Ladies and gentlemen, please welcome your host for today's event, the Corporate Vice President and General Manager of the Graphics Business Unit at AMD, Matt Skynner.
Good morning. Good morning, and welcome. Welcome to those of you here in Sunnyvale, our special guests here, including our employees. Welcome to our AMD Austin site and our AMD Markham site, who are also watching live. Welcome to those on the web through the webcast. You're here today to see a presentation that talks about a new member of the FirePro graphics family. You'll hear today from David Cummings, the general manager for professional graphics business, from Raja Koduri, our vice president for visual computing. They'll not only talk about a new product, but talk about a solution. A solution that makes professional workflows better for engineering and for media and entertainment professionals. Before I get there, I'd like to talk a little bit about AMD. We're in the middle of a transformation.
A transformation from a company, in 2012, had 90% of our business focused on traditional PC markets. We've got to continue to win in those businesses and maintain our traditional PC strongholds. We've got to grow in desktop and graphics channels. We've also got to transform. Transform to a more diversified, balanced business, such that by 2016, 50% of our business will be in those traditional markets and 50% in new growth markets. Markets like semi-custom, like dense server, like embedded, and ultra-low-power client, and also in Professional graphics. Professional graphics is a major growth item for AMD. How are we doing there? In 2013, from a client point of view, we grew in desktop channels. We launched new APUs like Kabini and Temash, and our latest one, Kaveri, in Q4. A great new addition to the APU family. It's an awesome product.
In embedded and semi-custom, we were proud to launch with Sony PlayStation 4 and with Microsoft Xbox One. In dense server, we're leveraging differentiated SeaMicro server technology. We are the only company to design both x86 and Arm-based 64-bit server solutions. Now we're powering Verizon's public cloud. In graphics, we've had a great reception for our 9 and our 7 series products. We launched Mantle, a new API that allows game developers to program games more efficiently, and that gets a better gaming experience for our end users. It gets better performance. You're seeing this come to life with Battlefield 4 and Thief being launched. We had a great year in Professional graphics. Now, AMD's been in Professional graphics for over 10 years. Began when the former ATI purchased FireGL.
We serve multiple markets, professional solutions for tablets, for notebooks, for traditional workstations, and for server solutions. We design high-quality, reliable product that's focused on application performance. We're shipping with all the top OEMs for workstations, Apple, Dell, HP. We work with, optimize for, and are certified by the leading ISV applications. Our strategy is clear, simple, leading in content creation and analytics. We're going to win in client by leveraging our technologies, like our GCN architecture. We're going to embrace the cloud trend with our professional graphics server solutions. We're going to drive OpenCL to win in compute. None of that's successful unless we're working closely with content ISV developers to make sure the applications work best on our hardware. We're going to leverage these technologies like Eyefinity.
In Eyefinity, we're the leader in multi-display technology, and we can leverage that to win in display walls and in finance markets. We can leverage technology like our GCN architecture for compute to win with research, media and entertainment, and engineering. Got some good momentum in FirePro. I talked earlier it was a good year. It was the highest grossing revenue for FirePro in our history. Instead of me talking about it, let's hear from some of our partners, some of our ISVs and some of our end customers.
Hi, it's Dave Helmle with Adobe, and we're going to share with you today some new features in Premiere Pro CC. We're actually working on FirePro technology today. What type of advantages you'll see, let's take a look at something like a multicam. As I'm working with my multicam, a lot of the windows that are having to update here are actually being controlled by OpenCL. You need that graphics pipeline to actually play back, in this case, all five layers of video.
We can actually play in real-time with the vortex field magnitude. If I reverse the magnitude, it will going one way, or it could going to the other way, and we can actually tweak all of the parameter, as you can see in this case, and see the result very quickly.
Our company is a very small company. Even though we're the largest hang gliding manufacturer, we have very limited resources. We don't have a team of engineers. We don't have teams of flight pilots. We have a couple of individuals who have a lot of different responsibilities. The requirements to make, to design, engineer, and flight test an aircraft like this require that an individual be able to do a lot of different things. From one generation of this graphics display workstation FirePro card to the next enables me to, for example, model and rotate this assembly of a thousand parts with RealView, the ability to see all of the shadows, the transparencies, all of the materials applied to the hardware artifact.
