Intel 14th generation Core CPU exposed
Intel will release the Meteor Lake 14th generation Core in the second half of 2023, using Intel's 4nm manufacturing process (formerly 7nm) for the first time, introducing chiplet chiplet chiplet architecture for the first time, upgrading the new CPU/GPU architecture, supporting up to 6+8+2 cores, 22-thread design, and introducing L4 cache for the first time.
According to Intel, the 14th-generation Meteor Lake CPU will feature a new tiled architecture with chiplets across the board. There are 4 main tiles on the Meteor Lake CPU. There are IO tiles, SOC tiles, GFX tiles, and compute tiles. Compute tiles include CPU tiles and GFX tiles. The CPU Tile will feature a new hybrid core design consisting of Redwood Cove P-Core and Crestmont E-Core to deliver higher performance throughput at lower power consumption. Intel will ramp up production of its Meteor Lake CPUs in the second half of 2023.
At the Taipei Computer Show, Intel publicly demonstrated the 14th generation Core for the first time, a notebook, this time showing the VPU vision processing unit and AI work part, but you can also see some key parameters, such as 16 cores and 22 threads, including 6 performance cores (Redwood Cove), 8 energy efficiency cores (Crestmont), 2 ultra-low-power energy efficiency cores (Crestmont).
Among them, the first two cores are located in the CPU module, and the latter is located in the SoC module, which is used to coordinate system control and distribution, especially in ultra-low-power, standby state. Three cores, composed of 16 cores, 22 threads.
The core reference frequency is 3.1GB, and the standby is only 370MHz, which may come from the contribution of ultra-low-power cores.
The cache continues to be greatly enhanced, with a L1 cache of 1.6MB, a L2 cache of 18MB, and a Level 3 cache of 24MB.
Intel's 14th generation Core performance per watt or more than 50% higher than the 13th generation Core, as of the beginning of 2023, Intel terminal CPU has been fully updated to Core 13th generation, from I3, I5, I7 to I9 series, using 10nm process. With the blessing of heterogeneity, the energy-efficient core E-Core can not only strengthen the performance core P-Core, but also carry multi-threaded tasks, which means that the overall multi-threading capability of the 13th generation Core will be greatly improved, and the current multi-threaded performance is mainly used in the field of content creation. The flagship model Core i9-13900K/KF has a maximum turbo frequency of up to 5.8GHz, i7-13700K up to 5.4GHz, i5-13600K up to 5.1GHz, and the remaining three KF series have the same parameters as the models with K except for the core display.
The 13th generation Core offers DDR5 and DDR4 memory controllers, and the DDR5 memory frequency is upgraded to 5600 MT/s, and DDR4 is still 3200 MT/s.
Domestic desktop CPU chip
At present, a number of leading enterprises have emerged from domestic CPUs. We classify according to the CPU instruction set system: 1) Under the complex instruction set (CISC), the X86 architecture is the mainstay, and domestic representative manufacturers include Haiguang and Zhaoxin; 2) Under the Simplified Instruction Set (RISC), it involves ARM architecture, LoongArch architecture, Alpha architecture, etc., and domestic representative manufacturers include Kunpeng (ARM), Feiteng (ARM), LoongArch (LoongArch), Shenwei (Alpha) and so on.
Status quo of Loongson CPU technology route
Loongson adheres to the road of independent innovation and ecological construction. Loongson takes the road of market-driven technology, drives technological progress through the guidance of the market within the system, and then participates in market competition outside the system. Based on the MIPS architecture, Loongson has formed an independent instruction set LoongISA, and the microstructure and physical design of the processor core are all independently developed and obtained, and more than 500 patents have been obtained. Loongson learns the business and technical model of Intel+Google+Apple, creates an open industry chain ecosystem from end to cloud, and realizes cross-platform compatibility of OS.
Loongson's latest 3A5000 quad-core desktop CPU adopts 12nm process, single-core performance strives for 25-30 points, tape-out in the second quarter of 2020; The 3C5000 sixteen-core server CPU also adopts the 12nm process and supports 4-16 channel servers, and the tape out will be completed in the third quarter of 2020. Subsequently, Loongson will release 3A6000 series CPUs and 3D5000 CPUs, which will further improve processor performance.
Current status of Feiteng CPU technology route
Feiteng CPU was created by the research team of the National University of Defense Technology and started in 1999. At present, it is a domestic CPU enterprise with the highest proportion of state-owned assets (CEC through China Great Wall Holdings), and is a CPU national team focusing on national strategic needs and major projects. Licensed by ARM instruction set and integrated fully autonomous processor cores, forming a complete product line covering desktop, server and embedded fields. In terms of ecological construction, Feiteng, as an independent chip supplier, only provides chips, adheres to the open strategy, and has a peer-to-peer open cooperation relationship with all machine manufacturers, and there are many ecological partners at all levels of the full stack, users have more choices, and the security of the supply chain is guaranteed.
