Intel Core Past Life (ii)

The previous article tells of the emergence of the core architecture. It is obvious to know that the core architecture actually stems from the notebook processor architecture. Because of the technological trend in the year, because of the increase in the frequency of the negative effects such as heat and high power has made ordinary consumers dissatisfied. However, the increase in the frequency does not improve the processing efficiency, people really feel that the increase in the frequency is not a bargain. With the improvement of computer performance, the trend of multitasking is becoming more and more obvious. Multi-threaded parallel computing is also becoming mainstream. In this trend, AMD and Intel are taking the path of efficiency and multicore. Of course, with the help of AMD's Fast Dragon, Ben 4, and its predecessor, Pentium M, there are many improvements, such as increasing the number of execution units, reducing pipelining, adding new memory communication mechanisms, dynamic frequency tuning, virtualization, and more. I do not want to discuss these technical details, just want to let you onlookers in the birth of the core of the CPU to understand the trend of development.

Fourth: Core I series--AMD to struggle

At the end of the previous article, I mentioned that the first generation of the core to 45nm, improve a lot of performance indicators. Are laying the groundwork for the next generation of new cool core architectures. And this new core architecture is the emergence of the core I series. The emergence and follow-up of the I-series has allowed Intel to tightly control the X86 high-end desktop, notebook, and server CPU areas. Almost drove AMD's K8 into a desperate situation. K8 that year with 64-bit, high-performance ratio, primary dual-core support will be the four-frame hit a mess. In fact, K8 architecture in a lot of design is not bad, but cool core this later will K8 in all aspects of a comprehensive upgrade. It was K8 completely lost its brilliance. In 07, AMD launched the K10 framework to bring the native four core into the market in response to the core architecture. A new shared three cache, with the addition of an integrated memory controller, gives it some advantage over the core architecture and the core-to-memory data communication between cores. The actual results are a bit brutal, just released 65nm process K10 CPU is not better than the core Q series, and even more than a lot of Q series CPU performance is weak.

In 2008, the core architecture has fully stepped into the 45nm process. But in order to be compatible with some of the old chipset, the old technology is still used, for example, the North Bridge chip is still responsible for link memory and CPU communication, and does not support Intel's Hyper-Threading technology, but hyper-threading is in line with the market demand of multithreading; The CPU pins are still 775. But with 64 bits of support, large memory and a wider communication bus, so that the 775 pins somewhat overwhelmed; the core has a Q series of four-core CPUs but its 4 cores are similar to the Pentium D of the year, encapsulating two core duo together. Compared to older rivals AMD's K8 and later K10, core is not like them in some ways. So the new core framework is also in the hope of the emergence of the people. But I didn't expect Intel to bring so many surprises.

Fifth: The birth of the core I series--the beginning of the myth of core

At the end of 07, AMD sacrificed its original four-core, K10 framework. Although the actual performance is not as good as the Intel Q Series, but its excellent architecture, or let Intel feel uncomfortable. 08 Intel decided to come up with new products to strangle the K10 framework directly to the cradle. This is the Nehalem framework. The second version of the 45nm core architecture, carefully counted, is also the three core architecture version. The Nehalem architecture created the core I series CPU. The first product named "Core i7", the kernel code Bloomfield. This version of the core architecture finally integrates the memory controller, directly communicates with the memory bus, supports 3-channel 1333DDR3 memory, abandons the old FSB front-end bus, uses the same technology as AMD's HT bus, qpi the new serial bus, the bandwidth is higher, the native four cores, The four cores are similar to AMD's K10 design, with a separate level one or two cache that shares up to 8M of level three cache; Hyper-Threading technology returns to SMT, that is, the i7 of the year can support up to 8 threads of parallel computing; stunt Turbo technology is CPU-independent overclocking, improving energy efficiency The latest version of the SSE4.2 instruction set! At this point, the latest Nehalem architecture of the core architecture has completely defeated the K10 architecture, surpassing AMD, so far, AMD has lost all of its technical advantages, leaving only the low-end market with Intel. Throughout the emergence of the core architecture and the emergence of the core I series, it can be seen that AMD first led the intel,03 year K8 almost defeated the p68,06 core architecture with a K8 similar design concept, the main performance performance can exceed k8,07 year release K10 prepare to halt core, However, Intel08 released Nehalem the design advantages of K10 all included, and in the various technical parameters in a comprehensive way beyond the K10. Once again, I had to marvel at the power of Intel's technology reserves. In just a few years will be the original technology beyond their own competitors, comprehensive beyond and firmly grasp the technical advantages.

High-spec Nehalem processing can no longer be used LGA775 interface, Bloomfield Kernel i7 upgrade to LGA1366. From the beginning of the I-series, Intel has launched the upgrade mode, CPU interface and chipset chip frequently upgrade, the era of LGA775 with the end!

The i7 specifications of the Bloomfield Core are very high, belonging to Intel's high-end products of the year, and Intel, which is familiar with commercial operations, subsequently publishes various derivative versions to match the refinement of the market

Kernel Code Lynnfield version, four-core version, communication bus from QPI to DMI, although the bandwidth is somewhat reduced, but the CPU is more than QPI integration of some more functions, fully integrated North Bridge function (PCI-E Controller), and DMI is for the CPU and South Bridge direct communication bus. This version of the CPU has two versions, does not support hyper-threading, is named i5 supports Hyper-threading for the i7 LGA1156 interface.

The kernel code is the mobile CPU version of Clarksfield. The kernel generation is also a Xeon version of Beckton and Gainestown.

This time the first generation of Nehalem has been layout-formed! and formed the i7 top desktop CPU i5 in the high-end desktop CPU layout. I-series is definitely the highest technology representative of desktop CPU in the year!

