NVIDIA CES Conference

[Hatsune Miku 「 」]

Source: NVidia CES Conference

 

Announced the World's first mobile chip that can perform 4k 60hz video is named the Tegra X1.

Tegra X1 is based on the maxwell gpu for power efficiency and performance. 

 

Unreal Engine 4 is shown running Elemental on the Tegra X1 in realtime:

 

Tegra X1 is the first gpu for Nvidia to create a new format and achieve new performance:

 

The Tegra X1 is even compared against the Intel Core i7 exceeding its performance at less power:

 

NVidia is now adapting the Tegra X1 for automotive use where your car is going to become one of the world's most advanced computational devices.

 

Nvidia also unveiled the a new computer platform the "Drive CX" which can process 6.2 megapixels, running virtual machines, 4k displays and support every major os in the world and includes the new Drive Studio by Nvidia.

 

As see, the Drive CX is demo'd as powering the dashboard of the car dubbed the "Digital Cockpit"

The Drive CX is powering everything in the new infotainment system for automobiles. 

The map rendered has ambient occlusion, gpu rendered vectors, and uses 3D Geometry.

 

 

The GPU is also able to render materials and retexture parts of the digital elements.

The image below is using bamboo texturing in realtime.

 

My Favourite: Carbon Fibre

 

NVidia hopes to achieve "Ecosystem Unification".

 

To achieve environmental rendering, situation aware and software updates, NVidia dreams to have a car to have supercomputing capabilities using their new DrivePX.

The DrivePX can be run in parallel or standalone. Introduces computer vision and surround vision.

 

All inputs as NVidia imagines will feed into the DrivePX including cameras, sensors, etc...

 

Deep Neural Net Computer Vision is a new technology with the DrivePX to simulate computer learning dynamically.

So engineers no longer need to hard code certain features or surroundings.

 

GPUs are required for parallel processing. GPUs can now recognise more images than humans.

 

Neural Net diagram.

 

Nvidia's new technology can see many types of the same thing/class such as different signs and forms of pedestrians.

In Nvidia's example, it is shown that it can process a cyclist but cannot see the bicycle behind a police car, however once it recognizes the bicycle, it reclassifies as a cyclist. Such as differentiating between red and green, a car or an ambulance.

In rear view:

 

 

Nvidia's car technology is shown with the Audi car self-driving on the Formula 1 Track in Germany. There is no one inside the car.

 

Real-time simulation of Nvidia's parking garage.

 

Sucessfully parked

 

[MegaDave91]

All of these Nvidia Conference threads should be condensed into one.

[Imabigmac]

All of these Nvidia Conference threads should be condensed into one.

How many are there? I started one like 15 minutes ago. 

 

Of course my thread in "off topic" won't be nearly as popular. 

[TopWargamer]

For those of you who want to watch along... http://blogs.nvidia.com/blog/2015/01/04/live-nvidia-ces/

 

Seriously though, holy crap, the X1 can handle 4K at 60 hertz!!!

[kwongger]

JUST ANNOUNCED THE TEGRA X1.

 

OMG

[Imabigmac]

The car stuff is pretty cool. It is something that I would like to work on in the future. 

[ELSknutson]

its cool but Im wondering how stable its going to be

I wouldn't want my speedometer or entire car to shut down due to a chip failure

[The_Strict_Nein]

its cool but Im wondering how stable its going to be

I wouldn't want my speedometer or entire car to shut down due to a chip failure

Anything that goes into something like a car is tested and validated for years before it gets near a consumer.

Trust me, it'll be stable

[ELSknutson]

I'm glad i chose electrical engineering as my coarse of study

[thatguyjerry]

I mean can you really tell the definition difference on screens that small?

[ELSknutson]

imagine moding your gauges or center stack like you would in a racing game

[The_Strict_Nein]

I mean can you really tell the definition difference on screens that small?

More pixels = more space to put useful things at legible sizes.

You can fit more legible small things on a 1080p screen vs a 720p screen

[Imabigmac]

imagine moding your gauges or center stack like you would in a racing game

It would nice if we could get driving games with "advanced"  interior modes/views. 

[Hatsune Miku 「 」]

imagine moding your gauges or center stack like you would in a racing game

Modding is real as demo'd by Nvidia.

[Admire]

Thats what we need....more things for drivers to be distracted over.

 

 

Bitter old man leaves

[thatguyjerry]

More pixels = more space to put useful things at legible sizes.

You can fit more legible small things on a 1080p screen vs a 720p screen

I agree, but I'm just thinking about what this means for mobile platforms. I can't tell the difference between some high resolutions on my iphone, and if text got much smaller on these mobile screens I'm sure I'll have to start wearing reading glasses. Is there really a need to cram more pixels into mobile platforms?

 

I can definitely see the upside and utility for things like cars, increased legibility is definitely a huge benefit in areas like this. But the day I'll need or want 4k on a tablet, phone, or portable gaming device I'm not so sure about.

[GoodBytes]

Well the OBox (http://linustechtips.com/main/topic/284669-new-game-console-obox-yes-thats-the-name/), new gaming console, says that it can do 4K and uses Nvidia Tehra chip... I wonder if this is the actual chip used.

[The_Strict_Nein]

"GPUs can now recognise more images than humans"

Unless GPUs have also formed advanced artificial intelligence, this is utterly useless.

You need humans to give context to an image. No machine can outdo a human for context

[Imabigmac]

"GPUs can now recognise more images than humans"

Unless GPUs have also formed advanced artificial intelligence, this is utterly useless.

You need humans to give context to an image. No machine can outdo a human for context

I think they are truly trying to get at driver less car tech still. 

[patrickjp93]

So, Nvidia claims 100% better performance than an I7 for 16-wide math... Okay...now let's see 32-wide scores where Intel has native 32-wide hardware... More Nvidia propaganda...

[The_Strict_Nein]

I think they are truly trying to get at driver less car tech still.

The point I'm making is that even if a car could recognise an image a human couldn't (still trying to work out what that means) you'd have needed a programmer or an itllegence to say what to do with this information, which ultimately comes from a human.

[Dabombinable]

So my old Transformer Prime (Tegra 3) is low-end now

[Hatsune Miku 「 」]

The point I'm making is that even if a car could recognise an image a human couldn't (still trying to work out what that means) you'd have needed a programmer or an itllegence to say what to do with this information, which ultimately comes from a human.

Thats where gpu accelerated deep neural net comes in. It triggers a response to stimuli. In a example NVidia showed a car stopping over for a police car.

[Imabigmac]

The point I'm making is that even if a car could recognise an image a human couldn't (still trying to work out what that means) you'd have needed a programmer or an itllegence to say what to do with this information, which ultimately comes from a human.

But that is all part of the programming. They trying to develop a free thinking AI. If the cameras see a head poke out of a tree, yes it can't do anything. But if cameras are attached to robotic device the robot can be trained to go over and say high to the random person. We need humans to train computer AI but we also need humans to train humans.

[bry.yyyy]

i don't need this crap car this i need GTX 960 NVIDIA!!!