Play time is over. When Jen-Hsun Huang takes the stage for the opening keynote for GTC, it can mean only ONE thing.... :)
http://www.geforce.com/whats-new/art...c-2016-keynote
And watch it be like the LAST thing he mentions (of course)
Printable View
Play time is over. When Jen-Hsun Huang takes the stage for the opening keynote for GTC, it can mean only ONE thing.... :)
http://www.geforce.com/whats-new/art...c-2016-keynote
And watch it be like the LAST thing he mentions (of course)
First big announcement (aside from the nVIDIA SDK)...
iRAY VR / iRAY VR Lite
http://s19.postimg.org/85jk7bgov/GTC_Cap_1.jpg
http://s19.postimg.org/yf4mq42m7/GTC_Cap_2.jpg
Breakthru VR rendering technology to power the VR experience of the future.
Did a live demo with Steve Wozniak and a VR experience on Mars. Also demoed an amazing Mount Everest VR recreation.
But...what is all this going to run on...? :)
First up (hardware-wise)...
Deep Learning / AI / Hyperscalers...
Powered by Tesla M4 & Tesla M40
(The M4 is the little guy, M40 big boy)
http://s19.postimg.org/k9ytoatkv/GTC_Cap_3.jpg
http://s19.postimg.org/qc6gesi0v/GTC_Cap_4.jpg
Crazy demo of how Skynet is definitely going to kill us all.
They fed thousands of images from a specific time period, the deep learning neural network drew its own, new images (i.e. unsupervised learning).
Up next....and getting closer...
The Tesla P100.
http://s19.postimg.org/ixh4mew5b/GTC_Cap_5.jpg
As a comparison, its predecessor (Maxwell generation) contained 7 billion transistors...
The Tesla P100 contains 150 billion transistors.
http://s19.postimg.org/omxd6q2bj/GTC_Cap_6.jpg
And since we know that there are always two classes each GPU...one can only imagine what's next...
So out of all the major announcements today (five), what's left?
http://s19.postimg.org/dbupicvgf/GTC_Cap_7.jpg
The world's first Deep Learning Supercomputer...The nVIDIA DGX-1
http://s19.postimg.org/4i3t198hr/GTC_Cap_8.jpg
So you see the specs on the Tesla P100, this beast has EIGHT of them inside.
http://s19.postimg.org/e3xdhjznj/GTC_Cap_9.jpg
And remember all the talk about Pascal being "10x faster than Maxwell"? It's not 10x faster...it's 12x faster.
I'm already salivating at the possibility of getting to work with a client that will be looking to implement this soon :)
So yeah...he's being a super big tease about waiting to the last minute to give us what we really want.
And the cost of the DGX-1 is $129k.
That's actually pretty impressive. Making it available to leading Deep Learning institutions (Berkeley, Mass. General, etc...).
Use-cases for medical and molecular research industries (and gas, oil, geology, etc...). I know not super exciting to some, but I geek out over some of this stuff because I work with clients in those industries :)
http://s19.postimg.org/apjsf5tmn/GTC_Cap_10.jpg
And yeah, yeah, yeah....we already know about nVIDIA Drive and AI Car Computing...
http://s19.postimg.org/n5gi8wmyn/GTC_Cap_11.jpg
http://s19.postimg.org/c758qpydb/GTC_Cap_12.jpg
Boooooooooooooooooooooooooooring! ;)
Keep it coming, I'm hooked!
Sent from my iPhone using Tapatalk
Wow...and so all that for nothing...
No GTX announcement :(
Quite a shocker.
Well that was a total waste of 2.5 hours! ;)