Click to embiggen
It looks like Element 14 screwed up a single shipment, because some lucky soul just received an unreleased model of Raspberry Pi. If you can believe the silkscreen, it’s called the Raspberry Pi Model B+, and while we have no idea what the chipset is, the layout and peripherals look pretty cool.
From the looks of it, this new board features four USB ports, a new, 40-pin GPIO header, and more screw holes that will allow you to secure this to anything. The analog video out is gone, and the SD card connector – a weak point of the original design – might be replaced with a microSD connector. Oh, every Raspi case that has ever been made? They won’t work.
Without booting this Raspi B+ there’s no way of knowing what the chipset is on this new board. The smart money is on the entire SOC being the same: basically, what you’re looking at is the same as a Raspberry Pi Model B, only with a few more ports.
There is no clue when these improved Raspis will be available, but the word “soon™” will probably appear on the Raspberry Pi blog shortly.
Thanks [John] for the tip.
EDIT: [feuerrot] is smarter than me and mirrored all the images in an imgur album.
i think they combined audio and video jack. silkscreen says A/V beside the jack.
Hope so, how else do I get XBMC on my ol’ Dumont.
I really hope so or they are bringing the Composite video out on he header. I would be happy if it’s on the header as that is the main display for my Motorcycle Audio system.
also it seems they changed the regulator to a switching one.
It’s almost like they looked at the user feedback, found the most common problems, and fixed them… But that can’t be it, that never happens in the real world… :P
I especialy like that all I/O connectors are on 2 sides, instead of 4. Makes it a hell of a lot easier to neatly arrange all cables.
Even better, the offsets of all the jacks are much better aligned. The standard RPi had demented offsets, like the Muppet’s Swedish Chef arranged all the connectors, aggravating case design.
I hope they updated the main CPU though.
I see Samsung on the main chip, maybe an Exynos?
That Package on Package (PoP) RAM sitting above the SoC. It’s likely the same BC chip.
Will the same old Broadcom chip support the apparent extra GPIOs? There’s so many ARMs out now with dual cores and over 1GHZ clock speeds, technology’s moved on at it’s usual rate since the Pi was first created. Might be they’ve upped the spec, it would help it play media and emulators better. Emulators in particularly are a popular use for Pi’s. And some of them are so badly-coded as to be useless at emulating 68K chips at maybe a dozen megahertz, on a 700MHz ARM.
I can’t remember the offending emulators off-hand, but I remember me and my friend being severely underimpressed before we clocked it up to 950. That’s as far as we wanted to risk, with the teeny-weeny heatsinks we ordered.
More horsepower would also allow more rendering stuff to be done in software, make up for whichever mysteries and problems are still left in it’s graphics units.
That’s just the PoP RAM chip.
In the picture its just a PoP samsung k4p4g324 512mb of ram … exactly the same as the model b … the ethernet chip has changed tho
The new ethernet chip is a LAN9514, just a 4-port version of the LAN9512 present on the regular Model B. Still limited in throughput by having only a single USB connection to the SoC.
Would have been nice if they had used a regular 9-pin USB header for the 2 new ports instead of 2 USB-A ports (or even a single 4 or 5-pin header + 1 USB-A port), it would have given a USB bus directly available on expansion board stacks.
Did you notice the 3.5mm “headphone” socket silkscreen says “A/V”? Possibly a 4-pole socket with the composite video, left and right plus ground channels. This is pure speculation but If they were to keep backwards compatibility with regular headphones I’d say the third pole / ring from the tip would be the video.
Possibly, but those 4-pole plugs are a pain in the arse, and not nearly as ubiquitous as the normal kind. Not always as easy to get hold of. Still they could include an adaptor with each one I suppose.
Worse would be to lose composite altogether, because so many projects use it. Unless they’ve upgraded to a chip that doesn’t do composite. Shame tho cos of all the composite LCDs you can get, in absolutely every size, often with touchscreens attached.
Maybe the Pi could generate composite through a GPIO in software? Andre Lamothe got a 52MHz chip to do it, I think a PIC derivative, with the addition of an analogue delay chip to do the phase-shifting for colour. B/W text works on an Arduino. Wouldn’t like to hear it go that way, but maybe they’ll keep selling the old line as well for people who still want composite.
Or maybe they’re gonna do something with the onboard LCD connector at last? Launch a range of screens that work easily with it. Or at least maybe an adaptor.
