News

All news

Compute Module 3 Launch!

Way back in April of 2014 we launched the original Compute Module (CM1), which was based around the BCM2835 processor of the original Raspberry Pi. CM1 was a great success and we’ve seen a lot of uptake from various markets, particularly in IoT and home and factory automation. Not to be outdone by its bigger Raspberry Pi brother, the Compute Module is also destined for space!

Compute Module 3

Since releasing the original Compute Module, we’ve launched 2 further generations of much faster Raspberry Pi boards, so today we bring you the shiny new Compute Module 3 (CM3); this is based on the Raspberry Pi 3 hardware, providing twice the RAM and roughly 10x the CPU performance of the original Module. We’ve been talking about the Compute Module 3 since the launch of the Raspberry Pi 3, and we’re already excited to see NEC displays, an early adopter, launching their CM3-enabled display solution.

Compute Module 3

The idea of the Compute Module was to provide an easy and cost-effective route to producing customised products based on the Pi hardware and software platform. The thought was to provide the ‘team in a garage’ with easy access to the same technology as the big guys. The Module takes care of the complexity of routing out the processor pins, the high speed RAM interface, and core power supply, and allows a simple carrier board to provide just what is needed in terms of external interfaces and form factor. The module uses a standard DDR2 SODIMM form factor, sockets for which are made by several manufacturers, are easily available, and are inexpensive.

In fact, today we are launching two versions of Compute Module 3. The first is the ‘standard’ CM3 which has a BCM2837 processor at up to 1.2GHz with 1GByte RAM, the same as Pi3, and 4Gbytes of on-module eMMC flash. The second version is what we are calling ‘Compute Module 3 Lite’ (CM3L) which still has the same BCM2837 and 1Gbyte of RAM, but brings the SD card interface to the Module pins so a user can wire this up to an eMMC or SD card of their choice.

Back side of CM3 (left) and CM3L (right).

We are also releasing an updated version of our get-you-started breakout board, the Compute Module IO Board V3 (CMIO3). This board provides the necessary power to the Module and gives you the ability to program the Module’s flash memory (for the non-Lite versions) or use an SD card (Lite versions), access the processor interfaces in a slightly more friendly fashion (pin headers and flexi connectors, much like the Pi), and provides the necessary HDMI and USB connectors so that you have an entire system that can boot Raspbian (or the OS of your choice). This board provides both a starting template for those who want to design with the Compute Module, and a quick way to start experimenting with the hardware, and building and testing a system, before going to the expense of fabricating a custom board. The CMIO3 can accept an original Compute Module, CM3, or CM3L.

Comprehensive information on the Compute Modules is available in the relevant hardware documentation section of our website, and includes a datasheet and schematics.

With the launch of CM3 and CM3 Lite, we are not obsoleting the original Compute Module; we still see this as a valid product in its own right, being a lower-cost and lower-power option where the performance of a CM3 would be overkill.

CM3 and CM3L are priced at $30 and $25 respectively (excluding tax and shipping), and this price applies to any size order. The original Compute Module is also reduced to $25. Our partners RS and Premier Farnell are also providing full development kits, which include all you need to get started designing with the Compute Module 3.

The CM3 is largely backwards-compatible with CM1 designs which have followed our design guidelines. The caveats are that the Module is 1mm taller than the original Module, and the processor core supply (VBAT) can draw significantly more current. Consequently, the processor itself will run much hotter under heavy CPU load, so designers need to consider thermals based on expected use cases.

CM3 (left) is 1mm taller than CM1 (right)

We’re very glad to finally be launching the Compute Module 3, and we’re excited to see what people do with it. Head on over to our partners element14 (or Farnell UK) and RS Components to buy yours today!

171 comments

Great! Can’t wait to upgrade my Slice Media Player to CM3.

Exactly my thought. Haven’t used my Slice much at all since I bought it, but hopefully with a little more horsepower, it will feel snappier and become more versatile.

Sweet!

Still hoping to see a stereo camera setup with (nearly) no sync error… anyone?

From my testing with the CM1, while the drisf is /not bad/ even without any hardware genlock, that initial synchronisation (or lack thereof) is a killer. Variance can be milliseconds, and there is no consistency (i.e. you can;t do a one-shot calibration and offset).

Sadly, the problem is intractable: the hardware to allow for proper genlocking is not exposed on the camera modules, so there is no possibility of doing a proper sync. And the closed nature of the video core (and arcane nature of camera IC tuning) means that using another camera module is not really going to happen. Raspberry Pi is not a suitable platform for proper stereo video.

Is it really that significant? There has been a change a little while back which may have slowed it down, but it should only be one I2C transaction of a 24bit write between the two “start streaming” commands.
There are options that can improve this and maintain long term synchronisation – I’ll stick it back on the to-do list to investigate.

I built a 3D camera using two Pi 3’s with 8 Mpx Pi camera modules. The camera prototype is capable of capturing acceptable 3D images of moving objects such as birds in flight. Some stereo pairs are slightly out of sync and look weird, but most look great.

I’m interested in any information you you would like to share about stereo photography with Raspberry Pi.

If you are interested in more information about my project, I can send some VR images and pictures of the prototype device. I also have some excellent 3D pictures taken in a model railroad museum. The depth of focus is amazing.

I’ll be working on 3D video in the future.

Sounds like a lot of work has gone into the redesign.

I like the sound of the development board being able to act as a Raspberry Pi itself. There’s a big jump between in the complexity of creating a custom PCB for a compute module and this now provides more flexibility in designing around the compute module before committing to a PCB.

