Detailed explanation of the main functional parameters of the CPU

1: CPU frequency:

This is an indicator that attracts the most attention to novices, and refers to how the CPU core works. Clock frequency (CPU Clock Speed). It is usually said that a certain CPU has a frequency of MHz, and this MHz is the "clock frequency of the CPU". At school, I often hear some people ask, XXX Internet cafe’s CPU is 2.66G! XXX Internet cafe only has 2G, and some people use 2.66G Celeron to compare with 2.0G-2.66G P4. This is a sign of ignorance, and it is meaningless to argue with them:). Although the main frequency is related to the CPU speed, it is not absolutely proportional, because the CPU's computing speed also depends on various performance indicators of the CPU pipeline (described below on the pipeline) (cache, instruction set, number of CPU bits, etc.). Therefore, the main frequency does not represent the overall performance of the CPU, but increasing the main frequency is crucial to increasing the CPU's computing speed. The calculation formula of main frequency is: main frequency = external frequency * multiplier.

Two: FSB:

The FSB is the base frequency of the CPU and even the entire computer system, and the unit is MHz (megahertz). In early computers, the synchronous operation speed between the memory and the motherboard was equal to the FSB. In this way, it can be understood that the CPU FSB is directly connected to the memory to achieve a synchronous running state between the two. For current computer systems, the two can be completely different, but the significance of the FSB still exists. Most frequencies in computer systems are based on the FSB and multiplied by a certain multiple. This multiple can be If it is greater than 1, it can also be less than 1.

Three: Multiplier: Multiplier

The multiplier of the CPU, the full name is the multiplier coefficient. There is a ratio relationship between the core operating frequency of the CPU and the external frequency. This ratio is the frequency multiplication coefficient, referred to as frequency multiplication. Theoretically, the frequency multiplication is from 1.5 to infinity, but it should be noted that the frequency multiplication is based on 0.5 as an interval unit. The multiplication of the FSB and the multiplier is the main frequency, so any increase in either can increase the main frequency of the CPU. Originally, there was no concept of frequency multiplication. The main frequency of the CPU was the same as the speed of the system bus. However, as the speed of the CPU became faster and faster, frequency multiplication technology was born. It allows the system bus to operate at a relatively low frequency, and the CPU speed can be infinitely increased through frequency multiplication. Then the calculation method of CPU main frequency becomes: main frequency === FSB x multiplier. That is to say, the frequency multiplier refers to the multiple of the difference between the CPU and the system bus. When the external frequency remains unchanged, the frequency multiplier is increased, and the CPU main frequency will be higher.

Having talked about the main frequency factor, now let us look at other "things" that affect the CPU speed. Please allow me to call it a thing and say that the function is active

Four:

Assembly line:

Those who study geography should understand this. It is said in the first-grade geography book that it is equivalent to part of a mile. Let me make an analogy myself! For example: Take running and walking as an example. They are divided into two levels of assembly lines, that is, the left foot, then the right foot, and the cycle continues. At the first level, it can be said to be jumping with both feet. This is of course inefficient, right? . . . . . . This is the assembly line of life. When you step out with your left foot, if you find a pile of poop in front of you, you have to start over (assuming you must take 2 steps at a time). This is the CPU that will recalculate the errors that follow after the assembly line level goes up. . . . . Maybe I didn’t quite understand what I said. I’ll quote someone else’s words below. In the process of manufacturing CPU, in addition to hardware design, there is also logic design. Pipeline design belongs to the category of logic design. For example, let’s take a company The automobile factory uses four large groups to complete the four production steps in the process of producing cars: one group produces the car chassis, the second group installs the engine on the chassis, the third group assembles the car casing and tires, and the fourth group does painting. Installing glass and other things, this is called a four-level assembly line. (Nowadays, large automobile manufacturers do follow similar assembly lines to improve production efficiency). Assuming that each step takes 1 hour, then if we let 1 large group work on After finishing the chassis of one car, they immediately start producing the chassis of the next one. After finishing the engine of one car, the second group immediately starts assembling the engine of the next car. And so on, the work of the third and fourth groups is the same. , in this way, a Mercedes-Benz or BMW will be produced every hour, which is equivalent to the sequence execution of instructions by the CPU. But what if we also want to improve the production efficiency of the factory? Then we can Divide each of the above large groups into two groups to form an 8-level production line. In this way, each group (note the "group") only needs half an hour to complete its work, so each half hour A car will come off the production line every hour, which improves efficiency (it’s not easy to understand here, please think about it carefully and you will understand).

