The Microarchitecture of the Intel ® Pentium ®

This paper describes the first Intel Pentium 4 processor manufactured on the 90nm process. We briefly review the NetBurst microarchitecture and discuss how this new implementation retains its key characteristics, such as the execution trace cache and a 2x frequency execution core designed for high throughput. This Pentium 4 processor improves upon the performance of prior implementations of the NetBurst microarchitecture through larger caches, larger internal buffers, improved algorithms, and new features. This processor also implements Hyper-Threading Technology, which is the ability to simultaneously run multiple threads, allowing one physical processor to appear as two independent logical processors. This technology is another means of providing higher performance to the end user. We discuss how this processor not only maintains support for this key technology but also increases the benefit seen due to Hyper-Threading Technology. We also describe 13 new SSE3 instructions that have been added to the IA-32 instruction set and are implemented for the first time on this processor. These instructions can be used in multimedia algorithms, such as motion estimation, and for complex arithmetic. Additionally, two new instructions are added for improving thread synchronization. To conclude, performance data are presented that show the benefit of this Pentium 4 processor over prior implementations on key applications and benchmarks.