We’ll find out more when the Kiwis who have it get round to testing, I suppose. Could be an escaped prototype where they haven’t fitted the RCA plug yet.
anyone else notice what looks like 2 DACs behind it? i believe the SoC used has built in audio dacs .. maybe for higher quality or bidirectional audio? i cant imagine there doing any video with them …
Those are two different chips, one is marked “Z16T” (The T might be two seperate lines, and not a real T, Z16 is a BZX84-C3V9 zener diode) and the other something like “6S4VD” (picture is a bit fuzzy).
The SOC has the absolute worst audio in the world. The raspi audio utterly sucks, that is why most audio projects use a USB audio device or a daughterboard.
I think you’re on to something there. Good news is that if they do have analog video available, it wouldn’t be difficult at all to break it out behind/under the a/v jack to a composite plug which could be mounted to any case with enough headroom to clear it. Here’s hoping that analog video is actually available on that plug!
English spec sheet (linked below) shows NTSC/Pal as being available “via composite”. So??? There’s no composite jack in the pictures. That means that it’s got to be part of the audio jack (or on an unpopulated header) if it’s there at all. Hopefully some folks that order today will post back and confirm whether or not it’s available in any form once their units arrive.
Schematic linked below confirms that it’s the innermost ring of the A/V jack (first position, before ground). So, if you use a 4 pole plug, you’re golden. If you use a 3 pole plug it just shunts the analog video to ground. This means that if you want to break it out internally, you’re going to have to break the connection to the socket (so that it’s not shorted out any time a normal 3 pole plug is inserted) and then tie in the composite socket. Alternately, 4 pole to 3 pole + video adapters abound on the internet (if you’ve got digital cameras of a certain age and you kept the adapters, you’ve already got one!).
Also, some older Roku units came with them, FYI.
The way the symbol is drawn is that what usually the ground “sleeve” is video (pin 1 – *outer most*) and the next ring is ground (pin 2). L audio is the inner most aka “Tip”.
In some sense it might be better as the construction of the plug is that these are made of 3 hollow concentric tubes for each of the 3 rings and and a central solid piece for the tip. So there is the ground tube separating the video from the audio stuff.
Sort of, kind of. A four pole plug like this still has the same max dimensions, and the tip and ring still have the same length dimensions. What differentiates the four pole from the three pole is that they split the ground (shield) in length and insulated the two parts (so that one is now ring2 and the other remains as shield). The ground contact point for three pole jack is typically far enough forward that when used with a four pole plug, the ground contact touches the front part of the shield section. With the back part of the shield section isolated, it is used as the composite video contact and the four pole jack’s extra contact finger makes the necessary contact. The three pole jack is blind to the fourth pole and sees it as a standard three pole plug. This works the same way with some headphone/microphone combos that plug into computers, game systems, or phones.
What will sometimes hang people up is that manufacturers weren’t always consistent when it came to which half of the shield area was assigned to be the video/microphone contact and which half was assigned to be the ground contact. Some did it as Tip/Ring1/Ring2/Shield, and others did it as Tip/Ring1/Shield/Ring2. So, if the 4 pole plug you have doesn’t properly output video from the RasPi B+, it’s probably the alternate configuration…
If you’re feeling adventurous, you can cut the cable midway and then re-connect as needed so that the outputs are functional. Of course, this would put signal over the cable shield on one end and bring the ground into the inner part of the cable, but in most cases this shouldn’t be an issue. Alternately, the typical RasPi vendors should have the correct four pole configuration of cable, properly built and reasonably priced.
And of course you can buy some bare four pole plugs and some RCA jacks and some shielded three conductor cable and solder it up a proper cable yourself from scratch… more satisfying (keep the Hack in HAD!) and keeps the shielding intact.
Schmatic of socket at http://docs-europe.electrocomponents.com/webdocs/12fa/0900766b812faa1d.pdf
Looks like the server is feeling the HAD effect.
Looks so. Sometimes I wonder whether HaD is not just a cleverly hidden voluntary DDoS. :-)
Maybe it would be a bit more polite to make the links lead to a local copy of the page on HaD servers. Imagine people with cheap servers – one entry at HaD and their monthly bandwidth is gone.
Maybe it might be possible to get HAD’s CDN buffers to cover the sites it features, for a while after each article. If it’s possible to do that transparently for a site.
If not, then yep a mirror would be good, updated near-instantly, just keep in touch with the site’s changes. There’s some HTTP way of doing that right? For each site HAD could set up a mirror and have the main article link to there.