According to this reason, the CPU assembly line is not difficult. I understand, it just turns producing cars into executing program instructions. The principles are the same.

So here you can think of, if the pipeline is lengthened, can the efficiency be improved? When people applied this idea to CPU design, they discovered that because the pipeline is used to arrange instructions, it is very inconvenient. Flexible. Once an error occurs in the execution of instructions at a certain level, the entire pipeline will stop, and the error will be found step by step, and then the entire pipeline will be cleared and the instructions will be reloaded. This will waste a lot of money. Time and execution efficiency are very low. In order to solve this problem, scientists have used various prediction technologies to improve the accuracy of instruction execution. They hope to avoid the tragedy of emptying the pipeline while maintaining a long pipeline. This is what we often see. Intel's advertisement "This processor uses advanced branch prediction technology...", when you understand what I said above, you will know that it is so mysterious, but in fact it is nothing more than that.

What else has to be said is: a long pipeline will allow the CPU to easily reach very high operating frequencies, but how many of these 2G and 3G frequencies are truly effective operating frequencies? And the more stages there are, , the longer the accumulated delay is, because the work group will generate signal delays when handing over work. Although each delay is very short, the accumulated delay of the 20 or even 30-level assembly line cannot be ignored, thus forming This creates a very funny situation. Pipeline technology increases the frequency of the processor, but it creates a large efficiency gap due to its own defects, which offsets the advantages. High-frequency CPUs also bring high power consumption and high heat generation. Therefore, the longer the pipeline is not, the better

In recent years, Intel's Ben4 processor has gone through three stages of development. The earliest Ben4 processor used the (William) core, which only had a 13-stage pipeline. , the general frequency is not up to 2G, and the speed is average. The second generation Pentium 4 uses the (northwoog Northwood) core. This core has a 20-stage pipeline. Since the number of pipeline stages is more suitable, the first officer increased the speed of the Pentium 4, but It did not affect the execution efficiency. At that time, the Pentium 4 2.4A was a classic product, which suppressed AMD's Athlon XP series. Intel tasted the sweetness and soon launched the Prescott (Poseidon) core. This long-term The new core with a 31-stage pipeline brought Pentium 4 to nearly 3G speeds. This number was beyond AMD's reach, but people soon discovered that the actual operating efficiency of the new Pentium 4 was not as good as that of the old core Pentium 4. However, The frequency is so high, and the heat and power consumption are so high. Intel has "gloriously" gained the reputation of "high frequency and low energy" with this new core. At this time, AMD timely launched the "Athlon 64" series, a brand-new architecture. The 20-level assembly line, low heat and power consumption, and most importantly, low frequency and high efficiency, defeated the new Ben4 in one fell swoop and received high praise. Intel also swallowed a bitter pill of its own making: it was forced to stop 4G Ben4. The development of 4 has lost a lot of market share, and even President Barrett knelt down to the public at IDF05 to beg for forgiveness.

CPU Cache:

Cache Memory) is a temporary memory located between the CPU and the memory. Its capacity is smaller than the memory but the exchange speed is faster. The data in the cache is a small part of the memory, but this small part is about to be accessed by the CPU in a short period of time. When the CPU calls a large amount of data, it can avoid the memory and call it directly from the cache, thereby speeding up the reading speed. . It can be seen that adding cache to the CPU is an efficient solution, so that the entire internal memory (cache + memory) becomes a storage system that has both the high speed of the cache and the large capacity of the memory. Cache has a great impact on CPU performance, mainly due to the data exchange sequence of the CPU and the bandwidth between the CPU and cache.

The working principle of cache is that when the CPU wants to read a piece of data, it first searches it from the cache. If it finds it, it reads it immediately and sends it to the CPU for processing; if it does not find it, it reads it from the cache at a relatively slow speed. Read the data from the memory and send it to the CPU for processing. At the same time, the data block where the data is located is transferred into the cache, so that the entire block of data can be read from the cache in the future without having to call the memory.

