How is Chipageddon affecting you?

[mikeselectricstuff]

Don't forget Octopart often have wrong info from RS regarding stock. To the point where its not trustworthy at all (for RS).

I invariably find that RS stock shown on Findchips is a complete lie. Findchips should drop RS if they can't provide true information.

[exe]

13 weeks -as I recall- is basically the time to go through the production line. So there is like constant free capacity?

Well, I'm no expert, but I'd assume out of those 13 weeks it takes only a fraction of time for actual manufacturing. I'd expect there is a lot of queuing time in those 13 weeks. Like, "waiting to be tested", "waiting to be packaged", "waiting to be shipped", etc. So, probably, not "constant", but there is some free capacitance on average. Or they do their priorities, producing what brings them the most money, or if they have some contractual obligations with strict deadlines.

[AndyC_772]

It's just logistics. Swapping mask sets to make a new IC takes time, which is downtime for the expensive fab, so it's done as infrequently as possible. Each individual type of die may be made only a few times a year, in a batch that's big enough to meet whatever demand is forecast.

If a die is scheduled to be manufactured 4 times in a year, then it gets labelled as having a 13 week lead time. The lead time you actually see from order placement to delivery may well be different, depending on where you are within that 13 week period, and how long it takes to package a die and get it packed and shipped through distribution.

[tszaboo]

https://en.wikipedia.org/wiki/Semiconductor_device_fabrication
"The fabrication process is performed in highly specialized semiconductor fabrication plants, also called foundries or "fabs", [1] with the central part being the "clean room". In more advanced semiconductor devices, such as modern 14/10/7 nm nodes, fabrication can take up to 15 weeks, with 11–13 weeks being the industry average."
I know wikipedia as source...

Well, I'm no expert, but I'd assume out of those 13 weeks it takes only a fraction of time for actual manufacturing. I'd expect there is a lot of queuing time in those 13 weeks. Like, "waiting to be tested", "waiting to be packaged", "waiting to be shipped", etc.

I'm fairly sure they have teams organizing orders to make production as efficient and fast as possible, since time is money.

[peter-h]

I am pretty sure the above 13 week period is not at all related to the fab lead time.

The chips need to be tested, packaged, etc. That leadtime is probably a good few months.

[tom66]

I am pretty sure the above 13 week period is not at all related to the fab lead time.

The chips need to be tested, packaged, etc. That leadtime is probably a good few months.

Fab lead time is probably the biggest single factor since it's highly specialised.  Packaging and bonding is often done in another country, usually a country with cheaper labour.  Though in some areas you have things like ABF (film) or BGA packaging capacity being the choke point for instance for GPUs and the likes... I recall that was a choke point earlier in 2021.

[peter-h]

The "end user" lead time is not related to the manufacturing process. It is virtually certain that chips in general are made in batches and each batch goes on the shelf. Hence when you look at date codes you find these in discrete "quanta" with many missing codes.

Lead time is quoted to manipulate demand and to create fear. The typical cycle is that prices generally keep falling, slowly. At a certain point the business decides it has gone too far and the reps are instructed to spread the feared "a"-word: allocation. Then buyers go crazy and start placing orders. You get shortages and 26 week lead times. 6 months later there is a bloodbath and things are good for another year or two.

And it swaps around. One year you get tantalum caps. Another year you would get RAM chips...

Since I started in my own business in 1978 I've seen this many times. This time it lasted 3 years due to covid having distorted everything. But the bloodbath will be bigger, too, and much more satisfying to those who actually create wealth (manufacturers) than those who just skim a percentage off the top (resellers)

The famous LM2936 is now back to pre-bubble pricing. Maxim is still trying to hang in there with totally crazy pricing. Just had a conversation with their "chatline" and they are still not communicating in a normal human way; it is still using support tickets on their website <bang head > which I could not log into despite assistance They don't seem to know that a MAX489 has a drop in replacement at 1/3 of the price.

[Karel]

A clear sign that things have gone back to normal is when the RPI4 will be plenty available for the normal price.

[paulca]

A clear sign that things have gone back to normal is when the RPI4 will be plenty available for the normal price.

Not when the audiophiles are buying/selling RPISnakeOil "network streaming FIFOs" for £500.

[tom66]

A clear sign that things have gone back to normal is when the RPI4 will be plenty available for the normal price.

RPi has faced shortages well before Chipageddon.  There is an argument that it is a little underpriced given the demand they receive for the product but it was originally an educational 'toy' and it's since been taken up by loads of SMEs as an easy way to embed Linux into X (whether or not it really requires it.)

[TomKatt]

RPi has faced shortages well before Chipageddon.  There is an argument that it is a little underpriced given the demand they receive for the product but it was originally an educational 'toy' and it's since been taken up by loads of SMEs as an easy way to embed Linux into X (whether or not it really requires it.)

