User avatar
mahjongg
Forum Moderator
Forum Moderator
Posts: 13379
Joined: Sun Mar 11, 2012 12:19 am
Location: South Holland, The Netherlands

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Mon Oct 13, 2014 11:48 pm

mung wrote: Yet I also heard that Broadcom dumped its mobile soc division, what chips could be used for the next pi now?

Are Broadcom still developing the videocore architecture?
please stop flogging this dead horse, just to annoy!

The PI's SoC doesn't fall under the mobile SoC division!

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27433
Joined: Sat Jul 30, 2011 7:41 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Tue Oct 14, 2014 8:35 am

In addition, all those non-Foundation who have access to the Pi offices and see HW sign an NDA. So if someone is saying stuff on IRC, they are either breaking their NDA, or lying.

Since I haven't seen any of those dev boards in the Pi office myself (and I do go there more regularly than most), I suspect the second.

Also, I wouldn't use any of those chips in a new Pi anyway. The first and last have modems and are therefore expensive, the second is a power hungry beastie (if its the one I am thinking of), and I am not sure it has the full camera support.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

User avatar
DougieLawson
Posts: 40212
Joined: Sun Jun 16, 2013 11:19 pm
Location: A small cave in deepest darkest Basingstoke, UK
Contact: Website Twitter

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Tue Oct 14, 2014 12:35 pm

jamesh wrote:In addition, all those non-Foundation who have access to the Pi offices and see HW sign an NDA. So if someone is saying stuff on IRC, they are either breaking their NDA, or lying.
Or mung is phishing trying to get someone from the Foundation to reveal something in an accidental forum post.
Criticising any questions is banned on this forum.

Any DMs sent on Twitter will be answered next month.
All fake doctors are on my foes list.

Note: Any requirement to use a crystal ball or mind reading will result in me ignoring your question.

User avatar
RaTTuS
Posts: 10609
Joined: Tue Nov 29, 2011 11:12 am
Location: North West UK
Contact: Twitter YouTube

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Tue Oct 14, 2014 1:17 pm

mung wrote:I heard on irc that someone has seen Broadcom development boards for BCM21664T, BCM11311, and BCM28155 knocking around the pi offices.

Yet I also heard that Broadcom dumped its mobile soc division, what chips could be used for the next pi now?

Are Broadcom still developing the videocore architecture?

I imagine its only another 2.5 years before the new pi is due?
I heard from a friend from someone he knows that you are talking completely balderdash and piffle [or something]
How To ask Questions :- http://www.catb.org/esr/faqs/smart-questions.html
WARNING - some parts of this post may be erroneous YMMV

1QC43qbL5FySu2Pi51vGqKqxy3UiJgukSX
Covfefe

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27433
Joined: Sat Jul 30, 2011 7:41 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Tue Oct 14, 2014 3:35 pm

DougieLawson wrote:
jamesh wrote:In addition, all those non-Foundation who have access to the Pi offices and see HW sign an NDA. So if someone is saying stuff on IRC, they are either breaking their NDA, or lying.
Or mung is phishing trying to get someone from the Foundation to reveal something in an accidental forum post.
Given that the people in the know have never EVER revealed anything up front, rather a fruitless exercise if that is the case!
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

W. H. Heydt
Posts: 13618
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Tue Oct 14, 2014 6:01 pm

Minor data point....

At the Bay Area Maker Faire in 2013, I had a Cubieboard with me and showed it to Eben. I was somewhat surprised that he had never seen one before (my expectation was that he would have obtained various other boards to experiment with and see what features they had...but that was not the case).

It would not surprise me if various people doing hardware work on the Pi look at a wide variety of boards, either physically after obtaining samples, or by looking at spec sheets. Such research would *not* mean that the Foundation was (or was not) considering any particular SoC or feature (or feature set) for future use. It would mean that they are keeping up with their professional field.

So...a report that one or more development board are "knocking around" Pi Towers only means that the engineers are maintaining their knowledge in their field.

ShiftPlusOne
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 6259
Joined: Fri Jul 29, 2011 5:36 pm
Location: The unfashionable end of the western spiral arm of the Galaxy

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Tue Oct 14, 2014 6:51 pm

I've searched the IRC logs for the main raspberrypi-related channels and see no mention of any dev boards floating around.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27433
Joined: Sat Jul 30, 2011 7:41 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Tue Oct 14, 2014 7:06 pm

W. H. Heydt wrote:Minor data point....

At the Bay Area Maker Faire in 2013, I had a Cubieboard with me and showed it to Eben. I was somewhat surprised that he had never seen one before (my expectation was that he would have obtained various other boards to experiment with and see what features they had...but that was not the case).

