Upton: Raspberry Pi: Two things you thought you weren’t going to get [LWN.net]

Liz Upton reports that Raspberry Pi boards will be available by the end of the month. "There’s another big piece of news today. We’ve been leaning (gently and charmingly) on Broadcom, who make BCM2835, the SoC at the heart of the Raspberry Pi, to produce an abbreviated datasheet describing the ARM peripherals in the chip. If you’re a casual user, this won’t be of much interest to you, but if you’re wanting to port your own operating system or just want to understand our Linux kernel sources, this is the document for you." (Thanks to Paul Wise)

to post comments

Upton: Raspberry Pi: Two things you thought you weren’t going to get

Posted Feb 7, 2012 21:38 UTC (Tue) by tialaramex (subscriber, #21167) [Link] (2 responses)

On January 10th under the title "We've started manufacture" Liz wrote that by choosing a Far Eastern manufacturer they would achieve "3-4 week turnaround" meaning the boards should be here by now. This was in contrast to British manufacturers who'd quoted 12 to 14 weeks.

Many of us have experienced the phenomenon where you get given an over-optimistic timetable by someone who wants your business but actually doesn't have the capacity to serve you. It happens all the time in restaurants for example. It's not that they're lying, there just isn't an incentive for them to be more conservative.

So, don't hold your breath for a Raspberry Pi. When you see stacks of finished boards being loaded onto a plane, that's when you know the delivery date. Today we just have the latest estimate from a company who've already been wrong once.

Upton: Raspberry Pi: Two things you thought you weren’t going to get

Posted Feb 7, 2012 23:21 UTC (Tue) by felixfix (subscriber, #242) [Link] (1 responses)

There are many more steps in delivering physical goods than just manufacturer turnaround. Turnaround != delivery to end user.

Upton: Raspberry Pi: Two things you thought you weren’t going to get

Posted Feb 8, 2012 1:15 UTC (Wed) by paravoid (subscriber, #32869) [Link]

OTOH, the article actually says:
“I’ve got some bad and good news about manufacturing. The bad news is that it’s taking a little longer than we’d hoped, because the factory had some trouble sourcing a specific component. The quartz crystal package we had chosen when we thought we were manufacturing in the UK is readily available over here in Europe, and was the cheapest we could find; but it turns out that in China, that crystal package has been overtaken in price and size by a smaller, cheaper one, so the one we’d designed for has been a bit hard to find. The factory has sourced crystals now, so we’re all go.”

Upton: Raspberry Pi: Two things you thought you weren’t going to get

Posted Feb 8, 2012 6:31 UTC (Wed) by ringerc (subscriber, #3071) [Link] (15 responses)

I don't suppose that'd involve datasheets on the graphics engine and video decoder, too?

Update: Nope, no video data

Posted Feb 8, 2012 7:42 UTC (Wed) by ringerc (subscriber, #3071) [Link] (12 responses)

Update: Alas, there's no information on driving the GPU or GPU peripherals here. It's all about the ARM SoC peripherals like the SPI interfaces, UARTs, I2C interface, DMA controller, etc.

It's all very good stuff to know and some details, like the GPIO and SPI interface data, will be exceedingly handy for anyone who wants to do any kind of interesting embedded hardware work with this board. For example, the ability to drive a couple of the GPIO pins as PWM outputs is awfully nice and not something I would've expected to see on a system like this.

The lack of info about the GPU is sad, but on the other hand it isn't trivial to write good GPU drivers. Having even a good datasheet would only be of limited benefit. The fact that Broadcom, a historically very closed outfit, is moving to release any data at all is a really positive move that should be welcomed.

There's no mention of Ethernet, but that seems to be an external part on the Rasberry Pi rather than part of the broadcom SoC so that's no surprise.

Update: Nope, no video data

Posted Feb 8, 2012 8:23 UTC (Wed) by mjr (guest, #6979) [Link] (11 responses)

Yeah, it's a positive surprise they went this far. The best hope for a freely supported widespread embedded GPU though is probably the Lima reverse-engineering project for supporting ARM's Mali GPU series. They have some initial rendering successes and the GPU is reportedly a pretty clean design and easy to code for, though they still have their work cut out for them (and the cheap Mali boards will probably have to wait a bit longer too). Somebody actually supporting their GPU with free code, if even just on the Linux side and keeping their firmware blobs, would obviously be better, but...

The ethernet on the B model is indeed an external SMSC LAN9512. A pretty nifty chip for what it does: it hooks up to the SoC's USB interface (exposed directly on the ethernetless Pi A model), and contains a 3-way USB hub with the ethernet on one internal USB bus and two of the ports exposed externally: http://www.smsc.com/index.php?tid=300&pid=135

Update: Nope, no video data

Posted Feb 8, 2012 9:57 UTC (Wed) by GhePeU (subscriber, #56133) [Link] (6 responses)

There is a (relatively cheap) dev board with a Mali 400 GPU, the ST-Ericsson Snowball (http://igloocommunity.org/). It isn't as cheap as a Raspberry but it's a completely different beast (dual Cortex-A9 at 1GH, 1GB of DDR2 RAM, Ethernet, Wifi, Bluetooth, etc. etc.) and the smaller version doesn't cost much more than a Pandaboard ES.

Snowball

Posted Feb 8, 2012 10:17 UTC (Wed) by mjr (guest, #6979) [Link] (5 responses)

Yeah, that's not bad for what you get, though as you said, a bit in a different category than Pi. There'll probably be more variety in Mali boards too before the Lima driver (hopefully...) matures, towards the lower end too. For anyone willing and able to hack towards it Snowball seems nice now though ;)

Snowball

Posted Feb 8, 2012 14:17 UTC (Wed) by pboddie (guest, #50784) [Link] (4 responses)

The project to watch is actually this one: http://rhombus-tech.net/

Their chosen CPU for "mass market" purposes also supports the Mali GPU architecture, so I imagine that if they continue to make progress, we might see sufficiently open hardware in a similar price bracket in the not too distant future.

