Deep learning is becoming powerful tool to analyze big data. However, one of it's biggest bottleneck is the "extensive computation requirement".For example, Google had to use 16,000 parallel CPU cores and 72 hours of time to train it's famous deep learning application, "detecting cat from random image".DeepX's mission is to save this time and resources used for deep learning application by providing dedicated hardware processor for deep learning. Currently, DeepX has the world's fastest hardware processor implemented in single FPGA board, which runs 784 times faster than Intel CPU thanks to our proprietary hardware architecture. This also outperforms NVIDIA's latest deep learning processor, M40 by 10 times. They are also planning for ASIC solution which will bring DeepX several hundreds of additional performance gain. DeepX can change weeks of training time to a minute for deep learning application companies.

"We definitely need lots of help from professionals in various area especially we are hardware startup that requires extensive capital and network. We are thrilled to have TFG on our back to support us!" - Justin Jang, Founder. 

Leran more about DeepX:


DeepX's EIR: Hon Wong

Posted on February 15, 2016 .