What does 4GHz mean? What does 5GHz mean? What does 100MHz mean?

What does 4GHz mean? What does 5GHz mean? What does 100MHz mean?

Yes I'm serious. And don't give me hurr it's the clock cycle, I know that.

Why is a 2GHz processor not twice as fast as 1GHz?

Attached: cpu-clock-speed.png (603x309, 27K)

Other urls found in this thread:

pugetsystems.com/labs/articles/Specs-Explained-CPU-137/#ClockSpeed
youtube.com/watch?v=OFgxAFdxYAQ
ece.ucsb.edu/~strukov/ece154aFall2013/viewgraphs/pipelinedMIPS.pdf
twitter.com/NSFWRedditVideo

It's how fast a second it can do operations, more operations more faster.
>Why is a 2GHz processor not twice as fast as 1GHz?
Cache and RAM bottleneck.

A clock cycle is how many operations it can do per second. So for example, any math operation, it all boils down to binary addition, so the more clock cycles per second you can do, the quicker the operations can be done, so yes 2ghz operation is faster than a 1ghz one.

Now for different applications, more cores can benefit even with a lower clock speed, for parallel calculations. A gpu can do many things in parallel so it's better for graphic rendering etc

gotta realize each instruction to the processor takes more than one clock cycle to execute.

better architecture CPUs can do more per cycle

It is twice as fast but there's more to it.
As clock speed increases other factors become more important such as memory access.

If all you ever did was input a few floating point numbers and you keep multiplying the results with the same number you'd expose the clock rate.
But even then you can execute multiple instructions in a cycle because there's many ALUs and FPUs. So for instance an epi32 add on skylake is 1 latency (you get the result back before the next cycle) but it's 0.33 throughput. Meaning you could tell the CPU to do 3 adds per cycle and it'd be done with them the next cycle.

So CPUs are really fast now (in many ways). But usually they're stuck waiting for data because of memory access speeds. So something I'd love to see is how different cache sizes correlate with benchmark scores for real world applications. It'd be a good measure of how poorly written the software is i think.
Or maybe even look at cache misses directly. Does windows have a counterpart to perf?

>It's how fast a second it can do operations, more operations more faster.
and the first answer is already wrong

So a 4GHz CPU is potentially "truly" a 2GHz CPU?

No. The clock is running at 4GHz, but instructions are not performed with only one clock cycle. Many can take 5 or 6 cycles depending on instruction.

This graph is completely wrong buy any laptop cheaper than 800 dollars it will have a 2 core 1.6GHz processor

>What does 4GHz mean? What does 5GHz mean? What does 100MHz mean?
pugetsystems.com/labs/articles/Specs-Explained-CPU-137/#ClockSpeed

>Why is a 2GHz processor not twice as fast as 1GHz?
System bottlenecks. What thing is the bottleneck depends heavily on your workload.

> any laptop cheaper than 800 dollars it will have a 2 core 1.6GHz processor
Cheapest Ryzen laptops have CPUs clocked from 2GHz to 2.5GHz base speed, 2 SMT to 4 cores.

Clock rate normally describes the lowest latency of a single pipeline stage, normally the lowest latency between two dependent operations.

For instance let's imagine we have a processor A that can execute one instruction per clock (a scalar processor) at 1GHz and one processor B that can execute up to 4 instructions per clock (super-scalar) at 500GHz.

A chain of additions (dependent) will take the same amount of clock cycles on both but the scalar processor have a higher clock frequency and have double the throughput.

When doing a bunch of independent additions the super-scalar processor can do 4 times the work per clock compared to the scalar processor. As the super-scalar have a lower clock frequency it only have twice the throughput of the scalar but is still faster.

Reality is much more complicated but the same principle applies: performance = effective clock rate * effective instructions per clock.

Effective instructions per clock is commonly called IPC or Instructions Per Clock.

You meant 500MHz and 4x throughput right?

Did you even try to research anything before throwing out an answer?

Attached: .png (789x362, 36K)

>What does 100MHz mean?
It means my childhood, need for speed 2se, paint shop pro 3.12, 650mb of porn downloaded from bbs and burned to cd marked as "old programs"

Is a 64bit CPU (with all other specs the same) twice as fast as a 32bit CPU? I'm thinking cause it processes twice as much per cycle, but maybe I'm misinterpreting what 64 vs 32 bit CPUs are.

see youtube.com/watch?v=OFgxAFdxYAQ

It means that clock speed is mostly a useless metric

it's harder to do a Squareroot then Add two bits together no ?

It's harder to Multiply then to add etc.

It also depends how the ALU ( Arithmetic logic unit ) been optimized for some specific task.

That's why you'll see some instruction will perform better between intel and AMD even if they both support the same instruction set they have different UAL.

And that's just the base, like mentioned Depends on A lot of factors how many clock cycle you need to do a operation.

if you want to understand more i would suggest you read up on Pipelining in MIPS ( it's a RISC but it will help you understand why GHz aren't the best source of performance metric)

ece.ucsb.edu/~strukov/ece154aFall2013/viewgraphs/pipelinedMIPS.pdf

no 64 bit and 32 bit just mean how many bit it can handle per clock

You can do 64bit math on 32bit CPU it's just much longer since you need to "merge" two 32bit block into 1.
it's mostly about precision of math at that point.