Silicon and integrated circuits come and go, but when it comes to extended lifetime support from a company, it’s very, very hard to find fault with Microchip. They’re still selling the chip — new — that was the foundation of the Basic Stamp. That’s a part that’s being sold for twenty-five years. You can hardly find that sort of product support with a company that doesn’t deal in high-tech manufacturing.
While the good times of nearly unlimited support for products that are decades old isn’t coming to an end, it now has a cost. According to a press release from Microchip, the price of these old chips will increase. Design something with an old chip, and that part is suddenly going to cost you 7.5% more.
The complete announcement (3MB PDF), states, in part:
…in the case of extended lifecycle product offerings, manufacturing, assembly and carrying costs are increasing over time for
these mature technology products and packages. Rather than discontinue our mature product, Microchip will continue to support our
customer needs for product availability, albeit increasing the prices in line with increased cost associated with supporting mature
product lines….For all orders received after 31 August, pricing for the products listed will be subject to an increase of 7.5%
The PDF comes with a list of all the products affected, and covers the low-end ATtinys, ATMegas, and PICs that are used in thousands of tutorials available online. The ATtiny85 is not affected, but the ATMega128 is. There are a number of PICs listed, but a short survey reveals these are low-memory parts, and you really shouldn’t be making new designs with these anyway.
Microchip have always done this, releasing new lower-cost parts while slowly increasing price on older parts. They know that it costs time & money to redesign around newer cheaper parts.
It also costs more to continue to run lower volume parts and old fabrication processes…and even more to redesign old parts on new processes.
And while releasing new part they also release application note describing best ways to migrate from older part to newer. For products that are intended to have long life cycle Microchip is an excellent choice.
As far as I know every supplier does this. It’s simple supply and demand. And demands for these old chips goes down, and thus the supply is more expensive. Till the point where it becomes to expensive to buy and the jump to a new chip is made.
Example, 10 year ago, at my old job, we where buying 48Mhz CPUs for 50 euros a piece. While you could get a 200Mhz ARM for 10 euros. A redesign of the board was done, replacing not just that CPU, but all other old chips as well, and the whole board went from 360 euros a piece to 120 euros a piece.
The complete announcement link is broken.
Yep. We are shocked, Hackaday!
https://www.allchips.ai/wp-content/uploads/2018/08/MICROCHIP%EF%BC%8DNotice-of-Price-Increase-on-15-Yr-Old-Devices1.pdf seems to have the full text
Yes! I was for decades a huge fan of Microchip and they got design-ins for a lot of high volume product by me.
My consulting customers were very pleased they didn’t need redesigns every couple of years, and often as not, Microchip would come out with some newer more capable chip they could put on the same product with no changes or maybe a recompile – or could then add more features to the same hardware.
More recent developments – which include the advent of the Arduino and other small cpus with far better IO than the bad old days when only Microchip had really good IO, and the fact I no longer do embedded design for a living, have made me more agnostic. That and the fact the for quite some time it was pretty expensive to get decent tools for Microchip. Yeah, their crashy IDE for asm was free, but any good C compiler – and who was good changed from company to company requiring re-buys of the tools, made them a lot more expensive up front and more trouble. the initial microchip C compiler – and yes, I still want my $300 back – didn’t support taking interrupts and didn’t handle thread-safe variables if you worked around it – couldn’t do for example floating point in more than one thread.
These days it seems there’s less need for long life, but I remember them fondly, they really helped butter my bread.
Now, outfits like Intel that introduce “gee whiz” tech to try and appropriate whatever is hot – IoT just now – and then drop them when they don’t magically take over the market without even a tenth the effort Arduino put in (even worse than Digilent’s arduino clones where the drivers for the better hardware are “left to the student”) – means I never would, and never will, design Intel into anything.
And they wonder why their offerings fall flat, other than also being overpriced. Geez.
Remember when Intel was getting into the Single Board Computer market a while back, and were hoing to take it by storm?
That was an interesting couple of weeks.
*Going
Stupid fat fingers…
What was wrong with the original spelling? Seems appropriate.
“Remember when Intel was getting into the Single Board Computer market a while back”
which time are you referring to?
I think Intel’s Quark line.
The 8032 and 8051 chips on the other hand where very successful and derivatives of the MCS-51 made by other companies are still used in many devices.