It would not surprise me if various people doing hardware work on the Pi look at a wide variety of boards, either physically after obtaining samples, or by looking at spec sheets. Such research would *not* mean that the Foundation was (or was not) considering any particular SoC or feature (or feature set) for future use. It would mean that they are keeping up with their professional field.

So...a report that one or more development board are "knocking around" Pi Towers only means that the engineers are maintaining their knowledge in their field.
They do/did actually try out some stuff a while back. Mainly Allwinner based devices which turned out ot be pretty pants. But I've not seen anything like that around the office. They have more important things to do!
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

W. H. Heydt
Posts: 13618
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Wed Oct 15, 2014 12:06 am

jamesh wrote:
W. H. Heydt wrote:Minor data point....

At the Bay Area Maker Faire in 2013, I had a Cubieboard with me and showed it to Eben. I was somewhat surprised that he had never seen one before (my expectation was that he would have obtained various other boards to experiment with and see what features they had...but that was not the case).

It would not surprise me if various people doing hardware work on the Pi look at a wide variety of boards, either physically after obtaining samples, or by looking at spec sheets. Such research would *not* mean that the Foundation was (or was not) considering any particular SoC or feature (or feature set) for future use. It would mean that they are keeping up with their professional field.

So...a report that one or more development board are "knocking around" Pi Towers only means that the engineers are maintaining their knowledge in their field.
They do/did actually try out some stuff a while back. Mainly Allwinner based devices which turned out ot be pretty pants. But I've not seen anything like that around the office. They have more important things to do!
As with most things...it depends on what your priorities are. I use Cubieboards to be "servers" for Pis because (a) they have a SATA port (SATA 2), (b) they're a moderate amount faster (1GHz v7), (c) they have more memory (1GB), and (d) there is a dual core version (A20). For the purpose, GPU capability isn't a issue, so the relative inferiority of the Allwinner GPU isn't a problem nor is--most of the time--the relatively thin and widely scattered support. The rest of my SBC systems are Pis (around a dozen of them). I consider them an interim solution until the RPF decides to come up with a better equivalent...assuming I live that long and/or still need that sort of capability. (Being 65, there is always the distinct possibility that a given piece of potential future hardware won't arrive before I depart this vale of tears.)

mung
Posts: 506
Joined: Fri Nov 18, 2011 10:49 am

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 11:13 am

DougieLawson wrote: Or mung is phishing trying to get someone from the Foundation to reveal something in an accidental forum post.
I would never try such underhanded tactics, and feel insulted you would suggest such a thing! :lol:

mung
Posts: 506
Joined: Fri Nov 18, 2011 10:49 am

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 11:16 am

ShiftPlusOne wrote:I've searched the IRC logs for the main raspberrypi-related channels and see no mention of any dev boards floating around.
Oh so you know where all the real propeller heads hang out, please tell me the IRC chans.

I could tell you where I heard it but I was sworn to secrecy :roll: ooops.

ShiftPlusOne
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 6259
Joined: Fri Jul 29, 2011 5:36 pm
Location: The unfashionable end of the western spiral arm of the Galaxy

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 11:26 am

Ah, the old invisible dragon.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27433
Joined: Sat Jul 30, 2011 7:41 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 11:34 am

mung wrote:
ShiftPlusOne wrote:I've searched the IRC logs for the main raspberrypi-related channels and see no mention of any dev boards floating around.
Oh so you know where all the real propeller heads hang out, please tell me the IRC chans.

I could tell you where I heard it but I was sworn to secrecy :roll: ooops.
I'd try and keep it polite if I were you. (And btw, the guy who told you this stuff on whatever IRC channel it was? Talking out of his choccy starfish)

If you want information on future Pi's then it's quite simple. Build a time machine, jump forward ten years, and check the WIkipedia page.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

mung
Posts: 506
Joined: Fri Nov 18, 2011 10:49 am

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 11:35 am

Getting back on topic of what imaginations wants for the model C.........

I wonder if the vc4 is still the most powerful ARM gpu available, anyone have any details of what the fastest cutting edge will be for the next couple of years(vc4,mali, tegra)?

If rpi is going to be using media processor chips what is available, what do others think are important features for future media processing chips?

I think I heard something about DTV satellite boxes moving toward dual use game/media centers?

Is broadcom working on those sort of chips?

