2021-10-25 12:52 PM
I'm sure ST must have a better idea than we do. Distributors seem to be saying "not in the foreseeable future", which I take to mean at least the end of 2023. Obviously no-one has a crystal ball, but a guess would be better than nothing. It does seem a bit pointless discussing technicalities of a product that may or may not be available in a few years time ( except it keeps us design guys in a job).
2022-01-08 08:20 AM
My projects are only open-source hobby but over 2000 versions of one of them based on STM32H743IIT6 have been built around the world but no more as it simply isn't available.
I'm now focussed on the Raspberry Pi RP2040 . Widely available and less than £1 in 10-off quantities. In my application I'm getting > STM32F407 performance. Nice (largely bug free) SDK, dual core + 8 PIO and good amount of RAM. Limited by number of pins and timers, no FMC/LTDC but rock solid in operation and clocks up to 252MHz with no issue
2022-01-08 10:15 AM
Might be nice, but what are they going to tell you?
The planed demand is mapped out and reflected in current delivery times, they aren't short-circuiting that, all they can do is increase manufacturing capacity, and adjust product mix, and offer to pull in deliver to each customer with an order in turn.
The REAL failure here is for the Retail Distributors to have any depth of stock, and forward planning to keep their warehouses full, but it's not the 70-80's any more and they have very shallow shelves, and very broad lines.
Their customers show up randomly, with random demand, and low levels of commitment, which is very complex to predict and potentially leaves them with a lot of obsolete stock they can't move, and manufacturers don't want to rotate, or take back.
Start pounding of their doors, because that's the only visibility and champion for the smaller customers collectively.
ST could perhaps provide a short list of specific parts they are focused on delivering, and which mask sets they have duplicated. Some die per wafer score, and wafers per week score, along with weeks to service current pending orders on that die. The are probably business sensitive numbers, but dire times call for more transparency.
The inflection point here is where people default on delivery because they can't use the parts due to other shortages in the build, or they have overestimated sell-thru of their own products due to everyone else stalling out.
I'd definitely like to see a strategic plan from Mouser, Digikey, Arrow, Farnell, etc *and* ST. Especially from those senior managers involved in the 10 Year guarantees, and massive expansion of SKUs
2022-01-08 10:22 AM
Qualifying new parts and die is not a quick exercise. It's hard to know how far down the process they are as this problem has been apparent since the beginning of 2021.
Guidance on new parts, especially Value Lines variants on shrunk die and compatible footprints might be especially useful.
2022-01-08 10:28 AM
Having a smaller die, and code space from external QSPI is definitely a plus there.
2022-01-08 11:05 AM
The Ideas is the place here for suggestions for ST to gracefully ignore.
JW
2022-01-08 12:49 PM
I'm not trying to point a finger of blame at anyone for the situation, I accept we are in unprecedented times, a perfect storm, and the system is broken. I'm only interested in where we go from here. If ST could say something like "we don't expect to supply the STM32 retail market before 2026" or "we are expanding our production and hope to clear our backlog by Q1 2023" then at least we'd have some idea of the magnitude of the problem we are facing.
Right now, being of retirement age, I'm working on the assumption that what I have in stock is all I will get for the rest of my working life.
A short list of priority parts would be a good idea, also I guess we can help by designing our products for the smallest necessary device, and lowest necessary technology, hence the smallest die size etc., although of course price largely does that for us.
10 Year Guarantee? That turned out well didn't it.
2022-01-09 05:30 PM
.. hence the smallest die size
Unfortunately they frequently make one die, and sell it in a dozen variants, where for example they only spend tester time on 128KB of 1 or 2MB of flash on die. Thereby allowing 10x throughput on the test equipment allowing faux price stratification. It's not that die tested bad, also by disabling sub-licensed features like CAN, USB or ETHERNET, time and costs can be reduced.
The smallest package available for a given part might be a better tell of physical die size.
Definitely don't see this pushing to 2026, but if i want to see parts in 2023 I'd be ordering them now and not dithering.
2022-01-09 05:41 PM
Yeah, that's a place were ideas go to die...
Someone from ST senior management needs to show up with some answers and charts, and show some kind of leadership.