Frequently Asked Questions for Virtex Series FPGAs
Q. Why do you say, "Xilinx is redefining the FPGA"?
A. Until Virtex series, the measuring criteria for an FPGA has focused
on density and performance. Virtex series both significantly exceeds these
current standards and offers more. In developing a device capable of implementing
one million gates, devices need more than just gates and routing. System
features such as advanced clock management; direct interface to existing
and future voltage and signal standards; multi-granularity high bandwidth
memory integration; and power management are critical features for what
we call the Xilinx Virtex technology employed in Virtex devices. Offering
this technology across from 50,000 to one million gates redefines what
is possible with an FPGA.
Q. What do you mean by "system level integration capabilities" or
"system level features"?
A. Terms such as system on a chip (SOC) and system level integration (SLI)
have been used to define a device or design methodology that allows multiple
functions of a system to be integrated into a single device. Such devices
must offer not only the density required to implement the system, but
also the capability of integrating these multiple functions easily with
high performance. The system level features of Virtex devices allow this
ease of integration. Delay locked loops (DLLs) allows high performance
communication between the different functions. Select IO allows direct
connection to external devices, high-speed busses and backplanes. SelectRAM+
memory allows efficient integration of high bandwidth memory functions.
Thermal management capabilities provide constant monitoring of environmental
conditions to allow systems to run at peak performance.
Q. How can an FPGA address system-level requirements?
A. For any device to address system-level requirements, they must provide
high density, high performance and an ease of integrating multiple functions
in a single device. Virtex FPGAs address these requirements by providing
up to one million gates with a routing architecture that allows true 100+
MHz performance. In addition to providing the necessary resources and
routing, Virtex series offers key system features in clock management,
memory hierarchy, highly flexible IO, and power management allowing Virtex
to be the center of any high performance system.
Q. How are these requirements being addressed now?
A. Until the Virtex series, ASIC devices, specifically standard cell devices,
have been used to address the requirements for system-level integration.
Virtex devices will offer the time to market and flexibility advantages
of an FPGA to the system-level design community.
Q. Who are the decision-makers for designing in a Virtex device? Are
these different people than other FPGA decision-makers?
A. The Virtex series redefines what is possible for an FPGA. Current FPGA
designers will like the increased density, increased performance, and
increased feature set of Virtex devices. In addition, Virtex will get
the attention of decision-makers that have not had the option to consider
FPGAs in the past because of its limited functionality. At many companies,
a System Architect will separate key system components from glue logic
functions. Historically, FPGAs get their first consideration for implementation
in the glue logic functions. Because of performance, density or feature
requirements, the key system components are commonly earmarked for ASIC
design. With Virtex FPGAs, System Architects can consider the time to
market and flexibility advantages of FPGAs for all components in a system.
Q. What are some of the new markets and applications that Virtex will
enable for Xilinx?
A. Virtex devices will find homes in the traditional strong markets for
FPGAs: networking, telecommunication, high-end computing, industrial and
wireless markets. These are key markets for Virtex devices since the time
to market and flexibility advantages of FPGAs have always been valued
here. Within these strong markets, Virtex devices offer new application
possibilities for FPGAs. The Virtex family is already being considered
for applications that previously required ASICs. Some examples are SONET
data stream processing for OC-3, OC-12 and OC-48 standards, main processing
engine for massively parallel compute network, variety of protocol controllers
and switch matrix for next generation network equipment, and backplane
controllers for ultra high speed board to board communication.
Q. What has allowed you to bring down the cost for such a robust feature
set?
A. Virtex devices are designed for optimal efficiency using advanced 0.22
micron, 5 layer metal process technology. This delivers the highest number
of gates for any given silicon area. The new features introduced with
the Virtex architecture also were designed for efficient implementation
using the state of the art process technology. Combining the highest density
process with an architecture optimized for this process results in the
highest number of gates per die area even with all the powerful new features.
Q. What is the advanced 0.22-micron process and is it different than
the 0.25-micron process?
A. The 0.22-micron process is the second-generation 0.25-micron process
from Xilinx representing a 10 percent shrink over the first generation.
Xilinx is still the only programmable logic vendor shipping on 0.25 micron
and has extended its leadership by delivering devices on its second generation
before other vendors deliver their first. This aggressive process results
in the most efficient use of die area, fastest devices, and the lowest
power consumption.
