Geek On The Mountain

Processor Speed

I see confusion over processor speed all the time. I recently saw a discussion on an article noting that Intel is moving away from just cranking up the “speed” and I was reminded of the problem.

The word speed is deceptive. When I say speed here, I’ll be referring to the number of cycles per second that the processor is running at (Hz). On the other hand, speed could also refer to how fast things actually happen, but I refer to that as performance. The difference is pretty simple. My computer cycles about 1.5 billion times per second (GHz) (I’ll mention what exactly a cycle is in a moment), but it’s performance is something that I would actually measure using some sort of benchmarking software. I might measure anything from the number of arithmetic operations it can do per second to how long it takes to start up a certain program (though the latter in particular starts to deal with much more of the computer than just the CPU).

A cycle is simply the smallest unit of time in which the processor can accomplish something. A processor may be able to perform several operations during the course of one cycle. What exactly happens during each cycle differs with each type of processor. This is a very important point. Not all cycles are created equal. An AMD Athlon processor running at 1.5GHz will not be doing the same amount of work as a P4 running at 1.5GHz. The Athlon will actually perform better. Because of this, to a point, an Athlon will outperform a faster (think Hz here) P4. This can be deceiving because most people have been trained over the years to look for a faster processor for better performance, and it used to be pretty much dead true. For this reason, AMD went to a numbering scheme a couple of years back so that their slower processors would have the same numbers as the equivalently performing P4. It’s good to note as well that although the speed is usually safe to compare to the same family of processors (comparing a P4 to a P4 for instance), even that is not always the case. You might expect a 3.0 GHz P4 to perform about 7% (3.0/2.8 – 1) better than a 2.8GHz P4, but it’s altogether possible that different features were added to the faster version and the difference in performance is actually greater than 7%. (this was actually backwards when comparing P3’s to P4’s running at the same speed).

To give a very general explanation, as a processor runs faster, it produces more heat. This heat means that a more limited number of operations can be performed during each cycle (too much heat will cause the processor to fail (2 + 2 CAN equal 5… ). By lowering the speed, less heat is produced and more operations can be performed during each cycle (but with fewer cycles per second, this doesn’t mean it’s just doing a ton more work).

You might be wondering at this point how exactly you do tell if one processor is faster than another. I think that for most people, it doesn’t much matter because they’re not doing anything that they need tons of speed for. On the other hand, if you really need performance (gaming, video encoding, etc), like getting the most bang for your buck or are just currious, then you need to actually check out some benchmarks that people have ran on different processors. A good site to check out is Tom’s Hardware Guide. They have lots of benchmarks. There are also many other sites that do the same thing. Google is your friend. Of course too, a system’s overall performance isn’t just all about the CPU, so if you’re looking at a particular system you might search for reviews on it as well. On the other hand, the old MHz and numbering systems do provide at least a rough guide of what’s faster than the next thing. Just don’t think that since one has a slightly higher number than another that it’s necessarily faster.

My point here is that MHz does not equal performance and more importantly that you can’t just compare two different chips based on MHz alone (and perhaps not even by the “number” given by the manufacturer). This is of waning importance because focus is moving away from the MHz (or GHz) ratings these days. For most people it doesn’t really matter anyways. The actual performance of a system depends on a lot more than the performance of the CPU on it’s own. (the performance of the CPU as the rest of the system sees it depends on a lot of other things too..). That’s just something that always bugged me so I though I’d confuse everyone even more

Popularity: 2% [?]