transcribe

GTC 2013: NVIDIA's Tegra Roadmap (6 of 11)

NVIDIA · 11m · transcribed 23d ago
More from NVIDIA Business
𝕏 Share ▶ YouTube 📥 PDF 🤖 .md

Transcript

0:07 okay Tegra we we felt that the world was changing that we were going to put computers and everything the idea that it used to be in servers then now in some PCS but in the future computers will disappear it'll disappear in the sense that it'll just be everywhere it'll be literally everywhere and so we wanted to start investing in processors that has the same visual Beauty and the same

0:37 interactive capability that we're known for but in computers that that are are literally everywhere literally everywhere in cars in stores in glasses in watches it doesn't matter where it is in phones and tablets anything that has a display in the future will have a visual computer inside anything with a display and whereas when we first came into the world in 1993 1993 there were each year about a

1:10 100 million CRT displays sold 100 million CRT displays sold now now this year 2 and A5 billion High defin display sold each year and by 2016 2017 2018 about that time frame it's going to double again the number of displays that needs to be driven by little tiny processors is growing exponentially we're surrounded by this place and so we

1:41 started this road map called Tegra our first GPU our first Tegra I we won't say anything about because it didn't turn out that well we were just learning we were just learning and tgra 2 was the world's first dual core mobile processor Tegra 3 was the world's first quad core and we invented this idea called the energy saving core the observation was GE CPUs were so abundant why don't we create four that are for

2:12 high performance but create one that is extremely low power called a 4+1 architecture just about everybody in the future are going to use this type of architecture we invented it so that you could have very low power when you're just using it to watch movies listening to music read a book but when you go play a game or access a website you want to power on those quad CPUs get the job done and go back to

2:41 sleep Tegra 3 was the world's first quad core Tegra 4 introduced two new ideas we integrated for the first time a softwar defined radio modem that is completely processed in a new type of architecture that is called a deep EX execution processor running modem protocol stack in extremely low power and we introduced a second idea called computational camera

3:12 using mathematics using the CPU and the GPU to work together to take sensor information from your camera from your image sensors and create using image processing surprising effects like for example you you don't have to take two shots to create HDR you create one instantaneous shot to create HDR for example it could track your face it could track an object while the object is still moving so if

3:44 you ever try to take a picture of a of a kid you're constantly refocusing now you don't have to do that you simply touch it tracks and now as the object moves it stays in focus and while it stays in focus is changing in real time its exposure necessary for HDR all of that is made possible by our computational camera so the question is what do you do next you've done so much what do you do

4:16 next well our next Generation GPU our next Generation mobile processor Tegra is called Logan this generation was called Wayne now it's called Tegra 4 the next version is called Logan Logan has something that we've been dying to bring to the world for so long it is almost the reason why we came Logan incorporates for the first

4:46 time our most advanced GPU it's the world's first mobile processor with Cuda you're not the only one who's been holding their breath I'll tell you Logan incorporates a Kepler GPU it is full K to five because it's the most modern

5:19 GPU out of the box is openg GL 4.3 and it does everything that a modern computer ought to do does everything a modern computer Compu ought to do we should see Logan this year and we should see Logan in production very easily early next year that's called Logan well let me show you one more because I have a feeling several of you

5:50 in the audience have heard of Logan let me show you one more the Next Generation Beyond Logan has a peculiar name the next generation is called Parker Parker brings three ideas to the market three ideas to mobile processing the first is I know you've been waiting so have I it is the first processor with Denver world's first high performance

6:21 64-bit arm processor coupled with our next Generation GPU Maxwell and in order to put all of that in the hands in your hands in a phone in a tablet we're using finfet 3D transistors that's Parker you know the observation that is amazing here is and I don't know if you guys noticed the scale earlier the GPU was increasing about two times every other year

6:53 basically still sustaining very naturally Mo's law because parallel Computing has the benefit of not having to push for just frequency we can have the benefit of pushing frequency and the number of transistors and because parallel Computing is still in its nent stage there's so many new ideas we can still incorporate architecturally to improve its performance there's still a lot of learning ahead of us but it's still two times every other year what's really remarkable here is in five years

7:24 time we're going to increase the performance of Tegra by a 100 times 100 times now mors law would suggest about eight well that's a perfect example of disruptive technology it happened back in the good old days when the PC industry came along the rate of innovation in the beginning was really quite staggering quite staggering and so you should expect that every generation every generation the integr processor brings something that is just enormously

7:55 surprising from the past now I just can't wait I can't wait any longer all this talk I can't wait any longer I'm like you I want to see arm and Cuda now I want to see Cuda on arm right now and so maybe next time and so so the guys went off and

8:25 they said you know Jensen you want to see Cuda on armed right now well we can we're going to build a super super low power GPU a new GPU that you guys have never seen and we're going to combine it with arm and we're going to create the world's most powerful arm computer ladies and gentlemen let me introduce you to Logan's girlfriend Kayla Kayla is a really interesting little

8:57 thing Kayla has a Tegra 3 because it has a PCI Express we added to a brand new very low power GPU and if not for all of the connectors we wanted to put on here Kayla would be a little tiny thing now the thing that's really amazing is this Logan's going to be the size of a dime this is going to fit into a chip the size of a dime require no fan

9:30 no heat sinks and will likely be higher performance than Kayla but let's just take a look at what Kayla can do shall we all right bass show me Kayla all right so this is Kayla Kayla is running realtime raay tracing and this realtime raid Tracer is it's

10:04 you could see the transparencies the reflection the refraction the cosics I mean it's really quite amazingly beautiful this is remember this is Kayla okay it just launched smoke these are all the same demos that we've been showing throughout the years on these massive g-forces and massive Teslas oh wow you're killing me open GL geometry shaders the point here is the the entire modern software stack that we know of the most advanced

10:34 Computing software stack we know of Cuda 5 openg GL with programmable G geometry shaders Linux physx physics processing all running at the same time on this little tiny computer pretty amazing hey guys thank you very much good job

© transcribe · For agents Built with care and craft by Gokul Rajaram