It’s too late for my current project where I’m having to use an Arduino to provide extra I/O, but I’m sure it will be useful for future projects.

Thankyou, I have been putting off ordering a heap of CM’s in the hope I could get CM3’s

Do the camera connectors still need the conversion header, or can we connect them direct ?

It appears to use the same fine pitch connector as on the original CMIO, so that it can carry all four data lanes on CAM1 & DISP1.

I guess you’ll be able to use the Zero camera flex cable in lieu of the adapter boards.

I beleive this will be a game changer, and as you said, now knowledge can much more fairly compete with big guys…Thank you for your all efforts.

Hi, As a newbie I have a question…
This is an upgrade module for Raspberry Pi 3 to 2GB RAM and 2.4GHz CPU?
Thank you!

No, this is an entirely standalone product, designed for hardware developers.
There’s no way for any of the Raspberry Pi models to be “upgraded” with a faster CPU or more RAM.

Will this new module be more readily available than the Pi Zero that, well over a year after its launch, is STILL only sold one-at-a-time?

Yes, as (from what I can tell) this is being produced by element14 and RS Components themselves, who can afford to do much larger production runs than the Raspberry Pi Foundation, who are the only people producing the Raspberry Pi Zero.

Just out of curiosity, with the demand, why has that not changed? Lets face it, people want the zero, and they want more than one. I have a dozen or more uses for them and as a result of the restricted supply for many of those uses I will use a different platform (likely ESP32 or ESP8266). Heck I would be willing to pay $8-$10 USD for Pi Zero if they would remove the purchase restriction so I could order them affordably online. As it is now I can pay $20 and order online, or I can drive 82 miles round trip and buy 1 per household at $5. I understand the Pi foundation is non profit, but they sure seem to under-stripe demand a lot.

The Pi Zero is a fiasco…the one-per-customer limit with shipping costing more than the board more than a year after release makes no sense. Why not make a deal with a supplier who can produce and supply them in volume?

Would you be happy to pay double the price, or more, for that to happen?

@James Hughes YES! For goodness sakes, YES!

Pi Zero: £4.50.

Pi Zero ‘bulk order’ price by your suggestion: £9

Pi A+: £20 and twice the size.

If nothing else, it would one-shot-kill the scalping market that has made it so hard to actually get a Pi 0 at times.

The compute module is available to buy in single units, hundreds and thousands.

.don’t forget 10’s ;-)

Yeeeeeaaaahhhh!!! At least!!! Great to hear we’ve got CM3!
But price for IO board (for CM3), shot me down to the ground :( . I know it’ll be more expensive than even RPi (PCB is bigger) , but when I see price 96£ (@RS) I can’t help but think it’s rip-off (while complete Dev Kit with CM1 cost 65£ @ RS ; not to mention that farnell/element14 added about 25% to the price just for converting it to my currency – PLN ).

It would be great if RPF would tell publicly how much cost the IO board for CM3.

If you look back when the CM1 was released the CM development board was more expensive. It’s a complex piece of equipment albeit for a single purpose.

It’s a sure bet that the price of the development board will come down in future.

As Dougie says, it’s a dev board. If £96 is too much for you then this product is not for you. If you think you can create a product using this then how can you afford to get it made it if you bulk at the price of the devkit.

I’ve worked on arm dev kits for years. This is very cheap. Normally 100’s if not thousands of pounds for devkits. (proper ones, not maker targeted ones). And this is a proper devkit.

What’s more, the dev kit is a sort of catalyst. Use it to develop something and the carrier board is unchanged and ready for the next project. So CMs just kind of flow through.

The original price for the CM1 Dev Kit (CMIO, camera and display adapters, a CM1…) was $230.(The CM1 was $40, so the drop in board price is nice, too.)

If the CMIO3 is that offputting, get a CM1 Dev Kit for $75 and add a wire to it to provide more VBAT current to the module. Search around in the Compute Module sub-forum for how to do it.

Perosnally…I’m hoping that the WD SATA adapter will be able to supply enough current or–if it can’t–that WD comes out with a new board very soon.

And next up…we get to wait for the Pi3A (there is always something to anticipate…).

If you want best value, you can combine the I/O board, CM3 & PSU for a better “kit deal”
For an embedded design, the I/O board/or kit provide you with the basics to set up and be developing a powerful embedded solution very quickly

Nice, now, please start development of something like the Pinebook ARM where a compute modules can be clicked in, turning it into a full Raspberry Pi laptop!

That’s more of a third party development, not really of interest to the Pi team themselves. You can always make one yourself!

You could say the same of the touchscreen ;)

No, not really.

I would love something like this! I have been using a Motorola lapdock for a long time, but then the HDMI connection broke due to physical stress (plugging in and out and so on) – a solution with a CM would be great!

Great News!

Thanks for all the efforts

But i still can not see the justification of the price point of (even CM1) being 25$ while we can have Pi Zero @ 5$ with the same chip and several extra connectors on board.

I believe the cost is a significant factor why CM cannot find itself in many products, especially since a lot of SoM rivals are beginning to appear with lower price tags.

The Zero is a hobby/teaching board, the CM is designed to be put in product. It’s also a quad core, compared with the Zero and has on board flash (not the Lite). The Zero does not have guaranteed production, is only available in small quantities, and does not have a guaranteed form factor. Schematics are available for the CM, not the Zero. These are all things you pay more for.

It’s not that the CM is too expensive (it really isn’t when you look at the spec of it compared with a lot of other industrial devices), it’s that the Zero is too cheap!

Thanks for the reply James!

