On AMD’s Do or Die 2016
JUNE 23, 2016 • This is a critical year for Advanced Micro Devices, Inc. (AMD) Not only has the technology firm just released its 14nm Polaris GPUs, but it will also be introducing its much more competitive Zen Summit Ridge processors later this year including eight-core FX enthusiast CPUs. It is not easy competing directly with the likes of Nvidia Corp. and Intel Corp. but this year AMD is making some major changes to its product strategy.
DFC Intelligence has pegged the graphics card space as a major growth area over the next two years. This growth is driven in large part by the emergence of a game consumer that is increasingly spending on performance-based hardware including custom PCs and high-end accessories. Virtual reality is also expected to significantly increase the need for high-end PC systems. Arguably the most important part of a high-end entertainment system is the graphics processing power. In this space there are really only two major competitors, Nvidia and AMD. Nvidia has been the clear leader for the past several years, but after struggling recently, AMD is coming on strong and its stock price has soared in 2016 with anticipation of its new chipsets.
In the graphics space, AMD is launching new chips based on its Polaris architecture, which utilizes 14nm manufacturing technology. This is a major improvement over previous 28nm technology and not only increases performance but is allowing for a lower price point. On June 29, 2016, AMD is launching the RX 480 at a price of $199. This is significantly below the price of flagship cards from Nvidia, which has just launched cards based on its competing 16nm Pascal architecture at prices from $379 on up. AMD is taking a very different approach than Nvidia. The RX 480 is not as powerful as the competing Nvidia cards and AMD is clearly targeting a more mainstream consumer versus a high performance user. The bet is that the new Polaris chip offers a major improvement, and a price point of $200, that will be attractive to a large audience that may be priced out by Nvidia’s cards.
The other major initiative for AMD revolves around the Zen CPU architecture. Chips based on Zen could come as early as the end of this year. Zen is performance focused and clearly targeting a core gamer community. Already the Zen family is generating excitement in this segment, which potentially marks a return to form for AMD whose reputation has been lukewarm in recent years.
To better understand AMD’s aggressive turnaround strategy, DFC went to industry veteran and corporate vice president Roy Taylor to fill in the details.
DFC: Please give us the up-to-the-moment overview of where AMD is today? What has changed in recent years in your product strategy, execution and technology goals… and why?
Roy: If you look at AMD today you will see that we are focused on three areas. First a return to high-end compute performance with our upcoming Zen-based CPU families. AMD fans, including myself, cannot wait to get our hands on new CPUs and motherboards. When fans are excited so are we. Secondly, our Radeon Technology Group is laser focused on taking back the graphics lead. With our experience in low overhead APIs like DirectX 12 and Vulkan – remember we invented Mantle – and our lead in supporting Async compute we are seeing the fruits of our work in a plethora of applications, as well as in VR. Thirdly, AMD is leveraging our IP strength in new and innovative ways, for example as with the deal with THATIC. It is an exciting time to be working at AMD under Lisa Su’s leadership.
DFC: Your 14nm rollout is taking a different tack. Instead of leading with the traditional high-end GPU, AMD is introducing the Polaris 10 for desktops, which is aimed at mainstream consumers. Last April, you were quoted as saying this was a very intentional strategy to help boost the total addressable market for desktop and VR applications. How big is the opportunity for addressing the mainstream market first.
Roy: First off please note that the incredible performance/watt for Polaris 10 means that we are also planning a presence for it in notebooks too. Yes you are correct that the plan is quite deliberate. Our goal with the Polaris family of GPUs is to take market share. A large portion of the desktop GPU market is between $100-$300 and that’s Polaris’ huge market opportunity. Whilst everyone enjoys seeing breakthroughs at the high end, including ourselves, it is vital to make sure that we have solutions for the bulk of the market.
But that is not all. For VR to take off it is vital that we find a way to run VR at lower price points, expanding the total addressable market (TAM) for VR content makers, publishers and consumers alike. So deliberately targeting this category makes AMD a darling of the VR community.
DFC: Tell us more about Polaris 10, your RX 480 series, what is it capable of compared to the previous Radeon generation?
