It has been over six years since BeagleBoard.org released the original Beagleboard, one of the first open source ARM SBC's for hobbyist use. Now the team has released the details of the upcoming Beagleboard-X15 based on the TI dual-core Sitara AM5728 SoC (system on Chip). It features a number of processor subsystems, modern, high-speed serial I/O ports as well as GPIO ports.
A partial list of SoC features include:
Dual ARM Cortex A15 cores running at 1.5GHz featuring: 128bit NEON SIMD, Hardware Virtualization, Jazelle RCT and LPAE allowing up to 1TB of RAM.
Dual C66x DSP cores
Dual IPU's (Image Processor Units) each containing dual Cortex M4 cores (only one IPU is user programmable, one dedicated to video decode)
1080P 60FPS video decoder (using IPU)
Dual core PowerVR SGX544-MP2 GPU
Dual PRU-ICSS coprocessors (Programmable Real-Time Unit and Industrial Communication Subsystem)
Dual Gigabit ethernet ports via internal switch from single GbE MAC.
The board features:
2GB DDR3 RAM on board
eSATA port @ 3.0Gbps
HDMI, LCD and a Video in ports
Dual PCIe 2.0 5Gbps lanes supporting two x1 or one x2 ports, routed to expansion headers
Audio in and out ports via 3.5mm jacks and HDMI audio out.
GPIO headers
More to come ...
The board is scheduled to be released in late February, 2015. Unfortunately, the price has not yet been announced. The official wiki page is located here: http://www.elinux.org/Beagleboard:BeagleBoard-X15 It looks to be quite a powerful board with a lot of connectivity and processing power.
Also revealed in the LinuxGizmos.com article:
Meanwhile, changes may also be in store for the BeagleBone Black. In response to a comment on his Google+ post, Pantelic confirmed rumors that TI was interested in offering the BeagleBoard community a more expensive processor than the low cost, low-margin Sitara AM3359 Cortex-A8 SoC used on the BB Black. BeagleBoard.org’s manufacturing partners Farnell and Embest will “take over producing the BBB, but since TI is not happy to sell to them a low cost processor, it will run with a certain Broadcom CPU instead,” wrote Pantelic. “Remember, you heard it here first.”
(Score: 0) by Anonymous Coward on Tuesday November 11 2014, @10:42PM
To me, that name screams BINARY BLOBS and FOSS-HOSTILE.
The big gripe against the original Raspberry Pi was the inclusion of Broadcom "intellectual property".
Have any Soylentil Linux users had to reflash Broadcom kit in their PCs to get rid of the closed-source firmware?
-- gewg_
(Score: 1) by aiwarrior on Wednesday November 12 2014, @01:16AM
You would be hard pressed to find perfectly open source system on a chip drivers. I just bought an odroid o3 and the Mali drivers are semi open source. There is currently an effort similar to the noveu nvidia drivers with the same kind of limitations. Would you deny linux to experimenters in the name of purity or would you sacrifice some of the ideals so that your message would be heard by more?
(Score: 0) by Anonymous Coward on Wednesday November 12 2014, @03:36AM
hard pressed to find
I won't argue that point.
Wouldn't it be nice, however, if openness was higher on the list of the folks building/specifying these systems?
If subsystem manufacturers knew that they lost the order because of their closedness, perhaps they would alter their ways.
I'm reminded of this quarter-billion-dollar cock-up:
nvidia-loses-10-million-gpu-order-over-awful-linux -support [inquisitr.com]
("Longsoon" should be "Loongson" there as pointed out by Wootery.) [soylentnews.org]
Would you deny linux to experimenters in the name of purity
Not I. When the Linux novices run into a situation that is due to a lack of openness I *would*, however, like it if they knew to put the blame for that on the "intellectual property" maximalists and not on Linux.
Being able to tweak things because you have the source code can make a dreary day bright again. [googleusercontent.com] (orig) [goodbyemicrosoft.net]
-- gewg
(Score: 2) by Arik on Wednesday November 12 2014, @03:55AM
You would indeed. And the reason for that?