Snowball

Posted Feb 8, 2012 22:12 UTC (Wed) by kragilkragil2 (guest, #76172) [Link] (3 responses)

Yeah, that *looks* great, but the thing is RPi has much more convincing story/people behind it. History is filled with projects/companies that don't deliver.
RPi has convinced me, I believe them when they say that they will ship next month. Other projects need to do a lot of trust building before I take them seriously.
That rhombus-tech site is filled with great visions and lofty goals, but no real price or shipping date. I certainly don't believe the 15$ price.

Snowball

Posted Feb 9, 2012 14:11 UTC (Thu) by pboddie (guest, #50784) [Link] (1 responses)

Sure, Raspberry Pi has a few names you've heard of behind it - that potentially raises other "trust issues", depending on which names are the ones you recognise - but having read a bit more background and having met one of the main players in a different context, I don't see the Rhombus Tech people giving up so easily. I also respect the fact that they're surveying and documenting the hardware landscape rather thoroughly. You also see this going on, perhaps in a slightly less coherent or convenient fashion, around other open hardware initiatives, so even if the stated project doesn't work out there is a body of knowledge that others can work with.

As for the "convincing story" behind Raspberry Pi, aside from reservations about such devices being the right vehicle for teaching computing concepts (in what amounts to a replay of the microcomputing era), I see a lot more enthusiasm from bystanders for a potential source of cheap media centre hardware than for the stated educational goals. In addition, it's still a single vendor show: creating an open standard doesn't seem to be part of the agenda (which would arguably be one of the lessons to learn from the microcomputer era). Since the hardware aspect of Raspberry Pi is also communicated largely as a done deal, alternative initiatives seem more attractive to me because they potentially widen the competence of the community and make it more likely that we'll be able to maintain access to a range of devices friendly to Free Software, rather than depend on vendors cutting one-off deals.

Snowball

Posted Feb 9, 2012 19:28 UTC (Thu) by dlang (guest, #313) [Link]

if you can run linux on it, hardware standardisation is less important. In fact having different groups offer hardware with different capabilities, but all able to run linux, seems like a good thing to me.

Rhombus tech

Posted Feb 11, 2012 17:52 UTC (Sat) by justincormack (subscriber, #70439) [Link]

The Rhombus Tech project is much more of a community project. If you want to help make it happen then join in.

Update: Nope, no video data

Posted Feb 8, 2012 17:02 UTC (Wed) by josh (subscriber, #17465) [Link] (3 responses)

Personally, I look forward to seeing Intel's in-house replacement for the third-party Poulsbo/GMA500 chipset they licensed. From what I've seen so far, they plan to produce a simplified version of the Sandy Bridge / Ivy Bridge GPU, with low power usage and video acceleration, and make it as open as the existing Intel first-party GPUs.

Intel's in-house replacement for Poulsbo/GMA500

Posted Feb 8, 2012 21:18 UTC (Wed) by pflugstad (subscriber, #224) [Link] (2 responses)

Any links on this? Thanks.

Intel's in-house replacement for Poulsbo/GMA500

Posted Feb 8, 2012 21:43 UTC (Wed) by josh (subscriber, #17465) [Link] (1 responses)

Checking again, it looks like all the reports I've seen about that trace back to a single Phoronix article (http://www.phoronix.com/scan.php?page=news_item&px=MT...) which doesn't actually cite its sources. Sorry for unintentionally propagating unsubstantiated rumors.

Intel's in-house replacement for Poulsbo/GMA500

Posted Feb 9, 2012 1:46 UTC (Thu) by rahvin (guest, #16953) [Link]

That project died. Intel is going to continue licensing graphics cores from PowerVR (the one that made Poulsbo) for low power products like atom and will probably keep doing their bad inhouse designs for integration into the top end CPU's. After following Intel's graphics efforts for a decade I can say with fair certainty that they will keep announcing graphics products and then keep killing them before they see the light of day. They've done it more times than I can count. The low watt design didn't even last a year before they decided it was just cheaper to use PowerVR designs (of course they waited till after the Poulsbo backlash died down).

http://vr-zone.com/articles/exclusive-intel-s-cedarview-a...

As noted in the article, the SGX535 is an upgraded version of the SGX500 that IS paulsbo. Though that isn't the best link you can probably google a better article using the keywords "intel powervr SGX". I should also note that PowerVR designs are very common in android cellular phones, so the driver situation should be much improved from when the GMA500 came out, though I don't know for sure it will be better.

Upton: Raspberry Pi: Two things you thought you weren’t going to get

Posted Feb 8, 2012 8:09 UTC (Wed) by i3839 (guest, #31386) [Link]

Of course not. This documentation only tells you where and what some of the things on the chip are. There is a link to the actual pdf in the post itself, so go there, download the thing and have a look yourself. Still, it's 200 pages of information more than 0, so it's a good start.

If you want to know more than the very basics, you have to probably pay 10k for a dev kit and order 100k chips a year. Broadcom apparently doesn't do smaller orders, the only reason they're so nice to RasperryPie is because one of the founders works at Broadcom. Having all schematics and info of the RasperryPie is a bit useless if you can't actually buy the chips in low volume.

Upton: Raspberry Pi: Two things you thought you weren’t going to get

Posted Feb 8, 2012 17:02 UTC (Wed) by robert_s (subscriber, #42402) [Link]

You know, I suppose we'll at least get a binary GPU driver (never thought I'd say _that_) that'll be ripe for reverse engineering.

If only we had a bigger pool of graphics hackers to pull on to get this done.