PCI 64/66 fpga eval boards

"Keith" <krw@att.bizzzz> wrote in message 
news:MPG.1ede86986bc24a65989a95@News.Individual.NET...
> In article <4474ba59$0$294$7a628cd7@news.club-internet.fr>,
> John@nospam.com says...
>> Isn't ISA a dead end considering PCI, yet it is still used !!!!

Hopefully not in *new* designs. :-)  As far as I can tell, ISA only sticks 
around in industrial PC boards so that already working systems can be 
maintained, which makes perfect sense.  (There's at least one company out 
there still making replacement PDP-11 boards, after all...)

> ISA is still used because it's trivial to hack.  PCI is anything
> but.

Only if you're planning to actually implement the PCI interface yourself... 
which really only makes sense if you're planning to try to squeeze the last 
once of performance out of the bus.  For most designs, using interface ICs 
from the likes of PLX Technology make PCI pretty darned friendly to implement 
to (their chips have options to treat the "card" side of the bus as anything 
from reasonably sophisticated down to dumb-as-a-PIC).  Cypress has an IC that 
places a dual-port RAM (and a couple of FIFOed mailboxes) across the PCI bus, 
also making it trivial for even the "dumbest" logic to interface.

USB is quite similar -- unless you're after an education, for the vast 
majority of designs the various USB interface ICs (e.g., from FTDI) or 
USB-based microcontrollers with supplied low-level code (from Cypress, Atmel, 
Microchip, etc) let you worry about the unique aspects of your design rather 
than interfacing with Yet Another Bus.

For FPGAs or standard-cell logic, there are lots of USB and PCI cores out 
there from the usual suspects.

---Joel Kolstad

In article <1279jbqev8c163d@corp.supernews.com>, 
JKolstad71HatesSpam@yahoo.com says...
> "Keith" <krw@att.bizzzz> wrote in message 
> news:MPG.1ede86986bc24a65989a95@News.Individual.NET...
> > In article <4474ba59$0$294$7a628cd7@news.club-internet.fr>,
> > John@nospam.com says...
> >> Isn't ISA a dead end considering PCI, yet it is still used !!!!
> 
> Hopefully not in *new* designs. :-)  As far as I can tell, ISA only sticks 
> around in industrial PC boards so that already working systems can be 
> maintained, which makes perfect sense.  (There's at least one company out 
> there still making replacement PDP-11 boards, after all...)
> 
> > ISA is still used because it's trivial to hack.  PCI is anything
> > but.
> 
> Only if you're planning to actually implement the PCI interface yourself... 

What's so hard to understand about "hack"?
 
> which really only makes sense if you're planning to try to squeeze the last 
> once of performance out of the bus.  For most designs, using interface ICs 
> from the likes of PLX Technology make PCI pretty darned friendly to implement 
> to (their chips have options to treat the "card" side of the bus as anything 
> from reasonably sophisticated down to dumb-as-a-PIC).  Cypress has an IC that 
> places a dual-port RAM (and a couple of FIFOed mailboxes) across the PCI bus, 
> also making it trivial for even the "dumbest" logic to interface.

Sure, but that's hardly hacking.  I've used PLX chips and they're 
certainly not trivial to use.  Even given that the chip simply 
works, the board layout is 100x harder than ISA.  Programming is 
equally hard (comparatively).  Any damned fool can build an ISA 
channel card with a few 74xx gates.  PCI is *hard* even with PLX 
doing the heavy lifting.

> USB is quite similar -- unless you're after an education, for the vast 
> majority of designs the various USB interface ICs (e.g., from FTDI) or 
> USB-based microcontrollers with supplied low-level code (from Cypress, Atmel, 
> Microchip, etc) let you worry about the unique aspects of your design rather 
> than interfacing with Yet Another Bus.

Certainly, but USB ain't ISA either.

> For FPGAs or standard-cell logic, there are lots of USB and PCI cores out 
> there from the usual suspects.