If a site objected you could always stop doing it. Maybe allow advert-links to come through unbuffered, if that’s how a site’s earning it’s money, since they can generally cope with mega-traffic. I think most site owners would appreciate the attention and the large audience, and any mirroring that came with it.
Here’s what I’m hearing: “Let’s do something useful and innovative with wordpress.”
I’ll talk to the overlord responsible for this, but we’re going to do something cooler before doing a hug of death buffer/cache.
I don’t think that is a good idea, there are possible copyright implications, and maybe even the potential for fraud abuses by by mirroring content without prior permission – no matter how well meaning.
Well, so how Google doesn’t get sued then for caching pages? ;)
@RicoElectrico
Well Google are a huge corporation whom get away with plenty of legally suspect things without scrutiny, doesn’t make it legal. Besides, they have a protocol whereby you’re able to have cached versions of your site removed from their index.
@Greenaum
Yes, I’m fully versed in how the web works thank you very much.
There’s 2 options, do it anyway, and comply when people want it taken down, while being very friendly and decent with it. I shouldn’t think anyone could prove any harm done, and I think that’s the basis of getting sued for copyright, right?
Second option is ask the site’s owner first, and offer to mirror it. If they don’t answer in time, run the article anyway, with the offer still open.
Pretty sure you have to commit fraud intentionally, so as long as nobody does that, everything’s ok.
Just since the “slashdot effect” is so well-known, a solution to it would itself be notable in Internet publishing.
It’s also the case that probably 90% of the web content you read has run through a “content distribution network”, whereby sites are mirrorred, silently and invisibly, all round the world. Saves on intercontinental web traffic, and traffic in general. You ask for whichever IP address, and your provider returns the data you’re expecting, but it doesn’t actually come directly from there. There’s now a giganormous series of caches. One at each end of each intertube.
Came in a few years ago, look it up. It’s all “transparent” and “neutral” whatnot, so they think it’s covered legally. Most of the web needs it to run nowadays.
Yeah, it happened yesterday with the computer on BGA chip…it took more than a day for me to be able to read it… XD!
I dared people to make a giveaway on my blog appear on HAD, luckily for me, it doesn’t appear to going to happen (never intended to, just the extra “impossibru” bit on the gymkhana), but previous posts of mine have really scared the hell of me..and they’re not as cool as this.
The german version of the shop already offers Model B+:
http://de.rs-online.com/web/p/entwicklungskits-prozessor-mikrocontroller/8111284/
I ordered one. :-)
The datasheet is in English! Looks like it answers a lot of the questions:
http://docs-europe.electrocomponents.com/webdocs/12de/0900766b812decd9.pdf
Broadcom BCM2835 SoC
700 MHz Low Power ARM1176JZFS Applications Processor
512MB SDRAM
Boots from Micro SD card, running a version of the Linux operating system
Well it must be official. At least two posts have been made on the Raspberry Pi forums and they have both been deleted without comment.
Comment i made on the front page post asking for more info is also gone
The LAN/USB hub chip seems to have been upgraded from the LAN9512 to the LAN9514 to support the extra two USB ports.
And the USB and ethernet connectors are lined up!
Yay! Finally!
Aren’t those aligned with the earlier versions?
The connectors are properly aligned this time, hopefully the new regulator will allow for some extra margin for the crappy usb chargers
There is nothing wrong with most chargers, it’s the ultrathin (sometimes not even copper) cables. A SMPS will only be worse with these kinds of cables due to the negative input resistance of a step-down convertor.
negative resistance, that’s something i’ve never heard off, time to read on the subject, any tips on where to start?
It sounds more difficult then it is:
With a positive resistance, the current goes up when the voltage goes up.
With a negative resistanve, the current goes DOWN when the voltage goes up.
DCDC convertors are a good example of negative resistances. When the load on the output of the convertor stays the same (for example; 1V 1A), the input current depends on the input voltage (and the switcher efficiency, but lets assume 100%). If the input voltage is 5V, the input current will be 0.2A. If the input voltage drops to 2.5V, the input current will rise to 0.4A. If the total resistance of the system (bering the input resistance of your DCDC convertor plus the output resistance of your power source (for example batteries + wires)) is positive, there’s nothing to worry about. However, as soon as the total resistance drops below 0Ohm, things start to go catastrophically wrong.
was thinking buck/boost converter not just a buck smps
Even then it is a very real possibility that the RPi will require more power then the PSU plus cables can supply. A 5V PSU with 10Ohms of wires connected to it, can only supply 250mA at 2,5V to a RPi. That’s not enough, even with a buck-boost convertor.
