Bare RP2350 chips are now available.

[brucehoult]

https://shop.pimoroni.com/products/rp2350

I didn't yet check the professional distribution channels.

But if you just want samples, 10 for £8.80/£9.60 depending on A/B model seems not bad.

[brucehoult]

PiHut has packs half the size for half the price

https://thepihut.com/products/raspberry-pi-rp2350a-microcontroller

Can't see anything on Mouser. Digikey has listings but no stock.

[SiliconWizard]

Cool, I have bought a couple dev boards with the RP2350B (48 GPIOs) a couple days ago.
The RP2350A and B are now available in large quantities at JLCPCB too (a bit expensive there, though, for some reason).

[iMo]

Is it a new silicon revision?

[phil from seattle]

Nothing at DigiKey/Mouser.

Though, there is pricing on reels/cut tape at Digikey (0 in stock). RP2350B Q500 is $1.15 USD.  JLCPCB RP2350B Q500 is $1.24 so not that different. Pimoroni Q10 is $1.25 at current exchange rate - a decent price for a hobbyist quantity.

Chatter on the RasPi forums is that it is the A2 stepping and might be shelf clearing for the A3 when it comes out.

[phil from seattle]

RasPi announcement today - https://www.raspberrypi.com/news/rp2350-now-available-to-buy-a-high-performance-secure-microcontroller-for-your-next-project/

Includes pricing.  Hope JLCPCB gets the memo...

By the way, there is an odd entry in the JLCPCB parts library - RP2350-Tiny.  The footprint is LCC-23_18x23.5mm which doesn't make a lot of sense.  No datasheet nor stock.

[Foxxz]

RasPi announcement today - https://www.raspberrypi.com/news/rp2350-now-available-to-buy-a-high-performance-secure-microcontroller-for-your-next-project/

Includes pricing.  Hope JLCPCB gets the memo...

By the way, there is an odd entry in the JLCPCB parts library - RP2350-Tiny.  The footprint is LCC-23_18x23.5mm which doesn't make a lot of sense.  No datasheet nor stock.

Gonna guess its this....

What’s next? Stacked flash is coming soon

For smaller-footprint applications, our RP2354A and RP2354B variants — featuring 2MB of stacked flash memory — are in the final stages of development and testing with early-access partners. We’ll soon be ramping to mass production, and plan to make these variants available through our Approved Reseller partners later this year.

[brucehoult]

Is it a new silicon revision?

Upton said: "This is the A2 stepping, which is affected by Erratum 9. This may be mitigated in a future stepping, and we had considered waiting until we'd done that, but demand for the part has been crazy, and it felt foolish to delay availability for a documented issue that only affects a minority of potential users."

[SiliconWizard]

Is it a new silicon revision?

Upton said: "This is the A2 stepping, which is affected by Erratum 9. This may be mitigated in a future stepping, and we had considered waiting until we'd done that, but demand for the part has been crazy, and it felt foolish to delay availability for a documented issue that only affects a minority of potential users."

As a reminder, E9 is this one: "Increased leakage current on Bank 0 GPIO when pad input is enabled".

[tszaboo]

For smaller-footprint applications, our RP2354A and RP2354B variants — featuring 2MB of stacked flash memory — are in the final stages of development and testing with early-access partners. We’ll soon be ramping to mass production, and plan to make these variants available through our Approved Reseller partners later this year.

[/quote]
No built in flash for the RP2040 coming? Why?

[phil from seattle]

For smaller-footprint applications, our RP2354A and RP2354B variants — featuring 2MB of stacked flash memory — are in the final stages of development and testing with early-access partners. We’ll soon be ramping to mass production, and plan to make these variants available through our Approved Reseller partners later this year.

No built in flash for the RP2040 coming? Why?

Pretty sure they are trying to move everyone to the RP235x series.

[brucehoult]

Pretty sure they are trying to move everyone to the RP235x series.

Raspberry Pi has stated that they expect the RP2040 to remain in production until at least January 2041.

[phil from seattle]

Pretty sure they are trying to move everyone to the RP235x series.

Raspberry Pi has stated that they expect the RP2040 to remain in production until at least January 2041.

Sure.  Chip makers rely on design wins.  To get them they have to commit to the chips being available for a long time.  But, they typically want to move their customers to the newer (and likely higher margin) designs. The 2354 is a carrot to get them there.  I worked for a very large semiconductor manufacturer for 8 years. The design win thing was a mantra.

[SiliconWizard]

If you want to know the RPi's strategy about the RP2040, I guess you should ask instead of guessing. Who knows if they're gonna release RP2040 variants with a stacked Flash die. Maybe. Maybe not. I guess it will all depend on demand and their strategy. I'm personally not convinced of the point, as the prices of the RP2040 and RP2350 are very close (except at some distributors currently, but that's probably a matter of stock management) and I guess those more likely to stick to the RP2040 are those who already have a design around it. But who knows, market will tell.

