Actel Fusion FPGAs

I read a couple of threads on the Actel Fusion FPGAs from early this
year saying they were delayed and not to expect them for up to 6
months.  It is now 3 months later and I was wondering if anyone had
heard anything more from Actel on these parts.

I was looking at the web site for their combination Fusion and ARM core
they call the M7Fusion.  It is shown in two parts, the M7AFS600 with
7000 tiles free of 14,000 or the M7AFS1500 with 32,000 tiles free of
about 38,000.  I am not so familar with the Actel tiles.  Are they
roughly comparable to a 4 LUT with a FF?  Or are they more limited?

I don't suppose anyone would have an idea of the price of the M7AFS600
and M7AFS1500 devices?  I guess if they are not shipping samples yet,
they don't really have a handle on pricing.

I am still waiting for the data sheet to download.  Are these parts 5
volt tolerant?  I see the analog section is good for +-12 volts, I
would expect the digital section could do 5 volts!

rickman wrote:
> I am still waiting for the data sheet to download.  Are these parts 5
> volt tolerant?  I see the analog section is good for +-12 volts, I
> would expect the digital section could do 5 volts!

I got the datasheet and it is not 5 volt tolerant on the digital I/Os.
Very odd.  They use a process that can build a 12 volt tolerant analog
section and can't make the digital section 5 volt tolerant.  :^(

I don't know about the Fusion tile but the Actel Proasic 2 family tile
seemed a lot more limited than when compared to a xilinx spartan 3
logic cell say.

>From what I recall, the Proasic logic element could be a flop OR a
lut(i think only 3 input from memory) not both at the same time like
the spartan 3. Also no carry chain or XOR/AND arithmetic bits and
pieces that  the spartan 3 has extra.

Regards
Andrew

rickman wrote:
> I read a couple of threads on the Actel Fusion FPGAs from early this
> year saying they were delayed and not to expect them for up to 6
> months.  It is now 3 months later and I was wondering if anyone had
> heard anything more from Actel on these parts.
>
> I was looking at the web site for their combination Fusion and ARM core
> they call the M7Fusion.  It is shown in two parts, the M7AFS600 with
> 7000 tiles free of 14,000 or the M7AFS1500 with 32,000 tiles free of
> about 38,000.  I am not so familar with the Actel tiles.  Are they
> roughly comparable to a 4 LUT with a FF?  Or are they more limited?
>
> I don't suppose anyone would have an idea of the price of the M7AFS600
> and M7AFS1500 devices?  I guess if they are not shipping samples yet,
> they don't really have a handle on pricing.
>
> I am still waiting for the data sheet to download.  Are these parts 5
> volt tolerant?  I see the analog section is good for +-12 volts, I
> would expect the digital section could do 5 volts!

well, take PA3 price, multiply by 2 and then add 120 USD for ARM
license (in small qty)
its the best estimate for today.

so qty 1 price over 200 USD

Antti

rickman schrieb:
> I am not so familar with the Actel tiles.  Are they
> roughly comparable to a 4 LUT with a FF?  

No, a tile can be a 3 LUT _or_ a FF. I compared an Actel ProASIC and a
Spartan3 for my applications (controller in mechatronics) and estimated
a ratio of 6.5 tiles to one Slice.This is simple to comprehend then you
need two tiles to implement a 4 LUT and one for the FF, and this fits
roughly the ratio.
Furthermore you should know, Fuisons are ProASICs extend by some analog
circuitry.

Bye Tom.

Antti wrote:
> well, take PA3 price, multiply by 2 and then add 120 USD for ARM
> license (in small qty)
> its the best estimate for today.
>
> so qty 1 price over 200 USD

Something is wrong with that.  $120 per usage of a core that I can buy
as a chip for under $5 with tons of peripherals seems a bit rediculous.


Where did you get the PA3 price multiplier?  I was not initially
interested in the Fusion since typically combined analog and digital do
not do justice to either.  But I have an application where combining
some basic analog functions such as a multichannel ADC with digital and
potentially an MCU in a Flash based part would be very advantageous.
It doesn't have to be an ARM, pretty much any MCU could do the job.  I
could use the Fusion without the ARM if the price point is ok.  But a
2x multiplier seems a bit steep.

Personally I have always felt it was not a good idea to try to
duplicate a complex CPU in FPGA fabric.  Both Xilinx and Altera have
very nice CPU cores that are tightly coupled to their architectures and
so efficiently use the fabric.  But implementing an ARM in fabric is
not efficient either in size or speed.  It would be much more efficient
to add the ARM as a hard peripheral like they do on the Virtex 2P
parts.   But they seem to charge a price premium even then.  So the
simple CPU cores seem to be a good tradeoff if your processing is not
complex.  NIOS II seems to be especially well matched to the Cyclone II
chips!

the ARM license agreement that Actel has doesnt work at all for 'small
guys'

the ARM license fee per device is 1USD for qty 100,000 for smaller
quantities it scales down.

ARM enabled license fee (eg price for the pre-programmed encryption
key!) for PA3 in qty <=10 is about 120USD
it better than linear scale what would be 100,000 USD but its still
ridicilous. But thats the way the ARM deal was made.
to my knowledge Actel can not offer the ARM enabled devices in small
quantity without the scaled license fee

Antti

Antti wrote:
> the ARM license agreement that Actel has doesnt work at all for 'small
> guys'
> 
> the ARM license fee per device is 1USD for qty 100,000 for smaller
> quantities it scales down.
> 
> ARM enabled license fee (eg price for the pre-programmed encryption
> key!) for PA3 in qty <=10 is about 120USD
> it better than linear scale what would be 100,000 USD but its still
> ridicilous. But thats the way the ARM deal was made.
> to my knowledge Actel can not offer the ARM enabled devices in small
> quantity without the scaled license fee

Given these factors
a) High cost of license
b) high cost of the fabric to deploy the core

why would _anyone_ design this into a system ?

It seems like someone in marketing's idea, that never hit an
engineering or accounting reality check....

There are surely, much better ways for Actel to approach SoftCPUs

-jg

Jim Granville wrote:
> 
> Given these factors
> a) High cost of license
> b) high cost of the fabric to deploy the core
> 
> why would _anyone_ design this into a system ?
> 
> It seems like someone in marketing's idea, that never hit an
> engineering or accounting reality check....
> 
> There are surely, much better ways for Actel to approach SoftCPUs

to add some more examples :

... Such as this item, in todays news ?

http://www.eeproductcenter.com/embedded/brief/showArticle.jhtml?articleID=187203441

Well Actel is really trying only to fry the big fish only. They dont
care about small players at all.

PA3/Fusion with ARM in qty may make sense, but... well at those
quantities it may as well be ASIC or that thing from ST that offers
fixed ARM with user ASIC block.

When actel announced the PA+ I immediatly obtained the kit, only to be
disappointed, then waited for PA3 for 3 years to come?? And now waiting
for Fusion kit, it was ordered DEC 2005 (from stock delivery promise),
as of last info it may arrive in June/July. 2007 I hope

Antti