Test setup and comparisons
Here's a quick rundown of the test system should you wish to compare benchmark results with your own.- Intel Pentium 4 3.2GHz Extreme Edition ES 800FSB CPU (12x - 28x)
- Intel Pentium 4 3.2GHz ES 800FSB CPU (12x - 16x)
- AMD Athlon 64 FX-51 CPU (2.2GHz) RAM running with an 11 divisor (DDR400, dual channel)
- AMD Athlon 64 3200+ (2.0GHz) RAM running with a 10 divisor (DDR400, single channel)
- AMD Barton XP3200+ S462 CPU (2200MHz / 200FSB)
- EPoX 4PDA2+ i865PE Springdale with PAT-like BIOS (15/08/03) for both Intel CPUs
- ASUS SK8N nForce3 Pro150 motherboard for Athlon 64 FX-51
- EPoX 8HDA3+ VIA K8T800 for AMD Athlon 64 3200+
- EPoX 8RDA3G nForce2 Ultra 400 for the XP3200+ Barton
Other components
- ATi Radeon 9800
Pro (380/340)
- 2 x 256MB Corsair XMS3500C2, run at 2-6-2-2 @ DDR400 for
both P4, VIA K8T800 (SC) and nForce2 motherboards. - 2 x 512MB Legacy Electronics DDR400
ECC/Registered memory at 2.5-3-3-10
- Liteon 16x DVD
- Samcheer 420w PSU
- Samsung 181T TFT monitor
- AMD reference S940/754 cooler
- Akasa Silver Mountain cooler
- Thermaltake AX478 cooler with a 25CFM fan
Software
- Windows XP Professional
Notes
Three of AMD's finest and Intel's 3.2GHz Northwood will provide adequate comparison to the new Extreme Edition. The CPU arrived unlocked from 12x to 28x, giving the reviewer plenty of benchmarking options. The CPU was run at its native 3.2GHz / 200FSB (800MHz QDR) and also at an overclocked 3.6GHz / 200FSB (18x multiplier). It managed 3.6GHz on default voltage and reasonably quiet air cooling. Benchmarks were run three times (SETI once) and the higher and lower results were discarded. Tests were run at 1024x768x32 85Hz unless otherwise stated.
A quick rundown of a few key characteristics of each of our protagonists.
| CPU | Pentium 4 3.2GHz EE | Pentium 4 3.2GHz | AMD Athlon FX-51 | AMD Athlon 64 3200+ | AMD Barton XP3200+ |
| Clock speed | 3200MHz | 3200MHz | 2200MHz | 2000MHz | 2200MHz |
| L1 cache total | 20kb | 20kb | 128kb | 128kb | 128kb |
| L2 cache | 512kb | 512kb | 1024kb | 1024kb | 512kb |
| L3 cache | 2048kb | - | - | - | - |
| Memory bandwidth | 6.4GB/s | 6.4GB/s | 6.4GB/s | 3.2GB/s | 6.4GB/s |
| FSB | 800MHz | 800MHz | Core speed | Core speed | 400MHz |
| Integer pipeline length | 20 | 20 | 12 | 12 | 10 |
| CPU Die Size | 231mm² | 131mm² | 193mm² | 193mm² | 101mm² |
| Transistor count | 167 million | 55 million | 105.9 million | 105.9 million ? | 54 million |
| Manufacturing process | 0.13-mircon | 0.13-micron | 0.13-micron SOI | 0.13-micron SOI | .13-micron |
| OS Support | 32-bit | 32-bit | 32-bit/64-bit | 32-bit/64-bit | 32-bit |
| Voltage | 1.525/1.55v | 1.525/1.55v | 1.525 - 1.55v | 1.525 - 1.55v | 1.65v |
| Pin count | 478 | 478 | 940 | 754 | 462 |
We've left out any form of Hyper-Threading benchmarks out on the subsequent pages for a simple reason. The effectiveness of HT technology depends upon which applications are chosen. One can make it look ever so impressive in one light and mundane in another. The idea is a simple one. Fool a compliant operating system into thinking that there are 2 virtual processors currently available for application execution. The CPU can then run two tasks simultaneously, without having to wait up certain resources to become free (they're duplicated, after all). Running two tasks concurrently better utilises resources, and allows certain P4s to accomplish more work per clock cycle than their non-HT counterparts. Running two instances of SETI is a good example.
If you want to see our look on Hyper-Threading in general, head here
