Thursday, April 18, 2019

Working with GPUs for Accelerated Computing: An Open Source & Decentralized Approach

A Convergence for Computing

GizmoQuest Computing Lab is an Open Source, Independent laboratory which currently empowers Decentralized Bioinformatics Research, with the help of 3 Compute Servers and Workstations (specifications available below) to accelerate computational simulations and bring out results much faster with GPUs. The lab has been used to author and publish a book on GPU Computing:
The book has been enlisted at the advanced book list of the Python Programming Language and also at Stanford Libraries.

The Deep Learning Perspective For The Book

Figure: How deep learning is a subset of machine learning and how machine learning is a subset of artificial intelligence (AI). Photo by Avimanyu786 / CC BY-SA

Here is a look at our liquid-cooled computer in action!


This laboratory caters to the growing requirements of accelerated research with a 300 mbps internet connection and is used for a wide variety of research applications. It provides modern computational and facilities currently for all kinds of research. It especially provides support for GPU Accelerated Computing. Likewise, it remains busy 24/7.


The aim of this laboratory is to provide required Computing and Simulations facilities to creative learners, researchers and PhD students in order to meet their requirements of the growing trends in Bioinformatics and Science.
The main emphasis is to do project work in lesser time with the use of GPUs and deep learning. Collaborative research is highly encouraged, and Decentralized Science, or DeSci, makes this very easily possible! How? Read on.

Primary Hardware Configuration


AMD Ryzen 7 2700X Processor at 3.7 Ghz
AMD Radeon VII 16 GB HBM2 GPU (Vega 20 Architecture)
8 GB DDR4 RAM (tentative)
500 GB WD M2 SSD Blue
4 TB Western Digital HDD Red

Secondary Hardware Configuration


Intel Core i7 4770K Processor at 3.5 Ghz
Nvidia RTX 2060 12 GB GDDR6 GPU (Maxwell Architecture)
32 GB GSkill DDR3 RAM
4 TB Western Digital HDD Green

Tertiary Hardware Configuration


AMD Athlon 3000G at 3.5 Ghz
Nvidia RTX 3060 Ti 8 GB GDDR6 GPU x3
Nvidia RTX 3090 24 GB GDDR6X GPU
32 GB Corsair DDR4 RAM


Current Line of Work

Our primary goal is currently on building a platform for computational research through DeSci with NuNet. We believe in convergence, be it Science or Art! Our founding member, Avimanyu Bandyopadhyay is pursuing his PhD in Bioinformatics at Maulana Abul Kalam Azad University of Technology since January 2016 and also working as a Systems Engineer at NuNet.

Once NuNet's Decentralized GPU infrastructure is live, we look forward to onboarding all the above machines to help build a globally decentralized computing economy for researchers!



Future Projects/Ideas on a Decentralized Network

In the field of genomics, a simple modern run of DNA sequencing can take up to 1 TB per run according to computational biologists. So a 4 TB storage would take care of such requirements. Also, as an example, the transfer rate of the Radeon VII happens to be 1 TB/s with 1:4 FP64 performance at 3.46 TFLOPS (great at the consumer level)!

This Mini Open Source Lab will be open to all diverse computational fields and not just Science. One great example would be Computational Archaeology. Such ideas could help preserve heritage sites such as the Kailash temple at Ellora!

We're very much eager to explore the field of Computational Psychiatry and Computational Psychology especially through Video Games.
The BIG 5 Domains of Play have been explored and explained many times by Jason VandenBerghe at Game Developers Conference (GDC) events.
Once our work is completed on building a Decentralized Computing Infrastructure, we look forward to exploring the above and more!

Working Hours (Includes Manual/Automated)

9:00 AM - 12:00 AM IST (Subject to Change)



No comments:

Post a Comment