What's interesting behind this question though is that apparently, many people who were kind of sold on the RP2040 seem definitely not on the RP2350, possibly due to its errata, or the currently (relatively) "inflated" price on LCSC/JLCPCB, or a combination of both. Possibly also, as always, the perspective to have to update their knowledge and developments (if any existing) to the RP2350 which brings a number of changes for sure (but via their SDK, there's not a lot of work to port  an existing firmware to the RP2350.)

[brucehoult]

apparently, many people who were kind of sold on the RP2040 seem definitely not on the RP2350, possibly due to its errata

Personally the GPIO thing doesn't bother me at all. Just pretend it doesn't have an internal pull-down mode. Done. Plenty of chips don't -- AVRs for example, at least the ones I've used ATmega328 and ATtiny85. Floating or pull-up only.

If I was designing my own boards rather than using standard ones, then I'd be much more concerned about the weird thing with having exactly the right external inductor and holding your tongue right wrt orientation or whatever. It all seems pretty amateur.

Having near twice the performance between the higher clock speed and the full Thumb2 implementation seem pretty useful. And the FPU! And the rather nice RISC-V core, which has pretty much the same performance as the M33, though fewer features (FPU, TrustZone). More RAM is nice too. And more PIO.

If the prices are close then why wouldn't you? Assuming the RP2040 isn't already overkill for your application.

[SiliconWizard]

Yes the internal buck thing looks nasty, I don't know if they are going to fix it?
Otherwise yes, it's a great improvement compared to the 2040 (a lot more performance, double the RAM, more PIO SMs, more GPIOs with the B variant, and even the high-speed TX interface.)

[WillTurner]

apparently, many people who were kind of sold on the RP2040 seem definitely not on the RP2350, possibly due to its errata

...
Having near twice the performance between the higher clock speed and the full Thumb2 implementation seem pretty useful. And the FPU! And the rather nice RISC-V core, which has pretty much the same performance as the M33, though fewer features (FPU, TrustZone). More RAM is nice too. And more PIO.

If the prices are close then why wouldn't you? Assuming the RP2040 isn't already overkill for your application.

RP2040 was stellar when it first appeared for any number of reasons. Likewise RP2350 is magnificent. It looks to me like the entry point for RISK-V. But, by a similar argument, don't dismiss RP2040! It seems to me that RP2040 is the starting point to access the basic Thumb instruction set, and I have been playing with doing so with a self-hosted MMBasic environment, writing wrappers around the Thumb Format 0 to Format 19 instructions (Thanks Bruce!), without having to resort to the toolchain to access assembler. Kind of bare bare metal.
  What will swing the RP2350 deal for me is a Raspberry supported operating system which will provide a self-hosted compiler. With that, I may well concede defeat, and give up on assembly. So my point is that if you want bare metal access, the RP2040 is far more accessible. I could be wrong?
   

[SiliconWizard]

So my point is that if you want bare metal access, the RP2040 is far more accessible. I could be wrong?

I'm not sure I see why. The 2350 has more RAM, Cortex-M33 cores rather than M0+, more PIOs and a bit updated peripherals for some, but otherwise it's very similar. There's nothing inherently more complicated with the 2350. You can absolutely go bare metal to your heart's content.

The 2040 is going to be there for a pretty long time still, according to the RPi, so I'm sure prices will go down significantly for it in the coming months/years, so it could still be a good option if you don't need the extra performance, for cost-sensitive applications. But technically speaking, the 2350 is not significantly more intimidating.

[tszaboo]

So my point is that if you want bare metal access, the RP2040 is far more accessible. I could be wrong?

I'm not sure I see why. The 2350 has more RAM, Cortex-M33 cores rather than M0+, more PIOs and a bit updated peripherals for some, but otherwise it's very similar. There's nothing inherently more complicated with the 2350. You can absolutely go bare metal to your heart's content.

The 2040 is going to be there for a pretty long time still, according to the RPi, so I'm sure prices will go down significantly for it in the coming months/years, so it could still be a good option if you don't need the extra performance, for cost-sensitive applications. But technically speaking, the 2350 is not significantly more intimidating.

So let me get this strait: Your reason is that it has a bunch of other features, that are unnecessary for most projects, so why don't we use that? While the new one is more expensive, bigger, requires extra voltage rails, has broken GPIO structure. It's even more PITA to design a board with it than the RP2040.
Which could be alleviated with the built in Flash.
The main issue with the external Flash is that you need to recompile micropython or circuitpython every time you want to update the version, instead of just pulling a compiled binary. And if you don't use that, it's still an additional line item on the BOM, which may require firmware changes in case it's changed unlike an LDO or a button.