Xilinx FPGAs:Evolution and Revolution
Download
Report
Transcript Xilinx FPGAs:Evolution and Revolution
Xilinx FPGAs:Evolution and Revolution
Evolution results in bigger, faster, cheaper
FPGAs; better software with fewer bugs, faster
compile times; coupled with better technical
support.
Innovative architectural and circuit features with
advancements in design methodology - modular
team based design & internet based configuration
method: resulting in Revolution.
A Decade of Progress
Xilinx FPGAs : An Endless Journey
XC2000
7.5K, 50MHz
XC5200
0.5 micron
16K, 50MHz
50K-1M,
200MHz
4M,
250MHz
40K-8M,
420MHz
10M,
420MHz
1.5K
XC3000
XC3100
XC4000
0.5 micron
SPARTAN
0.35, 0.25 m
VIRTEX
0.22micron
SPARTAN-II
0.22 micron
VIRTEX-E
0.18micron
SPARTAN-IIE
0.18micron
VIRTEX-II
0.15 micron
VIRTEX-II
PRO
0.13micron
85K, 80 MHz
100MHz+
40K
200K, 200MHz
300K,
200MHz+
Xilinx FPGAs - Generic Features
High Performance at different voltages
Footprint Compatibility
- Devices within each family are compatible.
Low power consumption/high performance
Integrated Software
Technology independence
- EDIF, VHDL, Verilog, SDF interfaces.
XC2000
First FPGA Family from Xilinx.
Two members:
XC2064
1000 Gates
XC2018
1500 Gates
Ext. Crystal Oscillator.
No Tri State Buffers.
XACT 1.0 Development System.
XC2000
Max.Logic Gates
Max. CLB Flip-flop
Max. I/O Pins
Max. I/O Flip-Flop
1500
100
74
74
XC3000
Replaces TTL, MSI and other PLD logics.
Integrates complete subsystem into single
Package.
System clock Speed up to 50 MHz.
On-chip crystal Oscillator.
Low-Skew Clock Nets.
Over 20 different Packaging Options
Interface to popular design Environment like
Mentor, Cadence and View Logic.
XC3000 CLB
XC3100A
Ultra-high-speed Family with six members.
XC3195 in 22 X 22 CLB array size.
Compatible with XC3000.
Error checking of configuration bit stream.
XC4000 CLB
XC4000 Family Features
Synchronous Single and Dual-Port RAM
Internal Three-state buffers.
JTAG Boundary Scan
System performance to 80 MHz
0.5 µ SRAM Process Technology
XC4000 Sub-Families
Version
Max. logic
Gates
XC4000XL 3k-85k
XC4000EX 28k, 36k
XC4000E 3.0-25k
Max. I/O Voltage
448
320
256
3.3
5
5
XC5200
Low Cost FPGA Family.
System Features:
- Fast Arithmetic Functions
- High Performance Clock Network
- Highly routable
- Easy Pin Locking
- Fast wide Functions
- Three-state buffers
- JTAG
- Performance up to 50 MHz.
Resource Comparison
Xilinx Spartan/XL FPGAs
Xilinx 4000
Series
Heritage
Advanced Process
Technology
100 MHz+ performance
On-chip SelectRAM
Software v4.2i
Core solutions
Small die size
Low cost packaging
Low test cost
Total Cost Management
Total Cost Management
Leading edge process technology
Smallest die size of any FPGA with on-chip RAM
Focused package offering
Low-power architecture allows use of highest volume
plastic packages
Streamlined test flow
Lower cost test hardware
Built-in self test features and shorter test times
Optimized manufacturing flows
Spartan-XL Family
Advanced 0.35m Process
Transistor gates 0.35u
Allows 3.3 V supply
All other features 0.25u
Small size
Low capacitance
Performance
Chip
Combines 3.3 V operation with 0.25u benefits
Spartan Speed Grades
Higher speed grade = higher performance
XL-5
E-1
-4
E-2
-3
XC4000E
Spartan
XL-4
-3
-4
XC5200
Spartan-XL
What’s missing in Spartan?
No asynchronous RAM
Only RAM16(32)X1S, RAM16X1D, ROM16X1
No edge decoders
No DECODEx
No wired-AND
No WANDx or WOR2AND
Mode pins not usable as I/O
No MD0, MD1, MD2
Virtex - features
Densities from 50 K to 1M system gates.
System performance up to 200 MHz.
Multi-standard Select IO interfaces.
Built-in clock-management circuitry
- Four DLLs
- Four Low-skew global Clock Distribution Net
Hierarchical Memory System.
Dedicated Multiplier Support.
Spartan-II - features
Densities as high as 200K gates.
Streamlined features based on Virtex architecture.
Very Low Cost
LUT Distributed RAM and Block RAM support.
Dedicated Multiplier support.
