Turing Pie 2: Low energy clusters

We’re not in the habit of recommending the Kickstarter project here at Hackade, but when our desktop shows prototype hardware, we can’t help but play with it and write it for the readers. And that’s exactly where we find ourselves with Turing Pie 2. You may be familiar with the original Turing Pie, a career board that runs seven Raspberry Pie computer boards simultaneously. Which supports Compute versions 1 and 3, but a new design was definitely needed for Compute Module 4. Not only does CM4 support, the developers of the Turing machine have designed a 4-slot carrier board based on the NVIDIA Jetson pinout. . The entire line of Jetson devices is supported, and a common adapter CM4 works. Even around the RK3588 a brand new module has been planned, which should be quite impressive.

One of TP2’s design decisions is to use the Mini-ITX form-factor and 24-pin ATX power connection, which gives us the option to install TP2 in a smaller computer case. Even people in my electronics are planning a custom rack-mountable case. So if you want 4 or 8 raspberry pieces on a rack mount then this is for you.

Appeal – and risk

“Wait, wait,” I heard you say, “there are plenty of ways to rack-mount raspberry pieces!” Of course. The form factor options are simple, but the real magic is the rest of the board. Individually controlled power supply for four boards from a single ATX power supply creates a very clean solution. Need to reboot a hanging Pi remotely? There is a Baseboard Management Controller (BMC) that will have full power control over the network. This is the real killer feature: BMC is going to run open source firmware, and will power some very clever functions. UART Want to solve a boot problem? It is available from four nodes in BMC. Need a CM4 to push a new image? BMC will include image flashing function. The board features a Gigabit network switch that connects PIS, BMC and two external Ethernet ports, supporting all VLANs.

On the other hand, much of the BMC Wizard has not yet been implemented in the review unit. This is the biggest promise of the project and it can be distorted. Putting together a stable firmware with all the bells and whistles three months before the scheduled ship date can be a bit optimistic. I’m hoping for a functional firmware, with updates to refine the experience in the months following the launch.

Then there’s the expanded IO. The board comes with a pair of mini PCIe ports, 4 USB3 ports and a pair of SATA ports. It works through PCIe lanes open by various computer modules. Connected to Node 1 and 2 mini PCIe ports, Node 3 with SATA and Node 4 with USB3 port. On top of that, a convertible USB2 port can be dynamically assigned to any of the existing nodes. Oh, and there’s an HDMI output from Node 1, so more options, such as running a Pi CM4 8GB on a desktop machine. A late option added to Kickstarter bolts four NVMe ports under the board, one in each slot, although not every computer module has PCIe lanes to support it.

Now keep in mind that I’m testing a pre-production unit (more on that later), and all of the above aren’t actually working yet. Several changes have been made to the production board vs. my unit and the BMC firmware on this board is minimal. There are also supply-chain issues that we covered here in Hackaday, but have the advantage of being designed in the absence of TP2, so using hard-to-source parts should be avoided.

In case of use

Now let’s talk about what he doesn’t do. This may sound obvious, but Turing Pie 2 doesn’t give you a single ARM machine with 16+ processing cores. There is not enough magic onboard to make devices work like a unified multi-processor computer. I’m not sure there really is enough magic somewhere to stop this. However, what you will find is four easy-to-operate machines that are suitable for running light-weight services or docker images.

Looking for a platform for learning dockers and coubernets? Or a place to host GitLab, NextCloud, and a file server? Maybe you want to play Nginx as a front-end proxy, and there are several devices running behind it? The homelab-in-the-box nature of the TP2 makes it a useful choice for all of the above. And while you can’t reasonably do all of the above in a raspberry pie, a programmable cluster of 4 of them works pretty well. VLAN support means you can add virtual NICs to your nodes and create an internal network. With two physical Ethernet ports, you can even use your TP2 as your primary router, on top of what it can do.

Real-world testing

So what is the actual status of the project? My pre-production board is currently booting a Raspberry Pi CM4, a Pine64 SOQuartz module, an NVIDIA Jetson Nano, and a Jetson TX2 NX. The Jetson Xavier NX needed a minor board change, but once it was done it went like a champ. A pre-production board has the usual warts, such as extra deep switches everywhere and some jokes, such as Ethernet only coming to 100M for some devices. These are known problems, and are a good example of why you run a Rev 0 board test. The final product should work all kinks.

I’m observing a power draw, and the one I pull the most is just 30 watts of power. It offers a real-world use, an off-grid computer cluster. Mini-PCI ports should allow an LTE modem (or you can use Starlink if * path * is off the grid). Add a few cameras and install the ZoneMinder Docker image and you have a low-power video monitoring solution. Add an RTL-SDR dongle, and the rtl_433 software is listening to a solar-powered weather station, and you can track the weather at your remote location. Just for fun, I ran a Janus Docker image on one of the Raspberry Pi CM4s on my TP2. Janus is the WebRTC server we’ve integrated into ZoneMinder, and I’ve been able to stream 12 security cameras live to 1080p, using only about 25% of the available processor power, or a load of 1 on a four-core pie. It’s a testament to how lightweight Janus is, but also a great example of what you can do with a TP2.

What next?

Kickstarter is over, more than two million dollars raised, but don’t break a sweat, because you’ll soon be able to buy a Touring Pie 2. Orders will be handled via the Turing Pie website, stay tuned for details. There will be a few months until the final review of the board is completed and sent, hopefully with some killer firmware and everything will work just like the ad. Then there’s the alluring RK1 computing board, with RK3588 to 32 GB of RAM and eight-core arm goodness. It’s a little ahead, and it could be a second kickstarter campaign. I asked about mainline support for RK1, and was told that this was a preliminary target, but they weren’t sure about the exact time. There’s been a lot of excitement around this particular chip, so wait for the community to work together to get all the bits needed for mainline support.

Turing Pi 2 and RK1 using NVIDIA Jetson SO-DIMM connector may have an unintended consequence. Imagine a handheld device built into the antimicrobial open source Jetson baseboard, which works with multiple computer modules. I mentioned the Pine64 SOQuartz: it’s not an officially supported board in TP2, but since Pine64 built it into the CM4 specification, it clicks on the adapter card and acts like a champ. There is an interesting possibility that one or two of these computer module interfaces will gain a considerable critical mass, given that it is widely used in devices. And if anyone thinks, using the TP2 CM4 adapter magically doesn’t allow CM4 to boot on the Jetson Nano carrier board. Yes, we checked.

So is Turing Pie 2 for you? Maybe. If you don’t mind juggling multiple single-board computers, and need a mess of wires, maybe not. But if the ability to have four SBC slots in a single mini-ITX case, with a BMC that makes life easier sounds like a breath of fresh air, then take a look. The actual test will be when the finished product will be shipped, and what form the support is in. I am cautiously optimistic that it will not be too late and that the OSS firmware will work on it. I look forward to getting my hands on the final product. Now if you forgive me, I think I need to set up an automated system to create aarch64 docker images.

Leave a Reply

Your email address will not be published.