Even 5Ohms is too much; that will only be able to deliver 2,5V at 500mA.
FYI, they are probably using a fixed voltage (3.3v/1.8V) version of PAM2306D. I recognize that layout and the symbol in their schematic pinout is identical.
BUT the part # is not labeled, so how “open source” is that. :P
BTW that PAM2306 chip can operate at 100% duty cycle, i.e. a “drop out” mode (i.e. directly passing regulation). So its output can still be 3.3V when the input drops down to 3.6V give or take.
chip is rated to 1A output current, Rds for their PMOS is 0.3R (typ) at 3.3V input.
^bypassing i.e. shorting the PMOS from input to output
linear voltage regulators are gone also.. I can’t wait to buy one of these board!!
rs part 811-1284 in stock
It’s not available on the Dutch RS website :( I will give them a call tomorrow morning.
in stock at the spanish
i’m suprised nobody linked to this yet
data sheet
http://docs-europe.electrocomponents.com/webdocs/12de/0900766b812decd9.pdf
connector schematic
http://docs-europe.electrocomponents.com/webdocs/12fa/0900766b812faa1d.pdf
mechanical schematic
http://docs-europe.electrocomponents.com/webdocs/12fa/0900766b812faa0e.pdf
Its the same old single core CPU, that’s the biggest weak point I will say…
http://docs-europe.electrocomponents.com/webdocs/12de/0900766b812decd9.pdf
because there is no such thing as “good enough”
you need to ask for the extra price model that goes to 11
sure, Arduino is good enough too, righT?
meanwhile Mediatek ships whole 4x1GHz ARM A7 +PMIC solution at $4
“meanwhile Mediatek ships whole 4x1GHz ARM A7 +PMIC solution at $4″
I’m assuming you’re talking about the MT6589 SoC? I can find no indication that the chip is $4. If anything, it’s closer to $30.
Its the same old 700mhz single core CPU, its a real pity the PI really could benefit a lot from more horse power. You can find the data sheet here:
http://docs-europe.electrocomponents.com/webdocs/12de/0900766b812decd9.pdf
because 700 MHz is not enough horsepower for anything at all
the apollo astronauts must clearly have faked their moon landing because how could they have done it with less than a beaglebone?
Yeah, and insurance companies managed to do all of their accounting on a first-run Univac, so why would anyone want anything faster? Maybe because people have seen what higher-powered ARM devices can do, and continuing to charge the same amount for tech that’s been handily superseded over the past two and a half years is ridiculous, you supercilious dickbiscuit.
They didn’t need high definition video or fancy graphical interfaces. We could do likely 95% of our computing tasks with the horsepower of a C64 if only we didn’t take into account the presentation. Save for some niche applications, computing power today is mainly used to present data rather than to create it.
Beaglebone? The lunar lander had a 16 core i7 computer with at least 128gig of overclocked ram RAID 0 SSD’s and a pair of the fasted Nvidia video cards available!
Nobody can do anything with less than that!
“Datasheet” is also available on German RS site :-) http://docs-europe.electrocomponents.com/webdocs/12de/0900766b812decd9.pdf
Datasheet for the B+: http://docs-europe.electrocomponents.com/webdocs/12de/0900766b812decd9.pdf
Hmmm, its for sale here, http://www.mcmelectronics.com/product/83-16318?scode=GS401&catargetid=530004080000291183&cadevice=c&gclid=Cj0KEQjw3IieBRDl1oG0gr_PweoBEiQAwGHVw7vxRSXw8icmfml6OJfKsqJIIkbmXLV8PmZnpee-wxAaAkL68P8HAQ
2A over a micro USB socket. They still don’t learn.
But they learned from the Beaglebone, and reserved pins for id EEPROMs, to enable automatic configuration.
I don’t think that I’ll buy one, I already have 4.
2A!? you must mean 2^-1A… There’s no way that the CPU, RAM and ethernet chip are dissipating 10W.
are you insinuating that the designers left no headroom to power external devices?
No, I am insinuating that 0,5A is enough in 99% of the use cases. The remaining 1% can use a powered USB hub, or feed power to the RPi via the GPIO header, that’s what I am doing for 5RPis equpped with servo’s for a pan-tilt mechanism.