We are here at the Helmholtzzentrum GSI in Germany, right in front of the data center where the JUQUEEN computer has been constructed. We decided to try to build the highest efficient computer and also the fastest computer available for the budget. Today, this is possible by using graphics cards, general-purpose graphics cards, which are integrated in server.
You saw those customers and partners talk about the importance of OpenCL, the importance of 4K. Technologies are changing. The demands from professional users are changing. We're entering a new era of visual supercomputing for workstation professionals. With that, what is driving that new era? Well, OpenCL is one of the things that's driving that new era. OpenCL allows developers to harness the power of the GPU to do more. ISVs like Adobe, Autodesk, Dassault, Blackmagic Design, Houdini are all switching to OpenCL. It's not only software vendors. There's hardware vendors that lead from an OpenCL point of view. One of those is Apple. We're proud to be a part of the Mac Pro. I love this machine. I love this machine because it's cool. I love this machine because it packs more than seven teraflops of performance in here.
I love this machine because it has not only one but two FirePro graphics cards in it. It's beautiful. Another technology is 4K. 4K Ultra HD displays. You saw the gentleman from Wills Wing talk about seeing those thousand parts in the hang glider, seeing the detail, seeing how those parts work. He needs more detail. Four times more pixels means you need more graphics horsepower. 4K is a huge driver. With technology changing, with OpenCL, with 4K, with demands from workstation users, we need a new class of GPU. We need it for this new era. Now I am pleased to announce the launch of the AMD FirePro W9100. Now I'd like to ask Mr. David Cummings, our general manager, I'll take him up on stage, and I'll pass the graphics card to him.
Thanks very much, Matt, and good morning, everyone. I am really excited to be here today to talk to you about the AMD FirePro W9100. It is AMD's new flagship professional graphics solution. It is powered by OpenCL, and it is really designed and optimized for real-time 4K. Let us kick off by looking at some of the capabilities of this new product. Matt talked about GPU compute. This GPU is the first GPU ever to break the two teraflops barrier for double-precision compute. In fact, we have over 2.67 teraflops of double-precision compute. The first time that has ever happened. Over five teraflops of single-precision compute, again, the fastest professional graphics card on the market in that domain. We have leading display technology. We can support up to six 4K displays on a single card.
Last but not least, we have an amazing 16 gigabytes of video memory on the FirePro W9100, the industry first 16-gigabyte graphics card. 16 gigabytes of fast GDDR5 that delivers over 320 gigabytes per second of memory bandwidth. Again, the fastest on the market. Let us take a competitive snapshot. Here, I am comparing the FirePro W9100 to the two nearest competitive cards. Whether you look at the K6000 or the K5000, it is clear that the FirePro W9100 has faster GPU compute power, has more memory capacity, and has greater display flexibility and display connectivity. What does that mean for the end user? Matt talked about a couple of the trends, some of the needs of professional users, how those are evolving. Those are the things that we look at when we design new products like the 9100. One of those trends was GPU compute.
Traditionally, GPU compute has been the domain of server-side HPC applications. We are seeing a big move, a big shift for workstation applications, whether it is in engineering or media workflows, to adopt GPU compute to get better results, better productivity for the user. The FirePro W9100 really excels in that area. Looking at some performance. On the left here, this is a single GPU comparison. This is using LuxMark. LuxMark is a popular OpenCL benchmark that is based on a ray tracing application. In a single GPU comparison, you can see the W9100 is clearly ahead of the nearest competitor. The story is even better when you add multiple GPUs. Many of today's professional applications can take all the compute power, all the memory capacity that you can throw at them.
The W9100, as you go from dual to quad GPUs in the system, the performance scales beautifully, as you can see there. Again, that is something that is really critical to the pro users going forward. Switching to another example. This is a SiSoftware Sandra benchmark, which stresses the double-precision compute capability of the GPU. Here we are seeing W9100 about 1.5x the nearest competitor. This is really critical for many of the applications in the scientific space, in the engineering simulation area, that will use the double-precision compute. We have a great solution for compute. Another trend that Matt touched on is the move to 4K. Whether it is for media professionals or engineering professionals, this is becoming really critical, a real key inflection point. Those professionals want to work with the content in native resolution in 4K, but in real-time.
