The Pentium 4 Bench Test 
Architecture and Performance Charting Of Intel's New Flagship

By Dave Altavilla - November 20, 2000
 

An eternity has passed it seems, since Intel has released a processor based on an entirely new architecture.  In actuality, it's been 5 years to be exact, since the P6 Micro-Architecture was introduced in 1995.  Against the backdrop of the perpetual design cycles of the Semiconductor Industry, in 5 years a lot can change.

However, mainstream processor base cores are not trivial to design.  Also, Intel has a pretty good knack of leaving themselves the hooks in the technology that will allow roadmap products at higher frequencies with more  features, to carry them forward with relatively minor iterations of the base product.  It seems as though Intel has milked the P6 core for all it is worth now and has finally hit the wall with respect to clock frequency.  The Pentium III will top out at roughly 1GHz. and faced with a growing competitive environment as well as the end user's insatiable need for speed, Intel needed to "reinvent the wheel".  Perhaps not totally reinventing it but a major overhaul was in order.

In the following pages we'll show you the inner workings of the all new Pentium 4 Processor, the Motherboard and Chipset that supports it and the performance levels that are to be expected across various software applications and operating systems.

Specifications Of The Pentium 4 NetBurst Micro-Architecure
.18 Micron, 42 Million Transistors and a little attitude adjustment

 

 

  • Introductory 1.4GHz. and 1.5GHz. Clock Speed with roadmap to 2GHz. and beyond

  • 400MHz. "Quad Pumped" System Bus

  • "Hyper Pipelined" Technology - 20 stage pipeline depth for greater frequency capability

  • "Rapid Execution Engine" -  ALUs run at twice the speed of the core frequency

  • 256K L2 Advanced Transfer Cache running at core processor speed

  • 8K L1 Data Cache

  • Execution Trace Cache - Caches decoded Micro-Ops readying them for execution

  • Advanced Dynamic Execution - More efficient speculative out of order execution unit feeding execution engines

  • Enhanced Branch Prediction Capability - Compensates for the deeper pipeline's higher likelihood of mis-predicted branches

  • Streaming SIMD Extension 2 (SSE2) - 144 New instructions including 128bit SIMD Integer Arithmetic and 128bit double precision floating point instructions in addition to SSE and MMX instructions.

Click images for full view
 



Things certainly have changed for the Pentium 4 and clock speed is only one of many aspects that are completely upscale versus its predecessor.  The way Intel was able to get the core speed up was by deepening the CPU's pipeline.  This allows for higher core frequencies but also increases the chance for missed directed branch predictions.  Branch predictions are the way most x86 processors today anticipate what instruction they will execute next.  In a addition, because the pipeline is twice as deep as say the Pentium 3, the Pentium 4 pays a higher performance penalty for those missed branch predictions.

 

In addition, a deeper pipeline also means fewer instructions per clock cycle can be executed.  This situation is more than compensated for as long as you can achieve the substantially higher clock frequency that the P4 has at 1.5GHz. versus a 1GHz. Pentium 3.  In order to offset some of these concessions Intel had to make with the deeper pipeline, they have also incorporated many new enhancements to their design including adding Execution Trace Cache for decoded Mico-Ops, better speculative out of order execution, an enhanced Floating Point Unit, Arithmetic Logic Units running at twice the speed of the core, SSE2 Instructions and a 400MHz. System Bus.

 

So, the Pentium 4's architecture sounds pretty beefy but what about the Motherboard, Chipset and other hardware that support it?  Enter our Pentium 4 "White Box" Test System.  Let's have a look...

 

The Pentium 4 White Box and D850GB Motherboard