Low powered USB and low power GPIO are one of RPi weak points. I don’t want yet another power supply for powered hub just to run a USB device like an additional UART. Most other SBC, even the lowest PcDuino just say 5V 2A and get lost with less than 2A.
Max power from a USB is 500mA. 4 ports at 500mA is 2A, the limit of uUSB. But leaves nothing for the CPOU. But then how often do you use 500mA from each port.
Keyboard, mouse, Wifi, something else. 2A should be fine. Assuming the new power circuitry can provide that level of power.
If they allow for 500mA per USB downstream connector (4), 4 x 500mA =2A is at the limit without any budget left for the board itself.
Interesting circuit they used for driving the MOSFET Q3. Tempted to check that out in LTSpice.
The uUSB connector is rated to 2A at 5v so I am not sure what your point is?
The point is that every USB has to guarantee 500mA @5V.
Add some 500-800mA for other things (CPU, HDMI…) and you get in the ballpark of 5V 2.7+A, like the good old Beagleboard xM, which sports four 500mA-capable USB ports.
Oh, well, the Belching and Raspberrying Pi has already a tradition of tragically underpowered USB ports… ha ha ha.
Hopefully they cleared up some of the USB bandwidth problems I heard the originals suffered from. Really nice to see more GPIO but to be honest I was mostly using I2C so I effectively had all the IO I would have ever needed anyhow.
I am a little disappointed that it doesn’t sound like they added a more powerful CPU. Would have been nice to see it get a boost from 700MHZ to 1GHZ or so.
The USB “problem” will still be there, just as the I2C silicon bug; they haven’t switched CPUs.
Hm, all those pins, and the chip couldn’t spare another USB controller!
Tho just to follow-up myself, you need a powered USB hub to run much of anything on it anyway. I can’t think of 4 host-powered devices I’d want to run on it, without maybe a hard drive, a Wifi stick, or a coffee-cup warmer.
I can. I have a project right now that uses TWO Bluetooth adapters so I can stream A2DP stereo audio to two helmet headsets while also routing telephone and GPS audio from TWO phones around in there. I would KILL to have an extra BT and a Wifi Dongle in the slots to give me the ability for some breathing room as well as automatic podcast downloading while the motorcycle is warming up in the driveway from the houses wifi.
You need to think of embedded devices and stop trying to use a RasPi like a computer.
there’s only one USB controller on the SoC, so more pins won’t help.
another USB controller is what you should desire
They seem to run reliably at 950MHz. My friend’s one certainly does. Then again a 1GHz one would probably run to 1.2, so I see your point.
Some of the tiny stick on heat sinks and a tiny fan and they run all day long at the overclocked speeds. it’s the fools that seal them up in the boxes and then stick them to the back of their TV set where even more heat is added to the mix that have problems.
Apparently the RPi foundation does not want to get the news out.. Just asked about availability on the forum and now my post has been removed..
*sigh*
Sounds like a play from the Apple handbook, I’m surprised it wasn’t stolen from a bar!
selling raw boards at manufacturing cost and pushing vendors to release specs are not things you will find in the apple handbook
True, but accidentally releasing information about upcoming products to whet people’s appetite is.
yes and now all the sellers are stuck with stock they can’t sell on a product with low margins as it is
You play the game, you take your chances. Well… that’s what everyone tells me when I buy a product and they release the replacement a week later.
Throwaway account (obviously). I can confirm this is the new raspberry pi model. It does have a new chipset, although I can’t confirm the identity or any other details. I got this information some weeks ago under NDA from a source at Farnell/Element14 directly.
I have some hi def press release pictures, but I’m not elligable to share them. Keep pushing, they have more information available.
I can confirm a/v is one connector, more ram, more gpio, more usb connectors.
Your source sucks.
Can you tell us whether the old model (A and/or B) will still be made and sold, or will be fased out as soon as stocks are gone?
More RAM is not true, however. The B+ had 512MB, just as the B.
Brief details.
http://docs-europe.electrocomponents.com/webdocs/12de/0900766b812decd9.pdf
I mirrored the Images to imgur, they should be able to handle more traffic: https://imgur.com/a/GyZgA
I have been banned from the Freenode IRC Channel #raspberrypi for breaking the news. This is exactly the type of poor behavior that I expect. I am also banned from the Foundation forums. Please protest loudly.
