Flag This Hub

Tegra 2 and Nvidia 's Strategy

By


See all 8 photos

Introduction

Nvidia is a company traditionally associated with 3D cards. It was the largest in the industry, defending itself from attacks from ATI, SiS, Intel and others.However, it began to decline with the purchase of ATI by AMD, which began investing more heavily in the development and manufacture of GPUs, making the nVidia lost space. To complicate matters, the discrete GPU market itself is being threatened by Sandy Bridge and the APU from AMD, although not as powerful as a dedicated GPU, offering a performance "good enough" for casual players, enough to convince many to stop buying a dedicated GPU.

Noticing the decline of its core business, nVidia started to invest on two fronts. The first is the HPC market (supercomputers), where GPUs offer several competitive advantages compared to CPUs Though things are going more slowly than investors would like, nVidia has already achieved some victories in the industry, as the Tianhe-1A, a supercomputer composed of 7,168 GPUs NVIDIA Tesla M2050 (combined with 14,336 CPUs), which offers 2.507 petaflops of processing crude.

The second front is the launch of SoCs based on ARM processors, for smartphones, tablets and other mobile devices, which can become the main breadwinner of the company in coming years.

The first onslaught was made in 2008 with the launch of nVidia Tegra, which combined an ARM11 processor with a GPU powerful enough, that besides the strong 3D performance was able to record and display with 720p video processing in hardware, support for cameras up to 12 MP, supports HDMI output and so on. Though he was a little weak in processing (due to the use of one instead of an ARM11 Cortex A8 SoCs or similar as in use today), he had a very strong set of features for its time. But he was being ignored by the makers of smartphones due to nVidia's decision in tow strong performance of the chip to Windows Mobile at a time when the interests of manufacturers were already in Android.

In early 2010, NVIDIA introduced the Tegra 2, a more powerful chip that combines two ARM Cortex-A9 processor with a GPU more powerful than twice that of the original Tegra, which offers a better performance than that of direct competitors (the 20 to 30% faster than the PowerVR GSX540 iPhone 3G and 25 to 100% faster than the Adreno Snapdragon) and is able to decode 1080p video via hardware. Despite the announcement of 2010, Tegra 2 finally coming to market just now effectively in early 2011 (due to delays in the final stages of development, production and project development by manufacturers of smartphones) but he has arrived as a strong competitor .

he Cortex A9 is a very different processor Cutting A8 used in devices today. At first glance, the main novelty is the support for multiprocessing, which allows the development of SoCs with two or four cores, but it also includes many structural changes.

With the Cortex A9, ARM took a path similar to the PC manufacturers in the late 90's, moving from an architecture in-order (as in the Pentium and Atom) to an architecture out-of-order (as in other current processors), keeping the processing of two instructions per cycle. The adoption of an architecture out of order provides a superior performance (instructions can be rearranged to make better use of available processing units) but then it increases complexity and reduces the efficiency of the chip.

To counteract these factors, the Cortex A9 is based on a much shorter pipeline, with only 8 stages (against 13 of the Cortex A8). The use of fewer stages greatly reduces the consumption of the chip, more than offsetting increases in other areas, but then limits the frequency of operation that he is capable of achieving. However, this is not a big problem if the Cortex A9, since SoCs for smartphones operating at frequencies lower at around 1.0 GHz

It also includes a buffer for the development of 16 new instructions and registers, with 64 KB L1 cache (32KB for data and 32 KB instruction) per core L2 cache and up to 1 MB shared between all.

As I said, Tegra includes two 2 Cortex A9, equipped with a 1 MB L2 cache and a GeForce ULV. It includes 8 color shader, but unlike current GPUs that use unified shaders, it adopts a simpler architecture (similar to that seen in the GeForce 7xxx) with four processing units, 4 pixel shaders and shaders for processing, combined with units for processing vertices, textures and so on:

Despite the use of dedicated units for pixel and vertex shaders suggest using an ancient architecture, this is not the case, since it includes a pixel cache and texture cache (which improve performance and reduce the number of memory access operations ), texture compression and other functions that just recently debuted on GPUs for desktop PCs. She also received changes designed to reduce consumption, including a large reduction in the number of pipeline stages

In a strategy similar to that adopted in desktop PCs, NVIDIA is working closely with game developers to be released titles optimized for Tegra 2, with better graphics and better physics effects or even the development of exclusive titles that will be marketed through the "Tegra Zone", plus an application store.

