CPU performance analysis of decision criteria

December 24 [Mon], 2012, 16:55
The internal structure of CPU can be divided into control unit, logic unit and the memory unit three most. CPU performance generally reflects its configuration of the computer performance, so the performance of the CPU is very important. The main CPU performance have the following:

A result,

A clock cycle completed order number is fixed, so basic frequency is higher, the CPU speed is fast. But as a result of all sorts of the internal structure of CPU is endless also and same, so there is no way to use basic frequency to hp pavilion dv4202 laptop cpu fan generalize the CPU performance. As for the frequency is the work frequency of the system bus, And the double frequency is refers to the CPU by and basic frequency discrepant multiples. Formulization is: basic frequency = external clock × frequency doubling. We usually say cy young 433, PIII550 is refers to the main frequency of CPU in.

Second, the frequency

The speed of the memory bus to the whole system performance, is very important, because the speed of memory lags behind the development of the CPU speed, in order to alleviate the bottleneck of memory, so bring appeared level 2 cache, to coordinate the difference between them, and memory bus speed is refers to CPU and level 2 (L2) cache and memory work between frequency.

Three, working voltage

Working voltage refers to the normal work of the CPU is the voltage. Early CPU because technology is backward, their working voltage is commonly 5 V, the development of the pentium 586, is 3.5 V / 3.3 V / 2.8 V, along with the CPU manufacturing technology and the improvement of basic frequency, the CPU working voltage is gradually falling trend, Intel newly released Coppermine has adopted the hp pavilion dv4203 laptop cpu fan working voltage of 1.6 V the. Low voltage power and can solve the problem of high fever too, this is especially important for laptop computers.

Four, random sequence implementation and branch prediction

Random sequence execution refers to the CPU allows multiple instruction not to procedures prescribed sequence points for the development of corresponding circuit unit processing technology. Branch is refers to the process operation need to change the node. Branch have unconditional branches and conditional branch, including unconditional branches need only CPU according to the instruction sequence implementation, and condition branch must be according to the processing results before deciding whether to run the program direction change, so need to "branch prediction" technology processing conditions is branch.

Five, L1 cache

The CPU inside the internal cache can improve the operation efficiency of the CPU. The built-in L1 cache capacity and structure on the CPU performance influence is bigger, but cache memory all by static RAM composition, the structure is more complicated, the CPU tube core area cannot too big, L1 level cache capacity might not do too big. The write back structure of the cache. It to read and write operations are can provide cache. Write the general structure of the cache, only the read operation effectively. In the more than 486 computer basic adopted write back type cache.

Six, L2 cache

PentiumPro processor L2 and CPU run on the same frequency, but the cost is high, so PentiumII operation is equivalent to the CPU frequency in half, and capacity for 512 k. In order to reduce the cost Intel company has produced a without L2 CPU called cy young.

Seven, manufacturing process

PentiumCPU manufacturing process is 0.35 microns, PII and cy young can reach 0.25 microns, the latest CPU manufacturing process can hp pavilion dv4217 laptop cpu fan achieve 0.18 microns, and will use copper wiring technology, can greatly improve the CPU integration and working frequency.

Eight, the coprocessor or call math coprocessor

Before 486 CPU inside, there is no built-in coprocessor of. Due to the coprocessor main function is responsible for floating point arithmetic, so 386, 286, 8088 and so on microcomputer CPU floating point arithmetic performance are quite backward, since 486, the CPU usually built-in the coprocessor, coprocessor function is not only to enhance the floating-point arithmetic. Now CPU floating-point unit often the multimedia instruction are optimized. Such as Intel's MMX technology, MMX is "multimedia extensions instruction set" abbreviation. MMX is Intel company in 1996 in order to strengthen the PentiumCPU in audio and video, graphics and communication applications and adopt new technology. Article 57 for CPU new increase MMX instruction, the ability to deal with multimedia improved 60%.

Nine, pipeline technology, superscalar

Assembly line is the first time in Intel 486 chip to use. Line way of working as industrial production assembly line. The CPU by 5 ~ 6 different function of the circuit unit composition an instruction processing line, and then will be a X86 instruction is divided into 5 ~ 6 steps again after the circuit unit respectively carry out, so can achieve in a CPU clock cycle to complete a instruction, thus increasing the speed of the CPU. hp pavilion dv4223 laptop cpu fan Super line refers to a certain type of CPU internal line more than usually 5 ~ 6 steps above, such as Pentiumpro assembly line is a step 14. The pipeline design step (level) number, the more complete it an instruction and the faster, so to adapt to the work main frequency higher CPU. Superscalar is a clock cycle CPU can perform a more instruction. This in 486 or before the CPU is difficult to imagine, only Pentium CPU just above the level with the superscalar structure; This is because modern CPU more and more adopted RISC technology, so would superscalar CPU.