Q. Which features are unique to Xilinx?
A. The system level features in Virtex devices include multiple delay
locked loops (DLL), multi-standard I/O, hierarchy of RAM capabilities,
and thermal management. Each feature by itself, although similar to features
in other programmable devices, is unique to the industry in the specific
implementation.
- Virtex devices are the first to provide high performance, working
clock management circuitry with its DLL. Others have introduced devices
with phase locked loops (PLL) that have not worked or only worked to
lower frequencies.
- Having four DLLs on a single device is also unique and allows for
complete system clock management as compared to single chip clock management.
- The multi-standard I/O is implemented such that even standards not
defined today are likely to be supported with the Virtex device. Others
have offered a finite number of options limited to different voltage
standards.
- With Virtex devices, both voltage and swing characteristics are independently
programmable.
- Virtex devices are the first to embed both distributed and block memory
that will allow the right type of high bandwidth memory. Distributed
memory has proven very valuable in high-speed data processing systems
where a small amount of wide data is needed. The block memory in Virtex
provides extremely efficient data buffering on and off chip and temporary
storage for processing kbytes of information found in graphics and networking
applications. Having full dual port, two independent read/write ports,
is also unique to the Virtex block memory. This will allow for bidirectional
buffers that do not suffer performance degradation seen with dual port
memories from others.
- Thermal management control is a first in the industry.
Q. Why is temperature monitoring important?
A. Virtex devices are inherently low power (milliwatts of quiescent power)
because of the 2.5V operation. The largest device has 75 million transistors-10
times the Pentium II-and the architecture supports performance over 100
MHz which can lead to tens of watts. By offering temperature monitoring,
Virtex devices can be designed to perform right up to the limits of the
packaging technology. No performance is compromised to account for worst
case conditions that may never happen.
Q. What is the DLL useful for?
A. Clock synchronization is critical in high performance applications.
The four Virtex delay locked loops (DLLs) provide a mechanism to remove
all clock skew in a system as well as minimize clock to output delays.
In ultra high performance applications, clock skew and clock to output
delays can easily be more than 50 percent of the critical path. With Virtex
DLLs these values can be reduced to nearly zero effectively doubling possible
system frequency.
Q. What are the advantages over PLL?
A. Compared to a phase locked loop (PLL), a DLL has a number of advantages.
The Virtex DLL is a complete digital implementation that simplifies the
processing and the future scalability of the DLL. The Virtex DLL does
not accumulate jitter because each clock edge in essence restarts the
circuitry. Finally, the Virtex DLL has been proven to operate at frequencies
over 200 MHz while competing PLL solutions in programmable logic devices
either don't work or are limited to frequencies below 100 MHz.
Q. What is the difference between all these I/O standards?
A. The different I/O standards supported by Virtex come in two flavors.
The common standards such as LVTTL, LVCMOS2, PCI33 and PCI66 operate at
either 3.3V or 2.5V and swing about a point that is common. The more exotic
standards like GTL, GTL+, SSTL, HSTL, AGP and CCT are specifically designed
for low power and high performance. The voltage levels and the swing points
differ and are designed for specific applications. GTL and GTL+ standards
are for low power high-speed backplane interconnect. SSTL, HSTL and CTT
standards are specific to certain high-speed memory devices. AGP is specific
to high performance graphics. To support all of these with single device,
Virtex SelectIO offers programmability to the output voltage waveforms
as well as programmability to the swing point.
Q. Do I need to order different Virtex devices for the different standards?
A. No, every I/O on every Virtex device is capable of supporting every
standard. To interface to a specific standard, a designer drops in the
type of I/O buffer required. For example, to implement a GTL input buffer,
a designer simply specifies IBUF_GTL. The configuration of the I/O buffer
is done during device configuration and the specific I/O detail is contained
in the bitstream generated by the Xilinx software solutions.
Q. Does the Virtex family support applications for reconfigurable
computing? Is it partially reconfigurable?
A. Virtex devices do support reconfigurable computing. Each device can
be reconfigured to provide different functions at different times. A specific
high-speed configuration port allows bitstreams to be downloaded into
Virtex devices at a rate up to 400 Mbit/second, which allows even the
largest device to be completely reconfigured in milliseconds. Virtex devices
will also provide for partial reconfiguration, meaning a portion of the
device can be changed while the rest of the device continues to operate.
