Raspberry PI 2 or Compute module

No, you have summed up the advantages/disadvantages pretty well although I'm pretty sure you can buy only the compute modules if you want to.

I am assuming that at some time in the future a Pi 2 core compute module will become available, the only question is will it be soon enough to help you ?

Unfortunately since the Rapsberry Pi Foundation has become a bastion of secrecy for no apparent reason (paranoia ?) it is impossible to say.

@Clem

A lot of clusters still use Ethernet (although it is more likely to be 10 GB/s and of course they also require routers) and a FPGA would be required to make a bus interface between

the Pi modules increasing the size, power, complexity and cost by a considerable amount while providing (much) less than half of the compute power per module.

No, you have summed up the advantages/disadvantages pretty well although I'm pretty sure you can buy only the compute modules if you want to.

I am assuming that at some time in the future a Pi 2 core compute module will become available, the only question is will it be soon enough to help you ?

Unfortunately since the Rapsberry Pi Foundation has become a bastion of secrecy for no apparent reason (paranoia ?) it is impossible to say.

@Clem

A lot of clusters still use Ethernet (although it is more likely to be 10 GB/s and of course they also require routers) and a FPGA would be required to make a bus interface between

the Pi modules increasing the size, power, complexity and cost by a considerable amount while providing (much) less than half of the compute power per module.

Gary,

usually the dev kit - is the producer really want to sell the product - is always distributed in the meantime as the product itself, as an alternative to simplify the engineers in their own projects. This occur with all I saw with Texas, Panasonic, NXP, Atmel and many more. Now are months that this B+ model is on the market, but there is not announcement at all of the product to be bought by itself.

Then also this super-secrecy and, sure, paranoid behaviour I think it is simply damaging the market itself. PI clones exists anyway, the conceptual architecture of this board is not so complex to be kept in absolute great secret and, despite I love it very much it is anyway in the average of many other devices. At this point, to have a reliable linux working on a well assessed machine in an industrial or anyway custom application it is the worth I spend few more money (but very few) and buy a Gizmo2 from gizmosphere, with more potential, open-hardware and really customisable in a lot of ways...

Enrico

I can't see any point in the RPI compute module because:

1) It's so obviously not the core business (as evidenced by the lack of  RPI2 version).

2) The RPI people are not a B2B supplier (so they don't understand about providing information, proper support etc).

3) The closed nature of the RPi and its processor mean that you can't just make the things yourself if you have to or if the numbers get big.

4) There isn't enough on the Compute module so you end up with an expensive interface to a support board, which itself may be pricey.

5) You still don't get proper Ethernet

6) You don't, for all that you have the huge connector, get PCIe or any other high speed bus

7) You can get all this stuff by other cost competitive routes without the supply and support issues.

So - use Pis if they suit (as they are) and you don't mind if you can't get them next year (and you don't mind having no idea at all what the development roadmap is)

MK

Michael,

I totally agree with all your points. As a matter of fact, it seems an isolated and not replicable experiment without too much sense as it is now.

What is blocking me using the PIs in serious projects is just this aspect, that we don't know what happens next year or next month ...

Enrico

@ Enrico

Yes I know what development kits are, I've been in the business for over 30 years now, and if that is the only thing that the compute model/expansion boards are supposed to supply then there is

no reason to have a separate compute module. There are no real clones of the Raspberry Pi because nobody can buy the BCM2835 SoC. You could make one with a compute module though.

There are a few good "look a likes" out there, Banana Pi, CubieBoard(s), ODROID-C1 but they all use different SoCs and are not completely hardware or software compitable.

@ Enrico and Michael

The idea is for you to use the expansion kit for development and then they sell you (lots of) compute modules that you plug into your final product adding whatever extra is needed. This is made

more obvious by the fact that you can't simply buy the BCM2835 SoC by itself, and a few have tried. Also, there are more than a few other manufacturers that use the "DIMM" plug in module form

factor and expansion boards for development for reasons described below. Another advantage to this for the Pi is the ability to use already available software.

@ Michael

1) It's so obviously not the core business (as evidenced by the lack of  RPI2 version).

