Interconnects occupy upto 90% of the area in Reconfigurable Architectures and affect the speed and noise of the chip. This presentations gives briefs about interconnects, particularly in context of Reconfigurable Architecture (eg FPGAs)
2. Introduction
Chips are mostly made of wires called
interconnects, which form the physical
connection between two transistors.
Interconnects are basically metal layers that
are fabricated along with transistors, and it
depends on the technology and foundry the
number and characteristics of the layers.
5. Introduction
Dominant contributor to die area & cycle time
Support large functional density
Parallel data routing
FPGA’s place most of the area into
interconnect
7. The Area considerations
4-LUT is roughly 600Kl2 . The flip-flop and 16:1 LUT
multiplexor make up very little of this area, easily
less than 20Kl2 . estimates the area of the 4-LUT
multiplexor with flip-flop as 13K l2
The majority of the area associated with each 4-LUT
(97%), goes into programmable interconnect and its
configuration memory
The area required for the full LUT, including its
configuration memory, is less than 10% of the area of
the 4-LUT cell
8. The Delay Considerations
Vendors lump interconnect timing with evaluation
time of lookup table
Interconnect typically accounts for 80% of the path
delay
9. The Noise Considerations - Crosstalk
A capacitor does not like to change its voltage
instantaneously
A wire has high capacitance to its neighbor.
When the neighbor switches from 1→ 0 or 0→ 1, the wire
tends to switch too.
Called capacitive coupling or crosstalk
Crosstalk effects
Noise on non switching wires
Increased delay on switching wires
Crosstalk causes noise on non switching wires
11. Local interconnect
Lowest level of interconnect
Used for very short interconnects at device level
Higher resistivity
Do not travel long distances
Withstand higher processing temperatures
12. Semi Global interconnect
Used to connect devices within the block
Mid length wires for communication within a
block is placed in this category
Used because complexity of interconnects is
increasing
13. Global Interconnect
Upper level of interconnects
Mostly made up of aluminum
Used to connect long interconnects between
the blocks, including power, ground & clocks
Low resistivity
Length increases as transistors per chip
increases
17. The programmable routing between the Logic Blocks
consists of fixed metal tracks
Fixed metal tracks run horizontally and vertically,
and are organized in channels
Each channel contains the same number of tracks
Switch Block defines all possible connections
between channels
Flexibility of each Switch Block is a key to the
overall flexibility and the routability
Interconnects Routing
18. Types of Switch Blocks
Disjoint switch block Universal switch
block
Wilton switch block
- Two types of switches are used in switch block
19. The Connection Block defines all the possible
connections from a horizontal or vertical channel to a
neighboring logic block
The connections in the switch blocks and connection
blocks are made by programmable switches
A programmable switch consists of a pass transistor
controlled by
-a static random access memory cell (SRAM-
based FPGA)
-an anti-fuse (anti-fuse FPGAs)
-a non-volatile memory cell (floating gate
devices)
Interconnects Routing
20. SRAM cells based FPGAs - can be reprogrammed by
the end user as many times as required and are
volatile
Anti-fuse based FPGAs - can only be programmed
once and are non-volatile
NVRAM based FPGAs- floating gate technology - can
be reprogrammed and are non-volatile
Types of FPGAs based on Interconnects
22. Parasitic Resistance and Capacitance
The metal wires and antifuse contribute significant
parasitic resistance and capacitance
The parasitic resistance and capacitance result in
large signal propagation delay and power consumption
23. Rent’s Rule
In 1960,E.F Rent, IBM employee stated that
- Interfacial growth is driven by functional
increase in semiconductor chips
- Number of interconnections increases as a
power function of number of circuits with N gates
- No. of i/p and o/p Interconnections
- No. of gates in circuit
p – Rent’s Exponent (It denotes the degree of
interconnection complexity)
- typically for logic functions 0.5 < p < 0.7
25. Rent's rule is an empirical result based on observations
of existing designs
Less applicable to the analysis of non-traditional
circuit architectures
It provides a useful framework with which to compare
similar architectures.
26. Test yourself
What are the different types of Interconnects in RC
architectures?
What is Rent’s rule? How do you correlate it with
Moore’s law?