They have huge models and assemblies they're working with. They want to interact with the entire assembly, be able to visualize it in high precision. What we're going to do now is to talk about this in a bit more detail, to dive into those workflows and the benefits the product can offer. I want to invite on stage someone that's really involved in both worlds here, and it's Raja Koduri, AMD Vice President of Visual Computing.
Thank you, David. I'm really happy and excited to be part of this launch. This product is really near and dear to my heart. I spent a lot of time on the engineering side here at AMD and have worked with various engineering teams designing products, designing complex GPUs, and complex computer systems. I also spend a lot of my time working with creative professionals on the media side, in the gaming industry and in the film industry, who create content. We have an incredible engineering team here at AMD that's involved in creating this product, W9100, and they are very creative as well. This slide delineates creative professionals and engineering professionals, but these days they are tied together more closely than ever. Matt Skynner talked about the product here, Mac Pro. He called it an engineering marvel, and he also ended saying it's beautiful.
To design a product to pack seven teraflops in a box like that requires incredible engineering design, simulation, and also incredible amount of art goes into a product like that. Today, we live in a world where there are more products being designed than ever in the history, right? Be it computers, cars, wearables, personal electronics, so many products are being designed today. We also live in a world where there is incredible amount of content being created. Every second, there is another YouTube video being uploaded, created by creators worldwide. Incredible world we live in where there's a lot of creation, a lot of engineering that's happening. Never before we have seen such creation happening in the world. One thing working with the creators, what they ask, they want more. They want more of everything.
They don't want the tools around them, the world around them, to limit their imagination. They don't want to be curtailed by memory capacity, they don't want to be curtailed by how much memory bandwidth they have. They don't want to be curtailed by compute. Over the years, we have taken it as a challenge, as GPU creators, to give them more. Let's take a quick look at how far we have come in the last 10 years. 10 years ago, we launched this incredible product called FireGL X3 in 2004, 28.8 gigabytes per second. In 10 years, over 10x increase in bandwidth. On the memory size, it was 256 megabytes. We have 16 gigabytes, 64 times increase in memory size. On the compute, on a single card from 21 gigaflops to over five teraflops, over 250 times increase.
You have heard David and Matt talk about being able to put 4 GPUs or ASICs, excited about 4 GPUs. That's over 1,000 times increase in compute in the last 10 years. This is all to enable the creators to not have any limits. Hardware specs aren't all that's required to enable this incredible creative workflows. We need to work with software vendors. We need to enable the end-to-end solutions for our users to get benefit of this incredible hardware. Let me talk through a few workflows that we enable with this incredible product of the FirePro W9100. First, in the engineering workflow, there are 3 distinct phases: design, simulation, and visualization. In the design phase, you're designing your product with very fine detail, like a car, like a computer, like a phone, and you want very accurate representation down to sub-millimeter level.
You have very fine detail geometry models. You need very high geometry complexity to be handled on your GPU. The W9100, with nearly 4 billion primitives per second, can render these hugely complex models really, really fast. You also need very accurate representation. Memory footprint is very, very important. 16 gigabytes of GDDR5 enables you to load very high dense models into memory. You also want to simulate your product's interactions with other products or the environment, like 2 cars colliding with each other or the thermal modeling inside a computer. You need incredible computational complexity running over OpenCL. Double-precision floating point is really, really important for ensuring good accuracy of your results. You also want to visualize all of this stuff, so you combine your graphics compute and the display capabilities to achieve very photorealistic rendering of your products.
You want multiple displays to have the flexibility to view the different aspects of your product. There are many software solutions that enable this workflow. We work with all of these folks to provide great solution to the users. A few cases here that I'll walk through that W9100 enables. SOLIDWORKS, leader in the modeling space. We have more accurate designs that can be visualized in this tool with the GPU-accelerated order-independent transparency. Our ISV engineering teams, our driver teams work really close with SOLIDWORKS to enable this. It uses the OpenGL 4.2 shader atomic counter extensions to perform this great rendering here. On the performance side, on W9100, with these optimizations, in loading a very, very complex model, we get over 180 frames per second, leaving enough headroom to perform other things like simulation and photorealistic visualization.
