Memory width on Spartan-3 boards

[Edmond Cote]

Hi Simon,

I've been investigating lately *exactly* the same issues you have,
however I finally made a decision and ordered the NuHorizon's board.

While the width of the SDRAM was a deterrent, simply placing two read
cycles back-to-back should solve the obvious problem.

For curiousity's sake, are you working on a commericial or academic
project?, I am also working on implementing a MIPS based
system-on-a-chip. If you pick up that particular board, maybe we could
help each other out, since the kit includes next to no documentation.

Please contact me by e-mail if you are interested.

Thanks,

Edmond

Simon wrote:

So, I have a 32-bit processor design which I'd like to move to the
implementation stage, and I was wondering which of the
currently-available boards have 32-bit wide memory... I don't really
fancy designing my own because I've never ventured near 4-layer boards,
and if I was making my own, I'd want one of the FG456 packages, and
presumably you'd have to get that professionally soldered anyway...

As far as I can tell:

Xilinx starter kit
------------------

Pros: 32-bit wide RAM
LCD/Leds for debugging
PS2/VGA outputs
Price: $99

Cons:
Only 1 MByte of RAM
Uses the '200 part not the '400
No ethernet PHY
Fixed oscillator freq. (the CPU goes faster
Not many user-IO's available


AVnet Spartan-3 evaluation kit
------------------------------

Pros: '400 part used
Two oscillators, socketed
Lots of IO's available
Ethernet, VGA, PS2, RS232, Leds etc.
Could potentially be a PCI card

Cons: Only 1 MByte RAM
Not clear if the memory is 32-bit wide
Price is $399


Memec DS-KIT-3SLC400-PAC
------------------------

Cons: Has no memory on-board, enough said.


Nu Horizons Spartan3 board
--------------------------

Pros: Uses the '400 part, but only in the '208 package
Has D2A and A2D onboard
Has Flash RAM
Has LCD (4x24) as well as leds,buttons etc.
Has spare oscillator socket for > 20MHz operation
Price - $164

Cons: SDRAM appears to be 16-bit wide
Only has ~20 user io due to '208 package



So, nothing is perfect [grin], The AVNet one may be the best of the
bunch, despite being the most expensive, so long as it has 32-bit wide
RAM. I'd really appreciate it if someone who already has the board could
tell me



For the record (in case any board companies are listening my ideal
board would be something like:

- FG456 Spartan 3 '400 part
- Lots of user-IO, some with pin headers not obscure connectors
- 32+ bit wide RAM, either SRAM or SDRAM. How about a DIMM
- Ethernet PHY
- Leds / buttons / LED (or LCD) display
- VGA and PS2 connectors
- PCI edge connector would be nice but not essential

If Xilinx can do theirs for $99, I think the above could be do-able for
$200 (or $199 in marketing speak). I'd bite your hand off

Simon

 

[Simon]

Edmond Cote wrote:

Hi Simon,

I've been investigating lately *exactly* the same issues you have,
however I finally made a decision and ordered the NuHorizon's board.

While the width of the SDRAM was a deterrent, simply placing two read
cycles back-to-back should solve the obvious problem.

Yes, but I have *plans* [grin] The idea is to use the burst-mode of the
DDR2 SDRAM to fill a line of i/d cache at once, taking a hit on fetching
the first entry in the line, but hopefully winning overall. For that to
work though, I want the access to RAM to be as fast as possible. It
would of course work with 2x16-bit access, but I want speed

For curiousity's sake, are you working on a commericial or academic
project?, I am also working on implementing a MIPS based
system-on-a-chip. If you pick up that particular board, maybe we could
help each other out, since the kit includes next to no documentation.

I've more or less decided to go for the BroadDown2 board - I think the
possibilities for adding stuff onto it are better, and I want SDI in and
analog-video --> video decoder chip (Phillips SAA7115 probably). I want
to put an MPEG2 decoder / encoder (one at a time) onto the FPGA as well
as the CPU core for general purpose stuff. It's going to need an SDRAM
controller, and probably a PCI interface too... Like I said, plans

So, I'm looking really at the 1.5M part rather than the 400k part, which
means lots of expense (Foundation, aaarrggghhh), and I'm going to try
and get as much done *before* I cough up the cash to decide if I can
actually *do* it or not...

As for motive, well at the moment it's purely curiosity on my part, but
if it all pans out well, there is a commercial possibility (I work in
the professional post-production industry, and there is a gap I could
fill...)

Simon.