Jon Peddie Whitepapers
GPU Developments 2016: no real surprises, more of Moore’s law
GPUs have found their way into just about every corner of science, industry, design, automotive, and entertainment. GPUs are the heart of the PC’s CPUs, the soul of SoCs, and the essence of AI, and DL. It is truly amazing how many problems can be solved using SIMD architecture. And yet for all its power and performance, there are those who would deny it its rightful place and accomplishment out of jealousy, failure, and spite. But denied it cannot be, there are too many things our lives today that depend on, need, and want that are delivered by a GPU to not pay respect.
The purveyors of GPUs, and GPU designs are few today, just nine: Qualcomm, Intel, Nvidia, and AMD design and manufacture GPUs, while ARM, Imagination, VeriSilicon, DMP, and ThinkSilicon offer GPU designs.
The GPU has evolved since its introduction in the late 1990s from a simple programmable geometry processor to an elaborate sea of 32-bit floating point processors running at multiple gigahertz speeds. The software supporting and exploiting the GPU, the programming tools, APIs, drivers, applications, and operating systems have also evolved in function, efficiency, and, unavoidably, complexity. The manufacturing of GPUs approaches science fiction with features below 10nm next year. They’re on a glide-path to 3nm, and some think even 1nm—Moore’s law is far from dead, but is getting trickier to tease out of the Genie’s bottle as we drive into subatomic realms that can only be modeled and not seen.
The notion of smaller platforms (e.g., mobile devices), or integrated graphics (e.g., CPU with GPU) “catching up” to desktop PC GPUs is absurd—Moore’s law works for all silicon devices. Intel’s best integrated GPU today is capable of producing 1152 GFLOPS, which is almost equivalent to a 2010 desktop PC discrete GPU (i.e., 1300 GFLOPS).
We acquire 90% of the information we digest through our eyes. Naturally we need abundant sources of information- generating devices to supply our unquenchable appetite for information. And the machines we’re building have a similar demand for information, though not alway visual information. In some case such as in robots and autonomous vehicles that’s exactly what they need. The GPU can not only generate pixels, but it can also process photons captured by sensors.
Scalability is the other big advantage the GPU has over most processors. To date there doesn’t seem to be an upper limit on how far a GPU can scale. The current crop of high-end GPUs has in excess of 3,000 32-bit floating-point processors, and the next generation will cross 5,000. That same design, if done properly, can be made with as few as two, or even one SIMD processor. The scalability of the GPU underscores the notion that one size does not fit all, nor does it have to. For example, Nvidia adopted a single GPU architecture forthe Logan design in and used it in the Tegra 1. AMD used their Radeon GPUs in their APUs.
It’s probably safe to say the GPU is the most versatile, scalable, manufacturable, and powerful processor ever built. Nvidia, which claims to have invented the term GPU (they didn’t, 3Dlabs did in 1993), built their first device with programmable transform and lighting capability in 1999, at 220nm. Since then the GPU, from all suppliers, has ridden the Moore’s law curve into ever smaller feature sizes, and in the process, delivering exponentially greater performance. Today’s high-end GPUs have over 15 billion transistors. The next generation is expected to feature as much as 32 GB with 2nd gen high-bandwidth memory (HBM2) VRAM and that will exceed 18 billion transistors.
GPUs have been one of the best, most significant developments in our lifetime.Permalink
GPU Virtualization in SoCs
Rick Tewell of Verisilicon presents a white paper on GPU Virtualization in SoCs, please find the executive summary below with the full 11 page paper available for download.
When you think about virtualization computing technology one of the applications that doesn’t typically spring to mind is the automobile. Yet, the automotive electronics industry is on an aggressive drive to adopt virtualization in the car in a big, perhaps even revolutionary, way. For a variety of reasons, primarily driven by cost and ever increasing electronics complexity, automotive OEMs and their tier one suppliers have a desire to have a single electronic control unit (ECU) driven by a single powerful system on a chip (SoC) controlling both the instrument cluster and the infotainment system.
The revolutionary step here is the combining of two major, historically and fiercely independent ECU platforms in the vehicle into a single ECU in an effort to decrease overall system complexity and cost.
