User avatar
Lob0426
Posts: 2198
Joined: Fri Aug 05, 2011 4:30 pm
Location: Susanville CA.
Contact: Website

Re: New version of Pi. When?

Tue May 28, 2013 7:16 pm

There was some speculation at using the BRCM11311 chip in a future model. It has integrated Ethernet. It also has V7 inst. set cores and uses the VideoCoreIV GPU. These would make it easier to carry the software already completed into the new design.

I am sure they (the foundation) have looked at other ARM SoC's but the majority of work that has been needed is towards the GPU (and the USB). So switching to a totally different GPU might mean a square one start over. Another item to remember is that several of them work at Broadcom, so they have access to internal documentation, that would only be available to a company that placed a large SoC order.

Again the foundation would not announce a "new board" until it is actually ready for sale. They do not want any rumors to hurt, or stop sales of the boards that are in production now. Look up the "Osborne Effect". And unfortunately there are many that would not buy now if there was any rumor of a new board in the works.
512MB version 2.0 as WordPress Server
Motorola Lapdock with Pi2B
Modded Rev 1.0 with pin headers at USB

http://rich1.dyndns.tv/
(RS)Allied ships old stock to reward its Customers for long wait!

User avatar
Jim Manley
Posts: 1600
Joined: Thu Feb 23, 2012 8:41 pm
Location: SillyCon Valley, California, and Powell, Wyoming, USA, plus The Universe
Contact: Website

Re: New version of Pi. When?

Tue May 28, 2013 9:07 pm

Lob0426 wrote:It also seems some of you believe the Foundation designs ARM SoC's. they do not! They use existing SoC's to carry out their design.
This may not be quite accurate. As we say in the intelligence analysis business, "There are indications that ... " a few features in the BCM2835 just happen to strongly align with particular Foundation needs. Sort of like my Dad described how, during the Depression, certain things would "fall off the back of a truck" at coincidentally convenient times for needy people. You wouldn't be able to do something like "accidentally" wind up with a quad-core Cortex ARMv9 CPU running at 2 GHz in a product that the customer specified should only have a single ARMv7 core running at 1 GHz. However, I can imagine how some things might "fall onto the masks" for the wafers fabricated with the SoC to be used in Broadcom customers' future products, as well as fortuitously later in some far-off-in-the-future Model C, Raspberry Rho, or whatever it might be called. Of course, this would only happen if the opportunity should ever arise to do so in a next-generation Broadcom product with improved performance over the BCM2835, and even then only with Broadcom management's blessing.

As anyone who's familiar with fab processes knows, anything integrated into a semiconductor device has to be done with extreme care and a heck of a lot of testing to ensure non-interference between various modules. Any additions have to be performed at the front end of the design process, not tacked on willy-nilly later. I'm bemused by comments I've seen to the effect that the ARM CPU was "tacked onto the GPU as an afterthought". I'm sure the Foundation's volunteer Broadcom engineers have rolled their eyes so much at seeing such comments that they now have better peripheral vision than a walleye fish! :lol:

The challenge may be to ride the coattails of a next-generation SoC that's not too advanced over the BCM2835 for compatibility reasons, but with at least the highly-desired improved characteristics of additional RAM capacity, increased nominal (non-overclocked) clock speed, separate buses for networking and USB (with SATA support and multiple cores being real nice-to-haves), full ultra-high-speed SD card support, etc. Broadcom certainly can't complain about the just the positive PR value of the Pi using their SoC, not to mention the good will their engineers have garnered for the company from hundreds of thousands of customers in a wide variety of communities that have no idea they've been using a product containing a Broadcom device even before the Pi. That's highly likely nowadays, given how many mobile devices, video devices such as DVRs and probably cable/satellite set-top boxes, DVD/Blu-ray players/recorders, etc., contain Broadcom products, but as they say in the PR business, there's no such thing as bad PR, especially when it's good PR! ;)
The best things in life aren't things ... but, a Pi comes pretty darned close! :D
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!

Heater
Posts: 12076
Joined: Tue Jul 17, 2012 3:02 pm

Re: New version of Pi. When?

Tue May 28, 2013 9:25 pm

Jim,

"...tacked onto the GPU as an afterthought...", that would be me. That rather offhand statement of mine did not exactly come out of thin air. It's derived from a statement by one of the foundation members about the history of the Raspi's Soc. Made in this very forum or in one of the many raspi videos I have seen. I might have misunderstood or exagerataed the story but I'm fairly sure it is not a total figment of my imagination. I wish I could recall where it came from and link you to it.