Roy: Jumping from 28nm down to 14nm FinFET gives some very obvious advantages: for example, the capability for less power and more performance. We are still committed to enabling Async compute through the range so users can expect to see sterling performance in DX12, Vulkan and VR on our products. It is important to remember that Async time warp is a critical feature in Oculus VR. So we expect to see great gains through our support of it.
DFC: How will you frame/explain Polaris 10 to consumers?
Roy: By moving to why over what. For the longest time the tech industry has been focused on what marketing. That is to say that the emphasis has been on what is in the technology and what it does. Those are important but we think the better question is why. Why did we make this and why should you care? Today’s AMD places the emphasis on the delivered experience – note our commitment to content and our collaborations with Microsoft, Square Enix, EA, and others. Polaris will be explained in the context of how it makes you feel and why you should care about that.
DFC: Will you be relying on your manufacturing partners to take the marketing lead role on Polaris or will you be aggressive, as well?
Roy: We are intent to share with the world how excited we are about our great new products. So expect us to be forthright about why we believe we can deliver the experience consumers want. This has been our focus in VR, for example. In the final analysis we cannot be successful if we do not get our message across. We plan to clearly do that. With respect to our partners they have a leading role in supporting this and we look forward to working with them on this mission.
DFC: Similarly, tell us for Polaris 11 for the notebook market. What advantages does Polaris 11 give notebook system makers?
Roy: Polaris 11 offers quite astonishing performance for the power consumption. If you wanted to do some serious gaming on a notebook – and I’m not talking about browser games, but actual AAA PC games – you needed something very bulky, thick. Polaris is changing that. Gamers who want to play the latest Doom game in all of its Vulkan glory whilst travelling will be able to do so for hours, running on battery power. Polaris 11 will deliver Async compute, DirectX12 and Vulkan in insanely great new platforms.
DFC: Next year, your high-performance Vega 10 GPU is expected to debut. How will Vega 10 compare to Polaris 10 – how much of a trendsetter will it be?
Roy: Vega 10 is going to surprise a lot of people. We know what it needs to do and we are preparing for it do so. I cannot wait for us to share it with the world.
DFC: How do you plan to establish Vega 10 as the high-performance GPU to beat?
Roy: We will be focused on HDR, DirectX 12, Vulkan and VR experiences. The why of Vega 10 is going to be key to its success.
DFC: Although it was not adopted as far as hoped, what did Mantle help you with in getting more out of DirectX 12 and Vulkan on your hardware?
Roy: Mantle became the basis for Vulkan, just as we promised we would, we made it available to the community – in this case Khronos. Vulkan is available in DOTA 2, The Talos Principle and Doom so the fruits of our labor are clear and we could not be more pleased. Our experience and knowledge base in low level API’s enabled us to help support Microsoft in the success of DirectX 12. This is a further credit to the team that invented the first low overhead API.
DFC: How is AMD getting the most out of its drivers where DirectX 12 and Vulkan are concerned?
Roy: First of all it is important to note the importance of having Andrej Zdravkovic run Radeon Technology Group’s software team. Andrej sits with me on the team reporting to Raja Koduri. I have come to admire Andrej’s leadership and skill in making our drivers world class. Under Andrej’s guidance the graphics software team has become maniacally focused on delivering drivers that surprise and delight our customers and users everywhere. Terry Makedon, our director for software has also delivered an outstanding new UI that represents how we feel about delivering a great experience.
We are getting the most out of our drivers by making sure that they not only deliver fantastic performance and reliability but are also packed with helpful features. Expect more from this awesome team. I am proud of what they have delivered yet this is just the beginning.
DFC: In years past you have said a major bottleneck for developers was the level of support provided them and the quality of tools presented them. Has this changed? What is AMD doing in this regard, plus what developments have you seen when it comes to polished and intriguing new tools?
Roy: My personal passion is tools. It was my passion in 2008 when I did my last interview with you and it remains my passion. The reason is simple: give people the tools to make great content they’ll do the job to do so. AMD’s CodeXL tool was recently updated and is good and getting better.
But the real big breakthrough for us is www.gpuopen.com where we have taken the step of making a comprehensive set of tools available on github with source code. We strongly believe that making this code available to the community creates a synergy of growth, innovation and development that benefits all developers and the community as a whole.