I meant solely from the BOM cost perspective, CM seems expensive compared to zero with 5x price tag. ( maybe Pi zero is being subsidised by the foundation as an educational platform, that I do not know)

Nonetheless, agreed that Zero is too cheap.

Cheap enough to raise this question of mine :)

The Pi Zero isn’t subsidized. It just has very low profit margins. So low that Farnell and RS don’t want to carry it. Note that a CM3 (at $30) is actually cheaper than a Pi3B while having the on-board nonvolatile memory. Even at the old (current at one vendor I checked) price of $40, the CM1 still has good value.

Go look at other SBCs that have on-board eMMC and comparable CPU specs to the Pi3B. You may be surprised.

Margins would be improved if they sold them in bulk! Right now, all of the margin goes to the shipping company.

@david ALbert. No they wouldn’t. The margin is paper thin. It not going to get bigger just by selling a lot of them. Price is the only variable that can be changed here.

Keep in mind the Zero was also specifically designed to be as cheap as possible to *assemble*, while this was not a restriction on the CM3, whose primary goal, I suspect, is reliability as part of a larger product.

Among other things, the Zero has a mini-HDMI connector in part because the full-size connectors require work to be done on both sides of the board, while the Zero is designed to be assembled from the front side only, if I recall correctly.

CM3 and CM3L with guaranteed availability until at least to 01/2023, and all the documents…
thank you very much!

Did you look in the documentation? All this information is in the datasheet…

Sorry, I thought you were complaining!

Bad Gordon…

you are wondering…
but you see, i will also give positive feedback, from time to time. :D

yes, it planed as to be a positive feedback – and not as complain

yes, it was planed to be a positive feedback and not a complain.

even when you could not beleave, but i also can give positive feedback – from time to time :D

Is the CM-3 compatible with the previous compute dev-board or is it necessary to get the new board due to power requirements?

The original dev board drives VBAT from 3V3 which is marginal for CM3 as it can take a lot of current under heavy load, but it will work OK if nothing else is using 3V3. We recommend you mod the board to drive VBAT from the 5V input for best results.

There’s a new raspbian release out since 2017-01-11, but no release notes. Might well be that the release notes were kept back not to reveal the CM3 support before time.

Now the news is out, you can release the release notes :-)

The only note a relevance is that we changed the FAT partition to be FAT32, this is because there is a problem when booting from an eMMC device with a FAT16 partition.

Yey! I was just hesitating about getting my hands on the old CM.

I know what my next acquisition will be =o)

Hey Gordon, is this currently compatible with slice?

Yes,

The Slice hardware works fine with the Compute Module 3. You should probably add a heatsink since everything is contained inside the box (I’ve gone a bit overkill and connected the CM3 to the case with a slug of copper!)

Gordon

I presume from that comment that subject to power drain and thermal requirements, the CM3 will work in the original Compute Module IO Board V1?
Sorry If I’ve missed that in the blog posting.

Are there any plans for the Slice software in terms of CM3 support? Is there a software stack (beta or stable) that will allow me to use Slice with CM3 as a media player?

Luckily, the CM3 (just like all other models of Raspberry Pi) is software backwards-compatible.

There is a LibreELEC ‘alpha’ build that works very nicely: https://forum.libreelec.tv/thread-1893.html

If the Compute Modules can be inserted into touch-screen units (whatever the manufacturer) … job done …

I would just take a suitable carrier board…

OMG Fabulous

Very exciting! Congratulations on a job well done.
Now with these specs, Quad cores and 64-bit, who’s going to release that 1U chasis that can slot hundreds of these babies in a for a cheap cluster? Or maybe I’ll have to do it myself…

What is needed is a motherboard that has a RAID disk controller,
graphics, NICs, and numerous (8-64) CM “memory” sockets. Perhaps, a RPi3 could be utilized for a lot of these functions,
via GPIO. The result should be multi-fold: 1) The teachers computer, 2) The schools computer, 3) Low cost server, 4) Scientific Compute Server, 5) many more…

Congrats for the Raspberry Pi Org!!!

A couple more data points….

For those complaining about the price of the Dev Kit at 98GBP (basically, $120), the US subsidiary of Farnell is listing their kit at $200, though it includes both a CM3 and a CM3L. Given the prices of the modules, that means they are charging $145 for the rest of the kit.

The other main US supplier, the subsidiary of RS is selling a kit for $136, but no indication that it comes with a module.

The Farnell subsidiary has a few of each in stock, with more expected in late April. The RS subsidiary doesn’t have any.

And the word doesn’t seem to have gotten to the Farnell subsidiary about the price change on the CM1. They still list it at $40 in unit quantity.

I ordered CM3 gear from both Farnell and RS France on Monday.

The CM3 dev board I received yesterday from RS France does not include the compute module, nor does it include documentation, the CMIO and camera display adapters and power supply.

The Farnell CM3 dev kit is complete and includes a CM3 and CM3L. The CM3 is mounted on the dev board carrier. Just received this morning although Farnell had indicated availability from 5th February only!

I also ordered a CM3 and CM3L separately from Farnell which arrived yesterday however I received two CM3s even though the packaging and delivery note had the correct references for CM3 and CM3L. Bit strange.

RS and Allied offer the full suite of CM3 products.

The CM3
The CM3 Lite
The I/O board
The CM3 Dev kit (Includes I/O board, CM3 & PSU)

I am so hoping the next Raspberry PI model ships in two models. One in the $35-40 range, and a second in the $99-$120 range. The world is ready for high-end ARM processing, and while the smaller SoC’s are fun, I think its time for some serious processing power.