This represents a watershed moment in the history of automobile electronics. To accomplish this complex task successfully, a strong virtualization platform is required to allow the running of a minimum of two independent operating systems on an ECU under control of a single SoC. There must be, at a minimum, one OS to drive the safety critical function of the instrument cluster and another completely separate (and often more than one) OS to drive the infotainment platform for in vehicle infotainment (IVI).
The process of virtualization requires that resources inside of the SoC be shared between the various and diverse operating systems running on the SoC. This is, of course, somewhat obvious. For example, if there are three operating systems running on a single SoC, and there is only one graphics processor (GPU) in the SoC and the applications running on these operating systems require graphics, then the only option is for the operating systems to share the GPU; precisely how to do this sharing is the big question. Tremendous effort has been invested in answering this question over the last several years and it is the automotive market with its unique requirements and concerns that is fueling a high percentage of this research and development. To say that there is technical consensus on the issue would be inaccurate at best.
In this paper, we will explore GPU virtualization as it applies to embedded systems with a special focus on the automotive market. We will discuss embedded systems GPU virtualization concepts, methods and current solutions with a forward look to new better performing hardware based techniques that will exist in the near future.Permalink
The Mobile Experience - WhitepaperPermalink
An Analysis of the GPU Market
Computer graphics is hard work. Behind the images you see in games and movies, or while editing photos or video, some serious processing is taking place. All the processing power you can muster is needed to push and polish pixels. And this task is only going to get more demanding as these applications get more sophisticated. Graphics Processing Units (GPUs), which do the heavy lifting in computer graphics, range greatly in size, price and performance. They span from tiny cores inside an ARM processor (such as Nvidia’s Tegra or Qualcomm’s Snapdragon), to graphics integrated within an X86 processor (such as AMD’s Fusion, Intel’s Sandy Bridge), to a standalone discrete device, or dGPU (such as AMD’s Radeon, or Nvidia’s GeForce).Permalink
Bigfoot Killer 2100Permalink
Power Management Opportunities in Graphics-Processors: Considerations for Regulators
Since its introduction over thirty years ago, the personal computer (PC) has been able to satisfy the market’s needs for entertainment, productivity, engineering and design, reaching shipments in year 2009 in excess of 300 million units worldwide, and 95 million units in the EMEA (Europe, Middle East, Africa) region1. The success of the PC is largely due to the users’ ability to adapt and configure it though open industry standards such as USB, WiFi, Bluetooth, PCI Express, DVI, and DisplayPort, to name a few. These standards allow the overall ecosystem to flourish and deliver purpose-built and innovative solutions for the consumer and businesses alike.
As a result of industry standardization, consumers can migrate their monitors, keyboards, and other devices from an older PC to a newer one, or they can upgrade their existing PC with a new graphics board, monitor, operating system, or other peripheral--capabilities that would not be possible without the use of open standards promoting interoperability. At the same time, there is no such thing as a “one size fits all” PC that responds to the requirements of all consumers.
Depending on their graphics requirements, customers can choose from a PC equipped with low- cost, low-performance integrated graphics processors and/or high-performance discrete graphics processors. The term “integrated graphics processor” refers to a graphics processing device that is typically integrated with the PC chipset or the CPU and that shares the PC system memory. The phrase “discrete graphics processor” refers to a graphics processing device that includes its own dedicated memory and memory controller. Discrete graphics processors today enable high- performance 2D, 3D, video, audio and display capabilities that are particularly popular with European consumers.
PC systems sold in Europe tend to be of higher performance levels than global averages, and accordingly tend to have a higher proportion of systems equipped with discrete graphics. The heightened demand for high-performance graphics enables an established ecosystem of regional system integrators to flourish, meeting local customer demand with differentiated graphics products, and stimulating competition in the European market.
A telling example of the need for open industry standards can be found in the critical arena of power management for discrete graphics processors. Computer manufacturers (“OEMs”) and component manufacturers alike, in their race for differentiation and sales, strive to manage power down to the lowest achievable levels—seeking to minimize cost of ownership and to maximize battery life in the case of laptops—while still delivering the needed performance and features to PC users. Power management technologies continue to evolve and rapidly improve over time. For example, it has been reported that the energy efficiency, or performance per watt, of discrete graphics has improved more than tenfold since 20052.
