Difference Between NVIDIA Tegra 2 and Tegra 3

NVIDIA Tegra 2 vs Tegra 3 | Nvidia Tegra 3 (Quad Core Processor) vs Tegra 2 Speed, Performance
 

NVIDIA, originally a GPU (Graphics Processing Unit) manufacturing company [claimed to have invented GPUs in the late nineties] recently have moved into the mobile computing market, where NVIDIA’s System on Chips (SoC) are deployed in phones, tablets and other handheld devices. Tegra is a SoC series developed by NVIDIA targeting deployment in the mobile market. In a Layperson’s term, a SoC is a computer on a single IC (Integrated Circuit, aka chip). Technically, a SoC is an IC that integrates typical components on a computer (such as microprocessor, memory, input/output) and other systems that cater electronic and radio functionalities. The target of this article is to compare two recent Tegra series SoCs, namely NVIDIA Tegra 2 and NVIDIA Tegra 3.

The two major components of Tegra 2 and Tegra 3 are their ARM based CPU (Central Processing Unit, aka processor) and NVIDIA based GPU. Both Tegra 2 and Tegra 3 are based on ARM’s v7 ISA (instruction set architecture, the one that is used as the starting place of designing a processor) and their GPUs are based on NVIDIA’s GeForce. The CPU and the GPU in both Tegra 2 and Tegra 3 are built in the semiconductor technology known as 40nm of TSMC (Taiwan Semiconductor Manufacturing Company).

Tegra 2 (Series)

Tegra 2 series SoCs were first marketed in early 2010, and the first set of devices to deploy them is some not so famous tablet PCs. The first deployment of the same in a smartphone came in February 2011 when LG released its Optimus 2X mobile phone. Following which a large number of other mobile devices have used Tegra 2 series SoCs, some of which are Motorola Atrix 4G, Motorola Photon, LG Optimus Pad, Motorola Xoom, Lenevo ThinkPad Tablet and Samsung Galaxy Tab 10.1.

Tegra 2 series SoCs (technically MPSoC, due to the multi-processor CPU deployed) had ARM Cotex-A9 based dual core CPUs (that uses ARM v7 ISA), which were typically clocked at 1GHz. Targeting smaller die area, NVIDIA did not support NEON instructions (ARM’s Advanced SIMD extension) in these CPUs. The GPU of choice was NVIDIA’s Ultra Low Power (ULP) GeForce which had 8 cores packed into it (it is not a surprise for a company famous for their multi to many core GPUs). The GPUs were clocked between 300MHz to 400MHz in different chips in the series. Tegra 2 has both L1 cache (instruction and data – private for each CPU core) and L2 cache (shared among both CPU cores) hierarchies, and that allow packing up to 1GB DDR2 memory modules.

Tegra 3 (Series)

The first SoC (or rather MPSoC) in Tegra 3 series was released in early November 2011 and yet to be deployed in commercially available devices. NVIDIA claims that this is the first mobile super processor, for putting together quad core ARM Cotex-A9 architecture. Although Tegra 3 has four (and therefore quad) ARM Cotex-A9 cores as its main CPU, it has an auxiliary ARM Cotex-A9 core (named the companion core) which is identical in architecture to the others, but is etched on a low power fabric and is clocked at a very low frequency. While the main cores can be clocked at 1.3GHz (when all four cores are active) to 1.4GHz (when only one of the four cores is active), the auxiliary core is clocked at 500MHz. The target of the auxiliary core is to run background processes when the device is in standby mode and thus saving power. As opposed to Tegra 2, Tegra 3 supports NEON instructions. The GPU used in Tegra 3 is NVIDIA’s GeForce, which has 12 cores packed into it. Tegra 3 has both L1 cache and L2 cache that is similar to that of Tergra 2 and that allow packing of up to 2GB DDR2 RAM. 

The comparison between Tegra 2 (series) and Tegra 3 (series) MPSoCs is tabulated below:

Summary 

In summary, NVIDIA, in the name of Tegra 3 series, has come out with a MPSoC with high potentials. It obviously outperforms their Tegra 2 series MPSoCs in both computing and graphics performance. The idea of a companion core is very neat, as it can be highly useful for mobile devices, as such devices are in standby mode more often than not and they are expected to run background tasks. How the mobile computing industry is going to utilize the potential, is yet to be seen.