Status on FPGA PCB

Kris> Does that mean you will be running 8 FPGA's 
240mkeys/sec ?

Guerric is looking at his core on the Spartan-3 (which is a
bit slower
than Virtex-4 parts) and without optimization, is about 1/2
the speed
of the Virtex-4 at 1/4 of the price, making it very
affordable.

Best case numbers for the Spartan-3 using pencil and paper
are about
250MKeys/second for BRam based designs, and I expect we can
probably
get a good piece of that with tuning over time.

I have some serious questions about how to handle heat on
Spartan-3's
which will require a prototype (or two) to resolve. I
suspect the
octal design I did for the prototype may not be stable, but
I wanted
to give it a try anyway. I am pretty confident that that
single sided
mesh design can be made stable, so this board is useful to
test both
configurations.

Kris> Also what was the cost to build this?

A little over a $100/fpga. Something around $150-200 for
one, and about
a grand for two by the time we add eeproms/flash, power
supplies, cap's,
board, and connectors in.

I'm excited ... I didn't realize high end Spartan-3's were
this cheap and
performance rich for the dollar when applied to
reconfigurable computing.

Have fun,
John
_______________________________________________
Hardware mailing list
Hardwarelists.distributed.net
http://lists.distributed.net/mailman/listinfo/hardware

A grand for 2 or 8?

Seems expensive if it goes from 200 for 1 to 1k for 2.

Cheers,
Kris

-----Original Message-----
From: John L. Bass [mailto:jbassdmsd.com] 
Sent: Tuesday, 5 December 2006 10:32 AM
To: hardwarelists.distributed.net; krisamy.id.au
Subject: Re: [Hardware] Status on FPGA PCB

Kris> Does that mean you will be running 8 FPGA's 
240mkeys/sec ?

Guerric is looking at his core on the Spartan-3 (which is a
bit slower
than Virtex-4 parts) and without optimization, is about 1/2
the speed
of the Virtex-4 at 1/4 of the price, making it very
affordable.

Best case numbers for the Spartan-3 using pencil and paper
are about
250MKeys/second for BRam based designs, and I expect we can
probably
get a good piece of that with tuning over time.

I have some serious questions about how to handle heat on
Spartan-3's
which will require a prototype (or two) to resolve. I
suspect the
octal design I did for the prototype may not be stable, but
I wanted
to give it a try anyway. I am pretty confident that that
single sided
mesh design can be made stable, so this board is useful to
test both
configurations.

Kris> Also what was the cost to build this?

A little over a $100/fpga. Something around $150-200 for
one, and about
a grand for two by the time we add eeproms/flash, power
supplies, cap's,
board, and connectors in.

I'm excited ... I didn't realize high end Spartan-3's were
this cheap and
performance rich for the dollar when applied to
reconfigurable computing.

Have fun,
John

_______________________________________________
Hardware mailing list
Hardwarelists.distributed.net
http://lists.distributed.net/mailman/listinfo/hardware