Slashdot Mirror

MOSIS. https://www.mosis.com/

Proof-of-concept doesn't have to be on the latest technology, which is undeniably expensive. Do a shared-wafer (https://www.mosis.com/) on some near-obsolete technology, and when the bugs are worked out it's time for scaling.

Ever heard of SiGE and MPW/COT? Who needs FPGA when you can go open source ASIC and produce an initial production run for under $50k, possibly even $10k? There's been some interesting research from places like CalTech and Berkley in to fully designed MIMO's even with integrated antennas in an SOIC that are in many cases nearly a decade old now.

You can now deal directly with TSMC for this, or go through third parties like MOSIS and CMP. AFAIK none of the groups that do this publish their rates, and I haven't done this myself, so my numbers come from forums like this one, rounding up the lower numbers that I have commonly seen.

I think you're talking about MOSIS.

http://www.mosis.com/

No, but anyone who's that interested CAN have one of these in their garage (or on their desk, more likely), and get their design fabbed by these guys fairly cheaply.

Sure, it's not quite as easy as hacking on open source, but hobbyist CPU design is definitely possible. Especially when you consider there ARE open source CPU designs out there.

Look here for fab costs.
http://www.mosis.com/

Packaging is crucial to making the thing work too, however.

CAD tools to convert the RTL into GDS is also very expensive.

I would suggest a service such as http://www.mosis.com/. They specialize in doing small volume production and prototype runs, reducing cost by placing several customers onto a single mask set. They even have an educational program, where educational and research designs will be manufactured for free.

Exactly; in fact, most of the design on OpenCores are most often used on FPGAs for that reason (fabrication is expensive). For a solution to that, try http://www.mosis.com/. They combine many low-volume projecs onto MPWs (multi-project wafers) and lower the cost to the point where college VLSI design students can afford to fabricate a few of their class project. Granted, I'm not aware of a way to move from this very low-volume stage to something like production at a foundry like TMSC or UMC, there is a huge price jump I believe, but I'm not sure that's a very big obstacle. If your design is useful enough to justify creating an ASIC in the first place (as opposed to an FPGA or microcontroller+specialized hardware), it's useful enough to fabricate a bunch of them, and if you're not dead-set on the latest and greatest process node, a mask set's cost is manageable as long as your production run is sufficiently large.

Another thing I would like to see, personally, is something like MPWs at TSMC. Somebody correct me if I'm wrong, but I'm not aware of a way to amortize the cost of a mask set over many small projects. This, of course, would tie the production numbers of all related projects today, but if you're only making a few hundred wafers, perhaps this doesnt matter.

You want MOSIS. Providing small volume chip fab services (via short ganged-mask wafer runs at flexible mainstream fab houses) for decades now, Mosis is exactly what you want if FPGA and a programmable microcontroller aren't what you really need.

foundry!=fab

Huh? Sir, are you sure you know what you are talking about? ;-) Yes, your inequality is correct, but in much weaker sense than you assume: since the dawn of modern CMOS (you know, Mead-Convay time, circa 1980) people referred to foundry as an outside place to make your chips.

Foundries have fabs, Intel also has fabs but (AFAIK) does not provide foundry services; IBM provides, e.g., SiGe foundry services, MOSIS is a foundry, etc.

Places where you get wafers are just "wafer suppliers", places which make masks are "mask houses"...

Paul B.

Darn... I thought this said openMOSIS.

I don't think anyone would mind a sourceforge for chip building (especially free nightly builds!)

More on topic and to the point - it is good to see that MOSIX tech is now available opensource (stable anyway). Now we have yet another viable option for speeding up our Beowulfs (MOSIX is generally run with PVM/MPI - not as a replacement).