Thanks, MicroMan. That's an interesting story. EV2 wafer bank seems like a brilliant scheme to keep defense industry supplied with supposedly "obsolete" chip designs when the semiconductor industry moves on.
I'm proud to say I helped establish the 68020 microprocessor in numerous applications. Today, an announcement indicates a UK company in cooperation with Freescale has stowed away 68020 die and is selling them in packages for the defense/defence industry. I'm guessing these cost more than $5. But this processor first hit the streets 30 years ago and was very popular! That's a long life! The press release gives some insight (not on US site): http://www.e2v.com/news/e2v-releases-new-cerquad-package-for-68020-facilitating-the-transition-from-the-original-freescale-plastic-qfp/ You can't really put that effort in with every chip ever made. Each has its own demand and natural lifespan.
@krutsch, totally agree. The apps processor segment is growing to be a really tough market -- whether for tablets or automotive markets. In fact, make no mistake. The margin for the automotive market isn't that much better either.
It is a topic to follow in the future. It looks to me like more and more players disappear from this application domain... Many player in the apps processor domain changed focus from tablets to auto infotainment and cluster with the hope to get better margins. The question is if the auto market for such applications is not too small. I am skeptical that automotive can sustain so many NPIs , some of the players will either target also other applications with the same device or will exit because the margins are bad.
@Krutsch, unfortunately, Renesas isn't saying much at this point. What's clear, though, is that Renesas isn't interested in doing everything to create a so-called "eco-system" around their chip for in-vehicle infotaiment system. What specifics the "partnership" entails, however, remains to be seen.
I read the CEOs comment about infotainment but I still do not get what he wants to say. Are they looking to compete in this market or do they exit? Just saying they need partners is vague affirmation to say the least... What partners do you need to make an apps proc. , besides what they already had.
@microman "It must be particularly hard for a man in a culture where one's word is his honor, like Japan," I consider myself someone who always honours his word, and it is one of the things I most admire about traditional Japanese. We should all do this. Re rolling your own CPU, I think the idea of having another company such as Rochester receiving the masks and be able to manufacture the parts exactly per the original even if for 2 or 3 times the price is probably a smarter proposition. Maybe the original manufacturer gets 5% as royalty and Rochester charges $15 for a $5 chip and suddenly you have a win-win situation for all. Even these companies that say they can't manufacture the part economically anymore they are basing that on the ridiculous premise that it needs to remain a $5 chip when it doesn't. Make one wafers worth an divide the cost of that between the parts on a wafer and sell them at that rate. We design a product that used a dual port RAM that Cypress no longer made andt hey said if we buy a wafer's worth they'l run one. They cost us $15 each instead of the original $2 but it meant we didn't have to redesign the boards and pay recertification costs it was a bargain. This was after it went EOL. That sort of committment puts Cypress on the top of the list of vendors in my books. (maybe it 2 wafers? I can't recall)
@Junko: I agree with you. The openness of Mr. Sakuta about the reality might draw the customers, otherwise unwilling to discuss EOL, to the discussion table to agree upon a migration path. Being in the industry for more than 15 years (as consumer of the ICs) I understand how difficult it is to handle the EOL for everybody in the chain: from the IC manufacturer to the system developer to the end customer. But when somebody is very frank in bringing the issue in front of everyone affected, the discussion could drive to a solution. The Renesas team should make sure that they have an easier migration path for their customers who are using those EOL products, so that the users of those MCU could migrate their hardware & firmware to the new chips with much less effort.
It must be particularly hard for a man in a culture where one's word is his honor, like Japan, to have to wrestle with these business decisions that are EOL. He's very bold to face the customers so directly. Maybe here in the US we've gotten used to the dismissal "it's not personal, it's business", or analysts like me hearing vendors say "we understand our customers need a consistent supply for 7 years, so we assure that when we say 'embedded'". Yeah... right! And with the fast pace of process evolution, when they remove the 150mm 1.5 micron equipment, you can't just run the same masks on 200mm 90nm. There're also the test systems and packaging, let alone the quality and reliability that may be inherent in very specific techniques that may not be replicated in some foundry, foreign land, or even Rochester.
As for rolling your own replacement, whether an ECU or MCU, the code would be a killer to get right – or even to read out of the original – let alone if you had to modify it for that new Bosch sensor or different cam shaft (old school). Porting code from an 'HC11 to an RH850 will take more than a week in the garage. As mentioned, getting the engine running smoothly and fuel burning cleanly is quite the effort even for the professionals. These chips aren't just timers. It hard enough trying to figure out what's really wrong when your OBDII throws a P0440 code. It'll be as frustrating to have a $25,000 classic Camaro being disabled from a bum ECU as a brand new $50,000 Infiniti waiting for its $10 MCU to come from the earthquake-shaken fab.
No easy solution. We made the bed we're laying in.