These impressive improvements in the energy efficiency of discrete graphics evolved from the PC’s basic architecture and the existing array of open standards. In recent years, proprietary techniques for power management have been introduced in a limited fashion, but the special case nature of the implementation and incompatibilities with industry standards rule them out for general application.
Switchable graphics, one such proprietary technology for power management, allows dynamic switching from discrete graphics to integrated graphics in some notebook PCs. While switchable graphics solutions have had some success in the market, the solutions are proprietary by nature; relying on custom software and hardware developed by individual vendors. Products with switchable graphics are specific to individual product offerings by specific OEMs, and in some cases available only within specific countries or regions. Support for switchable graphics is not possible using the standard model for PC motherboards that leverages the diversity and choice offered by the open standards-based PC ecosystem. In addition, switchable graphics solutions are not upgradable, and are confined to a small subset of product combinations, do not offer choices in operating systems, increase system complexity and deliver reduced performance and capabilities.
Innovation in the PC market flourishes in a regulatory environment that enables open, interoperable industry standards3. Regulations setting power allowances for discrete graphics that cannot be feasibly achieved by available technologies within the context of industry standards force the industry into use of non-standardized, proprietary technologies, such as switchable graphics architectures. This in turn results in limited performance, choice and connectivity for PC end-users. To ensure that consumers continue to benefit from the results of an open and competitive PC ecosystem; and to promote innovation, a diverse ecosystem of computer builders, and market competitiveness in their jurisdictions, regulators should promote the use of open, available, and established industry standards.Permalink
Visions and Predictions
Graphics industry leaders take a look at the year to come
As is customary in this and other industries, presidents, pundits, and pontificators look into the horizon and try to divine the future. We asked some of the most visionary people in the industry to share their views with us and give us a glimpse of what they see coming. We were surprised by the responses, and sadly couldn’t include them all. Here then are the ones we liked best. As you might expect, each contributor tended to see the world in his or her own terms but these people are all very engaged in their industries as well. Taken together, the things that they see as important—really are important because these people and the people who work for them and with them are helping to make these ideas a reality.
For two years now the cloud has been on everyone’s list of most influential technology. The beauty of talking about the cloud is that it is a vague term that has room for a variety of actual applications. So, you’re pretty much not going to go wrong when you go out on an limb and say “the cloud is going to be important.” However, there is another kind of cloud and the CAD companies have been talking about how important cloud point data is going to be. Ping Fu from Geomagic gives us her take and reminds us that success is about focus.
Through 2009, we’ve seen quite a few new ideas ooze to the top layers of the collective consciousness. Thanks to the movie business, stereoscopic 3D is re-emerging. Levy Gerzberg from Zoran, however, thinks that 3D is going to evolve a little differently than is commonly believed. Not surprisingly, Jen-Hsun Huang at Nvidia is excited about GPU compute possibilities. His company is making an audacious bet on GP GPU computing and they’re putting a lot of resources into trying to kick-start the technology. Like Gerzberg, Huang is also very interested in 3D and he’s looking at tablets. As this is being written, the tablet is on everyone’s mind as people expect to see Apple’s take on it early in 2010. Meanwhile, over at AMD where the company has quite a bit to be optimistic about, Dirk Meyer believes the new PC operating systems Microsoft’s Windows 7 and Apple’s Snow Leopard are going to have a big impact on the industry. Meyer sees them as the gateway for his Fusion product. Luxology’s Brad Peebler has wide ranging interests. He’s expecting to see advances in Augmented Reality, 3D cameras, and photos. Hossein Yassaie from Imagination Technologies is another challenger in the semiconductor market. He has big ambitions for Imagination and thinks multimedia will be everywhere enabled by a new class of embedded connected processors.
The JPR crew piped in to, and Andy thinks 2010 will be the year of Always-on Content and never out of touch smart phones. Interestingly, Jake and Andy take different sides on a couple of issues. Andy sees Blu-ray becoming as important for storage and archiving as it is for content. Jake thinks social networks and the cloud will help kill off physical media, while Ted is cautiously optimistic about on-line games, likes consoles and thinks social networks are over hyped. Jon’s wrap up is, don’t go for the easy forecast and pay attention to intimacy.Permalink