I would like to see a chip with decent on SOC usb3 support with buffering and shared memory, and a number of io processors similar to the beaglebone PRU that can handle realtime control (possible 4 mcu with direct access to gpio and main memory/dma:- could control modem, LNB tuner control, io devices, emulate sata/1394/any generic io in programmable software).

I would also like to see some well thought out support chips for level conversion and RF interfacing (miracast/bluetooth/bla bla etc).

Again I should say I really know nothing about such things, and I am sure there are huge amounts of work to do such things and probably obstacles that prevent such chips being made at low cost or suitable for the rpi, but its always interesting to hear what others more knowledgable than myself have to say and understand a little more about the design process.

mung
Posts: 506
Joined: Fri Nov 18, 2011 10:49 am

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 11:51 am

jamesh wrote:
mung wrote:
ShiftPlusOne wrote:I've searched the IRC logs for the main raspberrypi-related channels and see no mention of any dev boards floating around.
Oh so you know where all the real propeller heads hang out, please tell me the IRC chans.

I could tell you where I heard it but I was sworn to secrecy :roll: ooops.
I'd try and keep it polite if I were you. (And btw, the guy who told you this stuff on whatever IRC channel it was? Talking out of his choccy starfish)

If you want information on future Pi's then it's quite simple. Build a time machine, jump forward ten years, and check the WIkipedia page.
Seriously I am not intending to be impolite, I know its better to ask people to keep it polite before things get out of hand, but I normally find this forum quite polite and assume all is taken in humor, I see nothing impolite in any of the other posters comments. I intended no offence to anyone, just hoping for an interesting conversation. If you think I am being impolite I apologise even though I am not sure what was impolite about what I have said.

fruitoftheloom
Posts: 24530
Joined: Tue Mar 25, 2014 12:40 pm
Location: Delightful Dorset

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 12:25 pm

IMO any talk of future products not using the BCM2835 SoC is just hot air ;)

The RPF have stated that there will be a new model but as yet no date has been fixed, so it could be anytime in the next decade ;)
Thinking outside the box is better than burying your head in the sand...

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27433
Joined: Sat Jul 30, 2011 7:41 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 1:33 pm

fruitoftheloom wrote:IMO any talk of future products not using the BCM2835 SoC is just hot air ;)

The RPF have stated that there will be a new model but as yet no date has been fixed, so it could be anytime in the next decade ;)
Yup, definitely in the next decade.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27433
Joined: Sat Jul 30, 2011 7:41 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Thu Oct 16, 2014 1:35 pm

mung wrote:Getting back on topic of what imaginations wants for the model C.........

I wonder if the vc4 is still the most powerful ARM gpu available, anyone have any details of what the fastest cutting edge will be for the next couple of years(vc4,mali, tegra)?

If rpi is going to be using media processor chips what is available, what do others think are important features for future media processing chips?

I think I heard something about DTV satellite boxes moving toward dual use game/media centers?

Is broadcom working on those sort of chips?

I would like to see a chip with decent on SOC usb3 support with buffering and shared memory, and a number of io processors similar to the beaglebone PRU that can handle realtime control (possible 4 mcu with direct access to gpio and main memory/dma:- could control modem, LNB tuner control, io devices, emulate sata/1394/any generic io in programmable software).

I would also like to see some well thought out support chips for level conversion and RF interfacing (miracast/bluetooth/bla bla etc).

Again I should say I really know nothing about such things, and I am sure there are huge amounts of work to do such things and probably obstacles that prevent such chips being made at low cost or suitable for the rpi, but its always interesting to hear what others more knowledgable than myself have to say and understand a little more about the design process.
The videocore team has been disbanded (not the best commercial decision ever I don't think) so chances of a VC5 are remote, which means there are already SoC with more powerful GPU's. But the VC4 is still a decent bit of kit - 1080p encode decode, camera I/F fast 2D/3D, decent interfaces for LCD and cameras, so there is a good argument that it is good enough for the intended Pi market. The ARM is the weak point of the 2835 rather than the GPU.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

W. H. Heydt
Posts: 13618
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Fri Oct 17, 2014 2:21 am

Now for something completely different...

Given the way that SD card capacities are going, a future Pi upgrade should be able to boot from file systems other than FAT. This should include (besides FAT itself) exFAT and EXT4.

(I've been reading the thread in Beginners about the problems with using 64GB cards and remembering that a 1TB card was announced, IIRC.)

