CODEWITHSUNDEEP
assemblycpuclockrates
Celeritas
Celeritas

Reputation: 15043

In terms of instructions processed by the CPU what does the clock rate determine?

There seems to be a gap in my understanding between the CPU processing instructions and the frequency at which the quartz crystal oscillates. How exactly does the frequency of a CPU affect the speed at which a program is run? Is there a relation between the clock rate and the number of instructions processed in a given time?

Example: mov R0, #1. Assuming this instruction took one clock cycle would this take one osculation of the crystal or does it not work like that?

Upvotes: 0

Views: 321

Answers (2)

Earlz
Earlz

Reputation: 63835

Clock rate is a very VERY poor estimate for the speed of a processor. In the olden (pre 2000s?) days, clock rate was a pretty good estimate of speed. Higher clock rate always meant better performance. These days, however, this is not true. You can have a 4GHz processor that takes 20 clock cycles to do anything, and it'll be significantly slower than a 1GHz processor that executes everything in a single clock cycle. This instruction:clock-cycle ratio stuff has become even more complex over the past few years. Now, there is L1 and L2 caches to consider. If there is a cache miss, expect an extra clock-cycle or two... usually. And then there is out-of-order execution and optimization mechanisms that can cause a divide instruction to take anywhere from less than 1(kinda) to 10 clock cycles. And there there is also hyper-threading which means that when another thread is using a component, the other thread will have to wait extra clock cycles... I could go on and on with this.

Basically, don't. Unless you're working on simple embedded(ie, not ARM or x86) architectures clock frequency to instructions/second has little relation. Sure, two equivalent processors with one using a higher clock, should in theory yield a higher instructions/second, but it's anyone's guess as to how many more instructions per second.

Upvotes: 2

Hans Passant
Hans Passant

Reputation: 941455

It doesn't, it depends a great deal on the specific processor design. Roughly, in the late 70s and 80s, microprocessors needed multiple clock cycles to execute an instruction. RISC processors had their heyday in the 90s, their focus was indeed on using a single cycle for an instruction, simplifying the instruction set where necessary to get there. Largely supplanted by processor designs in the 2000s that have multiple execution cores, permitting retiring more than one instruction per clock cycle.

Clock circuits have similarly evolved. Modern processors uses a relatively low frequency oscillator and use an on-chip multiplier to get in the gigahertz range. And vary that frequency dynamically.

Upvotes: 2

Related Questions