They clearly meant for it to be a core business (otherwise why do it at all?), although I'm not sure how well this has worked out for them. As I said previously, there is competition and I am

not at all sure how much effort they are putting into it. If it really is to be a core business, a Pi 2 version will be required.

2) The RPI people are not a B2B supplier (so they don't understand about providing information, proper support etc).

As far as I know the only SoC information that is not already available (see reply to 3) is the USB interface in the BCM2835 since it is apparently not Broadcom IP. I'm sure if you are really serious

about using the compute modules that signing an non-disclosure agreement would get you the USB information needed. The SoC manufacturer is where most of the information/support must

come from as they are the experts and this is normally where it comes from. The compute module itself is just a PCB that provides connections to the SoC pins and the expansion board just

makes these connections more accessible for development purposes.

3) The closed nature of the RPi and its processor mean that you can't just make the things yourself if you have to or if the numbers get big.

There are supposed to schematics for the compute module and expansion board as well. I don't know if they are really available, I don't have much interest in the compute

modules. The BCM2835 documentation was released at least 2 years ago and the only thing missing is the USB interface. The GPU documentation was released about a

year ago making it the only ARM SoC that I know of where this information is available. The lack of schematics for the B+ (partial update schematics are available, they

include most of the differences between the full schematics released for the B and its revisions but they do not include any of the new USB circuits) and Pi 2 (nothing but

a big FU (pardon my French) from the Raspberry Pi foundation for this one) is a continuing source of irritation for me.

4) There isn't enough on the Compute module so you end up with an expensive interface to a support board, which itself may be pricey.

As I said earlier that is actually how the compute module it is supposed to be used.

5) You still don't get proper Ethernet

And never will without adding external circuits using the USB interface on the SoC that is not really designed for it (like they do on the Pis) and ends up giving you less than

stellar performance at 100 Mb/s. A FPGA or FPGA/Ethernet chip solution with better throughput using interface pins only available on the compute module is probably doable

but not without considerable effort and expense.

6) You don't, for all that you have the huge connector, get PCIe or any other high speed bus

The huge connector is not an issue. In fact it is very useful because it provides access to SoC pins and their functions that are not available on the Pi and the large connector "DIMM"

form factor is used in other SoC products as well for the same reason. In most cases not having PCIe is not really an issue although it would allow for providing real Ethernet. A lot of

(most?) ARM SoCs do not provide for this either. Other high speed buses pretty much require FPGA solutions in just about any ARM SoC that I am aware of.

Overall though I agree, if the Pi 2 has the I/O you need, there is really no reason at all to use the compute modules.

Hey Gary,

There are no real clones of the Raspberry Pi because nobody can buy the BCM2835 SoC. You could make one with a compute module though.

There are a few good "look a likes" out there, Banana Pi, CubieBoard(s), ODROID-C1 but they all use different SoCs and are not completely hardware or software compitable.

A clone is a clone, not an exact copy. I mean that there are lot of "variations" on the same theme. The Raspberry PI generated a sort of "gold run" and seem that many companies started a challenge to make something similar and - just based on what they say - maybe better. No replicas, and probably this is a good starting point. I don't like the too secret and closed things, it's something that tend blocking the knowledge sharing...

The idea is for you to use the expansion kit for development and then they sell you (lots of) compute modules that you plug into your final product adding whatever extra is needed.

To be honest I have always look with a certain suspect this compute board and now more than before I think that it is not the better choice anyway. It seems that all is done to avoid that engineers start developing. If there is only the development kit, I suppose that no one spend time and money basing a project to something like this.

Then the rest of your post seems we share almost the same vision. The general feeling is a sense of uncertainly.

Enrico

Not to get into a silly semantics argument but by the very definition a clone is an exact copy. There are considerable hardware and software differences

between the Pi and the "look a likes" (I know, I have several). The Pi is the one that most of the others were aiming at, some with more success than

others but none have really gotten close in terms of sales or community size. The original Pi did have a couple of disadvantages to all the look a likes

that came out later, they used more modern ARM cores, they ran faster, and they usually had more RAM but the quad-core 1 GB Pi 2 takes care of all

of them while maintaining the original ground breaking price. Unfortunately they have gone the wrong way with the lack of schematics and they have

always been very secretive about new versions or variations on old ones such as the compute modules/expansion boards.

