256-Core RISC-V Megacluster

May 7, 2024 by Julian Scheffers 16 Comments

Supercomputers are always an impressive sight to behold, but also completely unobtainable for the ordinary person. But what if that wasn’t the case? [bitluni] shows us how it’s done with his 256-core RISC-V megacluster.

While the CH32V family of microcontrollers it’s based on aren’t nearly as powerful as what you’d traditionally find in a supercomputer, [bitluni] does use them to demonstrate a property of supercomputers: many, many cores doing the same task in parallel.

To recap our previous coverage, a single “supercluster” is made from 16 CH32V003 microcontrollers connected to each other with an 8-bit bus, with an LED on each and the remaining pins to an I/O expander. The megacluster is in turn made from 16 of these superclusters, which are put in pairs on 8 “blades” with a CH32V203 per square as a bridge between the supercluster and the main 8-bit bus of the megacluster, controlled by one last CH32V203.

[bitluni] goes into detail about designing PCBs that break KiCad, managing an overcrowded bus with 16 participants, culminating in a mesmerizing showcase of blinking LEDs showing that RC oscillators aren’t all that accurate.

Continue reading “256-Core RISC-V Megacluster” →

Parallel Computing On The PicoCray RP2040 Cluster

April 9, 2023 by Joseph Long 28 Comments

[ExtremeElectronics] cleverly demonstrates that if one Raspberry Pi Pico is good, then nine must be awesome. The PicoCray project connects multiple Raspberry Pi Pico microcontroller modules into a parallel architecture leveraging an I2C bus to communicate between nodes.

The same PicoCray code runs on all nodes, but a grounded pin on one of the Pico modules indicates that it is to operate as the controller node. All of the remaining nodes operate as processor nodes. Each processor node implements a random back-off technique to request an address from the controller on the shared bus. After waiting a random amount of time, a processor will check if the bus is being used. If the bus is in use, the processor will go back to waiting. If the bus is not in use, the processor can request an address from the controller.

Once a processor node has an address, it can be sent tasks from the controller node. In the example application, these tasks involve computing elements of the Mandelbrot Set. The particular elements to be computed in a given task are allocated by the controller node which then later collects the results from each processor node and aggregates the results for display.

The name for this project is inspired by Seymore Cray. Our Father of the Supercomputer biography tells his story including why the Cray-1 Supercomputer was referred to as “the world’s most expensive loveseat.” For even more Cray-1 inspiration, check out this Raspberry Pi Zero Cluster.

Building A Cheap Kubernetes Cluster From Old Laptops

February 8, 2021 by Lewin Day 28 Comments

Cluster computing is a popular choice for heavy duty computing applications. At the base level, there are hobby clusters often built with Raspberry Pis, while the industrial level involves data centers crammed with servers running at full tilt. [greg] wanted something cheap, but with x86 support – so set about building a rig his own way.

The ingenious part of [greg]’s build comes in the source computers. He identified that replacement laptop motherboards were a great source of computing power on the cheap, with a board packing an i7 CPU with 16GB of RAM available from eBay for around £100, and with i5 models being even cheaper. With four laptop motherboards on hand, he set about stacking them in a case, powering them, and hooking them up with the bare minimum required to get them working. With everything wrapped up in an old server case with some 3D printed parts to hold it all together, he was able to get a 4-node Kubernetes cluster up and running for an absolute bargain price.

We haven’t seen spare laptop motherboards used in such a way before, but we could definitely see this becoming more of a thing going forward. The possibilities of a crate full of deprecated motherboards are enticing for those building clusters on the cheap. Of course, more nodes is more better, so check out this 120 Pi cluster to satiate your thirst for raw FLOPs.

Terrible Cluster Of PIs

October 11, 2017 by Al Williams 36 Comments

When we first saw [Ajlitt’s] Hackaday.io project Terrible Cluster we thought, perhaps, he meant terrible in the sense of the third definition:

3. exciting terror, awe, or great fear; dreadful; awful. (Dictionary.com)

After looking at the subtitle, though, we realized he just meant terrible. The subtitle, by the way, is: 5 Raspberry PI Zeros. One custom USB hub. Endless disappointment.