Indeed.  It seems that many Pi projects could be more economically implemented with a ucontroller of some kind.  And many other projects don't even use I/O, so you might as well just use a cheap SFF pc - at today's Pi prices, you could pick up some old Dell or HP mini box for cheaper.

[peter-h]

Another sign that distis are desperate to ship stuff they can invoice, and that few new orders are arriving, is that rescheduling drops on scheduled orders is becoming impossible.

One of UK's biggest distis (a German owned company) is now totally rigid. I've never seen that before.

[Karel]

Things are slowly getting back to normal.
My favorite devboard NUCLEO-L452RE and my favorite mcu STM32L452RET6 have plenty availability now.

[Sal Ammoniac]

I've been on Digi-Key's notification list for STM32F7 parts for 2 years now and finally was notified that one of these parts (STM32F767) was in stock. The email arrived at 8:01am today. When I checked the website at 8:02am the part was out of stock again. 

[AndyC_772]

I take it as a good sign that deliveries to distributors are starting to come through at all. There's bound to be a backlog of demand.

For I time I worried if we'd ever see some of the STM32F4 parts again, but now I can (finally!) buy them off the shelf.

Still looking Intel squarely in the eyes and asking about WTF they did to the Altera product line, though.

[jonovid]

rumours has it. from game channels
AMD is deliberately undershipping chips to help prop prices up.
https://www.pcworld.com/article/1499957/amd-is-undershipping-chips-to-keep-cpu-gpu-prices-elevated.html
and
https://www.thegamer.com/amd-undershipping-graphics-cards-cpus-keep-prices-high/

[SiliconWizard]

Other rumours also say that AMD's latest Ryzen series is starting to look like a commercial failure.

[KE5FX]

I've been on Digi-Key's notification list for STM32F7 parts for 2 years now and finally was notified that one of these parts (STM32F767) was in stock. The email arrived at 8:01am today. When I checked the website at 8:02am the part was out of stock again. 

Or the email cron job runs at 8AM while the stock database update runs at 9AM....

[PlainName]

Other rumours also say that AMD's latest Ryzen series is starting to look like a commercial failure.

I've mostly been an Intel chap but was thinking of migrating to AMD... what's the problem with Ryzen?

[BravoV]

Tesla’s plan to slash silicon carbide use sends some chipmakers’ shares down

Source -> https://www.cnbc.com/2023/03/02/tesla-plans-to-slash-transistor-use-chipmaker-shares-plunge.html

Are we expecting a tsunami of cheap SiC semiconductor in the market soon ? 

[SiliconWizard]

Other rumours also say that AMD's latest Ryzen series is starting to look like a commercial failure.

I've mostly been an Intel chap but was thinking of migrating to AMD... what's the problem with Ryzen?

Just rumours...

No specific problem, I think it's more of a healthy competition thing between Intel and AMD, and it's a good thing.
The Ryzen line has been released in 2016, so there isn't just "Ryzen". It's a whole range of CPUs across a number of generations.
With it, AMD had managed to seriously beat Intel's pants, not just in terms of cost as they used to, but in terms of sheer performance. That trend has reversed again with the latest 13th gen Intel CPUs, as the performance/price ratio is currently in favor of Intel's CPUs. Even the "modest" Core i5 13600K at around $350 is a small monster and beats the Ryzen 9 7900X which is nearly 50% more expensive, in a range of benchmarks.

No doubt AMD's gonna surprise us next time around though. That is good competition!

[Kjelt]

Dont forget the embedded processors.
Synology has choosen AMD Embedded Ryzen R1500/1600 and V1780B Processors for their NASses.

[tom66]

Other rumours also say that AMD's latest Ryzen series is starting to look like a commercial failure.

I've mostly been an Intel chap but was thinking of migrating to AMD... what's the problem with Ryzen?

No doubt AMD's gonna surprise us next time around though. That is good competition!

I love this competition.  I upgraded from an old Sandy Bridge CPU to a 3800X a few years back.  The CPUs in real-price-terms cost within 20% of each other but holy crap is the 3800X a beast.  My current work laptop has a 5800X in it, the battery life is    but for performance it is amazing!

[TomKatt]

Other rumours also say that AMD's latest Ryzen series is starting to look like a commercial failure.

I've mostly been an Intel chap but was thinking of migrating to AMD... what's the problem with Ryzen?

Not sure if it's Ryzen so much as what application you run.  We have a Threadripper 3970X system (32 Core / 64 thread) built to the hilt for rendering using an application named Keyshot and it beats the pants off an equivalent Intel system.   But try and run AutoCAD on it and expect strange hangs and crashes.  In fairness, I think AutoCAD has always favored Intel.  Still, sometimes you need to mate the platform to the applicatiion.

[Karel]

But try and run AutoCAD on it and expect strange hangs and crashes.  In fairness, I think AutoCAD has always favored Intel.  Still, sometimes you need to mate the platform to the applicatiion.

A graphics card (driver) issue seems more likely here.