It is this reading mechanism that makes the CPU read cache hit rate very high (most CPUs can reach about 90%), which means that 90% of the data that the CPU will read next is in Only about 10% of the cache needs to be read from memory. This greatly saves the time for the CPU to directly read the memory, and basically eliminates the need for the CPU to wait when reading data. In general, the order in which the CPU reads data is cache first and then memory.

The earliest CPU cache was a whole, and its capacity was very low. Intel started classifying caches from the Pentium era. At that time, the cache integrated in the CPU core was no longer sufficient to meet the needs of the CPU, and limitations in the manufacturing process could not significantly increase the cache capacity. Therefore, there is a cache integrated on the same circuit board or motherboard as the CPU. At this time, the cache integrated into the CPU core is called the first-level cache, and the external one is called the second-level cache. The first-level cache is also divided into data cache (Data Cache, D-Cache) and instruction cache (Instruction Cache, I-Cache).

The two are used to store data and execute instructions for these data respectively, and both can be accessed by the CPU at the same time, reducing conflicts caused by contention for the Cache and improving processor performance. When Intel launched the Pentium 4 processor, it replaced the instruction cache with a new first-level tracking cache with a capacity of 12KμOps, which means it can store 12K microinstructions.

With the development of CPU manufacturing technology, the second-level cache can be easily integrated into the CPU core, and its capacity is also increasing year by year. It is no longer accurate to define the first and second level caches by whether they are integrated inside the CPU or not. Moreover, as the secondary cache is integrated into the CPU core, the previous situation of frequency division with a large gap between the secondary cache and the CPU has also been changed. At this time, it works at the same speed as the main frequency, which can provide a higher transmission speed for the CPU. .

The second-level cache is one of the keys to CPU performance. When the CPU core does not change, increasing the second-level cache capacity can greatly improve performance. The distinction between high-end and low-end CPUs with the same core often differs in the second-level cache, which shows the importance of the second-level cache for the CPU.

When the CPU finds useful data in the cache, it is called a hit. When there is no data required by the CPU in the cache (this is called a miss), the CPU accesses the memory. Theoretically, in a CPU with a second-level cache, the hit rate for reading the first-level cache is 80%. In other words, the useful data found in the CPU's first-level cache accounts for 80% of the total data, and the remaining 20% ??is read from the second-level cache. Since the data to be executed cannot be accurately predicted, the hit rate of reading the second-level cache is also about 80% (useful data read from the second-level cache accounts for 16% of the total data). Then the remaining data has to be called from memory, but this is already a fairly small proportion. The current higher-end CPUs also have a third-level cache, which is a cache designed for data that misses after reading the second-level cache. In a CPU with a third-level cache, only about 5% of the data Needs to be called from memory, which further improves CPU efficiency.

In order to ensure a higher hit rate during CPU access, the content in the cache should be replaced according to a certain algorithm. One of the more commonly used algorithms is the "least recently used algorithm" (LRU algorithm), which eliminates the rows that have been least visited in the recent period. Therefore, a counter needs to be set for each row. The LRU algorithm clears the counter of the hit row and increments the counters of other rows by 1. When replacement is required, the data row with the largest row counter value is eliminated. This is an efficient and scientific algorithm. Its counter clearing process can eliminate some data that is no longer needed after frequent calls from the cache and improve cache utilization.

In CPU products, the capacity of the first-level cache is basically between 4KB and 64KB, and the capacity of the second-level cache is divided into 128KB, 256KB, 512KB, 1MB, 2MB, etc. There is little difference between the first-level cache capacity among products, while the second-level cache capacity is the key to improving CPU performance. The increase in L2 cache capacity is determined by the CPU manufacturing process. An increase in capacity will inevitably lead to an increase in the number of transistors inside the CPU. To integrate a larger cache in a limited CPU area, the requirements for the manufacturing process will be higher< /p>

Front-side bus:

The front-side bus is the data channel between the processor and the motherboard Northbridge chip or the memory control hub. Its frequency directly affects the speed of the CPU accessing the memory; BIOS can be regarded as It is a software that memorizes computer-related settings and can adjust related settings through it. The BIOS is stored in a chip on the board, and the name of this chip is COMS RAM. But just like ATA and IDE, most people confuse them.

