarchitecture
Power consumption
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 29 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Power consumption and clock frequency
As seen before, CMOS only consumes power when switching
Therefore, in modern gate technologies the energy consumption is mostly proportional to the clock frequency
P ~ f
Reduction of clock frequence means reduction of power consumption,
but as well a reduction of the system performance
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 30 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
The power supply voltage
The power supply voltage cannot easily be reduced under 1-0,5 V.
1999 2001 2003 2005 2008 2011 2014
voltage High Performance [V]
voltage Low Power [V]
0,5
1,0
1,5
year
1999 1,8 1,52000 1,8 1,52001 1,5 1,22002 1,5 1,22003 1,5 1,22004 1,2 0,92005 1,2 0,92008 0,9 0,62011 0,6 0,52014 0,6 0,3
Voltage High
Performance [V]
Voltage Low Power
[V]
year
The reduction of supply voltage implies a reduction of max. clock frequency.
P ~ U2
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 31 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Possible approaches:
• Reduction of external bus activities (stay local)
• Static Power Management (sleep instructions)
• Dynamic Power Management (control unit deactivates non-used parts of the microarchitecture)
• Increase of code density (saves memory space and cycles)
Power consumption and microarchitecture
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 32 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
Example: reducing memorypower consumption
This disadvantage of a good memory hierarchy is a high degree in energy consumption.
The on-chip data - and instruction caches need itself appr. 25 % of the total power of a processor chip.
CPU
kernel instructions
other instructionsInstruction
Cache
KernelMemory
(small and fast)Off-Chip memory
common address space
A cache oriented solution as an example for a power aware microarchitecture.
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 33 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
SIA-Roadmap
1999 340 23,820002001 340 47,620022003 372 95,220042005 408 1902008 468 5392011 536 15232014 615 4308
yearchip area
[mm²]transistors/ chip [Mio]
1999 1250 902000 1486 1002001 1767 1152002 2100 1302003 2490 1402004 2952 1502005 3500 1602008 6000 1702011 10000 1742014 13500 183
yearfrequency
[Mhz]power con-
sumption [W]
1999 2001 2003 2005 2008 2011 2014
chip area [mm²]
transistors/chip [Mio]
frequency [MHz]
power consumption [W]
year
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 34 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
The challenge (limit) of chip technology in the future
The production of chips with layoutstructures in dimensions under 0.1 µmbecame very difficult with respect tolithography.
Therefore, only about 10% of today’s chips are produced using latest technology
Chips of more than 300 mm2 chip areawill include one or more faults in average, caused only by the technologyprocess.
Increasing cost of chipproduction
$
Decreasing yield of chipproduction
Hier wird Wissen Wirklichkeit Computer Architecture – Part 4 – page 35 of 35 – Prof. Dr. Uwe Brinkschulte, M.Sc. Benjamin Betting
0,001
0,01
0,1
1
10
100
1000
10000
100000
1000000
1975 1980 1985 1990 1995 2000 2005 2010
Bandwidth in Byte/s
Clock in MHz
Tape/Disk in MByte
OS in kByte
Memory in MBit
Design in MTrans.
Development of different parameters incomputer architecture in the past