Sixth: The emergence of Apu and the Battle of multi-core

High-end, Intel's i5 and i7 have been proud, AMD has adopted a conceptual product, low-priced multi-core and Intel competition. At this time AMD is still the K10 framework, the process also upgraded to 45nm, mid-high-end, AMD Go core thinking, released the K10 framework of the 6-core CPU, performance is almost dual-core i5 appearance, Intel's 32nm process Westmere version of the CPU also only has the Xeon CPU and the top i7 in the 6-core version. In the low-end, AMD's four-core and three-core CPUs are still very popular with the market. Because Intel's 4-core CPU prices are generally expensive, AMD's products are much cheaper. Of course, AMD's four-core and three-core products may actually not have the intel of some dual-core or even the older generation of the four core performance is strong. AMD has acquired ATI, the display's core business, when AMD spoke about fusing the graphics card with the CPU. Unfortunately, the first generation of the graphics card core is Intel. In 2009, Intel showed the latest generation of core architecture Westmere with 32nm technology as an upgrade to Nehalem. One of the brightest eyes is the mid-end Clarkdale series CPU. Because this CPU first shows the products that display the core. While Intel's solution is to directly weld the display core to the CPU on a single substrate, neither is even the same production process. 2010 Intel finally released the Clarkdale CPU, and brought a new low-end series I3 class. And this time AMD suffer from no new products come out, high-end by the main cost-effective Yi Dragon second generation 6 core playing low-priced 6 core, low-end playing a variety of lower-priced four core and shielding a defect core of the three core in the defense of their own market. But it is also because of this shielding a core three-core CPU pull up the low-end market to open a nuclear game prologue. All kinds of DIY players and businesses have joined the nuclear war. By the year 2011, AMD has finally come up with the first APU product. For the first time, the true single-manifestation core is integrated into the CPU. Unlike Intel's first-generation core graphics, which simply encapsulate the display core with the CPU, the APU is really architecting the CPU with the graphics core. The two share access to memory, direct access to the PIC-E bus, the CPU works with the GPU, and AMD employs a converged acceleration technology that allows some CPU floating-point technology to be given to the GPU for computation and to increase computational speed. In addition to AMD's APU because it is truly integrated, the GPU can directly access the PIC-E bus, so the APU can also achieve mixed crossfire. Overall, it does go beyond Intel's core graphics technology.

With the launch of the APU as well as the bulldozer Chinese name bulldozer, the most important idea for bulldozers is to allow the CPU's modular design to provide the technical foundation for AMD's multi-core stack. AMD's idea is still to compete with Intel by the number of cores in the cross-over. The core of the bulldozer is directed at the 8 cores. Of course, bulldozers also have a very technical update, and the operating system did not target AMD this multi-threaded CPU optimization, so that it does not feel strong performance, but here is not the table. The 8 nuclear CPU battle has arrived.

In the same year, Intel also released its own new 32nm process-based core architecture Sandy BRIDGE,SNB also joined the multi-core contention. The enhanced kernel version of SNB supports up to 8 cores. Because Intel is useful for Hyper-threading technology, the SNB's CPU can support up to 16 threads! In the core graphics, SNB, like AMD's Apu, is finally complete with the CPU core design.

In just a few years the CPU has been upgraded from dual-core to 8-core, and even the GPU has been integrated. The rapid new era of multicore CPUs brought by this core architecture is almost over, and Intel has not exited the technology-updated product since AMD has no more threatening products after the bulldozer. Ivy Bridge, Haswell, and Broadwell later also had no previous generations of products. Basic is the support of existing new technology and the improvement of production process. With the advent of this extreme process of 14nm. The process has become more and more limited.

Conclusion:

From these years the development of the CPU can be clearly seen, high efficiency and low energy consumption, multi-threaded computing, fusion computing and so on are the development trend of the CPU. And the story of the core is the renewal and transformation of these technologies.

This is the end of the story. Throughout the core story, you can see that core is truly Intel's hero, pushing Intel again on the X86 processor manufacturer's dominance. Also seen, Intel's strong technical reserves and business capabilities. It has to be suggested here that AMD is indeed the initiator and catalyst of this processor upgrade, and that Intel may be slowly upgrading CPU technology to the residual value of each generation of CPU technology without AMD's advanced technology in the pursuit of energy efficiency, multicore, GPU converged computing.

While Intel has been X86 on the ground, with the explosion of the mobile phone market, mobile chip makers, represented by arm, have laid the groundwork for the Intel crisis. ARM employs a CPU that is a thin instruction set, although it is not as powerful as the CPU performance of the complex instruction set X86 in complex computations. But there is a natural advantage in energy consumption. And with the software optimization and chip integration, in mobile computing is the advantage of the heavy! Now a mobile phone CPU is generally heterogeneous mode integration of multiple ARMCPU core and coprocessor, integration of multiple GPU cores, integrated communication baseband chip, integrated power control chip, integrated memory chip and so on. So that the arm mobile phone chip in the function and performance balance, for example, now the mobile phone chip can be more current mobile phone actual use, dynamic adjustment chip CPU core operating state and core frequency, can be in the whole chip in hibernation is to let the coprocessor processing simple sensor data, Can CPU dynamic tuning work with a high-performance kernel or a low-performance kernel. In general, the CPU is more functional, and can be in the state of the phone, only the corresponding functional areas to work to reach the best balance of energy efficiency.

The development of ARM also confirms the trend of contemporary energy efficiency and convergence calculation! Just arm out of the shackles of X86, its development of more unrestrained and bold.

The trend is China's old adage, the core in its years because of the trend and the shine. And now under ARM's attack, whether Intel can continue to follow the trend, I do not know. But if Intel has no major technological breakthroughs like the recent generations of the core, perhaps the next story becomes arm's story.

Intel Core Past Life (ii)