While smaller than many of its competitors, NVIDIA packs a big punch with its Tegra mobile processors. The company has achieved success by paying attention not only to mobile processor performance, but also to how CPUs and GPUs affect system-level features and performance. With each successive generation of the Tegra family, the fabless design house has not only met user expectations, but stimulated desire for even more features in next-generation products. Matt Wuebbling, director of product marketing for the mobile business at NVIDIA, recently spoke about how the mobile processor market is being driven by information mobility and access to information everywhere.
Tegra history
Tegra first gained popularity when NVIDIA launched its Tegra 2 mobile processor. The landmark processor was the first dual-core mobile processor available in the tablet market. At that time, many people questioned whether NVIDIA, as a new player in the market, could make a dent in the mobile space at all. However, with Android gaining traction in mobile devices, NVIDIA saw the potential of the market and focused on differentiating its products with its GPU DNA and heritage from PCs. This allowed NVIDIA to achieve significant smartphone design wins for the first time in the company's history.
To succeed in such a competitive setting, NVIDIA had to show that it could not only differentiate its SOC (system on chip) from other products but also provide its customers and end users with a glimpse into the direction of its future products. The key message NVIDIA delivered to the market at that time was that in addition to bringing the expertise of technology and graphics to its CPU cores, NVIDIA understood that it was the mobile profile that dictated development on its processor platform, and any other added value needed to address innovation on the system level.
This system level philosophy was clearly on display when NVIDIA launched Tegra 3 one year later. Although the top line message with Tegra 3 was that it was the first mobile quad-core processor on the market, from the perspective of NVIDIA, it was not just about cranking out more cores. Tegra 3 also featured system-level innovations such as PRISM technology to help lower overall system power, and DirectTouch to improve the touch responsiveness of mobile devices.
At that time Google also came out with its first Google-branded Android tablet, the Nexus 7, which was built around the NVIDIA Tegra 3, further validating NVIDIA's processor development strategy.
NVIDIA's success in the mobile processor market was also based on its focus on the overall user experience, rather than just benchmarks. For example, with Tegra 3, NVIDIA developed and implemented a memory interface that featured a standard PC interface and DDR3 solution. At that time, mobile chips tended to feature LPDDR2 memory, which boasted low power but was also more expensive. NVIDIA realized that if it could implement a memory solution that was slightly higher in power but offered a significant cost advantage, it could enable some interesting devices that could use the Tegra 3 but at a reduced price compared with competing solutions. NVIDIA realized that for some customers, a strong feature set would be considered as long as there were some considerations given to pricing.
To help tablet makers bring a low-cost, premium experience to consumers, NVIDIA also introduced its KAI reference platform. KAI was a recipe that tablet makers could use to design and build low-cost, Tegra 3-powered Android tablets targeted at the $199 USD price point. KAI had some basic ingredients that could be spiced up with some options. Its basics included Tegra 3, a unique memory controller, DirectTouch and PRISM, as well as components integrated from leaders in the tablet ecosystem. The Nexus 7 was also the first tablet based on KAI.
Extending its product line through multiple market segments
Tegra 2 and Tegra 3 were focused on high-end superphones and tablets. With Tegra 4, NVIDIA is targeting multiple market segments. Launched in early 2013, Tegra 4 is NVIDIA's first offering of a family of products. The company still has discrete application processors that focus on high-end tablets and smartphones, but it also has a second product family called Tegra 4i, which integrates a modem with a single-chip quad-core processor. This allows NVIDIA to attack the mainstream smartphone market. Tegra 4i-based products are expected to arrive in the first quarter of 2014.
Tegra 4 family performance
As a smaller company in the mobile processor market, NVIDIA has chosen to focus its roadmap and technology in areas and market segments where it can make a substantial difference. As a result, NVIDIA delivers a superphone experience at mainstream price points. The Tegra 4i, for example, competes with solutions at the same price point, while featuring graphics performance that is substantially (three times] faster than that offered by the competition.
On the one hand, CPU performance is substantially faster. End users will be able to have 1080p displays because the graphics horsepower in these chips is so significant. They will also be able to get very fast load times for web pages, great gaming and some really amazing camera effects.
