AMD Dual core Vs. Intel Dual Core

Never thought about the big difference between AMD Multicore and Intel's multi-nuclear power, the book of Andrew S. Tanenbaum's operating system was mentioned a few days ago, and the book mentions a little bit about the difference between AMD's multicore technology and Intel Multicore technology, and the online search for information Finally, there is a basic understanding of the multi-core technology of the two.

1 Background

From the date of the birth of computers, the endless pursuit of stronger computing power has been driving the development of computer industry. The ENIAC can perform 300 operations per second, and it's almost 100 times faster than any other calculator in the past, but people are not satisfied. We now have machines that are millions of times times faster than ENIAC, but there is also the need for more powerful machines. Astronomers are learning about the universe, biologists are trying to explore the human genome, and aviation engineers are working to build safer and faster airplanes, all of which require more CPU cycles. However, even with more computational power, people will not be content with this.

The past solution was to make the clock go faster. However, the clock speed limit is now beginning to be met. According to Einstein's theory of relativity, the speed of electronic signals does not exceed that of light, the speed of light is approximately 30cm/nsec in the air, and the velocity in copper or fiber is about 20cm/nsec. For a 10GHz computer, the signal transmission distance will not exceed 2cm (1/nsec=ghz). For a 100GHz computer, this distance is only 2mm. In a 1THz computer, the transmission distance is insufficient 100μm.

It is possible to make a computer so small, but this is another problem: the faster the computer runs, the more calories it generates, and the smaller the computer, the more difficult it is to dissipate heat. In the high-end Pentium system, the CPU radiator is bigger than the CPU itself, hehe, it is incredible. In summary, a better chip manufacturing process is required from 1 MHz to 1 GHz, while a full strategy from 1GHz to 1THz is required.

As a result, multicore processors have emerged. Here's a quick introduction to multicore technology, especially the huge difference between AMD Multicore and Intel's multicore.

2 Multi-core introduction

With the development of chip manufacturing, the size of the transistors is getting smaller, which is likely to put more and more transistors in a chip, the experience-based discovery called Moore's Law, named after the first discovery of the law of the Intel company founder Gordon Moore. The Intel Core 2 dual column chip already contains 300 million orders of magnitude transistors.

An obvious question arises: "How do you use these transistors?" One option is to add a few megabytes of cache to the chip. This option is serious, with a few megabytes of on-chip cache chip is now very common, and with more on-chip cache chip is about to appear, but to a certain extent, and then increase the capacity of the cache can only increase the hit rate from 99% to 99.5%, and such an improvement does not significantly improve the performance of the application.

Another option is to put two or more complete CPUs, often called cores, on the same chip (professionally called die). Dual-core and four-core chips have become commonplace, and 80-core chips have been made, and chips with hundreds of cores are on the horizon.

The core (die), also known as the kernel, is the most important component of the CPU. CPU Center that bulge of the chip is the core, is produced by a certain production process of monocrystalline silicon, CPU all the calculation, acceptance/storage of commands, processing data are carried out by the core. All CPU cores have a fixed logical structure, and logical units such as primary cache, level two cache, execution unit, instruction level unit and bus interface all have scientific layout.

Multi-core (multicore chips) refers to the integration of two or more complete compute engines (cores) in a single processor (chip).

3 AMD multicore vs Intel multi-core

Two caches are typically designed in modern CPUs. The first level, called the L1 cache, is always in the CPU and is typically used to call the decoded instruction into the CPU's execution engine. For those frequently used data words, most chips are arranged with a second L1 cache. In addition, a level two cache, called the L2 cache, is often designed to hold the megabytes of memory used recently. The difference between L1 and L2 caches is timing. Access to the L1 cache does not have any delay, while access to the L2 cache is delayed by 1 or 2 clock cycles.

In a multicore chip. The designer must determine the location of the cache. This leads to a difference between AMD and Intel practices, as shown in the following figure. Figure (a) the Intel Multicore chip uses this method for a L2 cache that is shared by all cores. Instead, each core has its own L2 cache, AMD uses this approach, as shown in figure (b). But each strategy has its own pros and cons. For example, Intel's shared L2 cache requires a more complex cache controller. AMD's approach is having difficulty trying to keep the L2 cache consistent.

(a) four-core processor (Intel) sharing L2 cache, (b) four-core processor with independent L2 cache (AMD)

3.1 AMD Dual core processor

AMD is two cores on a single crystal (die), through the direct-connect architecture (that is, through the hyper transport technology to connect the CPU core directly with external I/O, not through the front-end bus) connected, more integrated. It is said that the AMD Opteron processor was designed from the outset with the addition of a second core, with two CPU cores using the same system request interface, SRI, Hyper transport Technology and memory controller, compatible with the 940-pin interface used by the 90 NM single-core processor.

From the client's point of view, AMD's solution enables dual-core CPU pins, power consumption and other indicators to be consistent with the single core CPU, from a single core upgrade to dual core, do not need to replace the power supply, chipset, cooling system and motherboard, only need to refresh the BIOS software, this for the motherboard manufacturers, The investment protection of computer manufacturers and end-users is very advantageous. Customers can use their existing 90-nanometer infrastructure to migrate to dual-core-based systems via BIOS changes.

AMD's technology architecture lays a solid foundation for the realization of dual-core and multi-core. AMD Direct-Connect architecture and integrated memory controller technology make each kernel's own cache available, with its own dedicated lane-through I/O, no resource scramble, and easier implementation of dual-core and multicore. 3.2 Intel Dual core processor

Intel encapsulates two cores that are placed on different wafers (die). Intel's dual core is simply packaged together using two full CPUs to connect to the same front-end bus. and multiple cores share level two cache, common use of the front-end bus, when the core increased, kernel processing capacity is enhanced, must encounter bus congestion problem. Therefore, such a two core will inevitably create a bus scramble, affecting performance. Not only that, but also for the future of more core integration buried the hidden trouble, because it will aggravate the processor contention front-end bus bandwidth, as a bottleneck to improve system performance, which is determined by the architecture.

4 Conclusion

AMD's solution is a true "dual core", and Intel can only be considered a "double core".