Re: high number of multipliers / low cost

ryan_usenet@xxxxxxxxx wrote:

Hello,
I need the highest possible number of multiplication operations per
second at low cost. I know that several factors affect the overall
performance, but since I have no idea which FPGA chips might be worth
to be considered, I'd like to ask what you think is the chip with the
lowest ratio

R=(prize of chip)* (delay time)/(number of multipliers)

18x18bit multipliers seem to be quite common, so lets assume this
design for the estimate.

For example for the Spartan XC3S1000 (~60$, 24 multipliers, 4ns delay)
I have
R= 10$ per (Billion multiplications/s). The Cyclone EP2C70 (~230$, 150
multipliers, 4ns delay) has R=6.13$ per (Billion multiplications/s).

Do other FPGAs exist that are maybe specialized for multiplication-
intensive tasks and which therefore are much cheaper?

Best regards,
Ryan


What is your application? There is usually more than one way to approach a problem. For example, distributed arithmetic is an elegant solution to handling a sum of products with constant or nearly constant coefficients that does a great job at compacting the area required for fabric multipliers. The precipitation radar design on my website gallery ( http://www.andraka.com/precip_radar.htm ) does about 82 multiply-accumulates per clock cycle at 133 MHz, which works out to over 10 billion multiplies/sec in an FPGA that has zero hardware multipliers (XCV1000) and is rather small and slow compared to current FPGA offerings. In order to find a possible alternative approach, you need to disect your algorithm to see if there are other constructs that might get you the performance you want in a reasonable amount of hardware.
.

Relevant Pages

  • Re: using FPGA in control field
    ... suggest, the FPGA is great for creating high performance control functions ... One piece of advice - if you can afford it, use an FPGA with embedded multiplier blocks and plentiful RAM. ... In pretty much all control functions you're MAC'ing process variables and coefficients. ... You can reduce this somewhat by creating long multipliers, or off-line calculation of LUT's for partial products, but on my next design, I'm moving to Cyclone II just for the embedded multipliers. ...
    (comp.arch.fpga)
  • Re: using FPGA in control field
    ... the FPGA is great for creating high performance control functions ... One piece of advice - if you can afford it, use an FPGA with embedded multiplier blocks and plentiful RAM. ... In pretty much all control functions you're MAC'ing process variables and coefficients. ... You can reduce this somewhat by creating long multipliers, or off-line calculation of LUT's for partial products, but on my next design, I'm moving to Cyclone II just for the embedded multipliers. ...
    (comp.arch.fpga)
  • Re: Spartan-3A XC3S1400A development board?
    ... Would anyone happen to know where if could purchase an FPGA development ... board with the Xilinx Spartan-3A XC3S1400A FPGA on it (or a socketed ... multipliers on the 1400A, but the starter board Xilinx sells through ... presents, it appears, 32 for your needs) and isn't geared toward I/O optimized needs like the Spartan-3A but logic optimized. ...
    (comp.arch.fpga)
  • Re: FPGA history
    ... The article describes how to build multipliers, state machines, etc. ... They could be configured using jumpers to ... Unfortunately I have no pointer to that either. ... FPGA technology advances during the recent years (say 10 years - or even ...
    (comp.arch.fpga)
  • Spartan-3A XC3S1400A development board?
    ... Would anyone happen to know where if could purchase an FPGA development ... board with the Xilinx Spartan-3A XC3S1400A FPGA on it (or a socketed ... multipliers on the 1400A, but the starter board Xilinx sells through ...
    (comp.arch.fpga)
Loading