Q. What is the utilization percentage for the Virtex devices?
A. The five layer metal process allows for massive amounts of routing.
The design for the routing of Virtex is to allow full utilization of the
million-gate device. Because of the segmented routing nature of the Virtex
device, all family members will have a routing structure that is designed
to support high performance for the million-gate device. Designers will
directly benefit from this routing with very fast compile times, high
performance from push button design flows, and full utilization of all
device resources.
Q. Does the gate-counting method differ from the traditional Xilinx
method? What percentage of RAM use are you assuming in these gate counts?
How do you account for the Block RAM?
A. The gate counting used on Virtex devices is consistent with the system
gate counting used for Spartan and XC4000 devices. The basics of the system
gate count equation are each logic cell used as logic provides 12 system
gates; each logic cell used as distributed memory provides 64 system gates
(16 bits x 4 gates/bit). Since larger devices typically require more memory
the percentage of logic cells used as distributed memory ranges from 5-30
percent across the family. With Virtex additional capabilities add to
the system gate count. Each DLL used provides 7000 system gates and each
logic cell used as a programmable delay element provides 112 system gates.
Finally, block memory adds to the gate capacity of Virtex and each bit
of block RAM memory provides four system gates. Up to 40 percent of block
memory usage is factored into the system gate count.
Q. Is there JTAG support in Virtex devices?
A. Yes
Q. Will all future FPGA families from Xilinx be derived from the Virtex
architecture?
A. Xilinx invented the FPGA and continues to drive the standard platforms.
Virtex represents the fourth generation of FPGA standards, preceded by
the XC2000, XC3000 and XC4000 architectures. New families will use the
Virtex architecture as their basis.
Q. When will software support be available for Virtex?
A. Virtex software support is available today. Xilinx and other EDA vendors
including Exemplar, Synopsys and Synplicity offer breakthrough technology
in performance via VHDL and Verilog synthesis. The latest software releases,
the Xilinx Alliance Series and Foundation Series 1.5 software, offer full
support for the Virtex family of FPGAs. New timing driven compile time
benchmarks have been set with these tools by offering typical rates of
200,000 gates/hour for timing driven place and route.
Q. What is "vector-based interconnect"?
A. "Vector-based interconnect" conveys the result of the massive routing
capacity offered by Virtex devices and represents the next generation
of the Xilinx segmented-routing structure. With five layers of metal,
the routing architecture of Virtex has been designed such that the minimum
amount of metal is required to implement a given interconnect. This means
that delays are closely estimated as a function of the straight line distance
(vector length) of the interconnect. The designer benefits of vector-based
interconnect are: 1) predictably high performance due to routing efficiency;
2) consistent performance across devices due to consistent routing architecture;
3) fast compile times due to the massive amount of routing tracks; 4)
and a cores-friendly architecture. Add period
Q. What make Virtex series so "cores friendly"?
A. The suitability of Xilinx FPGAs to a high performance, highly predictability
and easy to use core methodology is further extended by the features of
the Virtex architecture. While the vector-based interconnect allows more
cores with higher performance requirements to be implemented in HDLs than
in older architectures, the fact that the routing scheme is a next generation
segmented routing structure means that the benefits of Smart-IP are still
applicable. Smart-IP allows the core designer to leverage the absolute
maximum performance for the fastest cores without giving up predictability
or ease of customization. The Virtex memory capabilities further extend
the range of cores that can be easily implemented in an FPGA. SelectRAM
distributed memory is leveraged for small FIFO's, data buffers and memories
at the places that they are needed in the design. When larger memory blocks
are required, the new SelectRAM+ memory blocks provide a wealth of size
and bus width options to support single-port or multi-port memory for
larger storage needs. Add to this the ability to interface to high speed
SDRAMs using the SelectI/O and you have a memory hierarchy that gives
Virtex devices the capability of powering even the most demanding memory-based
applications. This wealth of enhancements ranging from the improvements
that allow faster multipliers to the system level features such as the
DLLs and clock management capabilities has resulted in Virtex being the
most core friendly architecture ever. Its feature set is unparalleled
and is especially important at the one million gates density level. It
will enable cores implementation that is not possible in any other device.
Q. When will the 64 bit/66 MHz PCI core be available?
A. Virtex is capable of delivering a 64 bit/66 MHz PCI compliant design.
We are in the process of productization of this core. Its availability
will be announced later.
Top of page | Return
to Virtex Press Kit page
|