It's incredibly faster than the fastest competitive card that's available today. On the simulation front, we work with leaders, Dassault Systèmes software SIMULIA. With W9100 card, you can model and simulate on the same system, which is really, really great. The double-precision floating point horsepower enables much accurate simulation. We are working with Dassault Systèmes to get fully optimized OpenCL implementation, and I'm expecting to see great results from them once they're done with their optimizations on the double-precision floating point capability of W9100. Last step, visualization. Said modeling, simulation, visualization. Autodesk VRED is a leading software in the engineering visualization world. We work with them to optimize their visualization on the GCN architecture. The GCN architecture allows for much heavier geometry to be loaded at much faster rate. They also make extensive use of our shading pipeline, and it enables smooth rotations.
The model you're looking at here on the screen, without the horsepower of W9100 and the 16 GB of memory, you cannot interact with it real-time. All the previous generations of professional cards, both from AMD and the competition, cannot render this complexity seen in real-time. This is incredible. Let's look at the media and entertainment workflows. The whole process starts with capturing the imagery using high-resolution sensors and some great camera technology that's around today. We have a revolution happening in the camera technology. 4K video is everywhere. You need to bring the data in really fast into the video memory, and that's where our DirectGMA technology comes into play. You can bring the data directly with no extra copies into video memory with the DirectGMA. You edit the video to remove the unwanted material, then you do the finishing step.
Finishing is really important in the professional video pipeline. That's what makes the difference between a consumer home video and a professional video. You composite with software like After Effects, then you publish. If you have some CG as part of your creative workflow, you use modeling and animation tools like Maya, 3D Studio Max, and Mari. We work with this entire workflow, software vendors in each step of the way to accelerate their flow using OpenGL, OpenCL, DirectX On W9100. Let's take a deeper look at the finishing step. Blackmagic DaVinci Resolve is the world's most advanced OpenCL-accelerated color grading system. The challenge for the AMD FirePro is to maintain real-time 4K video while doing various color correction effects. The Resolve uses a concept called nodes. Each node does some image processing.
They would like to have many nodes either arranged sequentially or parallelly. Nodes are like layers, but much more powerful. You have each node have several effects like blurs and some other image processing operations. The more nodes you can do, the more productive the color grading artist is. What can 9100 do? We have the blur effect as part of the nodes. You can do four nodes at 24 Hz with 4K interactive editing with the W9100. That's over 3 times faster than the nearest competition. When you do two cards, it scales up to eight nodes. You get to four cards, it doubles again. That's incredible. Let's hear from Blackmagic, their excitement around W9100 and working with AMD FirePro.
Blackmagic Design is a company that creates hardware and software for video production, and one of the important tools that we've created is our DaVinci Resolve color grading software. DaVinci Resolve is a professional color grading software which is used for image processing of professional video. We have folks that use it to do anywhere from simple color grading all the way to Hollywood professional movies that are being done for color grading, 3D work, and finishing of these productions. This is where things like working with AMD and OpenCL support for DaVinci Resolve is incredibly important for us because we want to be able to process those images and the effects as quickly as possible.
As we move into the new age of things like 4K Ultra HD, which is four times the basic data rate that we see with HD today, that's tremendous amounts of data that needs to be processed quickly. When we look at things like using the AMD FirePro W9100, that's a terrific card for us to be working with because not only does it have a tremendous amount of GPU muscle to be able to help with all this processing power, but the actual video RAM that's provided on there helps quite a bit as well because we're dealing with all these nodes, which are going to be the different color grades, blurs, and effects. Just the tremendous amount of data of working with 4K Ultra HD.
Another really important thing for us too with AMD is the fact that they are on two different platforms on both the Mac and PC. Our users want to be able to have choices out there. As a two-platform solution, OpenCL is tremendous for us. One of the really great things about our own software is we've been able to implement the fact that we're able to take advantage of all the cores and as much video RAM as possible in there. We have some folks that will buy a single GPU solution, put one of the FirePro cards in there, and get enough real-time for them to be satisfied with the work they're doing. We have other folks that may be doing a larger production, and you start talking about working in 4K Ultra HD or working with a 3D film.