Because you buy these SOCs anyway. Stop it. Please.
"I just bought an odroid o3 and the Mali drivers are semi open source."
So return them and get a refund.
"Would you deny linux to experimenters in the name of purity or would you sacrifice some of the ideals so that your message would be heard by more?"
I dont deny anyone to anything - not in my power if I wanted to, and I dont.
But I will tell you that if you say you are a 'linux experimenter' and you are buying hardware that is not suitable, not only are your problems of your own making in that regard, but you are actually causing problems for everyone else as well.
Manufacturers will start properly supporting linux when they see it as important to their bottom line, and as long as many 'linux experimenter's just shrug and buy their product anyone, that will not happen.
If laughter is the best medicine, who are the best doctors?
(Score: 2) by novak on Wednesday November 12 2014, @05:01AM
It's also worth pointing out that not being open source is totally different and not as bad as being a broadcom chip. Seriously. Many chips do not have open GPUs or open TCP/IP stacks but you can still get a full data sheet with several hundred or thousand pages on what the system is doing, and what every pin is. With people like Broadcom, you don't even get that.
The RPi data sheet is a good example of this, and that thing is the most open thing I've ever seen from broadcom.
novak
(Score: 2, Informative) by tftp on Wednesday November 12 2014, @12:54AM
There are tons of tablets [tablet.ninja] that are sold, with LCD and WiFi, for less than $100. This board has an advantage only if you need a specific I/O, like eSATA, that is only present on this board. For everything else you'd be better off with a tablet that may have USB OTG if you need USB devices.
(Score: 2) by Non Sequor on Wednesday November 12 2014, @02:14AM
Most important feature: 157 GPIO pins
You can wire this thing to 157 sensors, lights, etc.
Raspberry Pi has 17 pins.
This thing is ideal for data logging, automation, or robotics projects or just making some doodad that needs some computing power to compensate for your lack of electronics experience. A little electronics knowledge can go a lot further when you've got an overspecced general purpose computer in your parts list.
Write your congressman. Tell him he sucks.
(Score: 2, Informative) by tftp on Wednesday November 12 2014, @02:35AM
In most cases you don't want 157 dumb GPIO pins. Well, maybe if you are building a simple security system with switches... for most applications, though, you need something more - SPI, I2C, RS-232, and such. Sometimes you need timer capture pins; sometimes you need ADC and DAC; sometimes you need to generate PWM. Look at any common AVR to see the usual assortment.
I do not know how many of those 157 pins are capable of those more advanced functions. However I know that it's easy to create a simple USB peripheral with as many GPIO pins as you need - and with those advanced functions as well. You can do it with AVR, you can do it with PIC - there are several options. That USB only requires OTG - and that is present in many devices.
Still, there are times when you have to have direct access to a pin. For example, when that is an interrupt pin. Or when you are sending gigabits of SATA or PCIe over that pair. Then you must use hardware that supports those interfaces, and USB won't do it. But how many amateur designs require such things? A typical geek's project - a beer making rig - will do just fine being controlled over USB, from a tablet that is glued to the wall of the device. The board comes with no user interface, so it would be another expense and another complexity that a tablet saves you from. I'd think that the board has its use, but it has not as many such uses as a combination of a generic tablet and a generic USB perf board with an AVR32UC3x, with all its pins brought to headers. Such a breakout board would cost less than $20 in parts, and you can plug it *anywhere*. You can also replace it without much pain if your device manages to burn it. A combination of Android-based GUI and an AVR32-based MCU could be very, very powerful. You get the best of both worlds.
(Score: 2) by TheGratefulNet on Wednesday November 12 2014, @03:22AM
some of the tft touch screens have a 40 pin connector and its NOT just simple i2c or spi ;( you need tons of pins to talk to such devices.
btw, while we're talking about the beagle series, does anyone know of a stable distro (I don't want angstrom) for the BBB that does NOT crash and lockup on a daily or near-daily basis? the last 2 ubuntu distros by robert (forget his full name; but he's a digikey guy and he's pretty heavily involved in the bb-black) were not stable and I really need a system that will run for weeks and months without dying. the pi can do it, but the last few debian (ubuntu, certainly) efforts were pretty unstable.