LiquidVR, our software suite designed to reduce latency in VR was announced at GDC in February 2015. Since then it has been downloaded many 1000’s of times and has become a byword for smooth VR. All of the code for that is available as source. The community reaction to this move has been fantastic.
We also recently introduced FireRender, a free GPU rendering solution, again available with source code. We will have more new on tools and software to help content creators soon.
DFC: There is a lot of excitement with core gamers for you upcoming Zen CPU processor architecture – people who look forward to AMD again delivering a performance processor than can challenge Intel. Why do they have reasons to be correct?
Roy: The clue is in DirectX 12 and Vulkan. Some recent benchmarks have shown that the new API’s not only help modern GPU’s but they also greatly benefit CPUs with cores and threads. So gamers in particular should be very excited about what we have coming.
DFC: Zen is being touted as a new high-performance core design with a new high-bandwidth, low-latency cache system. How do these help Zen be a top-notch competitor and better scale between client and enterprise segments?
Roy: Zen is the name for the new x86 design. We’ve been using Summit Ridge to refer to the desktop PC version. We will have more to say on this nearer launch.
DFC: As we understand it, when Zen arrives in the fourth quarter it will do so in six- and eight-core versions. How much emphasis will be placed on packages tailored to core gamers in general, and VR applications in particular?
Roy: Whilst I cannot comment on unannounced products you can be sure that our commitment to the experience, to the why, will extend to our Zen based products too and this will very much include gamers and VR.
DFC: How is AMD targeting the VR market? We know you desire to help grow the number of VR users, please give us a rundown on how AMD plans to execute on the goal?
Roy: First of all we formed a VR team some time ago. From the moment that Raja and I tried it we both said this is it! and knew we wanted to be the leader in VR. So we made early investments in events like vrloasangeles.com where I have given the opening keynote for the last two major events and will be doing so again in August. As mentioned before we launched LiquidVR, and we then hired partnership executives for Oculus, HTC and Valve. From there we engaged heavily with the content creation industry, centered as it is in Hollywood, which led to our partnership with 20th Century Fox for the Assassins’ Creed movie VR experience.
We also joined with Dell to make ‘The Lonely Whale’ VR experience together with the actor Adrian Grenier. We also have VR test facilities to help optimize content for Oculus and HTC. I am also proud of our support for Dr. Skip Rizzo and his Bravemind project to use VR to treat PTSD. We are engaged in multiple unannounced enterprise VR projects. The list of our engagements is large and still growing. It also includes our support for academia, for example our support of VR development with the Chapman school.
Our plan is to be the best friend of the VR content creation industry and to continue to be the technology leader in this space.
DFC: How are you supporting VR hardware and software development?
Roy: We launched two new graphics products for VR development – for 360 VR where multiple high resolution cameras are used very large frame buffers are needed so that editing software can hold all the data in real time. For this we launched the AMD FirePro W9100 card with 32GB of RAM.
Secondly we saw that by introducing FireRender into 3DS Max and Maya we could benefit content creators by enabling GPU scaling. So then we launched the Radeon Pro Duo, the world’s first professional graphics card designed specifically for content creators. Both products have been immediate hits. As stated we also launched LiquidVR and continue to develop it.
We are also working in a consultative way with the content creation industry so expect more in this space later this year.
DFC: Can VR really be a mainstream consumer segment? How will VR get there?
Roy: Yes it can. But for it to happen we must, as an industry, do the following. First of all, we have to make it more affordable. Polaris will help that for the PC side. Second, we need some killer content that people want very, very much. This is why we are so intent on supporting the content creation industry. Third, we need headsets and controllers that are reasonably priced, available in very large volumes and which are seamless and easy to use. I trust our friends at Oculus, HTC/Valve and Starbreeze to do this.
DFC: You have had the opportunity to work with the two heavyweights in graphics technology. What does your experience with both tell you where the GPU and CPU business going? What are the challenges and opportunities?
Roy: Current VR headsets deliver roughly 2K resolution at 90fps with 10ms latency. That equates to around 8 TFLOPs of performance. Ideally, we could deliver 16K per eye at 120fps with little to no latency. That’s approx. 743 TFLOPs or 81 times more than we have today. Delivering that is both a huge challenge and an opportunity.