Well, the savings is usually in the range of $100K, not millions, and on a per-chip basis it's probably more like tens of thousands, and compared to the cost of fabrication and cost of labor for the average high-paid layout engineer, can actually be a minor expense. Where it helps us (MultiGiG), or at least has helped us in the past, is to augment a single Cadence license and a single Calibre license. I still refuse to use Cadence and do all my layout in Magic, then use a script to check it into the Cadence database. I have another script to take p-cells from Cadence and import them into magic. I should add that since I wrote the paragraph above, I have come to use Cadence regularly, sad to say, and although Magic remained as part of my digital flow, even that has been replaced by Cadence at MultiGiG. I still use XCircuit to do documentation such as wirebond diagrams, and Magic to render nice photographic-like die plots. Since my flow at work no longer uses these tools, I'm hoping to open-source a lot of my own code related to the digital flow. I will need to write a sea-of-gates router program, though, as the one I was using in my flow is not open-source.
Thank you for maintaining Magic and this opencircuitdesign website.
I am thinking about setting up a WikiNode? page to link to a few very closely-related wiki. You know, like the open-source PCB design wiki, the wiki that discuss FPGAs, the wiki that discusses a "open hardware" tablet PC design, the wiki dedicated to JTAG. Perhaps the "computer simulation" wiki.
So is the software you call "pcb" any relation to the gEDA software of the same name?
The PCB code posted on opencircuitdesign is a branch of the gEDA code that is implemented as a Tcl/Tk extension. Right now, it is largely just the same code, although the main branch developers have already implemented a few new features that make my version incompatible. But the bottom line is that there are many, many things I don't like about the original code base, and I don't like the Gtk-based format they're moving toward. I'd like to keep an I/O compatible version with a different interface. Due to limited resources, that's not working out too well right now, as the GUI is essentially the same but the I/O is already incompatible.
-- DavidCary? ~~~~