Because the motherboard directly affects the performance, stability, functionality and scalability of the entire system, its importance is self-evident. Buying a motherboard may seem simple, but there are actually many things you need to pay attention to. When purchasing, pay attention to the product's chipset, workmanship materials, functional interfaces and even ease of use. This requires a thorough understanding of the motherboard in order to choose a satisfactory product.

A bus is a set of transmission lines that transmit information from one or more source components to one or more destination components. In layman's terms, it is a public connection between multiple components, used to transmit information between various components. Bus frequency is often described in terms of speed in MHz. There are many types of buses. The English name of the front-side bus is Front Side Bus, usually represented by FSB. It is the bus that connects the CPU to the Northbridge chip. The computer's front-side bus frequency is determined jointly by the CPU and the Northbridge chip.

The CPU is connected to the Northbridge chip through the front-side bus (FSB), and then exchanges data with the memory and graphics card through the Northbridge chip. The front-side bus is the main channel for the CPU to exchange data with the outside world. Therefore, the data transmission capability of the front-side bus plays a great role in the overall performance of the computer. Without a fast enough front-side bus, no matter how powerful the CPU is, it cannot significantly improve the overall speed of the computer. The maximum bandwidth of data transmission depends on the width and transmission frequency of all data transmitted simultaneously, that is, data bandwidth = (bus frequency × data bit width) ÷ 8.

Currently, the front-side bus frequencies that can be achieved on PCs include 266MHz, 333MHz, 400MHz, 533MHz, and 800MHz. The greater the front-side bus frequency, the greater the data transmission capacity between the CPU and the Northbridge chip, and the greater the ability of the CPU to fully utilize the function. Today's CPU technology is developing rapidly, and the computing speed is increasing rapidly. A large enough front-side bus can ensure that enough data is supplied to the CPU. A lower front-side bus will not be able to supply enough data to the CPU, which limits the CPU performance. has to be exerted and becomes the bottleneck of the system.

The speed of the bus between the CPU and the Northbridge chip more substantially represents the speed of data transmission between the CPU and the outside world. The concept of FSB is based on the oscillation speed of digital pulse signals. In other words, 100MHz FSB specifically refers to the oscillation of digital pulse signals 10 million times per second, which affects PIC and other buses more. frequency. The main reason why the two concepts of front-side bus and FSB are easily confused is that for a long time in the past (mainly before the emergence of Pentium 4 and when Pentium 4 first appeared), the frequency of the front-side bus and the FSB were the same. , so the front-side bus is often directly called the FSB, which ultimately leads to such misunderstandings. With the development of computer technology, people found that the front-side bus frequency needs to be higher than the FSB, so QDR (Quad Date Rate) technology or other similar technologies are used to achieve this. The principles of these technologies are similar to AGP's 2X or 4X. They make the frequency of the front-side bus 2 times, 4 times or even higher than the FSB. From then on, people began to pay attention to the difference between the FSB and the FSB.

CPU process:

Refers to the width of the connection lines of the internal components when producing the CPU on silicon material, generally expressed in microns. The smaller the micron value, the more advanced the manufacturing process, the higher the frequency that the CPU can reach, and the more transistors can be integrated. Currently, both Intel's P4 and AMD's XP have reached a 0.13-micron manufacturing process, and will reach a 0.09-micron manufacturing process next year.

From the above we have learned about the logical structure of the CPU and some basic technical parameters. This article will continue to comprehensively understand the relevant technical parameters that affect CPU performance.

Introduction to CPU types in simple terms<1>

For a computer system, the role of the CPU is as important as the heart to us. When we buy a computer, we always ask first, is it 486 or 586, 100 or 300, MMX or 3D-NOW! , these all refer to the indicators of the CPU. The core role of the CPU in the entire microcomputer system is enough to be used as a standard for classifying CPU grades, which makes it almost synonymous with various grades of microcomputers. We can say that the performance of the CPU can roughly reflect the performance of our microcomputer system, and the importance of this to our choice is obvious.

