[rescue] Bigger Iron at Home (was: SNMP, Baby!)
Francisco Javier Mesa-Martinez
lefa at ucsc.edu
Wed Nov 12 17:16:31 CST 2003
On Wed, 12 Nov 2003, Joshua Boyd wrote:
> On Wed, Nov 12, 2003 at 02:15:42PM -0800, Francisco Javier Mesa-Martinez wrote:
> > it is more like a $100.00 part, still rather remarkable low price since
> > the part is relatively dense. Not particularly fast (50 - 70Mhz). Also you
> > need to take into account that is is a BGA part, not a socketed one, so
> > building a design around it can be a PITA uless you know what you are
> > doing. At least 6 layer board, plus it has a shitload of power ground
> > pins. The problem is that with the new denser high speed parts, we are
> > moving away from the wirewrap, and home etched boards of yore.... :(
> All I know is what I'm told around here, which is that Spartan3s are
> $15ish, and SpartanIIEs are $30ish. We are, BTW, using a 6 layer
> board. I get the impression that affordable auto-routers can do this
> level of complexity, but very badly.
I guess they may have differnt price levels, the quote I got from Xilinx
for their larger part Spartan3 is about 100.00, which is quite remarkable
since the previous XC4000 series chip we were using was a $1500+ part. We
are moving towards altera though since they seem to be more accomodating
with the educational market.
> That we are moving away from the wirewrap home-etched days of yore is
> all the more reason why we need good free software tools. We may not be
> able to sit down and actually build it ourselves any more, but
> considering the prices of some service bureaus, it doesn't seem out of
> the question to design it one's self and have them assemble the hard
I know, I have been loking for such a tool myself. Some of the foundries
seem to have pretty reasonable first tape prices ($300+). The problem is
that I am old school and I still enjoy the ocassional home edtched design
:), just wish I had more free time.
> > Eeerrr about the clocking, I really doubt you may get faster than 50Mhz
> > when synthesizing a complex processor in those FPGAs, I believe you need
> > Virtex parts to get in the 100 to 200Mhz range for such complex parts, and
> > that is with a very good placement and routing.
> 50mhz is a heck of a lot faster than early risc designs running a
> 10mhz. I certainly wasn't suggesting 100-200mhz. Not everyone needs to
> move faster, some of us are just interested in moving better as a
The earliest commercial RISC I can remember was teh R2000 at 12Mhz,
anyways I thought you were talking about the Apple's first riscs PPC (I
believe they were like 60Mhz parts).
For those kind of purposes, i.e. implementing a simple RISC CPU, I would
recommend you go the PIC route. The RISC CPU is already there, and most of
them come with socket pin parts, so you can get them into a normal design.
They are also very selfcontained some of them with their own clock and
> I was kinda thinking (for the purposes being argued here) of a 20-40mhz
> risc core, using part of the floor space for a graphics accelerator.
You'll need a very big part, a RISC core may take up most of the
floorspace even in larger parts.
> However, for the in cupboard model, we already established doing it
> oneself isn't practical.
> Thankfully, my music player idea should be able to be easily done using
> nothing worse than perhaps a TFP100 package, and at that, I still would
> want to pay someone else to do it.
> rescue list - http://www.sunhelp.org/mailman/listinfo/rescue
More information about the rescue