Joe Schmoe
Posts: 4277
Joined: Sun Jan 15, 2012 1:11 pm

Re: New version of Pi. When?

Tue May 28, 2013 9:54 pm

And unfortunately there are many that would not buy now if there was any rumor of a new board in the works.
It should be noted that that is almost certainly the reason why we see so many of these "When is the next version coming out?" and "What will the next version have?" posts. One can visualize somebody suggesting to their boss that they get on this "Pi bandwagon" and the first words out of the boss's mouth being "Find out when the next version is coming out. We don't want to buy now if they are coming out with a new version."
And some folks need to stop being fanboys and see the forest behind the trees.

(One of the best lines I've seen on this board lately)

OtherCrashOverride
Posts: 582
Joined: Sat Feb 02, 2013 3:25 am

Re: New version of Pi. When?

Tue May 28, 2013 10:11 pm

There was a video recently that I cant seem to find ["they" don't want you to know!] where Ebon [EDIT: I should say "Mr. Upton" instead. I do not portend a first name basis.] was asked the question about a new Pi. He replied that it would be at least a couple of years before that happened. He also stated they would continue making the current Pi until they could no longer source the parts.
Last edited by OtherCrashOverride on Tue May 28, 2013 10:41 pm, edited 2 times in total.

OtherCrashOverride
Posts: 582
Joined: Sat Feb 02, 2013 3:25 am

Re: New version of Pi. When?

Tue May 28, 2013 10:26 pm

Found the video:
http://fora.tv/2013/05/18/Raspberry_Pi_ ... Innovation
The question is posed @ 16:30 in the video.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22044
Joined: Sat Jul 30, 2011 7:41 pm

Re: New version of Pi. When?

Tue May 28, 2013 10:52 pm

OtherCrashOverride wrote:There was a video recently that I cant seem to find ["they" don't want you to know!] where Ebon [EDIT: I should say "Mr. Upton" instead. I do not portend a first name basis.] was asked the question about a new Pi. He replied that it would be at least a couple of years before that happened. He also stated they would continue making the current Pi until they could no longer source the parts.
Probably should use Mr Upton, since Ebon is actually spelt Eben.

(note, the Arm was in fact tacked on to a spare bit of die of the VC4 - there was no market for it when it was done, and I think the idea was we/others might be able to use it for something, but not sure what).
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

Heater
Posts: 12076
Joined: Tue Jul 17, 2012 3:02 pm

Re: New version of Pi. When?

Tue May 28, 2013 10:59 pm

jamesh,
note, the Arm was in fact tacked on to a spare bit of die of the VC4
Thank you James.

Jim is going to have to learn that I am not always wrong:)

OtherCrashOverride
Posts: 582
Joined: Sat Feb 02, 2013 3:25 am

Re: New version of Pi. When?

Tue May 28, 2013 11:02 pm

jamesh wrote:Probably should use Mr Upton, since Ebon is actually spelt Eben.
If you guys can spell "color" with a "u" then I can spell Eben with an "o" and attribute it to internationalization! :D

My apologies to Mr. Upton.

sdd
Posts: 64
Joined: Sun Oct 21, 2012 5:48 pm
Location: USA

Re: New version of Pi. When?

Tue May 28, 2013 11:06 pm

what I want to know is: how on earth do we as programers manage to make 700 million clock cycles per second look slow?

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22044
Joined: Sat Jul 30, 2011 7:41 pm

Re: New version of Pi. When?

Tue May 28, 2013 11:10 pm

Heater wrote:jamesh,
note, the Arm was in fact tacked on to a spare bit of die of the VC4
Thank you James.

Jim is going to have to learn that I am not always wrong:)
I should add that 'tacked on' does require an awful lot of work by some pretty clever people, so it's not like a skunk works project. Someone thought it would be a good idea and there was spare die space...

I found out today the the synthesis of the RTL (part of process of getting a design ready for tapeout) on the very powerful machines we use, is taking nearly 24 hours (for part of the chip!), and the FPGA devices we use for testing designs cost more than my house. That's why new chips cost so much to make...
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
Mortimer
Posts: 919
Joined: Sun Jun 10, 2012 3:57 pm

Re: New version of Pi. When?

