| |||||||||||||
Enough with the Pi HypeEnough with the Pi HypePosted Jul 23, 2012 21:23 UTC (Mon) by Cyberax (✭ supporter ✭, #52523)In reply to: Enough with the Pi Hype by dlang Parent article: Hardware Hacks: The Raspberry Pi is everywhere (The H)
They are.
>I've done custom development of products and the cost to produce a board and fill it with components is just not that high, even for small quantities (dozens), let alone larger quantities (hundreds or thousands).
>In my opinion, they should not be looking to make a profit on development boards, they should view them as 'loss leaders', selling them as cheaply as possible to encourage more developers to use them. I'm not saying that they should loose money on them, but they should be at or close to break-even.
(Log in to post comments)
Enough with the Pi Hype Posted Jul 23, 2012 21:30 UTC (Mon) by dlang (✭ supporter ✭, #313) [Link]
prototyping boards don't need lots of variants, they need one board that exposes all the connections for use/testing.
as for the rest of it, I think we will just have to agree to disagree. I know that they claim to be selling them at cost, but given the ability for others to produce the same thing at much lower costs, I really have to question their accounting.
Enough with the Pi Hype Posted Jul 24, 2012 9:22 UTC (Tue) by mpr22 (subscriber, #60784) [Link] Unfortunately, a significant number of SoCs have more functional units than can be simultaneously routed out of a package of the chosen dimensions, and one option on a given set of pins may not play nicely with the device that another option expects to find at the far end.
Enough with the Pi Hype Posted Jul 24, 2012 22:28 UTC (Tue) by dlang (✭ supporter ✭, #313) [Link]
route the pins out to pads so that the person using the system can wire whichever components up to those pins that they want to use.
It's not like the developer boards offer every possible combination of uses in any case to begin with.
Enough with the Pi Hype Posted Jul 24, 2012 19:43 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link]
All connections... Do you want a working PCI bus with that? Maybe an actual PCI slot or two? How about additional I2C controllers? Or two RS232 controllers?
There are simply too many choices. And chances are most of the components won't be present in most of the final devices.
Enough with the Pi Hype Posted Jul 24, 2012 22:30 UTC (Tue) by dlang (✭ supporter ✭, #313) [Link]
PCI is a rare case where there are strict requirements for wire lengths, so yes, if the chip has an on-board PCI bus, the developer board should have a slot.
But most of the low-end chips (<$30) don't have CPI bus connections, buying a >$1K developer board for a <$30 chip offends me.
Enough with the Pi Hype Posted Jul 24, 2012 22:35 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link]
So now you want two versions: with PCI and without PCI. How about versions with different storage controllers and storage types? Ditto for video controllers or SoC versions.
So you see where it's heading - vendors have to do several versions of developer boards. And they are not exactly cheap to make.
Enough with the Pi Hype Posted Jul 24, 2012 23:10 UTC (Tue) by dlang (✭ supporter ✭, #313) [Link]
no, I don't want multiple versions with all the different options.
First off, remember that I'm talking about the cheap chips, <$30 or so.
They normally don't have PCI.
I don't want a bunch of options for storage controllers on the board. Instead I want the pins available to wire to the storage controller of my choice. The exception to this is to have something like a SD card available, but it should be done in a way that if some different storage is wired up, jumpers can be cut to disable the SD card slot so the pins can be reassigned to other uses.
The more complex you make the board, the more 'extras' you put on it, the less flexible it is and the more likely it is that someone will want something else (and the more expensive it will be)
KISS, If your developer board has more external components on it than production systems using the same chip, you've done the wrong thing.
Enough with the Pi Hype Posted Jul 24, 2012 23:36 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link]
>First off, remember that I'm talking about the cheap chips, <$30 or so.
No we don't. Developer boards are also used for high-end stuff like industrial controllers.
>I don't want a bunch of options for storage controllers on the board. Instead I want the pins available to wire to the storage controller of my choice.
>The more complex you make the board, the more 'extras' you put on it, the less flexible it is and the more likely it is that someone will want something else (and the more expensive it will be)
Enough with the Pi Hype Posted Jul 24, 2012 23:55 UTC (Tue) by dlang (✭ supporter ✭, #313) [Link]
> Developer boards are just that - they are used for DEVELOPMENT. They are not intended to be used as general-purpose computers, but instead they are intended to mimic the target system as close as possible but no closer.
that can't possibly work if you are selling an embedded CPU that is going to be used in thousands of different systems. You cannot possibly imagine all the possible system designs that engineers are going to want to put your CPU into. As such, any attempt to mimic all of them is guaranteed to fail.
you can create a reference design of a particular type of thing, but a reference design for a set-top box is likely to be really bad for a tablet developer (and vice-versa)
A developer board is not supposed to mimic a particular target system, it's supposed to give you the minimum needed to get the chip up and working so that you can wire up the rest of the system that YOU (the engineer) are designing so that you can try various things before you have a custom board made.
In any case, needing to buy a >$1K board to mimic the board that will be installed in a <$100 retail product is still wrong.
Enough with the Pi Hype Posted Jul 25, 2012 10:42 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link]
>that can't possibly work if you are selling an embedded CPU that is going to be used in thousands of different systems. You cannot possibly imagine all the possible system designs that engineers are going to want to put your CPU into.
Again, you're not producing CPUs but SoCs. They are different - a SoC includes not only CPU but also RAM controller, various IO controllers, video chip, etc.
You can't realistically take a chip designed for WiFi routers and use it to develop a media device. So you need to provide multiple variants with different SoCs and different peripheral devices.
> In any case, needing to buy a >$1K board to mimic the board that will be installed in a <$100 retail product is still wrong.
Enough with the Pi Hype Posted Jul 25, 2012 18:46 UTC (Wed) by dlang (✭ supporter ✭, #313) [Link]
> Again, you're not producing CPUs but SoCs. They are different - a SoC includes not only CPU but also RAM controller, various IO controllers, video chip, etc.
but you don't need to make a bunch of different development boards for the same SoC chip.
And if you are not hooking up the peripherals, you just need to expose the pins from the SoC, and since the different SoC chips (with different options) tend to use the same pins for the core features, this means that one board can be used for many SoC chips.
> You can't realistically take a chip designed for WiFi routers and use it to develop a media device. So you need to provide multiple variants with different SoCs and different peripheral devices.
bzzz, wrong, people absolutely take chips designed for WiFi routers and use them to develop media devices and many other things (and similarly, they take chips designed for media devices and use them in routers)
> The first pre-production versions of cars cost literally millions of dollars, even if the final versions are sub-$30k. That's the same idea - mass production lowers the price.
have you looked at how cheap it is to produce small quantities of circuit boards nowdays?
if you have the layout in your computer, you can get boards done in single digit batches for $50-100 each
the automation in manufacturing is such that you no longer have to dedicate a production line to each board design, your parts stuffer robots can handle a wide range of boards one after another (within constraints). The cost per board of this arrangement is higher than with a dedicated production line, but it's at least an order of magnitude lower than it was to produce small quantity boards even 10 years ago.
but the prices of development boards have not dropped significantly in the last 20-30 years. Manufacturers have gotten people used to the high prices of development kits, and since they mostly come out of corporate budgets, there hasn't been a lot of pressure to lower the prices as the costs have lowered. There are some exceptions (TI has had MSP430 development kits for $20 for example), but far too many manufacturers still charge thousands for their kits.
|
|||||||||||||