Architecture & Design
Although all x86 processors are designed to produce the same results when they process the same data, they do not necessarily have the same design or architecture. It's like car engines. They may work on the same principle to produce energy from petrol or diesel, but internally, they differ in many ways. Similarly, each processor will have different design characteristics that make a large difference in its performance.
These design choices hint at the capabilities of a particular processor. However, you cannot rely on the processor's architecture to give you an accurate indication of its true performance. But it is still important to read up on its architecture because that can tell you of any potential limitations or performance issues.
Pipeline Length
Pipelines are an integral part of modern processors. Think of them as conveyor belts delivering data to a series of stations. Each station performs a small task on each piece of data as they chug along. Since the data has to pass through a few small stations (instead of just one do-it-all station), it takes a longer time for a single piece of data to be processed. However, it allows multiple pieces of data to be worked on in series - one after another, by each station. This greatly increases the performance of the processor.
The pipeline's length is measured in stages. Long pipelines have an advantage of allowing the processor to achieve a higher clock speed. However, in the event of a code branch, or cache miss (data isn't available in the cache); the pipeline has to be flushed and restarted, or put through a process called Replay (a Pentium 4 feature where the entire instruction line is relooped until the data is readied) or backed off (AMD's implementation).
Therefore, long pipelines + programs with lots of code branches = not good. Examples of such software are office productivity software, image-editing software and some video-editing software. Although Intel had an obsession with long pipelines in the Pentium 4 era, the current opinion is that a shorter pipeline is better than a long one.
Production Process
First, you should look for the lithographic process on which the processor transistors are built upon. The lower the number, the better. That's a gross generalization, but it holds true 99 percent of the time. For example, a processor built on a 45nm process is almost certainly faster and cooler than a processor built on the larger 90nm process. It's as easy as that.
Of course, there are innovations to the silicon process that can result in significant performance advantages, even on the same nanometer scale. An example would be the Dual Stress Liner (DSL) and Silicon On Insulator (SOI) technologies used in AMD processors, and the Strained Silicon technology used in Intel processors. Without going into those technologies, the rule of the thumb is that processors with these production features are better than those without.
The best way to ascertain whether a processor has these features or not is to check the information online. However, different steppings of the same processor can have different production features. These processors have new technologies added to them mid-stream. For these processors, you can still find out if they have a particular process technology - just check their stepping number.
It isn't hard to get a stepping number. Most e-stores provide the processor stepping number in the specifications page. At retail stores, you can ask the salesman to show you the box. It's usually printed on a sticker on the box. Just search online at the Intel or AMD website for the stepping number and you can easily find out if a particular processor has the production features you are looking for.
<<< Platform Flexibility, Performance Per Watt & Heat : Previous Page | Next Page : Architecture & Design - Clock Speeds, Bus Speeds >>>