Tue May 28, 2013 11:15 pm

jamesh wrote:I found out today the the synthesis of the RTL (part of process of getting a design ready for tapeout) on the very powerful machines we use, is taking nearly 24 hours (for part of the chip!), and the FPGA devices we use for testing designs cost more than my house. That's why new chips cost so much to make...
I take it you don't live in a shed!? :lol:
--------------
The purpose of a little toe is to ensure you keep your furniture in the right place.

OtherCrashOverride
Posts: 582
Joined: Sat Feb 02, 2013 3:25 am

Re: New version of Pi. When?

Tue May 28, 2013 11:17 pm

sdd wrote:what I want to know is: how on earth do we as programers manage to make 700 million clock cycles per second look slow?
The processor may be clocked at 700Mhz but other components like system RAM are not. IIRC, the RPi has LPDDR2 RAM that is clocked slower. In addition to the slower clock speed, there is a latency from the time the address is requested and when the data becomes available on the bus.

Heater
Posts: 12076
Joined: Tue Jul 17, 2012 3:02 pm

Re: New version of Pi. When?

Tue May 28, 2013 11:22 pm

James,

Sorry, yes "tacked on" was, as I said, the "off hand" or exageration part. We can put that down to British understatement.

I have been following the story of the development of the second version of the Propeller micro-controller by Parallax Inc. They just had a tape out run that cost 60000 odd dollars. Sadly none of the delivered devices worked due to a trivial error and they have to go around that loop again. I can imagine working with the ARM and GPU is a few orders of magnitude more complex and expensive.

Heater
Posts: 12076
Joined: Tue Jul 17, 2012 3:02 pm

Re: New version of Pi. When?

Tue May 28, 2013 11:32 pm

sdd,

It's not so bad. During the discussion of calculation pi on the Pi we find that running gmp-chudnovsky to calculate a million digits of pi is about 9 or 10 times slower on the Pi than it is on my 3GHz PC.

Given the differences in clock rate, memory speed and cache sizes that sounds quite reasonable. In fact I'm stunned we can get a million digits of Pi on any normal machine in these times.

Then with normal operating systems and apps we have to suffer the snail like pace of file system access to SD as opposed to a nice sata hard drive.

Many complain that Linux is bloated and slow. I don't buy that. Ask yourself what features you can live with out and remove them.

Graphics is a problem as we are all used to accelerated graphics on desktops now a days, the humble Pi (get it?) has to push all those pixels itself. At higher resolutions than were the norm in good old 1024x768 un-accelerated VGA.

Sill, looks like good progress in the use of the GPU is being made.

W. H. Heydt
Posts: 9985
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: New version of Pi. When?

Wed May 29, 2013 1:04 am

sdd wrote:what I want to know is: how on earth do we as programers manage to make 700 million clock cycles per second look slow?
Do you recall when it used to be said that Microsoft could chew up every processor cycle that Intel could manufacture?

W. H. Heydt
Posts: 9985
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: New version of Pi. When?

Wed May 29, 2013 1:06 am

jamesh wrote: I found out today the the synthesis of the RTL (part of process of getting a design ready for tapeout) on the very powerful machines we use, is taking nearly 24 hours (for part of the chip!), and the FPGA devices we use for testing designs cost more than my house. That's why new chips cost so much to make...
Perhaps someone should try to set up a cluster of Pis just to see how that process would take on it...

User avatar
Jim Manley
Posts: 1600
Joined: Thu Feb 23, 2012 8:41 pm
Location: SillyCon Valley, California, and Powell, Wyoming, USA, plus The Universe
Contact: Website

Re: New version of Pi. When?

Wed May 29, 2013 3:51 am

I would never attempt to speak for jamesh, but I can assure everyone that no one in their right mind can even contemplate "tacking on" anything, anywhere on production wafers. Yes, companies like Mentor Graphics and others provide incredible silicon layout software and libraries that takes much of the drudgery and error-creation out of mask development, but it is far from fool-proof and there are still many corner cases that have to be analyzed by humans and verified after pre-production wafer spins.

