Transcript
0:00 we've announced our Kepler gpus and we have 46 minutes to go and so there there's a there are a few jokes that I I heard recently I'd like to share them with you actually no we have something new the engineers have been working over the last 5 years on something that no one has ever seen before this is something really really special we had this the way that we work at Nvidia is we started with questions like wouldn't
0:31 it be great if wouldn't it be great if what if we did that wouldn't it be amazing if we start out just dreaming and at some point one of those dreams gain traction and we start working on it because we think it's going to be incredibly incredibly important today I want to share one of those ideas with you there's three Technologies I'm going to announce three more Technologies in the next few few minutes
1:01 first I'm going to announce the largest GPU and the most powerful GPU we have ever built much much much much more powerful than Kepler two I'm going to announce a GPU that is so small that we could put it inside an iPad and three I want to announce a GPU that we can all simultaneously
1:33 Share three new technologies today we're going to take the GPU into the cloud this is an Endeavor we've been working on for 5 years Engineers from all over our company working in small pieces and finally we've been able to put it all together we're going to put the GPU into the cloud now what does that mean well Kepler turns turn out has even more secrets than we thought Kepler is the
2:04 world's first GPU designed for the cloud kler is the world's first GPU we can deploy into cloud data centers all over the world there are three capabilities of Kepler that's important to make this possible the first is virtualized GPU a CPU is virtualized but a GPU is dedicated one application one GPU one user one GPU for the first time we have virtualized the GPU
2:36 second no longer does this GPU need to directly connect to a display it doesn't have to connect to DVI or HDMI it doesn't have to connect to lvds none of those physical connections to a display is necessary starting now Kepler could render and stream instantaneously right out of the chip to a remote location you choose and third Kepler with this Energy Efficiency
3:07 allows us to deploy Kepler in a massive scale so that we could literally Drive entire data centers with it we could put Kepler finally in the cloud at the core of the technology is this invention virtualizing the GPU whereas the CPU has an mmu virtual addresses come in goes through the mmu looks it up gives you the page table entry tells you what the virtual to physical address translation is it does it very very quickly every
3:38 single clock cycle in the case of Kepler we do the same for graphics commands for the very first time whereas we had many physical input command buffers every one of those command buffers are now virtualized every single one of those Graphics commands go through an mmu lookup and does a virtual physical address translation as a result of that we can now tell we can now discern we can assertain which virtual
4:10 machine were to send us a graphics command we would render the command from that particular virtual machine into the frame buffer of that virtual machine and at the end of it we would stream that frame buffer to that virtual machine everything is all done in Virtual space however one GPU can now be shared with countless users only can you only when you do this
4:43 can you integrate the GPU seamlessly into virtual machines and these virtual environments that data centers have become now the question is what's the application it's always great to build technology but what's the application well one application is an application we are desperately looking for in our own company there is a movement around the world for new employees to bring whatever device they want no longer can
5:14 we assign a corporate computer to them just as we no longer have company cars we now no longer have company computers employees could bring their own computer to work and yet it departments have to integrate these devices into the work environment in a secure and a logical way provide them the service and also most importantly the access to the company applications which tends to be Windows which tends to be Microsoft
5:44 Exchange these Company applications needs to be able to work simultaneously across all of these various devices whether it's iOS or Mac OS or Android or Chrome or Windows we now need to create a technology that virtual izes the Computing environment so that irrespective of your Computing device we could provide you with access to the corporate Enterprise applications and datas that you need while the architecture we all hope
6:16 for is the ability to put that computer in in this case a private Cloud to put that computer in the server room and to provide essentially streaming applications to your various Computing device if you were a task worker like a call center employee that technology is readily available today Citrix was the enabler of what we call today the virtual desktop infrastructure vdi
6:49 they've been the pioneers of this field started almost two decades ago and it's taken until now to really take off and the reason for that is because of the desire ire for security and more importantly the desire to support Enterprises with bring your own devices task workers were solved and now we need to solve the knowledge worker the power worker and the digital designers the people who ultimately create the digital Assets in our company
7:19 now their computers happen to be interactive Rich powerful computers so the question is how do we put these computers that require so much performance in the server room and make it available on every device well this is what an architecture looks like for Citrix this is the Revolutionary product called Zen desktop Citrix is doing fabulously I'm so delighted to see them do so great their sales the momentum around vdi propelled by this idea called BYOD
7:51 on the left the gray box is basically a hypervisor it virtualizes your server environment each one of the blue is another virtual machine and this is a virtual machine which is your PC and inside it is the operating system of your choice and inside that an application of your choice and inside that is a virtualized desktop called Zen desktop this is today's PC as an application however the big problem of today is that the graphics because it
8:22 has to be in a virtual environment is only done in software literally every one of these Zen desktops every one of these Citrix receivers is running a software PC with software GPU and every time you change your scene change your application interact with the application it renders it in software consuming all the CPU resource that's in the server room generate the pixels and streams it to you we thought wouldn't it
8:53 be great if we could virtualize the GPU if we could virtualize the GPU add a piece of technology called Kepler and secondarily develop a new GPU aware hypervisor software integrated with Citrix as a result these virtual machines seese a GPU except it's virtualized but it sees it as if it's their own every single virtual machine
9:25 now gets to have a GPU to themselves just as every virtual machine today has a CPU to themselves they now have a GPU to themselves what a fantastic capability if we can make it possible and with that we hope that partnering with Citrix we can accelerate the adoption of vdi across the entire Enterprise it no longer has to be limited to just task workers and knowledge workers it could actually move on and it and make it
9:55 possible to support power users designers literally literally anyone who has a workstation a PC anyone with a GPU wouldn't that be a wonderful dream