High-performance server chip: The latest product of Feiteng high-performance server CPU is the Tengyun S2500, which is mass-produced in 2020, integrating 64 Feiteng self-developed FTC-663 cores, the main frequency is 2.1GHz, the typical power consumption of a single chip is 150W, and it supports 2-8 direct connections, relying on this chip, it can form a computer system with 128-512 cores, which performs well in application scenarios such as large-scale cluster computing and high-performance data centers. In addition, the FT-2000+/64, which was mass-produced in 2017, integrates 64 Feiteng self-developed FTC-662 cores, with a main frequency of 2.0-2.3GHz, a typical power consumption of 100W, and a peak performance of 588.8Gflops, which can meet the performance requirements of cloud data centers and other application scenarios.
High-performance desktop CPU: Feiteng's latest high-performance desktop CPU is the Tengrui D2000 released in 2020, Tengrui D2000 can have up to 8 physical cores, the highest main frequency can reach 2.6GHz, the L2 cache is 8MB, the L3 cache is 4MB, support on-board memory, the memory specification supports dual-channel DDR4/LPDDR4, and the maximum frequency of memory is 3200MHz; In terms of compatibility, Tengrui D2000 is compatible with the previous generation chip FT-2000/4 pins, allowing customers to achieve in-situ plug-in and replacement of existing systems with seamless compatibility. Chip FT-2000/4, integrated with 4 Feiteng self-developed FTC-663 cores, main frequency 2.2-2.6GHZ, provides rich interfaces and security mechanisms, supports standby and sleep, typical power consumption is only 10W, and can be applied to embedded systems by "core reduction" and "frequency reduction".
Status quo of megachip CPU technology route
Megachip series CPUs are launched by Shanghai Zhaoxin Integrated Circuit Co., Ltd. The company was founded in 2013 as a joint venture between a subsidiary of Shanghai State-owned Assets Supervision and Administration Commission and Taiwan VIA Electronics. Based on the X86 architecture, the company has successfully developed and mass-produced multi-generation general-purpose CPUs, formed two product series of "Kaixian" and "Kaisheng", and realized the development and innovation of "from dual-core to eight-core", "from 1.6GHz to 3.0GHz", "from processor + chipset solution to SoC single-chip solution", and has the ability and conditions for independent evolution and development.
In June 2019, Zhaoxin officially released a new generation of 16nm X86 processor products in Shanghai - Kaixian KX-6000 and Kaisheng KH-30000. The new generation of megacore processor is the first domestic general-purpose processor with a main frequency of 3.0GHz, supporting dual-channel DDR4-3200 memory, using SoC design, including CPU, GPU and chipset, chip integration is further enhanced, and the performance and power consumption are increased by 3 times compared with the previous generation of products. KX-6000 adapts to mainstream operating systems such as win7 and win10, and the SPEC 2006 INT RATE score is 170 points, and the single-chip performance has reached the same level as the 7th generation i5, and the application experience has been greatly improved.
Status quo of Haiguang CPU technology route
In April 2016, Haiguang Information established a joint venture with AMD, AMD controlled 51% of the shares, authorized it to produce server processors, Haiguang Information was responsible for chip design and sales, and formed a mutually beneficial partnership with AMD. The Dhyana released by Haiguang is based on the AMD Zen core architecture, similar in performance to AMD EPYC processors, and the software ecology of Haiguang x86 series CPU+ Windows architecture is relatively mature, which can be compatible with most Windows and Linux architecture application ecology on the market.
Haiguang C86 CPU 3250 processor – cost-effective processor for multiple scenarios, using 14nm process, with 8 cores and 16 threads, main frequency 2.8GHz, maximum acceleration frequency up to 3.2GHz, L3 cache 16M, support 32 PCIe lanes, 2 DDR4 memory channels, the highest acceleration frequency reaches 3.2GHz; At present, the same C86 technical route of megachip products, the performance still needs to be improved, Haiguang 3250 can meet the user's requirements for higher performance of domestic C86 products.
summary
Driven by the joint drive of international environment, industrial policy and market demand, a large number of domestic CPU manufacturers have forged ahead, made breakthroughs in technology, performance, ecological construction and other aspects, contributed to the independent controllability, security and credibility of the CPU, and went further and further on the road of "easy to use" marketization. Although there is at least 3 generations of technology gap with mainstream Intel and other international first-class CPU chips, in the future, the "China Chip" business is high and the bird is flying, the sea is wide with fish leaping, the space potential is huge, and Chinese technology enterprises from chasers to parallelists, and then to leaders, it is worth looking forward to!