Sure, but FPGAs or standard-cell aren't a coupla 74xx buffers 
either.  ISA was a piece of cake for an idiot (and still lives in 
the ATA connector for those who want to play ;). 

-- 
  Keith

Thanks.

Thanks guys for these interesting posts...

Why is ISA still out there ? Look at the PowerPC architecture, there's ISA 
in it. Look at any PC - there's ISA in it (even if the bus does not come out 
on connectors on the mobo it's in the chipset). It's still there for legacy 
issues. Same reason we (well less and less it's true) carry on some of the 
initial limitations imposed by MD-DOS on our windows systems.
When a new std comes out, it usually sticks around some time. Look at the 
VME bus, still out there ! When those guys from the military make a new 
equipement, they have to be capable to insure that the technology will still 
be out there in 20 years (in most Xtreme cases).
Some never really did make it through (IBM_PS2 ... SUN_SBUS and a few 
others).

Rgds,
John

"krw" <krw@att.bizzzz> a &#4294967295;crit dans le message de news: 
MPG.1edee8d2bcf4487c9896d0@news.individual.net...
> In article <1279jbqev8c163d@corp.supernews.com>,
> JKolstad71HatesSpam@yahoo.com says...
>> "Keith" <krw@att.bizzzz> wrote in message
>> news:MPG.1ede86986bc24a65989a95@News.Individual.NET...
>> > In article <4474ba59$0$294$7a628cd7@news.club-internet.fr>,
>> > John@nospam.com says...
>> >> Isn't ISA a dead end considering PCI, yet it is still used !!!!
>>
>> Hopefully not in *new* designs. :-)  As far as I can tell, ISA only 
>> sticks
>> around in industrial PC boards so that already working systems can be
>> maintained, which makes perfect sense.  (There's at least one company out
>> there still making replacement PDP-11 boards, after all...)
>>
>> > ISA is still used because it's trivial to hack.  PCI is anything
>> > but.
>>
>> Only if you're planning to actually implement the PCI interface 
>> yourself...
>
> What's so hard to understand about "hack"?
>
>> which really only makes sense if you're planning to try to squeeze the 
>> last
>> once of performance out of the bus.  For most designs, using interface 
>> ICs
>> from the likes of PLX Technology make PCI pretty darned friendly to 
>> implement
>> to (their chips have options to treat the "card" side of the bus as 
>> anything
>> from reasonably sophisticated down to dumb-as-a-PIC).  Cypress has an IC 
>> that
>> places a dual-port RAM (and a couple of FIFOed mailboxes) across the PCI 
>> bus,
>> also making it trivial for even the "dumbest" logic to interface.
>
> Sure, but that's hardly hacking.  I've used PLX chips and they're
> certainly not trivial to use.  Even given that the chip simply
> works, the board layout is 100x harder than ISA.  Programming is
> equally hard (comparatively).  Any damned fool can build an ISA
> channel card with a few 74xx gates.  PCI is *hard* even with PLX
> doing the heavy lifting.
>
>> USB is quite similar -- unless you're after an education, for the vast
>> majority of designs the various USB interface ICs (e.g., from FTDI) or
>> USB-based microcontrollers with supplied low-level code (from Cypress, 
>> Atmel,
>> Microchip, etc) let you worry about the unique aspects of your design 
>> rather
>> than interfacing with Yet Another Bus.
>
> Certainly, but USB ain't ISA either.
>
>> For FPGAs or standard-cell logic, there are lots of USB and PCI cores out
>> there from the usual suspects.
>
> Sure, but FPGAs or standard-cell aren't a coupla 74xx buffers
> either.  ISA was a piece of cake for an idiot (and still lives in
> the ATA connector for those who want to play ;).
>
> -- 
>  Keith 

To each his own.

PCI is just a vehicle for me to get data from the CPU to my logic.  It
could be anything else for that matter.  Like it has been said earlier
in the post, there are many vendors with standard pci controllers that
cost less than a low end spartan.  I therefore, would like to see a
board that uses these of the shelf PCI semiconductors.  It woudl not be
very productive and cost effective to integrate 3rd party PCI core into
the fpga.  Not sure with PCI Express, but it certainly isn't as easy
and cheap to use as PCI.

