- Intel NetBurst Microarchitecture
The Intel Pentium 4 processor is built around what Intel has named NetBurst Microarchitecture Technology. The most notably and also most important feature of NetBurst is it's hyper pipelined technology which uses a 20-stage pipeline compared to a 10-stage pipeline to that of the Intel Pentium III and AMD Athlon. As mentioned in the introduction, Intel couldn't get much stability out of their sixth generation past 1.13GHz even when boosting the voltage, as the core was being pushed to far and to it's limits. To get stable clock speeds past 1.13GHz Intel needed to use a core based on a 0.13-micron fabrication process, this is still months off and Intel can't just wait around and let AMD keep producing faster and faster chips, thus is why NetBurst was created and implemented into the Pentium 4, you may even say Intel was forced to use NetBurst technology to keep up with AMD.
Basically, per clock Intel have reduced the number of operations that allows higher clock speeds. The less operations per clock can be made up with the high clock speeds the chip can produce, this is why Intel released the Pentium 4 at 1.3GHz and it will soon be available at 1.7GHz to keep it performing at reasonable speeds. This is also why we have the 20-stage pipeline of NetBurst, to allow the Pentium 4 to perform fewer operations per clock. The deeper the pipeline is, the more stages an instruction must go through before reaching the end of the pipeline, this is how fewer operations per clock is achieved by Intel.
One of the main reasons we see an Intel Pentium 3 beating an Intel Pentium 4 clocked at the same speeds is because of NetBurst. At "slower" clock speeds (eg 1.3GHz) NetBurst technology cannot be fully utilized to it's full potential. Because the Intel Pentium 3, which uses a 10-stage pipeline, performs more operations per clock, this is simply why the Intel Pentium 3 would beat the Intel Pentium 4 at the same clock speed if benchmarked against each other in non-SSE2 tests. However, this comparison won't be made for much longer as Intel will be releasing an Intel Pentium 4 clocked 1.7GHz, because of the limited core of the Intel Pentium 3 you won't see one reaching those clock speeds thus not being able to compete with it any longer.
There is however a down side to NetBurst technology, allow me to explain in the simplest way possible. Processors nowadays attempt to increase the efficiency of their pipelines by predicting the next operation. When a processor predicts the next operation correctly everything happens as normal. However, when an incorrect prediction is made when using NetBurst technology the processing cycle must start again at the start of the pipeline, which is of course much slower. With all this considered, a processor with a 10-stage pipeline has less performance loss for a mis-prediction compared to a processor with a 20-stage pipeline.