So the raspberry pi foundation were hiding something afterall. Why didn’t they post them to their blog? They are worse than NSA.
Really? Really? They’re worse than the NSA?
If you need more horsepower than a 700 MHz ARM11, pick a different option. There are tons of boards with Cortex A8s, A9s, and A15s out there. Personally I’d love to have a quad core A15 for $35, but I’m not going to flame the Pi folks for not building me one. Respect the gift horse please. These changes are very welcome even with an ARM11.
The guy who broke the story has now been banned from the Raspberry Pi forums and I just saw him kicked from the IRC channel.
And they are still deleting posts on the forum that make any reference to it.
Idiots….
They banned me long time ago when I pointed out RS basically stole my preorder money and kept it for 6 months _while at the same time shipping orders to newer customers_.
the guy who “broke” the story had been skating on thin ice for a while as shown by the fact he has been banned from the forums for many months for cause. The photos he published were not his copyright (afaik as they were taken by someone else who agreed to not say anything until informed that he could) & he was asked nicely to not ruin it for everyone else and not only did he ignore that request he went around being a jerk and pushing the info into as many places as possible. A post about it that he posted on another forum was removed by their mods and he was asked to refrain from reposting which he then went and did.
He was banned for being Jerk not because he was leaking.
You are stretching the truth for your own benefit. The pics were posted in public, copyright does not apply. Anyone can look up my posts on the forum (Bencom). E14 made a new forum and left it open. E14 had been denying the Model B+ existed for a while. I have no obligation to the Foundation to keep information discovered to myself.
Specially when they ban me for life with no reprieve.
The original images remain where they were posted on i.imgur.com
I have participated in the IRC channel helping other users since it’s formation pretty much without any incident at all. UKScone does not. Channel logs exist. my nick rikkib.
The statement about E14 asking me not to post again is untrue although I did not post again anyway. Repeating the same thing and hoping for a different result is pretty brain dead. E14 did not say anything. Others who saw the post questioned E14 about policy.
The E14 post was immediately indexed by google.
Social media did the trick and a few emails to a select few.
Mr Davis’ name calling gives readers an idea of the attitude of this man.
Do not believe UKScone’ FUD it is just that. He banned me because I posted the news and stole the Foundation’s thunder. People with half a clue knew there was something up.
No doubt I will never have access to the forums or the official IRC channel. Tongue in cheek of course. The IRC channel is not official just ruled by the Foundation and their hatchet guy UKScone.
I suggest that UKScone needs to do the right thing and stop what will be an ongoing bun fight by removing the two bans. I still wish to help my local community here in NZ and see more open debate. And of course I reserve the right to publish whatever I see fit on my web site. The Foundation is certainly not sending the right messages to the younger generation. When found out they deny and try to censor. That is not the kind of world I live in.
I consider UKScone comments offensive and defaming. He needs to retracted and apologise or allow my right of reply.
What the heck. They ban because someone got something in a mail. That’s just stupid. Mistake has already happened, it does not help anyone to start banning people in hope of they shutting up. Just sad to see this happening ;-;
My post re the Pi+ was accepted and promptly deleted. What are these AZZZZZZclowns afraid of?
Ooo,
ID Eeprom, noice…..wonder if one’s actually connected?
and 5 more GPIO’s
…… 26 GPIO’s!! Very Noice, up from 17 + 4 on the original
Also already advertized on Farnell (though linking to the old product):
http://de.farnell.com/raspberry-pi-accessories?isRedirect=true
Wow so many sensible design choices, even ports that line up. This should have been the original design.
Strange that they are being so fussy about the news when most distributors already have product pages up.
I wish they would make a Pi that was powerful enough to decode h265..
You still need a goddam microsb cable to power the thing,and (apparently) no way to power the thing through GPIO. Anybody want to shine some light on me? This just looks like bad design.
The rPi Model A and B can be powered over the 5V/GND GPIO pins, and I suspect the Model B+ will be no different in that regard.
Have fun using an external regulator,because the microusb regulator circuitry is inaccessible through GPIO.
What? I power a couple of them with 5v over the GPIO connector, no probs. External regulator? Yes, but how is that different? Are you feeding unregulated power to the micro-usb port now?
You do bypass a fuse, which is something you might want to account for, but not the regulator.
I hope they enable PoE.
As much as I would like that, that’s doubtful. PoE is expensive and complicated, and it’s main uses are outside the RPi foundation’s goals.