There's more needs to have more processing power. This is where you see people putting in multiple cards into either multiple slots or into other host bus and being able to get a tremendous amount of advantage over putting these solutions together. When we look at being able to tie in multiple cards, and especially new cards with the even more OpenCL power, with the more video RAM available, really it becomes a big help for us because of these new data rates and the workflows that people are now seeing out there. Being able to take advantage of all the OpenCL available, all that video RAM across multiple cards is tremendous for us.
That's incredible. As hardware designers, it makes us really, really happy when we see software folks like Blackmagic take full advantage of all the horsepower we pack in. You see, they scale not just with one GPU and all the shading horsepower and the floating point horsepower in it, two GPUs, three GPUs, four GPUs. Even more incredible is the leverage they are getting out of the GCN architecture. It's a huge thing for software vendors to take advantage of all the optimizations they do on a platform like Mac Pro with seven teraflops and two GPUs and leverage that on other platforms like Windows. Vice versa, all the work they're doing with W9100 with GCN architecture is transferable to the Mac platform. We have many tools that span many platforms. The creator workflows, they switch between various platforms on daily basis. This is incredible.
On the modeling and animation front, Maya needs no introduction. For a couple of decades, it has been the leading software. We have integrated OpenCL-accelerated physics solvers into Maya through the Bullet open-source engine. The challenge for AMD FirePro is to allow artists to quickly create, view, and edit all their simulations. Animators want to spend less time waiting and more time creating and making their dynamic simulations better. With the combination of the GCN architecture and the OpenCL acceleration, we give everything they want. Maya also has a very advanced GPU caching system called Alembic. With the combination of Maya's GPU memory caching system and the 16 gigabytes of GDDR5, Maya will enable incredible amount of complexity to be visualized in real-time. Never before possible amounts of geometry, amounts of textures can be loaded up and visualized. It's really, really exciting.
The other incredible tool in modeling and animation is the Foundry Mari. The OpenCL support in this tool will enable the artist to spend more time painting and less time waiting. This tool allows real-time preview with full displacement mapping, ambient occlusion mapping, specular highlights, real-time shadows, color correct rendering, all to happen in real-time because the power of GCN architecture and the OpenCL acceleration. You can load 32K by 32K textures, multiple layers of them, into the 16 gig GDDR5 memory. This is really incredible what the artist can do in the Foundry tool with W9100. Adobe needs no introduction. They are the leading tool for professional vendors. They have a full tool suite from end to end. Just looking at Premiere Pro, we have enabled hardware rendering for 4K. You get 12 times speed up on hardware rendering.
We also work with Adobe very closely on variety of their software suite, After Effects, SpeedGrade, Media Encoder, their Mercury Playback Engine. Let's hear from Adobe, their excitement about W9100.
Adobe and AMD have a great history of working together in an engineering sense and a marketing sense, and we're really, really excited by the new W9100 card. Our engineers have been working together. We've had it in the lab. We're seeing some pretty astonishing performance results from it. That's not surprising when you look at what this thing is. It's incredible. It has 16 gigs of video RAM. That's just insane. I can't believe that a card even has that. It can drive up to six 4K displays simultaneously. Like I say, you combine that with our tools, and we're just seeing this thing scream. One of the things we've worked on recently is the ability to support multiple GPUs when you're doing your output render, when you're doing your export through Media Encoder. People need to turn around projects faster than ever before.
We know that. They're working with bigger frame sizes, which takes more processing. If you're in a position where you need to create lots and lots and lots of exports of sequences with real-time effects and so on in them, being able to harness the power of multiple GPUs is valuable. You can, of course, have multiple W9100s. If you really need all that power, you could have two W9100 in a machine. That will give you literally about twice the speed of encoding that you would be able to expect with a single, and it gives you 32 GB of addressable video RAM, which I don't think anyone ever dreamed was possible.