Please don’t refer to an Arduino as a CPU, the Arduino boards contain MPU’s which are most of the time made by Microchip (previously Atmel)
MPU != MCU
Dumb question, but why would cost be increasing for mature products? The part is already designed, which should be a considerable savings even if tooling needs to be replaced.
Because it costs them to keep the old stuff in stock. They want to push you to use the newer parts.
1. Lower sales volume.
2. Needing the same fab stuff for new product.
3. Switching a fab to something else is downtime.
4. Stocking more parts that may never be bought.
Probably more…the NRE isn’t the big cost on really simple uPs anymore. That cost was amortized long ago. Keeping things in stock and fabs ready is more like an ongoing rent – as is support, sales and so on.
I’m not in the silicon business, but here’s how i’d imagine it goes:
New fabs not only make newer chips, they also make more/faster.
Keeping old fabs around to make old chips has not only opportunity cost, but it means the resources are less productive than they’d be if switched over to new fab.
Redoing an old chip to work on a new process takes resources similar to designing a new chip. So producing them on the old process makes sense, it just has a cost.
I think this policy is entirely reasonable, and it’s commendable that they continue to support extended lifetime products at what seems to be a pretty modest price premium.
I work in a Fab which produces discrete semiconductors and is probably one of the oldest fabs in operation (since the late 60s). Currently my oldest design is around 40 years old. And yes, most key process equipment is of the same age. We run those equipments far longer as they were ever intended to last. They are a nightmare in production. Long down times, poor yields, catastrophic spare part situation (OEMs being out of business for 30 years). Our service team regularly scraps eBay for key spare parts like amplifier tubes.
This all adds up to the running costs of such legacy products. At some point (say 10 years after a technology was introduced) you reached the bottom of productivity and maturity. From there things again start to become slowly more expensive and troublesome. From that perspective I fully understand to charge more for old legacy products, as your cost are also running away.
Designing old processes on current equipment is often nor possible. Especially if you have automotive costumers. They are very reluctant to process changes.
Microchip makes great silicon. But there tools are light years behind everyone else. I don’t know why people even bother using their products because of this.
Open source tools and old hardware go great together.
I disagree that Microchip’s tools are “light years behind” others. I’m a Microchip PIC32 user, but I looked at switching to ARM a couple of years ago. Cobbling together a free toolchain resulted in a less-integrated environment than MPLAB X, especially with the clumsiness of launching a debug session. Free commercial toolchains were severely limited in code size. The paid versions of those might’ve been competitive with MPLAB X, but since I wasn’t going to pay the money, I didn’t try them.
YMMV but that hasn’t been my experience doing arm development. I personally like Eclipse and setting it up for cross development wasn’t hard, or a compromise in my personal experience. If you are using STM32 parts then their packaged version of Eclipse does the work for you. CubeMX does the work of creating a makefile project if that is the direction you want to go
Does Microchip still use legal tricks to do an end run around the GPL in regards to GCC on the PIC chips? The one where the Microchip libraries (without which doing anything with the chips are a LOT harder) contain a clause that says you can only legally use them with the official Microchip GCC build (the one where if you want optimization options you have to pay them for it)
One of the below…. (in no particular order)
A) Demand is going down so production runs are smaller and more spaced out
B) The product costs more to produce than in the past because it requires older production technology
(go try and buy a brand new floppy disk and let me know how much it cost)
C) They are “trying” to reduce demand to make capacity for new products
D) You are are locked in, increasing the price only increases their profits, if you need it, you need it
This appears to be a valid link to the announcement:
https://www.allchips.ai/wp-content/uploads/2018/08/MICROCHIP%EF%BC%8DNotice-of-Price-Increase-on-15-Yr-Old-Devices1.pdf
In the opposite corner: Maxim.
They tend to cancel the production of chips that didn’t sell as well as expected on short notice, sometimes before the mass production phase.
And that’s the reason why I never consider using their stuff for anything else than hobby projects.
exactly! With some exeptions like the serial port buffer/level shifter IC’s with 232 in their name.
Regarding the price increase of Microchips older parts it makes much sense. And to be honest I am willing to pay the extra happily knowing that the used component will be available for the next decade. Because not every piece of electronic is a throw-away, some things are designed to last for 20 years (and sometimes used even longer). Knowing that important parts like microcontrollers will be available for a long time after their initial release is quite comforting during the design stage.