"It is now safe to switch off your computer."
(Score: 1) by tftp on Wednesday November 12 2014, @03:29AM
GPIO is all but useless for that because the refresh rate of any LCD beyond the postal stamp size is in tens of MHz. You cannot operate GPIO that fast. That's why some MCUs come with hardware LCD controllers that take care of this. The interface is not complicated, but it pumps a lot of data into the LCD. You need more or less 3 bytes per pixel, and there are X * Y pixels, and there is at least 25 Hz refresh rate.
I am running Ubuntu on my BBB, and it never crashes. I use this:
ubuntu@bbb:~$ uname -a
Linux bbb 3.8.13-bone20 #1 SMP Wed May 29 06:14:59 UTC 2013 armv7l armv7l armv7l GNU/Linux
(Score: 2) by TheGratefulNet on Wednesday November 12 2014, @03:38AM
ah, 3.8
kind of old, but if it stays up, that's what I'm looking for.
was hoping 3.13 or 3.16, but something really nasty happened after 3.8 (it seems); and I'm not sure if its the kernel or userland stuff.
thanks for the tip.
"It is now safe to switch off your computer."
(Score: 2) by Non Sequor on Wednesday November 12 2014, @03:30AM
To me, there is a certain elegance in a device that has so many pins and so much computing power that it allows someone who has no existing skills on how to economize those resources to pull off some project. It can help make stupid engineering work! And stupid engineering is my kind of engineering! I'm a math major, and I know all about what is possible but have a practical training in very little.
This thing is overspecced for most of my ideas, but it makes me want to dream bigger (and stupider).
Write your congressman. Tell him he sucks.
(Score: 1) by tftp on Wednesday November 12 2014, @03:58AM
Well, I'm an EE, and I would be wary of eagerly connecting anything to those pins. I'd think they are LVCMOS33 pins, which means that you need level converters on all of them, to the voltage levels that your controlled equipment requires. For that you need to build a separate PCB with translators on it - and, in practice, with TVS diodes and with EMI filters as well. A ribbon cable from the board to anywhere else is nearly guaranteed to burn it all. Some of my designs (for outdoor stuff) use optocouplers and isolated switchers.
As I said already, very few practical devices can be operated with GPIO. Take a DSP, for example. You want to build a music synthesizer, and you want to use Blackfin DSP [wikipedia.org] for that. (You can do this on this board too, but it may be not optimal for your processing.) Blackfin uses a flavor of I2S (or SPORT) to receive and send data. The GPIO is not capable of that, unless you are not interested in anything above 100 Hz :-) You need a hardware core within the MCU that will take care of I2S for you, using DMA. An AVR has such hardware; that's why AVR and PIC are so popular - they are loaded with fancy hardware that can reduce your CPU load a thousandfold. They have more hardware cores than the pins!
(Score: 2) by VLM on Wednesday November 12 2014, @01:40PM
Take a DSP, for example
Of course a lot depends on if you're trying to interface a DSP because you like DSPs or if you have an "audio related goal". I see its got onboard audio, and PCIe via headers and COTS PCIe sound cards work. COTS PCIe anything should work. Or you can hook up FPGAs (perhaps with DSP onboard or as an interface) to the multiple USB ports.
And of course DSP-ish computation that required dedicated hardware 20 years ago or serious hardware 10 years ago is just general purpose embedded computation in 2014. Doing cutting edge stuff that was computationally impossible 10 years ago is probably exactly the wrong place for a noob to start out, so that problem tends to be kinda self limiting.
And the SDR design pattern of figure out how wide of a bandwidth you can process and then design your A/D around that limitation, rather than designing your A/D and then trying to figure out how to process it. So if you can't process 4 megasamples/sec there is little point in worrying about the 4 megasample/sec hardware because you're only going to be able to process the data from of a "sound card class" USB device.
