nVidia Volta Titan V and Decentralized Deep Learning
Along with the new Nvidia Volta architecture, announced the release of the new nVidia Volta Titan V GPU cards recently and these are game changers. There’s no need to rehash why the Volta is different, so instead let’s discuss why this matters to what we do at Sugey One.
The internet as we knew it over the last 30 years of consumer growth is over. Instead, we believe between security concerns of IoT and governmental surveillance, societal challenges like cyber-bullying, and the generally harmful addictive nature of digital content, there is a growing desire on the part of many people to have a secure digital presence outside the standard web architecture. One could simply drop off the grid, or spend an enormous amount of time and energy staying vigilant through special software. However, it is easier (we love easy) to decentralize a digital presence into the home instead.
For Sugey, that means putting the data loops, the information collected by sensors, and the processing, or the computers acting on data, inside a closed loop within the home. It’s commonly known as a closed-circuit. The main bottleneck to this has always been processing power and energy. We’ll get to energy in the next part, but let’s quickly talk about processing power.
Running 20 – 30 FPS on multiple cameras plus audio required the processors to take one image every 33 ms, process it, and do something with it. Now, imagine 10 of those happening at a time and you can quickly see the processing bottleneck.
Whether it’s Google Home or Amazon Alexa, the average sensor will process a ton of data in real time and send that to a server. On a standard cloud based system, those images and audio at any point could be intercepted, or hacked at the source (with state-backed help for the biggies). People want the benefits of combined vision and audio, but they are going to want to know no one is judging them for stuff that happens in their home as well.
Closed loops and in-home servers are the best answer to both these issues.
Power per Watt
The other main challenge with home servers, or things like mining rigs, is the energy involved. Running 1000w – 2000w in a home 24/7 for a month can give someone a surprising electric bill. To be able to process all the sensors we use along with the Automated Learning, normally you would at minimum need 4 nVidia 1080’s running in SLI mode, approximately 1000w of power. It adds significantly to the cost of the product.
However, now, we can put one Volta nVidia Titan V in the brain of the system, and have twice as much processing power, with a QUARTER THE ENERGY at 250W. It’s almost a 9x multiplier! That’s insane! This is truly a major shift int he status quo for us here at Sugey.
We are very excited about the new nVidia Volta Titan V GPU cards. We look forward to seeing how they impact your businesses and lives, too.