User avatar
Jessie
Posts: 1754
Joined: Fri Nov 04, 2011 7:40 pm
Location: C/S CO USA

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Fri Oct 17, 2014 3:55 am

W. H. Heydt wrote:Now for something completely different...

Given the way that SD card capacities are going, a future Pi upgrade should be able to boot from file systems other than FAT. This should include (besides FAT itself) exFAT and EXT4.

(I've been reading the thread in Beginners about the problems with using 64GB cards and remembering that a 1TB card was announced, IIRC.)
+ F2FS. Fat will be common for a while even the AVR can read a SD formatted FAT.

W. H. Heydt
Posts: 13618
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Fri Oct 17, 2014 4:14 am

Jessie wrote:
W. H. Heydt wrote:Now for something completely different...

Given the way that SD card capacities are going, a future Pi upgrade should be able to boot from file systems other than FAT. This should include (besides FAT itself) exFAT and EXT4.

(I've been reading the thread in Beginners about the problems with using 64GB cards and remembering that a 1TB card was announced, IIRC.)
+ F2FS. Fat will be common for a while even the AVR can read a SD formatted FAT.
Wasn't intended to be an exhaustive list. Just a couple that came immediately to mind that would be appropriate.

mung
Posts: 506
Joined: Fri Nov 18, 2011 10:49 am

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Fri Oct 17, 2014 1:38 pm

fifacoinhuiy wrote:We all have made these dreams about a device that comes with things we need and want on board, but the chances to a dream as such to come true is one on a billion, not only because we all individuality need and want different things from our personal computer but because the primary goal of a device like that, is, the "low cost" :lol:
If enough people show support for an idea then its not a one on a billion. I am sure if 1 million people gave preorders for a product the chance would rise to better than 10 to 1.

Price comes down per item if there are large numbers produced and I think most silicon foundries production are almost totally automated so once the process is working (which can be phenomenally expensive) the running costs are fairly low in human resources.

I wonder what the costs of designing silicon are these days and if the foundation could either design the system itself or give input to Broadcom. Seems apple designs all its iphone cpu chips in house these days.

From what I heard the majority of the chip real estate on bcm2835 is used by the vc4, the ARM core is only a tiny area used, what difference would it make for broadcom to create a bcm vc4 chip that contained ARMv6 and v7 and other cores all on the same die with simple switch to choose which core is used (I thought the ARM was managed by the vc4 as a coprocessor). Would this reduce overheads for chip production and inventory for broadcom, is there a way users could buy licence for the ARM core they wanted to use and unlock a given core as they required?

That would allow backwards compatibility with older rpi and allow upgrades in performance if users wanted to pay for extra licences.

That is all assuming ARM or other cores can be turned off and have little power or interference problems when not being used and that bus multiplexing is not a major design issue.

Again I have no idea, but was an idea I had.

I have no idea if the foundation could justify chip development as necessary for education development, maybe if kids require speech recognition capabilities to make getting lesson notes easier or some other high processing requirement?

Also I am sure the current raspi is a dead end within a few years(it might go 10 years but I just get the feeling the form factor is too large), I would imagine mobile device integration with home automation and media centers will be expanding capabilities to mop up this market by giving some gpio options and running Linux.

If you have a Linux system you have exceptionally good access to learn programming on that system and develop and expand the system.

rpi has a great community which is also a very good selling point but eventually people will move away when better hardware comes along and they have learnt how to use computers and found they need something more.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27433
Joined: Sat Jul 30, 2011 7:41 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Fri Oct 17, 2014 3:17 pm

I think Eben has already said that the B+ will be the final design with the 2835. I think the form factor is about right though, if you want smaller there is always the Compute module, but smaller also means more expensive to make, with less places to put connectors.

The VC4 team was about 200 people. That should give some idea of the work involved in the GPU.

Cost to get a chip ready to go through a foundry is huge. Even when you have a design, a set of masks is $1M ish. So it's still very expensive to take existing IP (e.g. VC4) and bolt on new goodies. I think out of the range of the Foundation themselves.

I think multiple different cores on the same die would be too expensive - chips cost is dictated by the silicon area so you don't add anything you don't need. Just make a chip with the ARM core you want on it.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

W. H. Heydt
Posts: 13618
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Fri Oct 17, 2014 4:20 pm

mung wrote: I wonder what the costs of designing silicon are these days and if the foundation could either design the system itself or give input to Broadcom. Seems apple designs all its iphone cpu chips in house these days.
It's not so much the design costs (though they're not cheap) as it is getting the design into Silicon to find out if you got it right...and then fixing the design and trying again (and again, and again....depending on just how good the designers are).

