Tegra K1 Hairon 810 which is good?

　--CPU Contrast

and the adoption of Denver architecture Tegra K1 different, nvidia Tegra X1 chose the arm standard cortex architecture, and Qualcomm also abandoned the previous krait architecture, the same selection of the standard cortex architecture. The Denver and KRAIT architectures are Nvidia and Qualcomm based on ARM standard Cortex, respectively, and perhaps to allow the flagship processor to be listed as soon as possible, Nvidia Hairon has used the standard cortex architecture. The Tegra X1 Hairon 810 has the same choice in the processor core, with four performance-oriented cortex-a57 processing cores and four power-focused cortex-a53 processing cores, all of which can run at the same time.

In addition, the Tegra X1 Hairon 810 also uses 20nm process, the internal components are not only more compact and more energy-saving, but also equipped with a 1.6GHz of clock LPDDR4 memory, peak bandwidth up to 25.6gb/s. However, Nvidia has not yet announced the frequency of Tegra X1, and Xiao 810 of the highest frequency can reach 2.7GHz.

Although from the point of view of the processing core, Tegra X1 Hairon 810 gap is not big, but in fact, there are many differences between the two, such as the various optimization techniques and ultra frequency management, so the final performance of both may be very different. Of course, beyond these unseen distinctions, the biggest difference between the Tegra X1 Hairon 810 is the GPU on both.

　--GPU Contrast

Tegra X1 still continues the promotion strategy of NVIDIA on GPU micro architecture. It took Nvidia two years before the Kepler microarchitecture, which started on the GeForce 600 series, was applied to mobile processors, but now the Maxwell Micro-architecture took only a year to move from the launch to the landing of the mobile processor, and the pace of progress can be quite obvious. The GPU is equipped with up to 256 Maxwell architectural rendering cores compared to the Tegra K1,tegra X1 with 192 rendering cores.

The Xiao 810 is equipped with the Adreno 430 GPU and has 288 rendering cores built into it. Although the number is more than Tegra X1, but the frequency is only 600MHz, and Tegra X1 GPU frequency can reach 1GHz, so graphics processing performance is better than Xiao Dragon 810.

While Tegra X1 is a processor in the future that will appear on mobile devices such as mobile phones, the car is also one of the main applications of Tegra X1. With its powerful CPU and GPU performance, Tegra X1 can not only compute distances in the course of a car, but also process video signals from up to six cameras at the same time. Although Tegra X1 is also useful for running a variety of large 3D games, it is only possible to perform all of its performance in a car. If used on mobile phones or tablets, Tegra X1 has to reduce performance in order to conserve power, so the actual experience may not be much better than the high-end processors currently available on the market.

--Connectivity and multimedia technology

In addition to processing performance, smartphones and tablets also have a high demand for processors because of their multimedia and communication capabilities. Nvidia Tegra X1 supports playing 4K video and can output 4K video at 60fps frame rate via HDMI 2.0 interface, and also supports h.265/h.264 codec, JPEG image compression and decompression speed is five times times the original. In addition, Tegra X1 also supports EMMC 5.1, Bluetooth, Wi-Fi, NFC, GPS, and 2g/3g/4g communication baseband.

Xiao 810 is slightly less multimedia processing, only support eMMC 5.0, can only through HDMI 1.4 output or edit 4k/30fps video. However, Xiao Dragon 810 is equipped with a baseband to support LTE Cat. 9 Network, the highest data download speed up to 450Mbps.

Can the performance of--tegra X1 really exceed supercomputers?

Nvidia says the performance of Tegra X1 is more than the fastest supercomputer ASci Red at the time 15 years ago. ASCI Red has served 10 years in the US Department of Energy's National Laboratory and is the first supercomputer in the world to reach trillions of floating-point operations per second. Nvidia said ASCI Red covers an area of more than 150 square meters, power up to 500kW, while Tegra X1 can be encapsulated into the phone, power consumption is only 15W.

But in detail, Nvidia's assertion does not hold true, and its conclusion is based on a comparison between the 16-bit floating-point performance of Tegra X1 and the 64-bit floating-point performance of asci red. At present, 16-bit floating-point operations are sufficient to handle some image applications, but they are far from sufficient for high-performance computing applications. So the comparability between 16-bit floating-point performance and 64-bit floating-point performance is not significant.

It is important to note that the Maxwell Microarchitecture is not designed for supercomputers, while the 64-bit floating-point calculation rate is equivalent to about 1/32 of the 32-bit floating-point calculation rate, so if the 256 rendering cores in Tegra X1 can implement 32 floating-point operations per second in 512 billion-bit floating-point, Then the 64-bit floating-point calculation is only 16 billion times per second floating-point operations, far from the asci red trillions of levels. But the calculation speed of up to 1.024 trillion times in 16-bit floating-point computations still proves the power of the Tegra X1, but it is not comparable to a supercomputer.

--Conclusion: the main market is different, it is difficult to produce head-on confrontation

Admittedly, Tegra X1 has become the most advanced processor in the mobile application platform, but at the same time Nvidia says smartphones and tablets are still not using Tegra X1 's super performance, so Tegra X1 is actually more suitable for automatic driving and other applications that have higher performance requirements and are less limited to power consumption. Given this positioning difference, Nvidiategra X1 is hard to compete head-on with Xiao 810 in areas such as smartphones, while Xiao Dragon 810 will still be the choice for most of the flagship models of the 2015. Of course, because the main market is different, so nvidia to Xiao Dragon 8,101 Home Alone The situation is not worried.