There are four Raspberry Pi Zero boards that actually compute and one Raspberry Pi Zero W serves as a head node and network router. The total cost is about $100 and half of that is in SD cards. There’s a custom USB backplane and even a 3D-printed case.

At first, using five tiny computers in a cluster might not seem like a big deal. Benchmarking shows the cluster (with a little coaxing) could reach 1.281 GFLOPS, with an average draw of 4.962W. That isn’t going to win any world records. However, the educational possibilities of building a $100 cluster that fits in the palm of your hand is interesting. Besides, it is simply a cute build.

We’ve seen much larger Pi clusters, of course. You might be better off with some desktop CPUs, but — honestly — not much better.

Super Computing With Mini ITX Cluster

March 16, 2017 by Will Sweatman 16 Comments

[Colin Alston] was able to snag a handful of Mini ITX motherboards for cheap and built a mini super computer he calls TinyJaguar. Named partly after the AMD Sempron 2650 APU, the TinyJaguar boasts four, yes that’s four MSI AM1I Mini-ITX motherboards, each with 4GB of DDR memory.

A Raspberry Pi with custom software manages the cluster, and along with some TTL and relays, controls the power to the four nodes. The mini super computer resides in a custom acrylic case held together by an array of 3D printed parts and fasteners.There’s even a rack-like faceplate near the bottom to host the RPi, an Ethernet switch, an array of status LEDs, and the two buttons.

With 16 total cores of computing power (including GPU), the TinyJaguar is quite capable of doing some pretty cool stuff such as running Jupyter notebook with IPyParallel. [Colin] ran into some issues getting the GPU to behave with PyOpenCL. It took a bit of pain and time, but in the end he was able to get the GPUs up, and wrote a small message passing program to show two of the cores were up and working together.

Be sure to check out [Colin’s] super computer project page, specifically the ten project logs that walk through everything that went into this build. He also posted his code if you want to take a look under the hood.

Raspberry Pi Zero Cluster Packs A Punch

January 25, 2016 by James Hobson 91 Comments

If you could actually buy 16 Raspberry Pi Zeros, you might be able to build your very own Raspberry Pi Cluster for only $80! Well… minus the cost of the board to tie them all together…

A Japanese company called Idein is developing a Raspberry Pi module called the Actbulb for computational sensing and data analysis. In order to perform internal testing they decided to make things easier for themselves by developing a board to allow them to plug in not one, not two, but sixteen Raspberry Pi Zeros:

Since we will use Pi’s GPU for image processing, deep learning, etc. We need real Pis but not just Linux machines. Another reason. It can be used for flashing eMMCs of our devices via USB ports when we have to do that by ourselves.

Continue reading “Raspberry Pi Zero Cluster Packs A Punch” →

AppleCrate II Doubles The Cluster Computing Fun

May 4, 2011 by Mike Nathan 32 Comments

apple_crate_II

Back in 2004, Apple hobbyist/guru [Michael Mahon] built a cluster of Apple IIe main boards dubbed the “AppleCrate” as an experiment in parallel computing. Now that a few years have passed, he is back with a new iteration of the device, aptly named AppleCrate II.

AppleCrate II was built to address some of the design limits of his first cluster project as well as to expand his parallel computing capabilities. His gripes with the first model were primarily structural in nature. The new system is organized in horizontal layers, using metal standoffs between each main board, rather than relying on a shaky wooden superstructure to keep things together. He also found his previous 8-processor configuration a bit limiting, so the AppleCrate II has 17 nodes – 16 slaves and one main board dedicated to running the operation. The cluster even uses his own homebrew networking stack known as NadaNet to enable communications between the boards.

The project is pretty impressive, so be sure to swing by his site if you want to learn more. He has a ton of technical details there, as well as copies of all of the software he used to get the cluster up and running.

[via BoingBoing]