“Because not every piece of electronic is a throw-away, some things are designed to last for 20 years ”
For instance, the U.S. Postal Service’s LLV (Long Life Vehicle), those were designed to last 20 years, the last one (IIRC) was delivered 25 years ago, The USPS is still in the procurement process for its replacement, and in the meantime is cannibalizing junked ones and searching for Chevy S-10 pickups of that era (some of the same drivetrain/chassis components) for replacement parts.
They’re doing it right. For any professional product, its lifetime should be considered the time it’s actually actively used out there, not the time your marketing department has planned for it. That being said, any company that keeps cancelling products despite them having a active user base, I consider untrustworthy and try to avoid in the future if at all possible.
“…but a short survey reveals these are low-memory parts, and you really shouldn’t be making new designs with these anyway.” That’s an arrogant “new is always better” statement. If a part is suitable for an application, why not use it? Not everything needs an all-singing all-dancing MPU (and its 800+ page data sheet). Yes, 7.5% equates to real money, yet in the grand scheme of things, its an understandable adjustment and just plain trivial.
No, but if the 8k and 16k and 32k version of an otherwise-identical part are pin- and code-compatible, I don’t think it’s unreasonable to EOL the smaller chips and only keep the universal-donor in extended production.
Microchip did keep the 16C84 to 16F628 chips pin compatible.
Phew, glad our PIC16F690 did not make it on to this years list…
Kinda sad that SiliconExpert still calls it’s years to EOL to be about 5 years, guess Microchip does not want to commit any further into the future.
I really do like ST Micro for theyr 10 years of availability on the full STM32 portfolio they renew every year.
On many of the products using these old chips, I bet these components are way cheaper now, even with the extra 7.5%, than they were when these products were originally designed.
And for the old products being repaired to keep them going, a 7.5% increase is probably negligible too.
> Not everything needs an all-singing all-dancing MPU (and its 800+ page data sheet).
Suddenly the pictures of dancing hippo appeared in my mind. Well, well!
I read somewhere regarding medical,millitary, airplane one cannot change/upgrade the MPU since the license is only valid at the specific PCB/MPU, and it’ll cost a lot for a new license.
My first MPU was a PICF84, and I really love it. Mostly because I suffered.
Later it was superseded by C84, which I jumped on, but not 628 or so, which is pincompatible I think, later 16F1847 No problemo, I’ll buy it when I need it.
But what about the lifelength of a PIC16F84(A), when will it be useless?
mine are around 20 year old, stored into a ESD bag.
The FAA…
There was a story about 25 years ago of a company that supplied the FAA with PC’s with (IIRC) 200MX Pentiums.
A few years later when the FAA ordered replacements, the company tried to charge them more for newer versions of the Pentium. The FAA said something to the effect, “No, you agreed to supply us with 200MX Pentiums, so get on it!” and that company started scrambling to find New Old Stock.
Does Caterpillar still support their products for 40 years? (i.e. replacement parts)
Is there any way to find out ICs that are “actively” produced? (those who are still “cheap to produce”).
I’m not looking for the absolute cheapest ICs, this is just a small hobby for me. (Mainly 8 bits Atmel and some SAM dev board)
For sure the PDF linked above will help me, but without HaD I wouldn’t know in the first place about those extended lifetime products increase.
The IC manufacturers announce regularly the EOL of their parts, and before that, they tend to give an estimated EOL or put a notice such as “Not Recommended For New Designs” on the part datasheet. If in doubt is wise to check the data of the first release of the datasheet, that will give you a good indication of how long the part has been around.
Part vendors such as Digikey or Mouser are also useful for determining the availability of parts, not only for obsolete parts, but also if there are one that have very long lead times.
I have personally found that Flash memories are the worst when it come to lifetime; because technology is improving so rapidly, the parts, some less than 3 years old, become obsolete, and sometimes before the replacement part is made available to the vendors. This is a bit of a problem when you work in the defence sector and must guarantee a 20-30 year support for the product. a 7.5% increase in price is nothing compared to the NRE cost of the redesign of a 100% compatible board that will need a new certification.
elderly electronics tax!!
Rename is coming.
Microchip parts are instock with short lead times.
Have fun with ST.