This list is for discussion of the design and implementation of field-programmable gate array based processors and integrated systems. It is also for discussion and community support of the XSOC Project (see http://www.fpgacpu.org/xsoc).
|
Tom Cantrell wrote: > I seem to recall that Jan said his design simulated with better performance > on Spartan than XC4xxx - can't remember the explanation for the > non-intuitive result (Spartan is positioned by Xilinx to be 'lower end' > than XC4xxx). XC4xxx covers a lot of ground nowadays, XV/XLA/XL: Family: XC4xxxXV XC4xxxXLA XC4xxxXL Core Voltage: 2.5V 3.3V 3.3V Core Logic Cells: 8K-17K 1K-6K 128-6K CLB prop: 1.2/0.9 1.1/0.9 1.6/1.1 (CLB prop is max combinational delay, slowest/fastest speed grades. Most of a total path delay is elsewhere, but CLB prop is a decent relative index for comparison. A core logic cell is a 4-LUT+FF. XC4xxx and Spartans have 2 per CLB, Virtex and Spartan-II have 4.) So does Spartan-II/SpartanXL/Spartan: Family: XC2Sxx XCSxxXL XCSxx Core Voltage: 2.5V 3.3V 5.0V Core Logic Cells: 400-4K 200-2K 200-2K CLB prop: .7 1.1/1.0 1.6/1.2 And Virtex-E/Virtex: Family: XCVxxxxE XCVxxxx Core Voltage: 1.8V 2.5V Core Logic Cells: 2K-65K 2K-25K CLB Prop: .42/.47 .8/.6 So it depends on which XC4xxx and which Spartan we're talking about. Xilinx is positioning Spartan families for low cost in massive volumes, and low power, targeting consumer products like MP3 players (take a look at XAPP169). It's not necessarily quite so cheap in small quantities. Virtex is high speed, high capacity, targeting industrial applications like network routers. XC4xxxXLA is still a good choice for smaller industrial applications, like a PCI interface, but Spartan-II will probably replace it, just as Virtex is replacing XC4xxxXV. XC4xxxXL is in the tail end of its product lifecycle, which makes it affordable for hacker use. By the way, to give you an idea how exciting it is to have access to cheap FPGAs and tools these days, at work we were recently quoted around $2000 for an XCV2000E in small quantities this quarter. That's actually not unreasonable given the value the part delivers, and the fact that its price will come down steeply over the next couple of years. That's the kind of money FPGAs always used to cost. I remember when the XC4010XL in our XESS boards cost > $1000. A few years ago, before Student Edition, the Xilinx tools cost thousands, unless you were a promising customer, which was no good for the individual experimenter. HDL synthesis normally costs tens of thousands of dollars, for the latest and greatest industrial-strength tools like Synopsys Design Compiler. An FPGA board with Synopsys and Xilinx tools for a few hundred dollars is one hell of a deal. This is the beginning of another Golden Age of Hardware Hacking. --Mike |