The error that Pete Lomas made on the layout of the beta boards where a trace to one of the BCM2835 pins was missing is testament to how even the best of us can makes critical mistakes, and printed circuit board layout is a walk in the park compared with silicon wafer mask development. A war party of very agitated fab engineers is forming here in Silicon Valley to correct the misperceptions of ne'er-do-wells that bandy about such highly-inaccurate terminology, and I must warn you that they are talking about dumping tea in a harbo(u)r somewhere :lol:
Heater wrote:Graphics is a problem as we are all used to accelerated graphics on desktops now a days, the humble Pi (get it?) has to push all those pixels itself. At higher resolutions than were the norm in good old 1024x768 un-accelerated VGA.
If you're not already aware, the GPU is generating every single pixel in every frame of HDMI and composite video coming out of the Pi. As you're certainly aware by now thanks to my ranting ( :roll: ), the GPU is the real star on the Pi, not the effectively-traffic-cop ARM CPU. The CPU isn't really even needed, as the "tacked-on" comments hyperbolically suggest - it just makes it much easier to adapt traditional OS and application software to the SoC. In fact, we're held back by the moribund legacy code in things like X, not the GPU's capabilities. The Wayland/Weston, Pi3D, and Quake 3 demos show what life will be like once the cruft on the CPU side is transitioned to the GPU side - progress is being made, as you say.
OtherCrashOverride wrote:My apologies to Mr. Upton.
And it's Dr. Upton to you, sir, as he earned his PhD (Piled higher and Deeper) in Computer Science at the University of Cambridge in 2005. I wouldn't let him anywhere near me with a scalpel or even a stethoscope, though! ;)
The best things in life aren't things ... but, a Pi comes pretty darned close! :D
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22044
Joined: Sat Jul 30, 2011 7:41 pm

Re: New version of Pi. When?

Wed May 29, 2013 7:08 am

W. H. Heydt wrote:
jamesh wrote: I found out today the the synthesis of the RTL (part of process of getting a design ready for tapeout) on the very powerful machines we use, is taking nearly 24 hours (for part of the chip!), and the FPGA devices we use for testing designs cost more than my house. That's why new chips cost so much to make...
Perhaps someone should try to set up a cluster of Pis just to see how that process would take on it...
We actually use a cluster of very very fast rack servers for the synthesis and various stages in the development pipeline. They regularly get replaced with faster machines. I don't think the Raspi will cut it with the big boys!

And to reply to an earlier post, the memory subsystem, considering the Arm was bolted on as an afterthought, is actually one of the best on the market (I think the Apple system in their SoCs is better), so even though memory is more slowly clocked, it's very efficient, and explains why in real world bench marks the Pi performs well against higher clocked devices such as the A10.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
meltwater
Posts: 1014
Joined: Tue Oct 18, 2011 11:38 am

Re: New version of Pi. When?

Wed May 29, 2013 8:02 am

The way I interpreted "the bolted on an Arm" expression was more relating to the relative power, and size on the die between the two.
It would be excellent to see some die pictures one day to see the different sections which make up the die itself.

From the viewpoint of the RPI though, it does mean we get an ARM chip with a great GPU thrown in for free. The wayland desktop is a great start to making use of this, hope to see more use of it as time goes on.
______________
http://www.themagpi.com/
A Magazine for Raspberry Pi Users
Read Online or Download for Free.

My new book: goo.gl/dmVtsc

Meltwater's Pi Hardware - pihardware.com

Like the MagPi? @TheMagP1 @TheMagPiTeam

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22044
Joined: Sat Jul 30, 2011 7:41 pm

Re: New version of Pi. When?

Wed May 29, 2013 9:26 am

meltwater wrote:The way I interpreted "the bolted on an Arm" expression was more relating to the relative power, and size on the die between the two.
It would be excellent to see some die pictures one day to see the different sections which make up the die itself.

From the viewpoint of the RPI though, it does mean we get an ARM chip with a great GPU thrown in for free. The wayland desktop is a great start to making use of this, hope to see more use of it as time goes on.
The ARM actually takes up very little space compared with the Videocore and it's associated cache memory - hence it could be squeezed on.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
meltwater
Posts: 1014
Joined: Tue Oct 18, 2011 11:38 am

Re: New version of Pi. When?

Wed May 29, 2013 11:50 am

jamesh wrote:
meltwater wrote:The way I interpreted "the bolted on an Arm" expression was more relating to the relative power, and size on the die between the two.
It would be excellent to see some die pictures one day to see the different sections which make up the die itself.