If the high-end model could be 15-20 times faster than the RPI3b, this would revolutionize computing as we know it. Irrevocably so. And I for one would order 5 of them immediately, especially if they are software compatible with the RPI3b (either way I would order it without any hessitation).

Please RPI foundation, give us a high-end model.

Is there an ARM core design that can actually go that fast?

And in answer to my own question, there is not an ARM chip that can go that fast. Might not even be an x86 that can go that fast!

Might be able to double or triple current speeds with current tech, but even that is a stretch.

We see fast server grade chips (E5-2630v4 – burst to 3.5Ghz) at about 10x the single core performance of a single core in the Pi3 (1.2Ghz ARM). So a £5000 server (20cores/256GB) has similar total CPU to about 50-60 Pi3s (£1800), but dramatically better single core performance, RAM (up to 1TB), network (Nx10Gbps) and disk/SSD and considerably fewer connectors to get it up and running.

However, if you want to practice with distributed computing where network / cpu / io limitations are real, the Pi3 makes an astonishingly cheap testbed.

The foundations goal is a system that is cheap and ideal for learning activities. They have achieved that goal and some. There is no need for them to loose focus and try to produce something that tries to compete with the Intel based PC market.

That can be done by someone else whilst the foundation continues to change the world with their little chip that could. :)

Very glad to see that this is finally available right on time, per Element 14 posting the product pages about a month early, with today as the estimated initial ship date under the availability pop-up. Watching to see what will be available on Pi Anniversary of March 1 (or Early Bird special on February 28th? ;)

Time to spin up a GoFundMe page for my classes, school, and districts!!!

Thanks for the memories … double the RAM, and the option to flash, or not to flash, that is the question! :D

After checking out that link, I want one of those NEC displays. So many possibilities with all that Pi power and flexibility for a home entertainment center.

Congratulations, team!

My comments and questions: will (in the future) be any of the Broadcom chips produced on a finer process, for example 28nm? AFAIK the architecture of the chip is hacked to the max, it can’t address more memory physically, so a nice upgrade would be to have it a little bit more efficient. RPI3 can get hot easily. Also finer manufacturing processes increase speed, maybe we could have RPI3 at say 1.6GHz? And those compute modules in smart TVs wouldn’t have to throttle…

My other rant is about Pi Zero, it’s a great value, I’d buy it even if it had twice the price tag. What about adding the wifi chip from RPI3 and make it a little bit more expensive yet more attractive to partners? Price between RPI1 B+ and ZERO?

Maybe innovate the RPI1 model B+ and add the wifi chip to it too? Would be a nice IoT HUB where RPI3 is too power hungry… and expensive.

Chip design and process is down to Broadcom, not the RPF. As for future Zero models, who knows. It’s a question of development time and resource. I can think of better things to spend money and people on!

Well, since the Zero still uses the old Broadcomm BCM2835, which is going out of production, the RPF will have to either let the Zero go the way of the Dodo or it will have to upgrade to the BCM2837 of the RPi3/CM3.

I wouldn’t necessarily want a Zero with onboard WiFi/Bluetooth. An updated version with the BCM2837 and more RAM would be nice, though.

What I REALLY would like to see, is a RPi3 variant with onboard SMA or U.FL connector for an external WiFi antenna, ’cause the onboard antenna is really bad.

A RPi3 Model A would be nice, too.

Thanks for clearing up that mixup.

Nonethelss, a more powerful RPi Zero with more RAM and a modified RPi3, like described in my last post, would be very appreciated.

https://dorkbotpdx.org/blog/wramsdell/external_antenna_modifications_for_the_raspberry_pi_3

https://hackaday.io/project/10091-raspberry-pi-3-external-antenna

Not everybody has the fine motor skills to solder tiny smd components off and on a circuit board. And besides that, these modifications void the FCC certification, thus effectively making operation of the modified RPi illegal in many countries.

BCM2835 is end of life? Source? Google had nothing to say on the subject.

It’s not. :)

The zero is fine as it is, no need to add more skews and fragment the market. The RPi3 is too expensive??? What planet you on?

None of your suggestions make sense and you seem to be regurgitating phrases from PC gaming forums. You need to go away and understand the platform and the goal of the foundation.

Their goal is not to create media centers, that is just a happy product of their activities.

NO, I wrote “where …. is expensive”

One of our local power distributors seek a solution for its 200k customers – have a limited budget, where RPI3 with it’s onboard WiFi/BT would be too expensive, had unnecessary number of cores and RAM etc.

RPI0 would be fine price-wise (w/technological limitations), but’s not available in volume and as it seems, is a purely educational tool.

My other guess is at the lesser one-core models i.e. RPI1 B+ or A+, but the usage scenario might require WiFi and BT-LE, because of sensor integration (the whole thing should be a home gateway reading power usage and other power related customer patterns), essentialy a block in a smart power grid.

Hence the need to have basic RPI with WiFi/BT. A+ or ZERO model equipped with USB wifi/BT would be ok, but that wouldn’t have a single USB port free. And on ZERO it would be clunky to add WiFi dongle via a short cable from micro to big USB connector.

So the closest would be B+ or A+ with WiFi.

I know, connecting the solution via an ethernet cable is the best way, but again, how about BT LE? That’s meant as a basic way to connect to home sensors. Other sensors farther away would be connected via a different wireless technology (868 MHz , 433MHz, …)

I’ve seen an RPI customisation and manufacturing service on Farnell website. I wonder, whether they can add the WiFi/BT chip to the mixture, maybe with U.fl connector (my customer would then certify it itself regarding wireless emissions. The closest base RPI for this would be RPI1 B+ or A+. Maybe time to contact them…?

