0

This question already has an answer here:

Having seen various people test powerful emulators on their computers, I have come to surprise hearing about how one CPU can have triple the clock speed but still have worse single-threaded performance; or how one CPU can use 16 threads but one forced to use a single-thread can wipe the floor using 1 thread and tons of rapid context switching, even assuming a multi-threaded app.

Some of the best Intel CPUs can be as underclocked as much as possible and still run very demanding programs perfectly, and some of those Atom CPUs can be overclocked as much as possible and still not run them without chopping/lags, etc. Here's what I've noticed:

A 1.5 GHz Intel i7 5960x can run latest Photoshop and Adobe After Effects, including PS2/GameCube/Wii emulators and games fullspeed; an overclocked Celeron 450 @ 3.3 GHz can barely do the same, and it's technically more than twice as high in clock speed.

Also, a 4 GHz Atom N270 seems to perform worse than a 1 GHz i7 5960x. Why is this?

With these really new and expensive Intel CPUs, clock speed seems to mean very little, as they can be 1.9 GHz and run everything, whereas some 3/4/5+ GHz CPUs of yesteryears can't even.

So fast that even forced to use a single core while another CPU uses multi-core, they can do better.

marked as duplicate by Ramhound, bwDraco, Bob, Psycogeek, fixer1234 Apr 26 '15 at 3:02

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • The speed of a CPU is more then just the frequency of the clock. One major factor is how many instructions a CPU can be performed per clock cycle is more important which is the reason the i7 is faster then the Atom. – Ramhound Apr 26 '15 at 2:22
  • @T Gonza Please feel free to search the SU site for question(s) that already covered 99% of what your asking. I am also marking as a duplicate. Use the force (err search) Luke. – Psycogeek Apr 26 '15 at 2:50
2

CPU clock speed is only one of the many factors in "speed." Bus speed, memory speed, disk speed, instruction sets, cache, application design, etc, etc. All of these factors play a role in overall speed.

Intel Atom CPUs are not designed for intense calculations. They are designed for low power computing. This is why they often do not have fans or heatsinks.

The right tool for the job, as they say.

  • That's a black and white outlook. While an i7 can do better, an Atom can still do something! You can't simply write everything off in life as a binary -- you aren't a computer! Atoms can also run PS2 emulators and other modern programs, but usually with less responsiveness, speed, and lag. – user439239 Apr 26 '15 at 4:15
1

The clock speed provided by the specifications is the external clock speed, or data transfer rate to and from the CPU. In the Celeron and Atom CPUs, this is also the internal clock speed. For the i3, i5 and i7, the clock is multiplied to give a faster internal clock. The i7, with multiple cores, will out preform an Atom CPU with the same internal clock.

  • An i7 with one core can still outperform an Atom with multiple though, in nearly any case you can find. – user439239 Apr 26 '15 at 4:16
  • For the i3, i5 and i7, the clock is multiplied to give a faster internal clock. I would suspect... Any reference? – tod Mar 6 '16 at 8:23
1

There are two things that can limit performance among CPUs: the internal factors and the external factors.

Basically the internal factors are what is intrinsic to the internal core architecture of the CPU. The first thing to note is the number of cycles per instruction your CPU requires. For example, on floating point operations it went from above 50 or 60 cycles on the old CPUs to something like 3 or 4 now. That is also without considering new instructions that can for example perform 4 multiplications at the same time on one core.

Then comes the second major difference between many architectures is the pipelining. Since an instruction takes time to perform, modern architectures try to run more than one instruction at the same time if they are independent and use different parts of the CPU. Smarter CPUs can also ask for the memory access in advance and get less stalling. The Atom is by design and outside of the last designs pretty bad on this because a complex pipeline uses a lot of power. You might remember the Netburst architecture as a terrible example of what happens when you try to make a very big pipeline.

On the outside factors, the memory and caches size and architecture affects a lot the performance depending on the code. If I assume my benchmark test code fits in the cache of the i7 but not in the Atom, you can guess that the Atom will lose a lot because of that. The Atom uses smaller caches and has overall a higher latency when it requires a memory access. That's why it will perform pretty bad compared to an i7 when there are many memory accesses: the CPUs has to fetch from main memory all the time and this is slow.

  • Smarter CPUs -- smart term. right answer.. – tod Mar 6 '16 at 8:28