lowest-cost FPGA

Johnson,

From the NuHorizons web page FAQ:

"14. How much will Spartan-3 devices cost? Volume pricing at the end of 2004 will be under $2.95 for the XC3S50 and under $12 for the XC3S1000, and under $100 for the XCS4000 (based on

250K unit quantities)."

I suggest you contact your local Xilinx rep for pricing on 10K per year qty.

XC3S50 == 50K logic cells.

Aust> Hi there,

For lowest cost, the Altera MAX II "CPLD" can give you good old fashioned FPGA functionality without the bells and whistles. The EPM1270 (smaller than the Xilinx XC3S50) is the first introduced with others scheduled for 2Q

2005. The single-piece pricing on Arrow is ugly now but the later, higher volume price might be a better fit for your timeframe.

I know I can get the performance out of the Xilinx Spartan-3 but I also know some implementations can get a better overall solution with a different device.

I don't have a good feel for the lowest cost FPGA from Lattice, but I'd look at them as well. Perhaps someone else can fill in this blank.

"Austin Lesea" wrote

You presumably misspoke, here 50K = 50,000 system gates (a.k.a. marketing gates, dog gates, what have you).

An XC3S50 contains 768 slices, or 1536 4-LUTs and FFs, or 1728 logic cells by that curious derating understood and beloved (and believed) only by Xilinx marketing. The rest of us just giggle. See also

and its links.

The smallest 3S device with ~20K LCs is an XC3S1500 (26,624 LUTs+FFs, 29952 "LC"s). The 3S1000 is close (15360 LUTs+FFs, 17280 "LC"s).

To my knowledge, nothing has been announced that provides 20 KLUTs for $5 in any quantity. (Not to mention the configuration memory.) I think the closest announced EasyPath device is something like ~$13(?) for an XCE3S1500 in quantity with ~$75K(?) NRE.

But Moore's Law will take us there ere long. Make it your ASIC, indeed!

Jan Gray

Hello,

I think that you first must clarify whether you need FPGA or CPLD, which differe in the concept, capacity, routing resources, and price. CPLD gives your design more security (does not need external EEPROM) and guaranteed timing (important for high-speed designs), whereas FPGA provides more capacity but less security and non-guaranteed timing. For CPLD, I always go with Lattice M4000 series, providing excellent fitting/re-fitting resources and capacity. I would not claim on the CPLDs from other manufacturers since I missed the comparison tracking since '98, however I would definately recommend to consider Lattice (BTW, they have free design tools). For FPGA, I go with Xilinx due to a historical issue, however I believe that there is actually a smoall difference with Altera FPGA.

--

Regards,
Pavel

"Johnson"  wrote in message 
news:b1ac2406.0411291426.2b7e0d4a@posting.google.com...
> Hi there,
>
> Could anybody please let me know what is the approximate price of the
> lowest-cost FPGA or CPLD with about 20K Lggic Cells, or even less? The
> quantum will be around 10K per year. We will start a very simple
> application, and I hope I can find some FPGA less than $5.
>
> Thanks in advance.
>
> Johnson

Jan,

Yes, I blew it.

"About a 3S1000 to get close to 20K logic cells (has 17,280)" is one response I already got last night from a friend. Sorry to have mixed the gates thing. Something that I, too, don't like about how we count. I much prefer just counting look up tables and flip flops (at least that way I am less confused that normal).

Aust> "Austin Lesea" wrote

I wish you luck in finding an FPGA at that price in any size at that quantity. My experience is that unless you are using a large enough quantity to get a vendor to pay for a new fab, you will be paying about $10 minimum. I think this has to do with the costs associated with testing.

It seems that FPGAs will never truly be jellybean parts and I expect the FPGA vendors want to keep it that way since there is very little margin in such low prices no matter how efficiently they build them. By entering a low margin business model they could even become like the SDRAM vendors and loose a little on each unit they sell, trying to make it up in the volume!

One other point, using gate counts to estimate chip size is pretty pointless, even for a rough estimate. The variation in actual gate count achieved varies so widely and how you measure gate count varies so widely that the metric is pointless. I guess it could get you to within an order of magnitude... $5 vs. $50 :)

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Thanks a lot, guys, I really got lots of information and insight.

Johnson