Hi Keith,

"krw" <krw@att.bizzzz> wrote in message 
news:MPG.1edee8d2bcf4487c9896d0@news.individual.net...
> What's so hard to understand about "hack"?

Ah, sorry, I was thinking "hack" in the "relatively sophisticated endeavor 
given very few resources... and you might have to crack open a book or two..." 
sense (like the guy who did a full-up HID USB device using only bit-banging 
with some low-end AVR microcontroller) rather than the "Make magazine" sense 
(which is closer to "kinda cool things that, yeah, your beer-guzzling neighbor 
down the street won't be doing anytime soon, but that smart high school kid 
next door could probably pull off"). :-)  I see your point more clearly now.

---Joel

"John Aderseen" <John@nospam.com> wrote in message 
news:44755bcf$0$295$7a628cd7@news.club-internet.fr...
> Why is ISA still out there ? Look at the PowerPC architecture, there's ISA 
> in it. Look at any PC - there's ISA in it (even if the bus does not come out 
> on connectors on the mobo it's in the chipset).

To a certain extent pretty much everybody who ever had to hang a handful of 
peripherals off of a bus has designed their own bus "standard," it's just that 
some are more sophisticated than others and only a small handful ever become 
known outside of a given design gruop.  The guys at nVidia, VIA, SiS, etc. 
have been enjoying themselves coming up with new busses between their north 
and south bridge chipsets for years now... :-)

In article <1148542032.255347.136790@i39g2000cwa.googlegroups.com>, 
fpgabuilder-groups@yahoo.com says...
> To each his own.
> 
> PCI is just a vehicle for me to get data from the CPU to my logic.  It
> could be anything else for that matter.  Like it has been said earlier
> in the post, there are many vendors with standard pci controllers that
> cost less than a low end spartan.  I therefore, would like to see a
> board that uses these of the shelf PCI semiconductors. 

They're out there.  Look for prototyping boards.  I had one for the 
PLX-9054 some time back.  Programming them still isn't trivial (as 
ISA is).

> It woudl not be
> very productive and cost effective to integrate 3rd party PCI core into
> the fpga.  Not sure with PCI Express, but it certainly isn't as easy
> and cheap to use as PCI.

I came to the same conclusion a few years back.  Why use a hunk of 
an *expensive* (at the time about $1K) Virtex-E when a $25 PLX-9054 
works out of the gate, sorta. 

-- 
  Keith

John,

The link does not clearly mention PCI 32/33 or PCI 64/66.  Looks quite
similar to the one from http://www.4dsp.com/PCI.htm .

What's your experience been with it?

Thanks.
-sanjay

Sanjay,

I do not renember if this board handles PCI64/66. I can ask Hunt Engineering 
if you want. However, I am not really sure this is what you're looking for. 
This is a system I am throwing away for a few euros because I am tired of 
seing it sleep in my attick !
What exactly is it you are trying to do ?
Those TIM fpga modules from Sundance are great but they are tied to the TIM 
std which may not be the best way to start playing around with fpgas. 
Modules from alpha-data are great and are PMC modules (meaning you can stick 
them in your PC easily or in any cPCI system) but they are expensive.
Whatever it is I can point you to people from these companies if ever you 
are interested but I think that you should start of by describing your exact 
needs.

Thanks,

Rgds,

John


"fpgabuilder" <fpgabuilder-groups@yahoo.com> a &#4294967295;crit dans le message de 
news: 1148578538.513026.295340@g10g2000cwb.googlegroups.com...
> John,
>
> The link does not clearly mention PCI 32/33 or PCI 64/66.  Looks quite
> similar to the one from http://www.4dsp.com/PCI.htm .
>
> What's your experience been with it?
>
> Thanks.
> -sanjay
>