NXP chip supply - any news?
- Thread starter tyzjames
- Start date
joepasquariello
Well-known member
If there is good news on the T3.x, you'll see it on those product pages.
Yeah. Just checking if there might be any updates/insights from NXP into when it might be back in stock and the supply situation for 2023. The delivery for my order from Element14 has been pushed back multiple times since April 2021 and now they're basically saying they have no idea when it will be in stock.
KurtE
Senior Member+
The most recent posting I remember seeing was: https://forum.pjrc.com/threads/7165..._120-T3-x-like?p=317130&viewfull=1#post317130
PaulStoffregen
Well-known member
Heard another update on Tuesday this week. Situation is pretty much the same, only a small portion of the Kinetis chips we ordered in January & May 2021 will arrive in the middle of 2023. Or are at least scheduled to arrive... they've repeatedly been rescheduled/delayed, but the point is only a small portion even have a scheduled date. The rest have delivery dates about 12-13 months away, which in NXP-speak means they aren't actually scheduled yet.
At this time last year (December 2021), they were saying we would see "gradual improvement in Q2" (that's Q2 of 2022) and some would arrive in Q3 and the rest in Q4. Obviously that didn't happen this year.
We did very unexpectedly get some of the K64 chip for Teensy 3.5 several weeks ago, so technically some chips did come in Q4. But they weren't ever shown on any updates, they just arrived one day without any warning. Fortunately we had PCBs and all the other parts just sitting here waiting, so we were able to immediately put a small batch of Teensy 3.5 into production. They completely sold out over only 3 weeks.
I never did get any explanation why some K64 chips suddenly arrived (and I didn't ask too much... not wanting to look a gift horse in the mouth) but if I were to speculate, my guess is NXP probably made them for some far more important customer who decided not to buy them. These chips have been absent from the market since mid-2021 (PJRC had many extras so we ran out of Teensy 3.5 & 3.6 near the end of 2021 and Teensy 3.2 in January 2022) so my even wilder guesswork is as NXP starts making these chips again, probably a good portion of the orders will be canceled because people redesigned for different chips, and distributors will offer them to other customers who were previously told none would come anytime soon.
But whether NXP actually starts making the Kinetis chips again, and how many, is also a matter of pretty wild speculation. Most of the Kinetis chips are 90nm silicon. Before about 2012 Freescale was much more open about saying which silicon process node was used, but later they purged all that info from their website and documentation, so I just don't know for the later Kinetis chips. The L series might be 130nm or some process meant for lower leakage currents. I have long suspected K66 might be smaller than 90nm, partly by the higher clock speed, but also by the MCLK output having faster rising/falling edges which cause havoc to the SGTL5000 chip without a series resistor (added to the audio shield at Rev C). But the fast edges could also just be due to the HSRUN mode raising the internal voltage. Whatever process is actually used, I'm very sure it's older than 40-45nm (TSMC, Samsung, Global Foundries) or 55nm (Semicon China).
The reality of today's market is the main semiconductor fabs have converted as much of their production capacity as possible to ~45nm or smaller, and they're only building new factories to make the smaller size transistors, because the newer process nodes are much more profitable. Nobody is pouring investment into 90nm. So manufacturing of all the older chips gets squeezed into contracted production capacity.
My gut feeling, which admittedly is based on a lot of guesswork and sheer speculation, is the supply of older chips may never recover to what we enjoyed before the pandemic.
At this time last year (December 2021), they were saying we would see "gradual improvement in Q2" (that's Q2 of 2022) and some would arrive in Q3 and the rest in Q4. Obviously that didn't happen this year.
We did very unexpectedly get some of the K64 chip for Teensy 3.5 several weeks ago, so technically some chips did come in Q4. But they weren't ever shown on any updates, they just arrived one day without any warning. Fortunately we had PCBs and all the other parts just sitting here waiting, so we were able to immediately put a small batch of Teensy 3.5 into production. They completely sold out over only 3 weeks.
I never did get any explanation why some K64 chips suddenly arrived (and I didn't ask too much... not wanting to look a gift horse in the mouth) but if I were to speculate, my guess is NXP probably made them for some far more important customer who decided not to buy them. These chips have been absent from the market since mid-2021 (PJRC had many extras so we ran out of Teensy 3.5 & 3.6 near the end of 2021 and Teensy 3.2 in January 2022) so my even wilder guesswork is as NXP starts making these chips again, probably a good portion of the orders will be canceled because people redesigned for different chips, and distributors will offer them to other customers who were previously told none would come anytime soon.
But whether NXP actually starts making the Kinetis chips again, and how many, is also a matter of pretty wild speculation. Most of the Kinetis chips are 90nm silicon. Before about 2012 Freescale was much more open about saying which silicon process node was used, but later they purged all that info from their website and documentation, so I just don't know for the later Kinetis chips. The L series might be 130nm or some process meant for lower leakage currents. I have long suspected K66 might be smaller than 90nm, partly by the higher clock speed, but also by the MCLK output having faster rising/falling edges which cause havoc to the SGTL5000 chip without a series resistor (added to the audio shield at Rev C). But the fast edges could also just be due to the HSRUN mode raising the internal voltage. Whatever process is actually used, I'm very sure it's older than 40-45nm (TSMC, Samsung, Global Foundries) or 55nm (Semicon China).
The reality of today's market is the main semiconductor fabs have converted as much of their production capacity as possible to ~45nm or smaller, and they're only building new factories to make the smaller size transistors, because the newer process nodes are much more profitable. Nobody is pouring investment into 90nm. So manufacturing of all the older chips gets squeezed into contracted production capacity.
My gut feeling, which admittedly is based on a lot of guesswork and sheer speculation, is the supply of older chips may never recover to what we enjoyed before the pandemic.