4 DLLs
Virtex/Spartan-II CLB
1 CLB holds 2 slices
Each slice has two sets of
Four-input LUT
Carry & Control
Any 4-input logic function
Or 16-bit x 1 RAM
Or 16-bit shift register
Fast arithmetic logic
Multiplier logic
Multiplexer logic
Storage element
Latch or flip-flop
Set and reset
True or inverted inputs
Sync. or Async. Control
Virtex/Spartan-II DLLs Improve
Clock Networks
DLL1
Manage Multiple
System Clocks
Deskew
Clocks
on Chip
DLL3
Generate
Clocks
(Multiply,
Divide, or
Shift)
DLL2
Deskew
Clocks
on Board
Cascade
DLLs
DLL4
Convert
Clock
Levels
using
Select I/O
Delay locked loops synchronize on-chip and board level clocks
Virtex-E – what’s added?
Up to 4 million system gates
2-4X more Block RAM
8 DLLs
Differential I/O signaling (LVDS/BLVDS)
some new speed grades.
Spartan-IIE - features
Density 50-300K
Supports LVDS
4 DLLs
VCCINT – 1.8V
More speed grades than Spartan-II.
Less packaging Options.
Virtex-II
All Xilinx FPGAs contain the same basic resources
CLBs contain combinatorial logic and register resources
IOBs interface between the FPGA and the outside world
Programmable interconnect
Other resources
Three-state buffers
Global clock buffers
Boundary scan logic
Virtex-II devices contain additional resources
Block SelectRAM
Dedicated Multipliers
Digital Clock Manager (DCM)
CLB Tile
COUT
COUT
TBUF
TBUF
Slice S3
Slice S2
Switch
Matrix
SHIFT
Slice S1
Slice S0
CIN
Fast Connects
CIN
Slice Structure
Slice 0
LUT
Carry
PRE
D
Q
CE
CLR
LUT
Carry
PRE
D
Q
CE
CLR
Mult-AND and dedicated Multiplier too
SRL16
DDR Registers.
Fast Carry Logic
Digitally Controlled Impedance
Connecting Function Generator
F6
Slice S0
F5
Slice S1
F5
F7
Slice S2
F5
F6
Slice S3
F5
F8
CLB
Select I/O
Allows direct connections to external signals of varied voltages
and thresholds
Optimizes the speed/noise tradeoff
Saves having to place interface components onto your
board
Differential signaling standards
LVDS, BLVDS, ULVDS
LDT
Distributed and Block Select RAM
LUT
Slice
LUT
LUT
RAM16X1S
D
WE
WCLK
A0
O
A1
A2
A3
RAM32X1S
D
WE
WCLK
A0
O
A1
A2
A3
A4
RAM16X1D
D
WE
WCLK
A0
SPO
A1
A2
A3
DPRA0 DPO
DPRA1
DPRA2
DPRA3
Dedicated Multiplier Block
Eighteen-bit 2’s complement signed operation
Optimized to implement Multiply / Accumulate functions
Multipliers are physically located next to block SelectRAM
18 x 18
Multiplier
Output
(36 bits)
4x4 signed
~255 MHz
8x8 signed
~210 MHz
12x12 signed ~170 MHz
18x18 signed ~140 MHz
DCM
Up to twelve DCMs per device
Located on top and bottom edges of the die
Driven by clock input pads
DCMs provide:
Delay-Locked Loop (DLL)
Digital Frequency Synthesizer (DFS)
Digital Phase Shifter (DPS)
Digital Spread Spectrum (DSS)
Up to four outputs of each DCM can drive onto global
clock buffers
All DCM outputs can drive general routing
Challenges to accelerate Processing
performance
Very Large
Single Task
requires
Parallel
Processing
Types
of
Challenges
Multiple Tasks
Multiple Solutions
Parallel Processing Using
Multiple Processors
Parallel Processing in
Hardware
• High performance
• lower cost
• low complexity
Multiple Processors
on Multiple Tasks
• Specific task focus
• Scalable
Virtex-II Pro Addresses All
Processing Tasks
Fabric for parallel
Fabric for parallel
processing in hardware
processing in hardware
•Uunmatched Performance
•Uunmatched Flexibility
Up to four 300MHz PowerPCs
Up to four 300MHz PowerPCs
for multiple processing
for multiple processing
Virtex-II
Virtex-II
Pro
Fabric
Pro Fabric
Logic Capacity and Features
Virtex II Pro Leads all the way
Supply Voltage
XC4000 and Spartan families use a 5V supply.
The-XL families use 3.3 V supply.
Virtex and Spartan-II use 2.5V supply.
Virtex-E uses 1.8 V.
Virtex-II and Virtex-IIPro uses 1.5 V
Xilinx Development System
XACT 1.0; 2.0
... 6.0
Xilinx-NeoCAD
Merge
M 1.0
M2.0i....M5.1i
THE FUTURE…….
In 2005, FPGAs will be built on 70nm-Cu
process; will implement 50 million system
gates; with 2 billion transistors on-chip; with
10 layers of copper metal; with embedded
processors running at 1 GHz clock rate; with
direct interface to 10 Gbps serial data.
URLs
www.xilinx.com
www.fpga-faq.com
www.optimagic.com
www.datasheetlocator.com
Newsgroup comp.arch.fpga