The Pi compute modules were not by a long shot the first to use the DIMM form factor and in some cases it makes sense to use them from a product

design engineering stand point. In a lot of cases, it does not so I wonder if there will be enough interest in them to keep them in production. Without a

Pi 2 version I doubt it. Unless the compute module has some kind of hardware interface you must have that is not available on the Pi 2, go with the Pi 2.

That much is certain.

Just to disambiguate and make clear:

Not to get into a silly semantics argument but by the very definition a clone is an exact copy.

I wrote meaning the exact opposite. Then replica is wrong anyway because it is considered a replica a copy (identical) by the same creator.

There are two things that I don't understand in your posts.

The first is why you insists focusing the attention that the like-PI are not the PI, I told the same, and this is what I usually see on a lot of cases, last but not least the case of the Chipkit that aims to put itself beside Arduino with something more. No matter if they reach the goal or not.

The second is why you insiste about the DIMM form factor, this is an appreciable solution (maybe the only one). Then my initial thinking was to verify with the other users if there was really something incredibly better as I read in this alternative model or this is - for a series of aspects - the worst "model" of the PI series. If I have to decide, I'll adopt the PI2 without hexitation, it is sure. also in a project that is commercial.

Remain the same perplexity I read everywhere (you too) of the uncertainly determined by the lack of a clear line of the product, the future projects and I can assure you that also the past was a bit obscure.

     The Pi foundation is very prone to misinformation because of a cloudy crystal ball. Before the Pi 2 release I see http://www.makeuseof.com/tag/raspberry-pi-featureinterview/ which shows Eden not worried about competiton in 2013 after ramping up production.  As late as April 2014, Eben was claiming no need for a Model C because of software incompatible problems for the current Model B/B+. But then in February 2014 with the RPi 2 he admits maybe being wrong and back tracks his original comments with https://medium.com/@Xataka/raspberry-pi-present-and-future-of-the-minicomputer-revolutionizing-the-sector-6aa711252cb6.

This shows a different view with with Linux being open, but M$ coming in with the Windows 10 IoT core. Interesting!

C

"I wrote meaning the exact opposite. Then replica is wrong anyway because it is considered a replica a copy (identical) by the same creator."

Check out the definition of clone, it means an exact copy. The term replica is rarely used in electronics. Actually the PCB does not have to be an exact copy as long as the schematics, and connector layouts are.

"The first is why you insists focusing the attention that the like-PI are not the PI, I told the same, and this is what I usually see on a lot of cases, last but not least the case of the Chipkit that aims to put itself beside Arduino with something more. No matter if they reach the goal or not."

Your the one who brought up clones, not me. I just pointed out that there are no clones (exact schematic copies) of the Pi. Almost all of my replies are about the Pi, with a few references to other Pi like hardware (my entire reply to Michael is about the Pi, nothing else)

as needed to respond to your posts.

"The second is why you insiste about the DIMM form factor, this is an appreciable solution (maybe the only one). Then my initial thinking was to verify with the other users if there was really something incredibly better as I read in this alternative model or this is - for a series of aspects - the worst "model" of the PI series. If I have to decide, I'll adopt the PI2 without hexitation, it is sure. also in a project that is commercial."

No idea what you mean here. The Pi compute module is in a DIMM form factor so that you can plug it in to the hardware support you designed for the product you want to produce. The expansion board is only supposed to be used for

hardware (support) and software development of this product, it is not meant to be used in a commercial product.

There is clearly a language barrier here.

Gary Stewart wrote:

 

Check out the definition of clone, it means an exact copy. The term replica is rarely used in electronics. Actually the PCB does not have to be an exact copy as long as the schematics, and connector layouts are.

In electronics, a clone can mean something that is functionally identical, even if not an exact copy.  Here's definition #2 from Wiktionary: "A copy or imitation of something already existing, especially when designed to simulate it."

For example, a "PC clone" is a computer that's functionally identical to an IBM PC.  It's able to run the same software, the external interfaces are the same (though they may be placed differently), and you can use the same add-on cards.  It does not have to use the same chips, e.g., it can use an AMD CPU in place of an Intel CPU as long as they both run PC software.  The clone may have better or worse CPU performance.