Beyond that, there's a big difference between Apple--a company with tens of billions of dollars to spend however it wants--and the RPF for which a million dollars is a sizable amount of money.

The reason that chips are so cheap is the scale of production of identical devices. If you can sell 100 million of *anything*, the R&D part of the cost gets lost in the noise. Bear in mind that Pis sell between 1 and 2 million units per year . The cell phone businesses sell more like a billion units per year. The SoC in Pi is cheap because Broadcom sold millions of them before the Pi ever came along...the R&D costs were already amortized off. If the Pi switched to a custom designed SoC, the RPF would have to absorb the entire R&D cost.

The net result of all this is that the Pi--and any future Pi--will use "obsolete" SoCs.
I have no idea if the foundation could justify chip development as necessary for education development, maybe if kids require speech recognition capabilities to make getting lesson notes easier or some other high processing requirement?
(a) No, they couldn't justify the cost (and, indeed, they couldn't *afford* the cost) of chip design. (b) The thought of an entire class of kids speaking to their computers....the mind boggles at the din.
Also I am sure the current raspi is a dead end within a few years(it might go 10 years but I just get the feeling the form factor is too large), I would imagine mobile device integration with home automation and media centers will be expanding capabilities to mop up this market by giving some gpio options and running Linux.
Want to tell that to my '486 system?

In what way do you think the form-factor of the Pi is "too big"? What would you discard in order to make it smaller (you'd have to discard *something*)?

The Pi has the ability to communicate with home automation now. It can support, via dongles, HATs, or other USB devices any RF interface available and can certainly expand to any new comms designs.

But that is all beside the point. What would those features have to do with the goals of the RPF?
rpi has a great community which is also a very good selling point but eventually people will move away when better hardware comes along and they have learnt how to use computers and found they need something more.
This strikes to the heart of why the Pi was developed in the first place. It ignores the fact that we *used* to have fairly open platforms that kids could learn on. Home computing devices became increasingly closed and inaccessible. That is, they became "black boxes". The Pi is a deliberate step away from that. Sure, the current crop of kids will grow beyond the Pi...but there will be a next generation behind them that needs the same or similar learning tools. Just because you have outgrown flash cards and simple arithmetic drills doesn't mean that no one will ever need those tools again.

jdb
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 2466
Joined: Thu Jul 11, 2013 2:37 pm

Re: RaspBerry Pi(e) Model C (imaginations & wants)

Fri Oct 17, 2014 7:37 pm

W. H. Heydt wrote: It's not so much the design costs (though they're not cheap) as it is getting the design into Silicon to find out if you got it right...and then fixing the design and trying again (and again, and again....depending on just how good the designers are).

Beyond that, there's a big difference between Apple--a company with tens of billions of dollars to spend however it wants--and the RPF for which a million dollars is a sizable amount of money.

The reason that chips are so cheap is the scale of production of identical devices. If you can sell 100 million of *anything*, the R&D part of the cost gets lost in the noise. Bear in mind that Pis sell between 1 and 2 million units per year . The cell phone businesses sell more like a billion units per year. The SoC in Pi is cheap because Broadcom sold millions of them before the Pi ever came along...the R&D costs were already amortized off. If the Pi switched to a custom designed SoC, the RPF would have to absorb the entire R&D cost.

The net result of all this is that the Pi--and any future Pi--will use "obsolete" SoCs.
This guy gets it.

The Pi currently sells at volumes that are an order of magnitude below that which would cause a fabless chip design company (not necessarily Broadcom) to wake up and pay attention. Thus we are constrained by chips that are already on the market.

With significant software engineering effort already expended behind the current VC4 architecture, we would effectively throw that away if we suddenly migrated to a feature-equivalent chip from a different manufacturer. We'd also take a huge risk in terms of future upstream software support, community involvement etc.

Having an "obsolete" chip is actually beneficial in this case - there are no blue-chip customers that care too much about the internal architecture being publicly documented. They will have moved on to the latest and greatest new shiny toy. This allows us to insert various levers that open up bits of the chip that were previously undocumented without too much resistance.

The Pi's original laudable design goal was to be a cheap and cheerful minimal computer that could be used to learn the basics of programming. As it happens, with the hardware inside the Pi you can learn pretty much all of the advances in computer science in the last 30 years from this device - the 3D, codec and vector engines are still considered to be state-of-the-art. People whining that it doesn't do X fast enough don't know how to write software properly.
Rockets are loud.
https://astro-pi.org

Return to “General discussion”