![no use](https://cdn.opencores.org/img/pils_lt.png)
![no use](https://cdn.opencores.org/img/pil_lt.png)
![no use](https://cdn.opencores.org/img/pil_rt.png)
![no use](https://cdn.opencores.org/img/pils_rt.png)
Cooperative Structured ASICs
by Unknown on Sep 1, 2004 |
Not available! | ||
Hi.
This probably isn't the best place to post this idea, but what the heck. Where else can you go? I woke up this morning thinking that we can help solve some the major problems facing the ASIC industry if we could all get together to do it. Some of these problems include: - NREs for ASICs are way too high. .13u NREs are around 600K. 90nm are probably about double that. - Minimum volumes per ASIC design per year are several times higher than the NRE. Basically, if your market wont take several million dollars worth of a particular ASIC, you can't buy advanced silicon. - The ASIC design tools to build such an ASIC are also crossing into multiple millions of dollars. - Even if you had all the millions, any chip big enough to require advanced sillicon will take 100 engineers a year to design and verify. Structured ASICs in theory provide a solution to these problems. However, the likes of LSI and NEC still require over a million dollars per year in volume per ASIC design, and the tools aren't any cheaper. NREs are still over $100K, and they kill you on the part costs. So, is it possible for ASIC designers to take things into their own hands? What if we had a web site where ASIC designers could agree on common platform Structured ASICs? For example, if four designers could agree on a Structured ASIC platform, they could split the NREs and get the fabs interested through combining their volumes. As an example, consider the economics of the following Structured ASIC platform: - 500K usable gates - 727K bits of DP SRAM - .13u - 240 pads (I/Os being around 200) - Support for LVDS, PCI Express pads - Two PLLs - A small amount of configurable analog, for medium speed A/D, etc To get a main-stream ASIC supplier interested, plan on having a bank account with about $10M in it. The NRE is $600K if you get it right the first time, and minimum volumes better be at least $1M/year. Tools alone can cost you a million. The same project as a cooperative Structured ASIC could look much different. Suppose that four companies got together to build a platform. Each company puts in: - $150K NRE - Commits to buy $250K worth of parts for at least a year Together, they could get the thing built. Since it's a Structured ASIC, the NRE's to customize the chip for each vendor is minimal (think numbers like $50K). The base could tape-out when the first vendor is ready, and then ECOs could be used to produce parts for the other vendors as they reach a tape-out point. In addition, once the masks for the base is built, it could be used for any other vendors in the future that want in on this deal. As more people use the base, the original vendors could be paid back for their portion of the original NRE. I'm thinking that a good way to do this might be through a cooperative, like REI, where the customers are the owners. Someone has to own the masks. Why not the customers? Where I work, we write all the tools and develop all the IP needed to make this technology work. No more than four key guys at my company would need to sign onto this plan to make the technology fully available. The other part would be finding the first four guys to share a Structured ASIC, lining up the fab, and creating the co-op that would own the masks. What do you think? Bill |
Cooperative Structured ASICs
by Unknown on Sep 2, 2004 |
Not available! | ||
Where I work, we write all the tools and develop all the IP needed to
make this technology work. No more than four key guys at my company would need to sign onto this plan to make the technology fully available. The other part would be finding the first four guys to share a Structured ASIC, lining up the fab, and creating the co-op that would own the masks. What do you think? Bill, I think it would be quite hard to persuade four companies, since they would have to agree on time. There is a possibility that first run would not work, etc. I would say it is much wiser for a company like yours to produce structured ASIC and carry all the mask costs and NRE and then sell the customized structured asics with e.g. $100k NRE. But of course in that case you have to raise several milion $ to push technology through. best regards, Marko |
![no use](https://cdn.opencores.org/img/pils_lt.png)
![no use](https://cdn.opencores.org/img/pil_lt.png)
![no use](https://cdn.opencores.org/img/pil_rt.png)
![no use](https://cdn.opencores.org/img/pils_rt.png)