You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
This is the europractice site that shows all of the semiconductor foundries to make custom chips. Notice TSMC and GF are represented. Not IFS. https://europractice-ic.com/technologies/asics/
No, that is the Tower Semiconductor 65'sh process, probably out of their Japanese acquisition several years ago. I would like to see the a finfet shuttle, like what TSMC or GF has been doing for awhile. 2 years ago, Intel announced IFS. 6 months later, Pat claimed Intel "I'll make them as many Intel 16 [nanometer] chips as they want," during his visit to an auto industry trade show in Germany.
I have shown you how convenient is to see IFS's competitor's shuttle info. Semiwiki is loaded with Inteleons. Can one of you lead me to the site?
I did try contacting Intel. In fact, I emailed the VP of marketing of foundry services almost exactly 1 year ago. He gave me the run around. 9 months before that, we taped out on TSMC 16ffc, supported by Imec. Their support was amazing and almost always got back to me within hours. Not only technical support, but even NDAs so that we can download PDKs within 24 hours.
Hey Intel marketing, do you monitor this site? You should. Is IFS ready to deliver a PDK to customers? Please provide a link to your metal stack choices.
Formative? Still? Then they shouldn't be asking for taxpayer money. IFS has nothing officially to offer small US companies. GF (14) and TSMC (16) has documented their offerings since 2016. How does Intel not have a clue to what they will offer in their 16nm MPW? While TSMC and GF customers and wannabe customers are hammering away on circuits, Intel can't even post their MPW specs and schedule. Perhaps they know that their features cannot compete with TS16, so why post them? Dunno.
Formative? Still? Then they shouldn't be asking for taxpayer money. IFS has nothing officially to offer small US companies. GF (14) and TSMC (16) has documented their offerings since 2016. How does Intel not have a clue to what they will offer in their 16nm MPW? While TSMC and GF customers and wannabe customers are hammering away on circuits, Intel can't even post their MPW specs and schedule. Perhaps they know that their features cannot compete with TS16, so why post them? Dunno.
I remember seeing an older semiwiki article that was talking about the mindset changes needed for foundry, and one of the mentions was small vs big customers. They talked about how big customers are more demanding and many small customers require alot of extra manpower to support. Some thoughts I have to add on is that Pat often talks about how IFS will make IDM better. If that is one of the cornerstones of IFS, then big customers like Mediatek are a great way for intel to work on improving their manufacturing and support skills. I don’t know how big IFS is at the moment, but I wonder how many customers they could reasonably support (see Samsung foundry as an example of a company that can only support a few large companies).
Formative? Still? Then they shouldn't be asking for taxpayer money. IFS has nothing officially to offer small US companies. GF (14) and TSMC (16) has documented their offerings since 2016. How does Intel not have a clue to what they will offer in their 16nm MPW? While TSMC and GF customers and wannabe customers are hammering away on circuits, Intel can't even post their MPW specs and schedule. Perhaps they know that their features cannot compete with TS16, so why post them? Dunno.
I suppose the devil's advocate argument could be that GF has had two decades and TSMC over three to figure out the foundry business. Meanwhile your poor experiences seem to be during the first year of IFS's life. Definitely not a free pass to offer poor service, but I suppose not unexpected. From what I've heard in the early days of GF, the Chartered part of the company did reasonably well, but the AMD parts floundered and took years to adopt the first class foundry culture they have today.
IFS is in an amazing position. Most companies don't have their own fabs anymore. Only 4 own the entire market, and Intel is the biggie in the most desirable location. The incompetency must be overwhelming. Pat must have felt that he needed to buy an entire Isreali foundry to find one Andy Grove and somebody to make a website.
You need to bring Maxwell back from parole and have him meet with middle management, maybe after hours. They gotta go. Then get some guys crossing the border to bury them. That would be a good start. Then he can open the conference doors. If more than 3 people are in the meeting room, then bang bang with his silver aluminum bat. Nail the managers of those inefficient meetings
As far as the MPW, give the shuttlers a few sizes to choose from. An AI program can choose where each rectangle goes. Treat it like a lost liter. You won't lose though. Gina will fund it. They fund everything.
Was that cranky? I am providing real solutions to real problems.
Mosis... they are still around? Universities only, correct? What's up with the login requirement?
Can you please provide a snapshot of their metal stack. 13 layers would be good. That would be competitive with TSMC16 and GF14. This will tell us if this is BS. A yearly calendar snapshot will also be nice.
More crankiness.... Intel should only get US funding if they make MPW shuttles available to small US companies. If they don't make shuttles, then we are funding either monopolies (what else is new) and their CPU division, in which AMD should raise hell and ask for funding).
