Technical force blowout! In depth analysis of Intel 11 generation core processor

With the official release of Intel’s 11th generation smart core processor, “tiger lake, Xe graphics card and other information that have been rumored for a long time” are also fully presented to users. As the second generation product on Intel’s 10nm process node, tiger Lake integrates many new features, which is the most powerful release of Intel’s technical power in recent years. < / P > < p > first of all, the tiger Lake processor released at this stage is tiger lake-u series, that is, the low-power processor suitable for lightweight notebook computers. Desktop and standard pressure H-series processors will be launched in the next year. < / P > < p > · new design of willow Cove microarchitecture · new Intel iris Xe sharp core display · new AI features · new media and display engine · new hardware enhanced security features · integration of thunderbolt 4 · integration of new PCIe 4.0 interface < / P > < p > in addition to these significant new features, Intel tiger Lake processor also supports Wi Fi 6 wireless network, supporting 7-28w wide thermal design, power consumption adjustment and other characteristics. The details are as follows: < / P > < p > in the initial sequence, Intel has brought 9 different types of processors, including core I3, core i5 and Core i7. Among them, Intel Core i7-1185g7 processor has reached 4.8ghz single core Rui frequency capability, which is close to standard voltage and desktop level processor Rui frequency capability. < / P > < p > after understanding the basic information of Intel 11 generation core, we will explain the features of Intel 11 generation core processor in detail from four aspects of architecture, technology, AI and performance. < / P > < p > the previous generation ice lake processor adopts sunny Cove microarchitecture, while the new tiger Lake adopts willow Cove microarchitecture which is further optimized on sunny Cove microarchitecture, which makes tiger more flexible On the basis of 4 cores and 8 threads, Lake processor can achieve 4.8ghz single core frequency, and its overall power consumption can be controlled between 7-15w and 12-28w, leaving a huge controllable space for OEM. < / P > < p > as we all know, the weakness of ice lake lies in the relatively low main frequency and Rui frequency, which directly affects the performance release of ice lake CPU. Tiger Lake improves the Rui frequency capability by optimizing the micro architecture and makes it match the standard pressure processor. It can be seen that Intel has a very clear understanding of ice lake’s short board. Tiger Lake made up for this. < / P > < p > in order to improve the overall performance, tiger Lake’s bus improves the bandwidth from the basic structure and effectively reduces the delay. The new memory controller supports LP4 / x-4266 and ddr4-3200. The maximum capacity can support 32GB and 64GB respectively, further eliminating the memory bottleneck. In terms of graphics, we will introduce it separately in the following articles. The new Intel iris Xe core displays up to 96 EU units, and supports the new Intel DL boost: dp4a AI acceleration instruction. < / P > < p > based on the new display engine, tiger lake can support four 4K displays or one 8K display output, and the new encoding engine can support 4k60fps 10 bit and 8k30fps 10 bit video coding. In addition, tiger lake also integrates a new ipu6 image processing engine to achieve smooth playback of 4k90fps video. < / P > < p > in order to improve power consumption and performance, Intel also built a new GNA 2.0 neural network accelerator for tiger lake, which is one of the three AI intelligent engines of tiger Lake platform. < / P > < p > we can see that the larger package on the left corresponds to 12-28w TDP high-performance processor, which is mainly used in mainstream lightweight notebook computers and Athena plan products; the smaller package specification on the right corresponds to 7-15w low-power processor, which can realize fan free design. < / P > < p > in summary, tiger Lake SOC supports the modern standby function that can make the computer in low-power standby / wake-up state at any time; it has audio DSP chip to support low-power voice wake-up in standby state, so that the computer is in low-power standby state at any time; it has a new GNA 2.0 neural network accelerator, support dp4a instruction, improve its AI performance up to 5 times; support super-resolution technology; support high-performance AV1 decoding; support avx2 instruction set and Intel DL boost deep learning acceleration function; support IPU 6 high-quality image processing engine. < / P > < p > the seven underlying features enable tiger Lake processor to have a more comprehensive performance in vector, scalar, graphics computing, video streaming media encoding and decoding, artificial intelligence acceleration, etc., which is an almost perfect answer sheet delivered by Intel for the second generation of 10nm process products. Let’s have more expectations for the follow-up tiger lake-s and tiger lake-h! < / P > < p > those who pay close attention to the PC industry are familiar with and unfamiliar with Intel Xe architecture. Before tiger lake was released, Intel had talked about the basic technical features of Xe architecture on many occasions. < / P > < p > 1. Xe LP: for nuclear display, entry-level independent display, standard TDP 5-20w, up to 50W. < / P > < p > 2. Xe HP: for mainstream and feverish consumer markets, data centers and AI fields, with standard power consumption of 75-250w. < / P > < p > we can see that the Xe architecture, which is built by Raja koduri, the God of graphic architecture, has high scalability. It can be seen that the general graphic architecture from general PC to supercomputer platform can be used from scratch to fever. < / P > < p > the Intel iris Xe core display comes with tiger lake, which is the Xe LP level graphics card. The highest execution unit is 96, the core frequency is 1.35ghz, it can run under 1080p 60fps, and can run in 1080p 30fps. Compared with the iris plus core display of the previous generation ice lake, the overall performance is increased by 2 times, the number of execution units