Besides the acceleration of video decoding and support for 1080p video via hardware, Tegra 2 also includes support for acceleration of videos and Flash animations using the GPU, similar to what we have on the desktop with Flash 10.1. This solves the problem of low performance and alleviates the problem of very high battery consumption when viewing pages with Flash, enabling smartphones offer a user experience closer to that of desktops.

Also included was an ISP (image signal processor), which improves the performance of the camera, allowing the chip to process images in 12 megapixel camera in only 3 seconds (from 90 to 10 seconds on other current SoCs). Besides reducing the interval between catches and allow the devices support higher resolutions for video capture, it lets you capture several images in sequence, one possibility is explored in Optimus 2X LG, which offers a burst mode, where the phone takes six pictures in rapid sequence.

The first device with the Tegra 2 coming to market is the LG Optimus 2X, which takes advantage of greater processing power to deliver a series of improvements to the current crop of devices:

In addition to better overall performance, the second core allows it to decode 1080p video even via software (for free codecs via hardware acceleration) and offer a navigation experience well ahead of other devices. You can see some benchmarks of this post Anandtech (http://www.anandtech.com/show/4140/this-just-in-lg-optimus-2x ), which puts the Optimus 2X well ahead in relation to performance in navigation processing in javascript, numbers will be similar to other devices based on Tegra 2:

Via hardware acceleration for Flash videos greatly reduces battery consumption and good performance of the GPU makes it a potential competitor for the consoles too. It offers an HDMI output (with the possibility of using a TV or monitor to display the desktop, albeit more limited than that in Motorola Atrix due to the lower bracket on the software) and an 8 MP camera capable of recording Full-HD video (1920x1080) at 24 fps thanks to the ISP included in the SOC.

The adoption of two Tegra is surprisingly strong, both on smartphones as in tablets, which is not surprising, because in spite of the essential function (equivalent to the combination of processor, chipset and 3D card in a PC) represents only the SoC a small fraction of the cost in the end smartphones or high-end tablets (20 or 30 dollars on a device that costs over 500). This makes any new product with real technical advantages to be quickly adopted by manufacturers.

As he did in the first Tegra, Nvidia Tegra is guiding two strongly to a single operating system, but this time instead of choosing the late Windows Mobile opted for Android.

The emphasis on nVidia's Android is more than justified, since with the exception of him and the IOS, all other mobile operating systems are in decline, including Symbian, BlackBerry OS, Windows Mobile and even Windows 7 Phone, which has faced weak sales. Because Apple uses a chip in the iPhone itself so much as the iPad, the only opportunity for a manufacturer that is now entering the market is even Android.

Another factor behind the strategy from nVidia is the idea that smartphones will begin to be used in place of the PC for light tasks, connected to monitors and keyboards, or smartbooks "stupid", which will be only one set of screen and keyboard , leaving all the processing on behalf of the smartphone. This is an idea we are talking about in my article of January, on the Motorola Atrix, which is precisely based on the Tegra 2: 

Click thumbnail to view full-size

For nVidia, the general purpose PCs will be replaced by smartphones connected to external monitors by the end of the decade, with PCs being restricted to the segment dedicated professional games and heavier. Even disregarding this possibility, the projections show continued rapid growth for the number of smartphones sold in the Android next year, surpassing 300 million units annually by 2015 (Gartner). Given the advances of two Tegra, and the promise of a third Tegra quad-core 28 nm operating at 1.5 GHz for the end of the year, Nvidia has a real chance of becoming a leader in the segment, leaving behind losses in market GPU. Even if they get only 20 or 30% share of this market, already sell more than Tegra GPUs for desktop PCs.

Comments

No comments yet.

Submit a Comment
Members and Guests

Sign in or sign up and post using a hubpages account.



    Like this Hub?
    Please wait working