I have no better information but based on experience it seems to be at least an order of magnitude high. It would be interesting to understand it better but it's likely top-secret sauce that only the deep state would leak.
Speaking of bunny suits, they use upholstered chairs that must make a lot of dust from the cloth and the foam rubber filling.
Why would they put dust-generating cloth and friable foam rubber in clean room furniture? Materials science should be up to generating fabrics and padding that don't shed.
Clean rooms need flexible material for bellows and the like. Admittedly, we and Bell Labs looked at electroformed metal bellows to isolate lubricated moving parts inside our electron beam microfabricators, but that's an even more demanding environment.
-- Bill Sloman, Sydney
Probably for some chunk of the fab. The NRE for the chip design itself is likely to be quite a bit less. The masks could well be expensive. In the 1980's the design of an ASIC cost about $50,000 and the masks closer to $200,000.
More, bigger and higher resolution masks could be more expensive, but $250 M seems improbably high.
-- Bill Sloman, Sydney
I wonder what it costs to design a new car model and get the factory ready for production? A google search says it can be billions, as much as $6 billion if you start from scratch. Makes the cost of building a chip sound pretty cheap. Even the cost of the fab is a lot less than that.
Heck, one web page said it is less complex to engineer an airplane! Go figure.
-- Rick C
The cost of designing a new car model from the ground up, includes all the tooling needed to make the car. The mask cost is just the marginal cost of producing yet another part using the same tooling. If you added all of the engineering for the fab, mask, and processor design, you're right up there. Then there is the cost of the design system and all the software that goes with it.
And the Boeing site says that designing industrial electronics is trivial. ;-)
And some cardboard and styrafoam boxes on the table?
Note the gorgeous piece of equipment with the red circle around it.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Den tirsdag den 6. juni 2017 kl. 06.46.28 UTC+2 skrev rickman:
I'm guessing a lot of the of a car design is planning and logistics of the production, when a factory makes 500 a day of something that complex it has got to be planned and executed to perfection
Lasse Langwadt Christensen wrote on 6/6/2017 6:59 PM:
Are you suggesting the design of the car itself is relatively simple? Not to trivialize it, but I think the planning and logistics of auto manufacturing is not so difficult. Once you have ironed out all the details of what you wish to build, the processes for building it are very well understood. I think it is all the optimization and testing of the auto design that drives the cost. I believe it takes a lot more time to produce the detailed design of a part than it does to set up a CNC machine or injection mold to make it.
I believe I read somewhere that they spend 4 months just deciding all the features they want in a car before any part of it is designed. I believe the tooling for making a car is done in less time than that.
-- Rick C
I did but it's hard to read the front panel.
"For those who migrate beyond 16nm/14nm, it will require deep pockets. In total, it will cost $271 million to design a 7nm chip, according to Gartner. In comparison, it costs around $80 million to design a
16nm/14nm chip and $30 million for a 28nm planar device, the research firm said."-- John Larkin Highland Technology, Inc lunatic fringe electronics
Actually, I misread what you originally wrote. I can see NRE, including _design_ (and masks) being $250M. I read it as the fab NRE, not the complete end-to-end cost.
John Larkin wrote on 6/8/2017 8:56 PM:
If that number is in the ballpark, I see the end of advances in chip processing for economic reasons. This means selling a million copies of one design adds $271 to the cost of each chip. Even selling 10 million of a design adds $27 to each chip. CPUs, Flash and DRAM are the only devices that sell those sorts of quantities and I suppose the support chips. Of those, only CPUs can afford that sort of cost hit, but then I guess they sell a lot more than 10 million of a given DRAM or Flash. Certainly there aren't many other devices that are made in these quantities. Do FPGAs sell
100 million chips per design?Extrapolating these numbers would seem to indicate the 10 nm node costs around $147 million per design and the 5 nm node will cost $500 million. The next generation beyond 5 nm will cost nearly a billion dollars for the total NRE per design.
Sometimes IC design reminds me of building pyramids. To this day there are many ancient structures that we can't be sure how they were built with the available technology. I wonder if in a few thousand years if people will still understand how we were able to make the chips of today with the current technology. I know I'm impressed.
-- Rick C
I'm not. The process of automating a chip design still seems to need work - and once that work has been done it will get to be a whole lot cheaper.
-- Bill Sloman, Sydney
Am 09.06.2017 um 05:57 schrieb snipped-for-privacy@ieee.org:
Yes. The letter Z in Zilog was there because it is the last one in the alphabet and they were the last integrated logic company that was founded because of cost, as they wrongly assumed.
OMG, the Z80 had ~ 7500 transistors.
And the numbers given don't make a difference if you just use the 5nm process to make an ultrafast decision FF, a simple serdes on a multiproject wafer, an FPGA that just replicates the same cells ad nauseam or a 64 core Pentium with 3D graphics.
cheers, Gerhard
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.