is increased from 64 to 96, the core frequency is increased from 1.1GHz to 1.35ghz, and the L3 is up to 3.8mb. The performance is 1.5 times higher than that of the AMD ryzen 7 4800u. < p > < p > Intel Xe core display optimizes the EU unit, uses 8-bit shallow vector to achieve the resource utilization of floating-point unit, which is the underlying reason for the significant improvement of iris Xe core display performance. < / P > < p > from the SOC structure of tiger lake, the display engine and media engine in Xe graphics architecture are the most important weapons in creative design, video coding and decoding, and 4K / 8K video display of Intel 11 generation core. For example, the display engine supports hdr10, Dolby horizon, 360hz refresh rate and adaptive sync technology in addition to supporting four 4K and one 8K display output. < p > < p > media engine enables tiger Lake processor to support AV1 hardware decoding and 12bit end-to-end high-speed video pipeline. It gives tiger Lake processor more powerful media functions. < / P > < p > in addition, it is worth mentioning that in addition to working hard on the architecture, Intel also pays special attention to driver and software level optimization, completely reconstructing the Dx11 driver and making the load lower. At the same time, it also supports game sharpening, configuration optimization, and real-time game tuning, so that the Intel iris Xe core display achieves the graphics performance comparable to that of the mx350. In general, Xe LP nuclear display was significantly improved compared with Gen 11 nuclear display in terms of basic specifications. In addition to the above features, it also has efficient thread control features, supports VRS variable coloring rate, and processes up to 24 pixels and 48 textures per clock cycle. The floating-point performance of fp16 and fp32 is improved by 84%, and the int8 integer processing capability is introduced for the first time, and the performance reaches 8.29tops. < / P > < p > the reason why tiger lake can effectively integrate willow Cove microarchitecture, Xe LP core display, and many IO units, the most important underlying technology is the new 10nm superfin transistor technology. < / P > < p > in short, superfin technology further optimizes the underlying transistor design, not only redesigning the transistor, but also redesigning the metal stack. < / P > < p > first of all, Intel increases the drive current by adding new high-performance transistors and improved gate technology, which enables higher charge mobility, lower source drain resistance and lower capacitance. < p > < p > Intel uses this new transistor technology in high-frequency sensitive IP, such as processor core, high-speed bus and memory subsystem, and also uses existing high threshold voltage transistors in non high frequency critical IP, such as type-C and PCIe, to make it more efficient. These technologies increase the operating speed of transistors and reduce leakage, which in turn reduces the operating voltage of these transistors. < / P > < p > Second, in terms of improving the metal stack, Intel’s engineers have greatly improved the low and middle layer resistance and used a large number of through holes. At the same time, two additional high performance layers are added to the top layer of the transistor to achieve higher peak frequency. In addition, by improving the capacity of MIM capacitor, tiger Lake processor can be competent for higher load tasks and provide fast and stable power supply response. < / P > < p > as a result, tiger Lake processors can release more CPU and GPU performance than the previous generation products, the root cause is that the new superfin transistor technology brings excellent bottom optimization. < / P > < p > many people have no idea about the significance of AI in PC application. In fact, in simple terms, AI technology can help PC more intelligent control performance, extend the service life, and help users to complete related tasks simply and quickly in practical applications. < p > < p > Intel DL boost: dp4a is one of the symbolic instructions of Xe core. It uses 32-bit accumulation to calculate a 4-bit shallow vector dot product, thus speeding up 8-bit integer inference. This powerful 4-element vector point product expansion accelerates the inference for artificial intelligence after optimization, and it is also applied in deep learning. < p > < p > Intel DL boost: vnni is not the first time vnni has appeared on core platform. Intel continues to use this technology on tiger lake. Vnni provides the processor with excellent artificial intelligence performance or vector neural network instructions, which can accelerate the algorithm based on convolutional neural network. It integrates the previous three independent instructions, namely one vector multiplication and two vector addition, into a vector dot product instruction, thus providing 1.7 times higher performance than the competitor. < p > < p > Intel Gaussian & nervous accelerator, or gna2.0, is a low-power AI accelerator built by Intel. It is specially optimized for workflow load, such as dictation, translation or dynamic noise reduction, which can help users improve the experience. For example, background denoising in video conference is a typical application scenario. In addition, the biggest feature of gna2.0 is that it is an independent IP block, so it can work normally when the CPU or GPU is busy performing other workload, and it can achieve 1 billion operations per milliwatt per second in a more power-saving and lower power consumption state. And its peak computing power can reach 38 billion times per second! < / P > < p > by delivering the correct AI workflow workload to GNA, the CPU computing load can be further alleviated, which is the greatest value of GNA technology. < / P > < p > whether it is infrastructure, advanced transistor technology, AI technology, graphics performance, the most intuitive embodiment is to see the real chapter in practical application. At the end of the article, let’s take a look at the official test results of Intel tiger Lake processor. < p > < p > in video editing tasks, Intel Core i7-1185g7 takes 2 times faster to complete the full workload than AMD ryzen 7 4800u