1. What is CPU?

CPU is the abbreviation of "Central Processing Unit" in English. Its literal translation in Chinese is "Central Processing Unit". The main function of CPU is to perform calculations and logical operations. Its physical results include logical operation units and control units. and storage units. The logic operation and control unit includes some registers, which are used for temporary storage of data while the CPU is processing data. Here, we do not need to understand the complex principles of the CPU. We just have some necessary understanding of it from the selection of performance parameters, which is very helpful for understanding, purchasing, and configuring computers.

2. The main performance indicators of the CPU:

Clock frequency: That is the clock frequency of the CPU's internal core. The unit is generally megahertz (MHz). This is the parameter we are most concerned about, and what we usually refer to as 233, 300, 450, etc. refers to it. For the same type of CPU, the higher the main frequency, the faster the CPU and the higher the performance of the entire machine. Due to different internal structures, different types of CPUs cannot be directly compared by their main frequency. Moreover, the actual performance of a CPU with a high main frequency is also related to the size of the FSB, cache, etc. CPUs with special instructions are relatively The degree depends on the degree of optimization of the software.

FSB and multiplier: FSB is the external clock frequency of the CPU. The relationship between the main frequency of the CPU and the FSB is: CPU main frequency = FSB × frequency multiplier. The FSB is provided by the computer motherboard. The FSB of the 486 is generally 33MHz or 40MHz. The FSB of the Pentium motherboard is generally 66MHz. The motherboard supports 75 and 83MHz. At present, Intel's latest chipset 440BX can use clock frequencies of 100MHz or even higher. In addition, VIA's MVP3, MVP4, APPLO PRO and other non-Intel chipsets have also begun to support 100MHz FSB. Due to their sophisticated technology and advanced technology, some motherboards can be stably used overclocked by more than 1/3, making them the first choice for overclocking enthusiasts. Intel's next-generation motherboard chip will support 133MHz FSB, and AMD's K7 will even use 200MHz FSB.

Introduction to CPU types in simple terms<2>

Internal cache (L1 Cache): Made of extremely fast SRAM, the cache is packaged inside the chip together with the CPU. It is used to temporarily store the latest instructions and data during CPU operation. The access speed is the same as the CPU main frequency (generally called full speed). The capacity of the L1 cache is generally in KB. The L1 cache works at full speed. The larger its capacity, the easier it is for the most frequently used data and results to enter the CPU for calculation as soon as possible. When the CPU is working, the fewer times it exchanges data with the slower L2 cache and memory. Compared with computers, The operation speed can be improved. 486 is much faster than 386 because of its integrated internal cache. The earliest 486 generally had 1K to 8K L1 Cache inside. The current Pentium II's L1 Cache generally has 32K, while Cyrix and AMD chips have 64K or more inside.

Level 2 Cache (L2 Cache): A cache integrated outside the CPU. The general capacity of L2 Cache is 128K~2M. The larger the capacity, the higher the overall performance of the system. The general L2 Cache runs at half the system FSB or the main frequency of the CPU. Later, the L2 used by the Pentium Pro processor ran at the same frequency as the CPU. Because the chip yield was too low and the cost was expensive, the L2 Cache of the Pentium II later ran. At half the CPU frequency, but the capacity is increased to 512K. The current Xeon processor uses a full-speed L2 Cache, and the capacity is increased to between 512K and 2M in order to improve performance. The performance of Celeron processor without Cache drops a lot.

MMX technology: It is the abbreviation of "Multimedia Extended Instruction Set". MMX is a new technology adopted by Intel to enhance Pentium CPU in audiovisual, graphics and communication applications. This technology adds 57 new MMX instructions to the CPU, and increases the L1 cache in the CPU chip from the original 16KB to 32KB (16K instructions + 16K data). Therefore, the MMX CPU is faster than ordinary CPUs in running programs containing MMX instructions. program, the ability to handle multimedia is improved by about 60%. Even programs that do not use MMX instructions can achieve a performance improvement of about 15%. MMX has become a basic standard for selecting CPUs. Currently, most CPUs have MMX technology. In addition to Pentium P55C (Pentium MMX) and Pentium II CPUs, there are also K6, K6 3D (K6-2), MII, 6X86MX, IDT C6, etc. CPUs that do not support MMX instructions can be ignored.