Making our own motherboard for a compute module is the other possible way. But I doubt the price of this solution would meet the target.

Please note the intent: the power distribution company has a goal, and my first guess what to recommend to them was Raspberry Pi with it’s hardware and software ecosystem.

They can pay some company to make a specialised non-expandable product for them, which will end up an wastebaskets in 5 years’ time.

My idea was to select a base model which would be distributed to the customers. Operating system would be on a microsd card as usual, so the upgrade if anything on the software side went wrong would be cheap. Whoever wanted to pay a premium (the base model is expected to be paid by the company) could get a RPI3 with no modifications. And the children would get maybe a first touch with RPI for free (just change the SD card and go), the thing doesn’t have to end in trashcan, it can be reused. The software ecosystem and openness allows anybody to expand it. So the children could get a home gateway which they can hack.

So I’m trying to choose what the base model would be.

Lada, that power distributor should instead talk to V2COM (v2com.mobi), they have what is needed with very affordable cost (and wireless communication that can include Zigbee, Wifi, GSM, LTE or even some proprietary mesh protocols). They have an office in Austin.

(I am in no way related to V2COM, just got to know them very well)

I want Raspberry Pi Zero.

Then buy one, it isn’t hard these days, unless you live in some faraway land.

Thank you for an excellent progressive evolution of the CM with full backward compatibility. This is a big step up for small systems manufacturers and adventurous experimenters. Should be ideal for for robotics and quadcopter control systems.

Richard

Love this. Thanks very much to the entire RPi team!

In the past, the CM Dev Kit has always been out of my price reach, as high as $499 here in the states with the CM Module being seen at a minimum of $299. Seeing the CM Module going for $30 is great. I’m just worried about the price of dev board itself. I would love to get a couple kits to further my robotics program.

Where in the US are you shopping??? The Original Flavor CM dev board with a CM has been available from RS Electronics for the better part of a year via their export site for about $54, and the CM3 I/O dev board is priced at around $175 with a CM3, when it actually becomes available to ship in volume (by April?).

The highest price I have ever seen for the CM1 Dev Kit is $230. I’ve seen them as low as just a bit over $50. At present, they’re $75 (or thereabouts).

It looks like Farnell and RS are going in different directions with the CMIO3. You can either get a “full” kit that includes a CM3 and a CM3L (for $200), or you can get everything except the modules for less. Depends on your source. Newark/Element 14 and MCM are going the $200 route.

I know that I’m being lazy, but if I were to use this over a Pi 3 could I use it as a media center?

Also in the same vein of laziness, how better performing than a pi 3 is this?

It’s the same SoC as the Pi3, so performance is the same.

When will new Model A be out? That’s what I expect.

I hope not. No need for it to be updated. It’s lower power usage is ideal for robot projects and other solar / battery projects.

You don’t need quad core and a gig ram to update a mqtt server with weather details.

My old A+ with 256 megs of ram can run the OS in ram (no swap), the SD is read only! And I still have plenty of space to run the web server.

But a new A model with integrated WiFi/Bluetooth would be neat, wouldn’t it?!?

We don’t know about WiFi/BT (though it is certainly possible), but there is supposed to be a Pi3A in the works. Eben mentioned it in the same interview–last March–as the plan for the CM3. The Pi3A is supposed to be launched after the launch of the CM3. One might even hope that it will be official for the 5th Anniversary…

Those factors suggest that the A+ will continue to be a viable product after the Pi3A comes out. After all, you can still buy B+ boards even after the Pi2B (and Pi2B2) and Pi3B.

Its the right time to look at a RPi-0 upgrade.

BTW, till date, there’s not been a single competitor to RPi-0

Please upgrade RP-0 also

Why update the Zero when it has no competitors?

Why did the RPF improve on the initial RPi (A and B) before any competitors were available?

‘Cause there is always room for (evolutionary) improvement!

There is an old proverb that is attributed to french philosopher Voltaire and goes like this “That which is better is the enemy of that which is good”.

Translated freely, this means “never stop improving on your already good product”. And that’s what the RPF hast done with all incarnations of the RPi, so why stop with the current version of the RPi Zero?

Why update the Pi Zero when it has no real AVAILABILITY?

I (and I suspect many like me) would happily pay more per Pi Zero if we could buy them in quantity. Here’s a novel idea: $5 for one, $15 for two, $50 for five (and per five upwards).

Pi Zero ‘Reginald Perrin Edition’, anyone?

If you are willing to pay more for the Zero, why not just buy the A+? It’s three times the price, but you get a much easier to use device.

That aside, it would be a logistical nightmare to implement your suggestion.

The Zero isn’t attractive purely because of it’s price – its the smaller size /form factor that makes it much better than an A+ in many cases.

Limited availability means I don’t consider the Zero for projects, where I’d happily pay 2-3x

Western Digital are making some nice custom Compute Module carriers to fill this gap:
http://wdlabs.wd.com/products/media-stick/

Slightly more than 4x, actually! :)

You can get as many Pi Zeros as you want, so long as you are willing to do it slowly (I’ll get to a way that you can do it quickly). They are generally in stock at Adafruit, PiHut and Pimoroni. Order one from each every day. That will get you 3 per day at about $10 each (counting shipping).

If you need a bunch all at once, and are willing to pay what it takes, WD does not appear to be limiting order quantities for the PiDrive Node Zero…though WD wants $45 each, but it does include a Pi Zero.

If the effective price of the RPi Zero is 10£ instead of 5£, why not split the difference, raise the price to 8£ and give Farnell and RS enough of a margin for them to want to manufacture and sell the Zeros and thus make large numbers available to everyone?