I think one of the most incredible things about the latest generation of AMD GPUs is that they actually let customers drive a large number of 4K displays in real-time, and that is something that's becoming increasingly important in this industry. I said at the beginning, it's kind of astonishing how many projects that we're working with our customers on right now are going 4K. We're even hearing people go 5K, 6K, and beyond, and the Japanese market's asking for 8K. Anything that can drive these enormous frame sizes and can keep the quality is very important. Adobe and AMD have worked very closely for a very long time.
We support the GPUs, we support the APUs, from the Radeons right the way through to the FirePros, and it's really just all about making sure we work closely together for the best support possible, for the best performance possible, so that our mutual customers who are out there being creative, who want the fastest turnaround and the best performance, know that Adobe plus AMD is just a great combination.
Insane, astonishing, incredible. These are the kind of words we love to hear from ISVs and users when we introduce new graphics cards. You see Adobe's excitement with W9100 and all we are doing in the graphics space. Now, enabling great workflow is one part. Enabling great systems to be built is another part. To tell you more about that, I welcome back David Cummings, the General Manager of our Workstation business unit.
Thanks a lot, Raja. That was great. I want to talk to you in this final section of today's presentation of the next generation of workstations. At AMD, we believe that these new requirements, the maximum graphics capability of products like the FirePro W9100, really need a new generation of workstation to be able to support scalability through single up to multiple GPUs in a system, up to four W9100s in a system, and to deliver the uncompromising 4K in real-time with the great performance and the display flexibility. With that in mind, I'm pleased to announce the Ultra Workstation. AMD's introducing this new program, which really brings the visual supercomputer to the user's desk. What we've done is lay out a set of standard system requirements, and we're working with key partners worldwide to bring this to market.
We're working with leading workstation system integrators across the globe and targeting high-end professional workflows. We have a number of systems that can fit into different workflows. For example, a single GPU solution for digital content creation or some of the engineering workflows, a dual GPU solution for digital video editing or post-production, all the way up to a four GPU system, typically for 4K color correction, as we saw from Blackmagic earlier. I'm really excited to announce our launch partners. We're working with leaders across the world, Supermicro, BOXX Technologies, TAROX, Wortmann, Silverdraft, Mouse Computer, and Versatile. These are the leaders in their regions. We're working with them to bring this program to market in conjunction with the launch of the FirePro W9100. What we'd like to do now actually is to show a video.
We have one of the partners, Silverdraft, who will talk about their experience, what they like about working with AMD.
Right now, we are actually inside MobileViz, which is a, I would say, a supercomputer on wheels. It is based with our, what we would call our Devil's Playground, which is our supercomputer technology inside a mobile platform, and this rig or this environment can be used to go on set for production and kind of do whatever the artist can possibly dream of.
With FirePro graphics, we also feel that the cards are just exceptional when they run processes, particularly for entertainment. It's difficult for us to even argue for another solution because we just feel so strongly about how effective the AMD technology is.
Part of this whole devil and demon strategy was that we were looking for best-of-breed components and best-of-breed companies that would get what we were doing, that were really aligned with what we were doing. When we went up into the CPU and GPU landscape, there are not a lot of players in that landscape, but there was one absolutely right player for us in that landscape, which was AMD. AMD has a lot more power, a lot more bang for the buck, and maybe even more important than those, are much more aligned creatively with the way we're thinking as an industry.
Thank you, David, and thank you, Raja, for your excellent presentations. You heard David and Raja and our partners talk about the importance of 4K, talk about OpenCL, talk about the new demands from professional graphics users, and they need a new solution that delivers, and the AMD FirePro W9100 delivers. It delivers with over two teraflops of double-precision. It delivers with an industry-first 16-gigabyte frame buffer, so those users can do more. It delivers with real-time 4K. It delivers with the ability to have six 4K monitors connected to a single graphics card. AMD is still the only company that offers that. With those six monitors, they can see more. It delivers in a solution, a platform solution, with the Ultra Workstation that David announced. We're very happy with that to work with our system integrator partners around the world.
Now I'd like to thank you all for being here today on the web and here at Sunnyvale, and I'd like to say stay tuned on amd.com for more on the AMD FirePro W-