If you go through the history of Intel processors you will find that starting form Intel 4004 (first 4 bit complete CPU on one chip) to Intel 80386 to Pentium P5 (micro-architecture) the processors where in order execution processors (executed instructions in the order they are provided). As processors speed continued to increase developers felt the need to utilise cpu time efficiently so they developed out-of-order execution, speculative execution etc.
In out-of-order execution if one instruction requires operand to be fetched from memory then cpu can execute next instructions in the instruction queue which are ready for execution. The processor logic is designed such that instructions are re-ordered after execution of all instructions in queue is over. Thus processor clock cycles are not wasted for waiting for operands to be fetched from memory. But this kind of design requires complex logic designand more chip area.
Speculative execution and branch prediction helps processor to execute instructions which are more likely to be executed next. After execution of these the results are discarded if they are not needed (depending on branching condition) otherwise they improve efficiency of processor greatly.
In out-of-order execution if one instruction requires operand to be fetched from memory then cpu can execute next instructions in the instruction queue which are ready for execution. The processor logic is designed such that instructions are re-ordered after execution of all instructions in queue is over. Thus processor clock cycles are not wasted for waiting for operands to be fetched from memory. But this kind of design requires complex logic designand more chip area.
Speculative execution and branch prediction helps processor to execute instructions which are more likely to be executed next. After execution of these the results are discarded if they are not needed (depending on branching condition) otherwise they improve efficiency of processor greatly.
Intel Atom processors are based on the Bonnell micro-architecture. It translates x86-instructions (CISC type instructions) into simple internal micro-operations (i.e., effectively RISC style instructions) before execution. This keeps logic design simple and consumes lesser silicon chip area.
As Atom processors need to be very power and cost efficient they do not use out of order execution.They are designed to increase performance per watt ratio. Even so they give better performance than 5 years or more old Pentium processors at same clock speeds. That is why they are used in low cost netbooks.
No comments:
Post a Comment