One camera feature is NVIDIA's Chimera computational photography architecture, which melds CPU, GPU and image sensor processor abilities. The technology allows user to achieve results that cannot be done with an Apple iPhone 5 or Samsung Galaxy s4. One example of the technology is always-on HDR (high dynamic range). Competing HDR solutions require the processing of multiple exposures, accounting for ghosting and a limited ability to record video because it is hard to capture movement and HDR at the same time. However, with Tegra 4, every photo taken is in HDR. It captures scenes that have bright lights in the background or foreground. And it can take video in HDR, burst photos and photos with flash. With all of these features, pictures overall are substantially better when produced on Tegra 4-based devices.
These features won't only be found in high-end tablets and smartphones. NVIDIA expects this technology to get into smartphones at a sub-US$200 level or even sub-US$150 price points.
GPU performance of Tegra 4
NVIDIA has always been known for its GPU performance, and with Tegra 4, the company does not disappoint. Tegra 4 is about 6x the performance of Tegra 3. With performance gains like that, developers who had never looked at mobile devices before are starting to take it seriously.
This is primarily being driven by a level of GPU performance in devices that has never been seen before, with overall performance now approaching console performance. In addition, in the past it took developers a significant amount of work to port their games. Now, using NVIDIA tools, developers can take console-type content and put it on a mobile device without having to do a significant amount of rework.
Changing usage models: Tablets becoming more like PCs
When people first started using tablets, they wanted to access certain applications, mainly web and email, but the key concern was with power consumption. However, as the mobile market has matured, mobile users are now saying they want to do all the same things on tablets that they were doing on notebook PCs.
In terms of performance, the Tegra 4 CPU is 2-3x faster than the Tegra 3. That in itself represents some pretty amazing improvement. However, power consumption in mobile devices remains a general concern for all processors. With Tegra 3, NVIDIA introduced an architecture called 4-Plus-1 that features four high-performance cores and one low-power core in the CPU. Tegra 4 is the second generation product with 4-Plus-1 so NVIDIA has implemented more innovation to help fire up CPU cores when needed - for instance when switching tasks, having apps like music intelligently run in the background, or when downloading files to the device on a home network or cloud.
Based on ARM architecture, the NVIDIA Tegra 4 can intelligently fire up the cores needed to handle these important tasks, but more importantly making sure those cores are intelligently stopping when those tasks end. One of the advantages ARM architecture has over x86 architecture is the ability to very quickly deliver a lot of performance when needed and then quickly turn off those cores - not necessarily turning off all the cores immediately, but downscaling or down stepping them - when needed.
Modem technology
Another innovation with Tegra 4 is the efficient, software-defined modems - general-purpose modem processors that process algorithms in software. Because they are general-purpose processors, NVIDIA is able to deliver modem chip solutions that are 40% the size of a standard conventional modem. The reason is conventional modems will have to support legacy modem interfaces in different specific hardware blocks but NVIDIA can do that in software. NVIDIA's modem team has shipped 20 million modems on this technology with 95 different carriers around the world, so it's a proven technology and doing very well. One of the things that NVIDIA recently demonstrated is that it could upgrade its modem silicon from category 3 LTE to category 4 LTE on the same silicon, simply by flashing new software to the device. In this way, NVIDIA modems do not need a new tape out of hardware to introduce support for next-generation telecom services.
The takeaway is that NVIDIA is able to deliver a very efficient product, so it enables very small chips - about 40% the size of conventional modems. It's also flexible. It allows NVIDIA to not have to spin new chips when it wants to support new network interfaces. The advantage for end users is having new high-performance features in end devices and having it at a compelling price point.
The future of Tegra
The next chip on NVIDIA's roadmap, a processor that the company has started to briefly talk about, is a chip called Project Logan. Logan is a code name (there is no official name) for a processor that for the first time includes NVIDIA's desktop and notebook GPU capabilities. NVIDIA is basically bringing its Kepler GPU into mobile, meaning DX11, OpenGL 4.3, CUDA-capable GPUs will be available in mobile devices for the first time.
According to Wuebbling, this is exactly why NVIDIA is in the mobile processor business. This is what NVIDIA wants to bring to its products - 20 years of GPU technology capabilities that can be leveraged by more people and in more mobile devices. NVIDIA sees the mobile industry experiencing torrid growth and users are always demanding more.
People are demanding higher resolution displays and better games. More importantly though, Wuebbling argued that the market doesn't even know what it requires, but it better be prepared to deliver once users realize what they want. Wuebbling pointed out that 10 years ago, if you looked at PC gaming, most users said that graphics can't get any better. Yet every year it gets better.
You can see that same dynamic now going on in mobile. According to Wuebbling, we are at the very beginning of what mobile devices can do.
NVIDIA Tegra 4