From the viewpoint of the RPI though, it does mean we get an ARM chip with a great GPU thrown in for free. The wayland desktop is a great start to making use of this, hope to see more use of it as time goes on.
The ARM actually takes up very little space compared with the Videocore and it's associated cache memory - hence it could be squeezed on.
I guess there was a bigger impact on the pin count though, with all the GPIO from the ARM.

Worked out well considering it wasn't designed for it, of course it was designed for it then the cost wouldn't have been viable.


On a personal note, kinda hope the SD-card slot gets upgraded to a metal one at some stage.
______________
http://www.themagpi.com/
A Magazine for Raspberry Pi Users
Read Online or Download for Free.

My new book: goo.gl/dmVtsc

Meltwater's Pi Hardware - pihardware.com

Like the MagPi? @TheMagP1 @TheMagPiTeam

adlambert

Re: New version of Pi. When?

Wed May 29, 2013 11:56 am

sdd wrote:what I want to know is: how on earth do we as programers manage to make 700 million clock cycles per second look slow?
The RISC OS people reading that will be saying:

"Slow? What do you mean slow? I don't understand; the Raspberry Pi is EXTREMELY fast...."

Heater
Posts: 12076
Joined: Tue Jul 17, 2012 3:02 pm

Re: New version of Pi. When?

Wed May 29, 2013 12:26 pm

Jim,
I would never attempt to speak for jamesh, but I can assure everyone that no one in their right mind can even contemplate "tacking on" anything, anywhere on production wafers.
In my right mind I said "tacked on". James being also in his right mind said "bolted on as an afterthought" he should know better than either of us. Now it matters not if they used tacks or bolts it's just a way to say that the GPU was designed as a stand alone chip, with no ARM, the ARM being added later as a means of making the thing marketable. A story which James confirms here. These casual expressions in no way imply that adding the ARM was a trivial job hacked up over a lunch break.
If you're not already aware, the GPU is generating every single pixel in every frame of HDMI and composite video coming out of the Pi.
Actually I'm not aware of any such thing and I don't believe it to be true except for actually playing videos. For other graphical apps no.

For sure the HDMI/composit output is driven by the GPU but the arrangement of pixels is put together by code running on the ARM and written to bit maps which eventually end up in the frame buffer displayed by the GPU..

Why do I think so? Well when I boot a Pi I see a little picture of a penguin and my text based virtual console. Linux has been doing that forever on machines with no GPU. I have no reason to beleive the GPU is accelerating any of that.

So far, I understand that if draw graphics in my program, say a Qt4 app, it is drawn pixel by pixel, by code running on the ARM, to the frame buffer with no acceleration from the GPU.

Further it is commonly stated that GPU acceleration would not speed up your web browsing experience, for example, that implies to me there is no GPU acceleration currently.

I'm looking forward to getting GPU acceleration in Qt5 and webgl.
The CPU isn't really even needed, as the "tacked-on" comments hyperbolically suggest - it just makes it much easier to adapt traditional OS and application software to the SoC
What? There is no way Linux or perhaps any other OS is getting ported to the GPU and the ARM becoming redundant. I'm very sure the GPU is not a place where you want to run such an OS, it's performance would suck and you would be wasting a nice GPU.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22044
Joined: Sat Jul 30, 2011 7:41 pm

Re: New version of Pi. When?

Wed May 29, 2013 12:39 pm

Actually, it wasn't tacked on the make the chip marketable - the VC4 was already being sold to Nokia for their Symbian smartphones without the ARM. (a coprocessor). Nowadays of course, everyone wants SoC's rather than CoPro's, so the decision to add the ARM has proved to be worthwhile. I don't think there was a customer at the time the ARM was added, which makes it a real punt in the dark.

AIUI. The basic linux framebuffer is simply a single dispmanx image, which the GPU composites in real time to the output (whatever that may be). The camera preview is also a dispmanx image, which is why it can be superimposed over the top of the framebuffer (and have specific opacity etc) without the overheads of moving the data to the framebuffer.

I think Wayland also uses dispmanx images for *each* window (rather than just one as the case with the framebuffer), thereby offloading all the compositing from the ARM to the GPU. That should also make it easier to put OGLES, OVG, WebGL, camera output etc in to specific windows on the display.

As an aside, the GPU already runs a RTOS (threadX) and is quite capable of running generic programs (we runs a lot of test apps solely on the GPU, not using the ARM at all).
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

Return to “General discussion”