3D instruction technology: MMX instructions solve the bottleneck of multimedia operations, but they only speed up integer operations and are powerless for 3D graphics processing and games that require large-scale floating point operations. In response to the growing 3D processing requirements, supporting 3D instructions will be as important as supporting MMX instructions. Currently, the only CPU that supports 3D instructions is AMD, using 3D-Now! Advanced CPU technology can greatly accelerate three-dimensional processing speed, thereby bringing gaming and graphics processing to a new realm. Intel's upcoming MMX2 instruction set will be more powerful. These instruction sets must rely on software optimization support to fully utilize the CPU's performance. Manufacturing process: unit is micron. The current CPU manufacturing process is generally 0.35 micron, and the latest PII and K6-2 can reach 0.28~0.25 micron. In the near future, the CPU manufacturing process can reach 0.18 micron or even 0.13 micron. The micron level of the CPU directly determines the limit frequency of the CPU. The working frequency of the 0.35 micron CPU generally does not exceed 250MHz, while the 0.13 micron copper core technology chip can work stably at 1000MHz

Types of C P U

CPU

There are three types of CPUs available on the market: INTEL, AMD, and CYRIX/IBM.

INTEL is well-known as the leader. Its CPUs certainly have the best performance, but they are also the most expensive, especially those with high clock speeds. For applications that have high performance requirements, such as graphics processing, 3D games, etc., INTEL CPUs perform most prominently.

CYRIX/IBM's MMX CPU is extremely low-priced, has the best integer performance, and is weak in floating-point operations. It has a high market share in the general low-end commercial and home assembly machine market.

AMDK5 is no longer popular, K6 is more popular, its integer and floating point performance is between INTEL and CYRIX, but its price is not far away from PENTIUM MMX, and it seems that buyers are not interested in the retail market. many.

The CPU market is going through the Romance of the Three Kingdoms. INTEL, AMD, and CYRIX are one strong and the other weak. However, INTEL stands out because the situation is changing.

AMD's K6+ and K63D technologies are developing rapidly. A batch of new CPUs and motherboards will be released in 1998, which will compete with Intelpii and capture 30% of the market share. After being acquired by National Semiconductor, CYRIX has a strong backing. It has recently published the technical details of several high-performance future CPUs, and floating-point operations have become its strength. In 1998 and 1999, there will be a fierce battle in the CPU market.

Types of CPU

In terms of CPU selection, we have never faced so many choices as we do now. Intel is the dominant player, and AMD is the most powerful challenger. Cyrix, which has greatly expanded, and IDT, which has just been launched, have nearly 25 types of CPUs in total. In order to allow everyone to better choose CPUs, we have made an evaluation of these 25 types of CPUs.

Processor Selection

In order for you to make the best processor choice, you must first determine what you want it to do. For example, you often use commercial programs. , and you are concerned about the price of the processor, then non-MMX chips (Pentium, K5, 6X86) are your best choice, but chips of this level will soon be withdrawn from the market. Although today's MMX CPUs don't give you a lot of benefits, newer MMX CPUs have enhancements in other aspects that allow them to achieve higher performance when running all programs, and the price of MMX CPUs has now dropped to A very reasonable position, MMX CPU should be your best choice. In the sixth-generation chips, if you often use programs like those running in Winstone, and you also want to have MMX technology, then AMD's K6 and Cyrix's MII are more valuable. The same grades of K6 and MII can provide Winstone scores that are very close to PentiumII, and their price is more than half cheaper than PentiumII, but their MMX performance is much worse than PentiumII (this will be strengthened in their next generation CPUKII and Mxi) , but it is much better than a CPU without MMX. After testing, Pentium II has been proven to be the best. Whether in commercial or high-end applications, it has shown its due performance in MMX and floating point tests. Even more proven to be unparalleled. If you are a high-end user or a 3D game fan, then don't worry, Pentium II is your best choice, whether it is 233MHZ or the recent 450MHZ. Because PentiumII's dual bus technology and powerful floating point performance will enable you to get the best performance in your application.