Then head over to the local Microcenter. This is exactly what they do for in-store purchase:

1 at $5.00 each
2-5 at $9.99 each
6+ at $12.99 each
http://www.microcenter.com/product/463189/Zero_v13_Development_Board_-_Camera_Ready

What’s that you say? You don’t LIVE near a Microcenter>

Well poo. Then storm down to your local electronics store and yell to the manager:

“You, sir, go open a line of credit with this Raspberry company and sign a contract that GUARANTEES you will buy 1000 Pi Zero’s every single month for the next 5 years. With that, the company can confidently invest and hire people dedicated only to get these Pi Zero’s out the door to us genius h#ck3rs since it is now known there will be a steady stream of orders. Convince him, sir, if you do this you will become a rich b@$t*rd!”

So to all the complainers, the gist is that not everything rests solely on the Foundation. Go rant to your vendor instead, not here where it won’t help much. But if Microcenter is currently doing this, then anyone else can as well. People simply need to put their $$ where their mouth is.

Or, you know, go pack your bags and move to the nearest Microcenter.

Favourite comment of the day. Thank you.

Since I’m German and you British seem to have become a a bit xenophobic, I doubt I’d be welcome to live and work in GB even with my MSc in computer science.

So moving near a Microcenter is not an option.

But excuse me for threatening to buy a whole bunch of PRi Zeros for my Home Automation project.

How can this be so difficult? Obviously the current price of the RPi Zero is so low that nobody aside from the RPF is willing to even manufacture the damn thing.

This leads to the one RPi Zero per customer per order limit at every online shop, which in turn leads to having to place several orders of one RPi Zero each, causing a huge amount of shipping fees.

Raising the price of the RPi Zero just a little would fix this problem (by allowing RS and Farnell to make a profit) and make a high volume of RPi Zeros available to everybody.

That’s logic 101.

As it happens, we employ people of many different nationalities, from Europe and from much further afield.

We don’t employ people who have attitude problems, though, wherever they happen to come from (and although we’re quite used to people making uninformed assumptions about business models, I will admit that we find it a little aggravating); and we don’t tolerate rudeness and ad hominem attacks in these comments.

The rule here is a simple one: if you wouldn’t say it around my kitchen table, you don’t get to say it here. Accusations of racism (sorry – “xenophobia”) are out of order. Please reel it in.

On top of that, Microcenter is only in the USA, not the UK.

Quote on RPzero availability:
“Or, you know, go pack your bags and move to the nearest Microcenter.”

Or do something really modern: start a kickstarter campaign! Get enough people with sufficient numbers of RPzero’s in (think 20k and higher) and contact the RPF for a big batch order. Understand that the US$5 price tag is at the factory gate in Wales, so the unit price will be higher once shipped, but you’re no longer dependent on the likes of Farnell and RS ;-P

Best of all, you get to deal with the nice Foundation people *and* if you promise to behave, they might even invite you to their ‘new’ offices for a coffee with a raspberry. Not a pie though, I guess they’d be fed up with that by now :-D (on second thought…. ;-) )

Dutch_Master, if you are talking about just a large one time order of 20K Zeros, then this may not get the results hoped for. In fact this may cause such a “shock” to their current production system, i.e. to quickly hire and train temporary staff, set up new expensive equipment and look for additional parts. It may not be possible to complete this order if the expectation from the buyers are for immediate Zeros in hand.

So the keyword is “commitment”. As in over a long term time frame. If instead the campaign were to give a legally binding promise of 20K over 2 years (800units/mth), depending on the total cost to get it running that scenario could possibly make sense to invest in the necessary resources to add to the production line.

I hope not. No need for it to be updated. It’s lower power usage (compared to it’s big brothers) is ideal for robot projects and other solar / battery projects.

You don’t need quad core and a gig ram to do balloon flights and capture videos.

The zero is a rare product, they got it right first time!

“PLEASE DON’T EVER CHANGE IT” :)

Some people would say that they got it right the *second* time (third, if you go by the version number) with the addition of the CSI connector.

What would you upgrade and still be able to have them sold for $5 each?

If CM3 is really intended for “professional” or “industry” usage, what about secure boot / trusted boot features? As long booting by USB is always an option image access / modification can’t be prevented at all or am I wrong?
Where can I find security related details to the new CM3?

Thanks

Jens

USB-boot is only available if it’s been specifically enabled https://www.raspberrypi.org/documentation/hardware/raspberrypi/bootmodes/msd.md
And even then it only gets chosen if the SD-card / eMMC itself is unbootable.

And of course if you’re that concerned, you could just design your CM3-using-product to not expose the USB port.

We are showcasing the first NEC Displays with the integrated Compute Module 3 in just a few weeks time in Amsterdam. (The show is called Integrated Systems Europe taking place at the RAI Amsterdam on 7-10 February 2017.)

We will show amazing applications for Flight Informations Boards, IP-Streaming for Signage, Digital Menu boards for Restaurants, Digital Posters for Retail Stores and even Interactive Whiteboards for education and meetings. Hope to see you there.
https://www.iseurope.org/

Looking forward to getting more details (specs, pricing, availability) on the new displays. The last news about what’s to come was a press release back in October.

And pricing on the CM3-16.

Would you mind to give some details about how security support is implemented by the CM3 and available linux distributions?

With this being a new product launch and all, any chance I can sign up to get a demo unit shipped directly to my house, to kick the tires as we say. Preferably the 98″ model.

