Raspberry Pi Foundation Updates Compute Module With Faster Processor

Subject: General Tech | January 22, 2017 - 12:11 AM |
Tagged: Raspberry Pi, compute module, Raspberry Pi 3, broadcom, iot

The Raspberry Pi Foundation is introducing an updated Compute Module that puts the single board computer for embedded devices more in line with the performance of the newest hobbyist Raspberry Pi 3.

The Raspberry Pi Compute Module 3 is a pin compatible successor to the Compute Module 1 (there is no CM2) that, according to the Raspberry Pi Foundation, offers twice the RAM and 10-times the CPU performance.

View Full Size

Note that while the Compute Module 3 may be able to be a drop in upgrade / replacement for devices powered by the first generation CM1, it uses more power, puts out more heat, and is 1mm taller so while it is pin compatible it may not work in all devices if their module slot space, power supply, and airflow / heatsinks are not up to the task.

The Compute Module 3 is a small single board computer with a SO-DIMM connector that can slot into embedded and IoT products. It is powered by a Broadcom BCM2837 with four ARM Cortex-A53 CPU cores clocked at 1.2 GHz and a dual core VIdeoCore IV GPU clocked at 400 MHz. The processor is paired with 1GB of RAM. As far as onboard storage, the Compute Module 3 will come in two SKUs: the CM3 with 4GB of eMMC or a CM3 Lite without pre-installed eMMC and solder points for manufacturers to add their own eMMC or micro SD card slot. The VideoCore IV GPU supports 1080p30 decode of H.264. Users wanting hardware decode of H.265 and/or 4K support will have to look elsewhere. As is usual with Broadcom, exact specifications of the BCM2837 (especially their GPU) are kept close and quiet, unfortunately.

The exact ports and I/O from the Compute Module 3 will depend on the device and what manufacturers implement and wire to the connectors on the SO-DIMM slot. However, looking at the CMIO3 development board (96 Euros, $116 USD) shows that the CM3 supports GPIO, USB, micro USB, CSI (camera interface), DSI (display interface), HDMI, micro SD, audio, and networking. 

The Compute Module 3 can run Windows IOT Core or any number of Linux distributions compatible with ARM processors.

The Compute Module 3 is $30 while the “lite” variant without eMMC is $25. A kit including the development I/O board and both CM3 SKUs is $200. NEC has already announced it will be using the new Compute Module 3 in their digital signage and displays. Other applications include Smart TVs, home automation, and industrial control systems as well as hobbyist projects and robotics.


Source: Raspberry Pi

January 23, 2017 | 07:40 AM - Posted by Anonymous (not verified)

Hope we get pi4 soon

January 23, 2017 | 10:06 AM - Posted by Anonymous (not verified)

From the wikipedia entry there appears to be some info available on the VideoCore IV graphics IP. (1)

"On 28 February 2014, on the day of the second anniversary of the Raspberry Pi Broadcom, together with the Raspberry PI foundation, announced the release of full documentation for the VideoCore IV graphics core, and a complete source release of the graphics stack under a 3-clause BSD license. [10][11]" (1)


"VideoCore" [See section labled Linux support]


Note: the link in the wikipedia refrence(PDF) appears to be broken so maybe there is more info at xorg/other linux development website that has the proper documenation that wars released.

January 23, 2017 | 02:40 PM - Posted by Anonymous (not verified)

Looks like ASUS(1) is getting in on this market also, and it supports 4K. Maybe in the future they can update from the Mali/Midgard to a Mali/Bifrost micro-arch. I'd also like to see a 64 bit A 73 based system as that will really be great paired with a Mali/Bifrost GPU in the hobby class of system.

The GPU in the ASUS system is better but the CPU in not so great, that Gigabit Ethernet is also nice, as is the wifi.
Really once the A73(Artemis) paried with Mali/Bifrost garphics SOC based systems start to make inroads into this market then things will get really intresting for graphics uses on these types of devices.

"The 4K video support—including H.265 decoding—in Asus’ board is thanks to the Mali-T764 GPU lurking inside its quad-core, 1.8GHz Rockchip RK3288 SoC, as first reported by Hexus. In addition to 4K, the board also supports 192kHz/24-bit audio. The PC comes with 2GB of RAM, Bluetooth 4.0, four USB 2.0 ports, 1 HDMI out, microSD port, gigabit ethernet, and 802.11b/g/n Wi-Fi. For hobbyists, the Tinker Board packs a 40-pin internal header with 28 GPIO pins, CSI port for the camera, and DSI port.

The 2GB RAM is particularly nice considering the Raspberry Pi Model 3 B comes with just 1GB. As for software, the UK sales page says it supports Debian Linux and (conveniently) Kodi—the popular media player." (1)


"The Asus Tinker Board is a powerful Raspberry Pi rival that plays 4K video"


January 23, 2017 | 07:40 PM - Posted by Anonymous (not verified)

We don't need too many similar products. The POINT of this is to learn how to program and learn the lower-level hardware methodology such as Instruction Set, IO, bus etc.

Raspberry Pi has built up software and documentation useful for tinkerers and school courses.

So we really just need a good, limited ecosystem of hardware and tools.

January 24, 2017 | 12:06 AM - Posted by Anonymous (not verified)

You can stick with the Pi for high school level school work, but others are doing more than that level and no school level below college needs more than the first/second generation Pi! There are others developing for more intensive applications that require more power for their workloads that need the latest generation Pi and other makers’ products for their needs. There are the folks that are developing cluster computing systems around the Pi and other systems that need the extra processing power that the latest generation Pi/other competing systems provide.

Those low cost cluster systems made from other makers besides the Pi foundation really need and benefit from the gigabit Ethernet and other features for a very low cost method for creating/developing skills for cluster computing systems. By using several dozen/more of these small tinkerer SKUs some college level experience can be had using Linux Kernel based OSs to produce the working software that can then make for some very low cost cluster training systems without breaking the bank! And the software/learning skills developed using these systems is directly transferable to the large multi-million dollar systems that are very hard to get access to even among the colleges that have their own supercomputing facilities.

The more competition the better for the low cost hardware with the best feature sets and maybe by the time the next generation of these systems comes around there will be A73 CPU and Mali/Bifrost GPU based SOC SKUs so folks can learn on the latest GPU IP from ARM Holdings/others because the Mali/Bifrost GPUs are very different from what came before(On the same level as for example AMD’s transition from Terascale to GCN).

You do not have to ever leave the first generation Raspberry Pi for any basic learning needs but there are others than need a little more processing power for robotics projects that need a little more GPU processing power and AI power for the robotics tinkerers that are much more advanced. These systems are getting better because the hardware makers have to compete for offering feature levels for their systems to remain relevant otherwise the Raspberry Pi folks would have never updated beyond the original offering.

That ASUS tinkerer system would make for a nice candidate for a render farm running Blender 3d under Linux with the level of GPU processing power that it provides over what the Pi offers.

The Arm A73 provides plenty of 64 bit processing power on a very low power metric compared to even the 32 bit A17 based CPU in the ASUS system. I’m really looking forward to what that ARM Mali/Bifrost GPU micro-architecture is going to bring to the market once these types of systems begin using the latest ARM Mali-Bifrost GPU IP for compute acceleration as well as graphics.

January 24, 2017 | 12:58 AM - Posted by Anonymous (not verified)

The point is to have as many clowns as possible playing N64 emulators at a decent framerate asap, so they can complain Dreamcast is too slow afterwards.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.