With information gleaned from different session at GPU Technology Conference (GTC) 2013, we pieced together a short update on Tegra's developments so far from the unveiling of the Tegra 4 at CES 2013 to its successors in the form of Logan and Parker respectively.
To reiterate, the successor to Tegra 4 is named Logan and will be released in late 2013; while its successor, the Parker system on chip (SoC) will feature a 64-bit Denver CPU, coupled with a Maxwell GPU. Parker will feature 3D FinFET transistors that will allow the SoC to conserve power without affecting its compute performance. From the performance levels of its roadmap, NVIDIA CEO Huang claimed that the performance of Parker will be 100 times that of Tegra 4; however, according to Moore's Law, its stated performance should be only eight times. He added that by breaking Moore's Law, this is known as disruptive technology due to the nascent state of GP-GPU computing where the industry can continue pushing fabrication technology on both fronts of operating frequency and transistor count.
True to the pioneering spirit of early adopters, he also announced the availability of Kayla that is a development kit that features a board with a Tegra 3 chip and a Kepler GPU, which is hosted on an MXM module on the board.
This kit is not for sale to the general public but it is made available to developers through the company's value-added resellers (VARs). This is to ensure they get a headstart on application development for ARM-based Tegra SoCs with full CUDA implementation on their GeForce graphics cores. To demonstrate this point, CEO Huang showed a Kayla system, operating on Ubuntu, and running a full CUDA ray-tracing program, a PhysX smoke particle program and an OpenGL rendering program simultaneously. With the launch of Logan, the entire Kayla system will be squeezed onto a chip that size of a ten cent coin!