Hi Mento,
where are you from?
Could you kindly have a look on our website
https://www.nec-display-solutions.com and contact the sales office in your region. You’ll find the details under “Company” -> “Contact”. You can discuss demo options with the local contacts.
Sorry to say that larger sizes than 55″ will only be available later this year. ;-)
We are starting with three sizes: 40″, 48″ and 55″.
We hope these size also work for testing.

Broadcom tend to manufacture on demand most of these chips, so as long as you keep ordering high volumes and as long as the lithography processes are still mainline they can produce whatever they need. Given the high number of units that the RPG must be ordered, Broadcom wouldn’t easily EoL anything.

Is the cpu/ram on this faster and more powerful than the RPi3?

The same.

I’m happy with CM3, it has all what I need.
With two minor exceptions:
– it would be good to have already a proposal for an heatsink
– I can’t find a source for the SODIMM connector enough cheap: near 3€ is 10% of the CM3!!!

I see a number of posters who are having a bit of a clarity issue with regards to the CM3.

The CM3 is *not* the “next Pi”. It’s a Pi3B stripped of all the on-board connectors, reduced in size, given a general interface (edge) connector, and–for the CM3 but not the CM3L–4GB of eMMC flash added.

Think of the CM line as a sideways step, not a forward one.

It needs to have Backscatter WiFi capability. Backscatter wifi uses orders of magnitude less power than regular wifi, is cheaper to implement (no analog RF), and crucially has the same range as regular wifi. There is some loss of bandwidth but it’s well worth it given the saving in power and cost.

Actually I meant Passive WiFi, not backscatter.

WiFi is the sort of feature that belongs on the carrier board. Add it if you want it, otherwise not.

Do you think the updated schematics and BoM will be uploaded soon for the new CM3/IO board?

If you re-read the article, you’ll see that schematics are already available :)

Since I have also asked about the BOM, I thought it is obvious that I am asking for the actual schematics (which are normally distributed under design files, as Orcad schematics and BOM as an excel file)

The article links just to pdf schematics.

Let me re-phrase then, If _design files_ are uploaded please let me know.

I’d like to see someone come up with a RAMPS style board with a socket to pop in one of these as the main controller. $ GB should be more than enough to control the printer driver board and support an touchscreen LCD and Octoprint …..

They do however badly need just a breakout board that is more reasonably priced without any bells and whistles

The carrier is a dev board, and probably one of the cheapest ARM dev boards you can buy. As it’s a dev board it need to expose all the functionality.

Plenty of space in the market to design and build and sell what you suggest. That’s what I would do.

Congratulations to the Raspberry Pi Team! A new year, a new product. Last year was a great year for Raspberry Pi with all the exciting new projects the community has created and heading into 2017 I look forward to seeing what CM3 will have in store!

Nice update

I’m using a v1 Compute Module IO Board, will the v3 compute module (not lite) work in this or do I need to get the updated v3 board?

See the earlier comments about VBAT.

How does the eMMC compare to a typical SD card in terms of read/write speeds ?

Tested recently using https://github.com/hglm/flash-bench.git:
Values are MB/s, R=read, W=write, SEQ=sequential, RND=random

SD Card | RSEQ | WSEQ | RRND | WRND
———————–+——+——+——+—–
CM3 eMMC 4GB | 22.0 | 17.0 | 9.5 | 6.5
Samsung EVO 16GB | 21.4 | 11.9 | 4.9 | 2.0
Transcend Ultimate 8GB | 21.5 | 14.5 | 5.9 | 1.4
Sandisk Ultra 32GB | 21.1 | 8.4 | 4.9 | 1.1

Hi.

Does anyone have eagle or kicad library with so-dimm dd2 socket? I am trying to design my own carrier board but I cannot find that socket anywhere. Or maybe ddr3 so-dimm socket – I am considering different SoM?

You might be able to find (or ask for) useful info here:
https://www.raspberrypi.org/forums/viewforum.php?f=98

I developed a KiCAD script that generates Dual Inline schematics: https://github.com/Zinglish/kc-di

So far it doesn’t do footprints, only schematics. I am actually looking at using CM3 in a project of mine so I might update the script to do footprints too at some point

Hows that?
https://github.com/JoanTheSpark/KiCAD/blob/master/footprints/Conn_TE-DDR2-SODIMM-0.6-200P-doublesided.kicad_mod

If you got any questions, show up at kicad.info

It would appear Farnell are having some difficulty telling a Compute Module 3 from a Compute Module 3 Lite. Ordered the former, got the latter in packaging declaring it to be a full blown Computer Module 3. Slice upgrade will have to wait a while longer….

I guess it’s an easy mistake to make, with both of them saying “Compute Module 3” on the silkscreen. Only way to tell is by looking for the eMMC on the back of the board.

I spoke to them about this yesterday (it’s been happening to a lot of their customers); they’re on top of it now. There was a receiving error in their warehouse on the first delivery of both varieties of CM3, so the containers got mixed up. If it’s happened to you, get in touch with them and they’ll sort it out immediately.

Can I put this SODDIM into my PC where my memory lives and it will work or will I need to write a driver for that?

What gave you that idea ? :-(

You will probably blow up both your PC, AND your CM if you do that!
The CM is NOT a memory module! It just uses the form factor and availability of connectors.

The idea came to ensure that people do not do it.
Thanks for the answer.

Wouldn’t it be great to slot one of these into a Dr Who sonic screwdriver! Question is if it would fit…

No ethernet and usb hub controller, no power supply on board. very “useless” device. At the end also way much to expensive for what it offers.

For a more industrial device, for us it was important to have a easy USB connection, which this module offers, but then we have again to implement the USB Ethernet controller. The Problem with normal RPI and PI Zero is that big 2,54mm connector, which would be somehow OK if it would have minimum 4 pins USB downstream, all 4 USB ports are lost to put them on a Motherboard. Overall a good idea but in my opinion a lost opportunity to bring a PI into the industrial world.