More crankiness.... Intel should only get US funding if they make MPW shuttles available to small US companies. If they don't make shuttles, then we are funding either monopolies (what else is new) and their CPU division, in which AMD should raise hell and ask for funding).
Intel is promising to make their services available to small companies, and they're using Intel Capital to provide $1B for early stage startups. (See the link I provided.) How quickly any of this happens is undefined, but they're saying all the right stuff. I expect they're going to get off to a slow start, for numerous reasons. If it was that easy to become a world-class foundry there would be a lot more of them.
As for design companies like AMD and Ampere... the CHIPS Act is specifically focused on domestic manufacturing. It would take more legislation to address design companies, and that portion of the semiconductor industry doesn't appear to need any government interference at all.
Obviously we shouldn't be funding AMD, nor Intel's design division as well. Also, I am not talking about a new foundry. I agree with you there. I am talking about their existing 22-14'ish divisions that were created slightly before TSMC and GF finfet release. They want funds, then shuttle it ... for companies they don't own. Why does it take so long to add a shuttle service for one of their existing and competing foundries.
Is it a foundry technology problem, the lack of seriousness of being a longterm foundry service, or upper level incompetence? I assume seriousness of being a foundry service longterm. TSMC, GF, Skywater, Micron, and and other non-application specific products seem to be a better place for the CHIPs act spending.
Obviously we shouldn't be funding AMD, nor Intel's design division as well. Also, I am not talking about a new foundry. I agree with you there. I am talking about their existing 22-14'ish divisions that were created slightly before TSMC and GF finfet release. They want funds, then shuttle it ... for companies they don't own. Why does it take so long to add a shuttle service for one of their existing and competing foundries.
The CPU and Chipset teams never used shuttles to my knowledge. Chipsets were always on a different process (N-1) than CPUs (N). The original foundry effort based on 90nm process was shutdown ages ago. As for why it takes so long, from my experience (in design teams) it looked like managing a shuttle process in the foundry was complex and took a lot of complex process and experience.
Is it a foundry technology problem, the lack of seriousness of being a longterm foundry service, or upper level incompetence? I assume seriousness of being a foundry service longterm. TSMC, GF, Skywater, Micron, and and other non-application specific products seem to be a better place for the CHIPs act spending.
IMO, probably none of the above. Building a competitive foundry company takes considerable person-years developing BKMs and efficient processes. Especially with fab organizations used to having just a few internal customers who had no practical alternative until recently.
I must be missing something. They have the fab already making the process (hopefully competitive). Give the customer the choice of 2x3, 2x4, 3x3, 3x4, 4x4 dies. Run the DRCs checks upon each customer submission. The customer name/logo must be there. Add the scribe lines. Create the GDS2s to make the masks. Submit to fab. 5 styles of wafer is cut. Dies are given to customer. Perhaps Mr. Ng can tell me where I am misguided. I am not a foundry guy. I just submit GDS2s.
Intel can do whatever they want to do, unless taxpayers dollars are involved.
MOSIS are very responsive and helpful. I talked to them about something I wanted to make in late 2021 and they were polite and did not care one bit that I was just one person wanting one cheap shuttle project.
Just ask them for a contact for IFS to give you rules for the stack and info about the PDK you will need.
As for the MOSIS connection to universities .. MOSIS do not provide PDKs. You need to obtain PDKs relevant to the fab. Some universities have deals for that, ways for their on-site students to access EDA and they have a few PDKs included in the deals which may include ones that line up with MPWs. I found the same thing with OpenROAD and eFabless - they have not done anything to make modern-node PDKs available (If you want 130 or 180 nm then there are PDKs more available) except via the 20th century concept that only kids resident at universities need to learn anything modern.
Odds are that Intel have not got that sorted out yet, either. It would be wonderful if they made a PDK more widely available.
Sorry cliff, I can’t really give you a good answer on that. Running a foundry is way beyond my expertise. It would be like me asking you to do a chuck rebuild I ordered; it wouldn’t be pretty.
Based on my understanding of how foundry works and looking at the years long journey GF had to go on to offer foundry services on par with TSMC/UMC, I don’t know if it is realistic to expect IFS to offer customer service on par with TSMC during their first year of operation. As bluestone mentioned IDM and foundry are VERY different. It seems intel leadership is saying all the right things to actually offer excellent foundry services, but it will probably take time to be fully realized. Heck Samsung still seems to have major deficiencies on the foundry services front and they’ve technically been at this for longer than GF.