Processor Introduction

AMD’s K5 is the company’s first independently produced x86-class CPU. Due to problems encountered in the development of K5, K5 has postponed its launch date and has been Limited to entry-level CPUs, K5 was eliminated by the market by the end of 1997. The following is the specification sheet of K5

Processor Performance Clock Speed

(MHz) Bus Speed

(MHz) PCI Bus Speed

K5 PR75 75 50 25 1.5

K5 PR 90 90 60 30 1.5

K5 PR 100 100 66 33 1.5

K5 PR 120 90 60 30 1.5

K5 PR133 100 66 33 1.5

K5 PR 166 116.66 66 33 1.75

K5 The performance is better than the Pentium with the same frequency, so AMD uses the PR rating system for the K5.

However, AMD pinned its hopes on its K6

K6 was modified from the original NexGen company's 686 (AMD acquired NexGen company), thus obtaining NexGen's in-chip Technical strength in research and development. K6 has MMX technology and more on-chip high-level cache (32K instructions, 32K data) than K5, which can process more instructions in parallel and run at a higher clock frequency.

According to commercial application tests under Windows 95, AMD's K6 is positioned at a stronger level than Pentium MMX at the same speed, and this positioning has been confirmed by our tests. The performance of AMD-based K6/233 in Windows95 commercial tests is quite close to that of PentiumII/233, but it still lags behind by several percentage points. But it easily beats the Pentium MMX/233, which achieved a commercial Winstone 97 score of 7 to 9 percent higher than the corresponding Intel chip. Since the K6 has a larger L1 cache, it can achieve more significant performance improvements than the Pentium MMX as the frequency increases.

Where the K6 falls slightly behind is in running applications that require the use of MMX or FP (floating point instructions). Because the number of chips designed for these functions in K6 is smaller than Intel. For example, AMD's MMX unit can only process one instruction at a time, while Intel's MMX unit can process two instructions.

FP and MMX performance depends on response time and throughput. Reflection time is the time it takes for an operation to start and end. It is a decisive factor for an independent calculation whose result is necessary for further processing. Throughput indicates how fast several new operations can be processed; in a multi-pipeline multiplier or floating point unit, two or more operations can be overlapped; thereby speeding up the speed, but at the same time Reaction time is increased again. All Intel CPUs have fully pipelined MMX and FP units, so a new operation can begin on every clock cycle - even if the results of the operation may not be available until several clock cycles later. When performing a long sequence of calculations, as is common in some multimedia applications, throughput takes precedence over reflection time.

AMD's K6 has a shorter response time than Intel's processor when performing certain MMX operations, but their throughput is the same when processing a single operation, and the shorter response time cannot make up for it. It cannot handle the shortcomings of two MMX instructions at the same time. (However, Intel's MMX unit only has one multiplier and shifter, so it cannot perform MMX and FP operations at the same time. Moreover, only one MMX instruction can access the memory or shaping register at the same time.) When running FP, K6 is running a certain These instructions have a shorter response time than Intel, but it can only start an operation every two clock cycles, while Intel's chip can start a new operation every clock cycle, and the result is that it is performing many FP operations At that time, the throughput of K6 could only reach about half of that of Intel processor.

These weaknesses are reflected in K6's poor performance in ZD 3D WinMark 97, synthetic floating point, Photoshop and other tests. In the above tests, the K6/233 was much slower than the Pentium II/233, and sometimes even slower than the Pentium MMX/233. Especially in the test of ZD 3DWinMard. Therefore, for graphics design or 3D gaming fans, AMD's K6 may not be your choice, at least until KII (K6 with 3DNow! technology) is released. Since AMD has not provided more information about KII, for the time being, All we know is that KII comes with 3DNow! Technology - An instruction set used to enhance the floating point computing capabilities of the CPU. Another thing is that KII will be produced using .25 micron technology and will use a 100MHZ FSB. The frequency when released will be at a clock frequency of 300MHZ. Later we will also see a CPU named KII+ by AMD, which is an enhanced version of KII. , integrate the 512K L2 cache into the CPU, and the frequency when released will be at a clock frequency of 350MHZ.

IBM Company and Company

After Cyrix merged with National Semiconductor, it finally has its own chip production line, and its finished products will become increasingly complete and complete. Cyrix's 6x86 was the first Pentium-compatible processor it put on the market. It uses the PR rating to rate the CPU. Its PR-133CPU runs at 120MHZ, but its performance is the same as the Pentium 133.

But its hair