If you actually read the article, you’ll understand that you’re supposed to use a carrier board with it, which gives you the flexibility to add whatever I/O and connectivity you want.

Who in the industry is using joe-average usb connectors anyway?
Real men use M8 or M12 connectors in those environments.

Great news that the CM3 family has arrived.

Shame many comments are about the ZERO which is totally unrelated.

The price point of the CM3 family is extremely competitve, obviously the whingers do not realise it is differing user base than the Raspberry Pi family.

Looking forward to the RPi 3A…

I’m with you, I’m not sure everyone understands what the compute module is all about. This is aimed at a totally different market. I’m thinking this will give all the students of pi a huge advantage when they move in to the world of work!

Can’t wait for RPI 3A … Like totally can’t wait!

LOL, you say it’s a shame about all of the totally unrelated Pi Zero comments (clearly indicating pent up demand). But, then you go on to say you can’t wait for the Pi 3A which is also a totally unrelated product. Something must have been lost in the ASCII translation.

Oh well, I’ll put in a plug for the Pi Zero 3 then. :-)

The text says that the CMIO3 can accept an original Compute module, but how much support does it mean? We’ll be able to extend the memory of a CM1 using the SD card reader of the CMIO3?

“External” SD-Card interface via SO-DIMM pins is ONLY available on “CM3 Lite”, not on the “CM” module (with internal 4GB eMMC), right?

So will this be a simple way to have a cluster, assuming an appropriate SO-DIM board?

I have a number of industrial applications that I’d consider the CM3 for. For example, I have 120 Eurocard subracks containing 16 x 16-isolated-channel digital I/O cards. It has a PSU module and an I/F card that interfaces the I/O cards to an ancient industrial data bus which I want to phase out. I have huge investment in the cabling and the I/O cards are reliable, so a PLC is not a cost-effective alternative. But redesigning the I/F card would work. I could replace it with a Eurocard containing glue logic, an Ethernet port and a CM3. But why would I? I have to design the Eurocard and may as well incorporate the processor and memory directly onto the new pcb. I would eliminate a connector thereby adding reliability. Sure, I can see the CM3 as being useful as a development breadboarding component, but of no advantage when I go to even small-scale pcb production. What am I missing? Are the processor and memory components not available (or too costly) in quantities of 100-200?

The processor is not available in small quantities.

And I guess anything < 300k per year would qualify as small quantity in Broadcom dictionary :)

Even if it were available, the CM handles the high speed design part, so you would have faster time to market by designing a simpler breakout board with the needed peripherals for your application.

I think there is a customisation service (by element 14 ?) so you can get the chip with custom design, but even that is not a feasible option for quantities like 100-200 since there will be an NRE cost attached.

Thanks BG, that makes sense!

Does the CM3 have built in Wifi and BLE like the Raspberry pi 3? I couldn’t locate the Broadcom BCM43438 IC on the reduced schematic.

thank you in advance!!

Nope. Having those as standard would increase the part cost, power consumption and introduce potential security issues. Anyone building a device around the CM can add their own SPI or USB wifi components to that baseboard if they need it in their product.

Is there any documentation regarding addition of the BCM43438 on the baseboard? What do I need to keep in mind to maintain compatibility with the drivers currently available for Windows IOT and Raspbian?

Initially I thought I’d just check the RPi3 schematic and hook up my BCM43438 to the same pins, provided they’re available on the SO-DIMM connector, but as ali has noted above, this part of the schematic is missing.

Would be exciting to have the PIN assignment according to the SMARC 2.0 Standard. Are the any considerations towards this direction?

Any idea where to pick up a right angle connector for it like in the Dev board? I would like to use it in handheld devices instead of ripping off the big components of a pi3 but it would require parallel mounting rather than the standard perpendicular sodium slots.

Is the original compute module io/dev board compatible with the new compute modules? I own the original dev kit and would like to try the new modules without having to buy an updated io board. Is this possible?

Comments are closed

News

Latest posts

OnlyTrains: for fans of trains, and only trains

Meet Brian Corteil: CamJam event runner

This massive nose sniffs things then prints a description of the smell

Paragraphica creates "photos" based on your location

Next Post

Book-wrangling at Wordery with a Raspberry Pi

Previous Post

Community Profile: Tim Richardson and Michael Horne

Share this post

171 comments

graycrow

JanW

szapo

psuedonymous

6by9

Dan

Segen

AndrewS

AndrewS

Richard Ahlquist

David Albert

James Hughes

Gray

Texy

Przemyslaw W

Dougie

Richard

W. H. Heydt

W. H. Heydt

James Hughes

James Hughes

AberDerBart

BG

James Hughes

BG

W. H. Heydt

David Albert

James Hughes

Carl Jacobsen

beta-tester

Gordon Hollingworth

beta-tester

beta-tester

Daniel Hrynczenko

Raspberry Pi Staff James — post author

Kiste

Chris Evans

JanW

AndrewS

Raspberry Pi Staff James — post author

Krell Freelance Technology

W. H. Heydt

Ellis T

Robert Cromer

W. H. Heydt

James Hughes

James Hughes

Pete Stevens

Richard

Jim Manley

Heyer

Lada

James Hughes

pfgsnoopy

pfgsnoopy

Lada

pfgsnoopy

Gray

Richard

Lada

Lada

Marcelo Bello

aaa

mahjongg

Elfen

Jim Manley

W. H. Heydt

Randal