(Score: 2) by novak on Wednesday November 12 2014, @03:43AM
Yeah, ok, some of the features on this board aren't all that easy to use, but if you've got a system with the power they've outlined (Dual 1.5 GHz cores) you're talking more about something that can compete with a desktop. Sure, it's not equal to any sort of remotely new x86/x86-64, but it's way above being some tiny embedded device.
This thing has dual image processing units, DDR3, a video decoder- you're talking about polling events on a scale of maybe once per minute? You could do that with literally any microprocessor ever made, although depending on the sensor count you might need some external chips for I/O.
You need to think bigger, maybe a robot vision processor with facial recognition or something, because this guy can do roughly 2,999,999,999 more operations per second than you'd need for that project. It IS overkill for a project like that, and you'd be right to pick up a small AVR instead.
novak
(Score: 2) by cafebabe on Wednesday November 12 2014, @03:09AM
My local hackerspace went a large batch of shift registers to get more IO pins. In some ways, this is preferable because mishaps only cause a cheap interface chip to get fried.
1702845791×2
(Score: 2) by LoRdTAW on Wednesday November 12 2014, @03:15AM
I keep hearing this. A tablet running android is useless as an embedded system. They are about as close as you can get to a black box running a crummy OS with zero I/O. And USB is a joke for anything real time.
(Score: 1) by tftp on Wednesday November 12 2014, @03:44AM
It all depends. If you need a hard real time system, separate the real time functions into an external MCU. These cost a few dollars - no reason to solder wires to the LPT port as we did decades ago. Talk to that real time system via any channel of your liking, and have a wonderful and entirely not-real-time GUI on a separate control panel. Your TV is a hard real time design in A/V datapaths; however there is very little real-time in its remote control function, and nothing at all in the sleep timer.
But if you want to do it all in one box... put your real time code into a Linux driver, and insmod that driver into the kernel. Short of you planning to run busy-wait loops, it will do what you need, within limits (such as you cannot guarantee interrupt latency. If you need that, see above.) As a side benefit, the driver will expose a userspace interface that you can use from Android.
If you want to go deeper and achieve the ultimate performance, replace Linux with a task scheduler of your choice, or with a free/commercial RTOS. Porting of an RTOS usually requires only connecting a timer. But if timesharing is not good at all, write your application and boot it with a bootloader.
My point is that a tablet's hardware is the same hardware that you get on boards like this one. There is no magic. You aren't getting anything extra from a large board, except a few I/O connectors that are soldered directly to pins of the MCU. If that matters, go with the board. For many that does not matter, and they can use the tablet's hardware with their choice of software - from the bare metal code to Android and Java.
(Score: 2) by VLM on Wednesday November 12 2014, @01:52PM
Software is extremely expensive (devs don't work for free), and sometimes you can replace thousands of person-hours of dev time and RTOS development and latency debugging and tens of thousands of lines of code with a $1 PIC microcontroller as a smart interface/buffer in a small tight loop.
Its very expensive and difficult on modern hardware and modern OS to multitask while emulating cycle-perfectly what a $1 microcontroller can do natively for about a buck. Also there's no demarcation point between systems quite so reliable as a hardware bus LOL.
(Score: 2) by LoRdTAW on Wednesday November 12 2014, @02:34PM
You really didn't have to go through all this to prove some kind of argument. I just find it annoying that when a low cost high performance embedded hobbyist board comes out someone has to come along and claim "an R-Pi/tablet/arduino can do the same thing for less". Those people clearly don't understand embedded design and should not make such comments. Those solutions are all in different classes and they all have their advantages and disadvantages. If the Beagleboard-X15 is too costly for you then by all means, go buy something cheaper. Is the X-15 way over powered? Then buy only what you need. There is no argument because embedded design is a horses-for-courses practice. You buy what you NEED. You balance performance, quality and cost.
If a tablet is what you want to use, then use it. Just don't tell me or others that it is a comparable solution to higher performance embedded boards because it is an outright lie.