Under the microscope: Building a distributed 'supercomputer'
This physical tutorial gives you a scheme for building your very own computer 'cluster', allowing you to help with computational science such as Folding@home.
Folding@home relies on graphics processing units (GPUs) as the 'computers' to run their modelling software that simulates interaction dynamics at the molecular level. This tutorial helps you build your own computational (GPU-based) platform, giving you the power to help them do vaccine development better (and potentially helping in other important ways, too).
Supercomputer hardware
Within recent times we have seen a dramatic increase in power available in GPUs, an increase that is more cost-effective than compared to traditional CPUs. Along with this, the computational overhead of using GPUs has also decreased on systems such as NVIDIA's CUDA framework. Until CPU development catchs up, now is an optimal time to invest in a GPU-based 'cluster' to do something worthwhile, especially now crypto-mining has lost it's financial imperative.
How to build your own 'supercomputer'
1) Buy used hardware. The sneaky buyer may be able to snag a bargain on second-hand GPUs from existing crypto-miners. They are probably wanting to dump their hardware, as there is no financial benefit to mining anymore (due to the 'merging' event that occurred recently). Ideally get some NVIDIA RTXs around the 3060 Ti level to maximise your cost-efficiency.
2) Build your supercomputer. My experience led me to believe that adapting the mining 'rig' default configuration is almost ideal. But, change the motherboard (MB) from a mining type to an enterprise-y MB from last decade (I'm using an HP Z420 board). As these motherboards don't have all PCIE x16 type slots you will need to buy adapters. Besides the motherboard, the hardware layout is identical to building a mining rig with an open-air chassis.
3) Configure the platform. Install Linux, then GPU drivers, then Folding@home software. Dell supports Linux, while I had no issue running Ubuntu 22.04 LTS on my HP hardware. Ubuntu even has an install guide. The command-line Terminal program can be used to install the most recent NVIDIA drivers available, and also both Folding@home's client and core software. For power efficiency reasons, you can set the GPU's clock speeds to be below their maximum.
Conclusion
The main points covered within this article included suggestions: to find cost-effective GPU hardware, for building the actual cluster computer (which is similar to building a mining rig), and setting up the computer for molecular dynamics simulations via Folding@home.
Once your computer is online and ready to run such software, we believe you will be helping to solve one of the biggest problems of our time - potentially helping to save millions of people all over the world from dying unnecessary due to viruses such as COVID-19.