|
The Test
Setting up and stability
The ASUS P4T was tested in the following system configuration:
- ASUS P4T
- Intel Pentium4 1500 ([H]ardOCP)
- 2 x 128 MB Rambus RIMMs (+ 2 c-RIMMs)
- ATi Radeon AIW 32 MB DDR
- IBM 372050 DPTA
- Hitachi 10 X DVDROM
- Alps Electric Floppy Drive
- AMD PCI Etherpower
- Windows 98
Setting up the system from a freshly formatted drive went flawlessly. All drivers were found on the ASUS CD and installed without any problem. I didn't expect any issues regarding stability under standard conditions and none were found. One thing should be mentioned is that only the standard ATX power connector was used in a standard Antec KS282 case featuring a 300 W power supply. Under no circumstances was there any indication that the PSU was insufficient for the P4 configuration, despite the fact that no auxiliary power was used.
|
[an error occurred while processing this directive]
Overclocking
The P4T offers the possibility of changing the multiplier in the BIOS. As already mentioned, this possibility applies to unlocked Intel engineering samples but any standard P4 will be locked, thus, preventing the settings from taking any effect on the clock speed. In other words, boring! The P4T, however, offers the possibility to overclock the FSB up to 150 MHz. Please keep in mind that at this speed the data rate would be 600 MHz which is not necessarily realistic.
A new PCI divider
For the first time, the ASUS P4T offers a 1/5 PCI divider. We take this as indication that Intel will be moving towards a high speed SDR interface with their budget chipsets, currently limited by the PCI bus. I wouldn't be surprised if the next revision of the i815 chipset were to feature the same 1/5 PCI divider to boost memory and FSB frequencies up to 200 MHz.
Back to the P4T
At 150 MHz, the RIMMs would run at 1200 MHz data rate which is not very realistic, particularly in a dual channel setup. The workaround for this latter problem was already mentioned:
- Use PC800 RIMMs and set the RDRAM frequency to 3:1 mode. This way the memory bottleneck can be eliminated and the FSB can be overclocked to the limitation of the QDR protocol or the clock limitation of the P4, whichever comes first.
At standard 4:1 mode, the highest FSB speed at which marginal stability was still present was 115 MHz, resulting in a total of 1725 MHz CPU clock speed at the default 15 x multiplier. Stepping back to 14 x to eliminate the CPU clock ceiling did not improve the overclockability of the FSB.
 CPUID screenshot at 15 x 115 MHz with the RDRAM frequency set to default (4:1) At this setting, stablity was good enough to run benchmarks but occasional lock-ups were encountered. Note that CPUID is somewhat out of its league here.
With the RDRAM frequency set to 3:1, however, and the multiplier at 14 x, 120 MHz FSB was no problem. Keep in mind that this setting results in 480 MHz FSB data rate, which is quite phenomenal. At this speed, no problems were observed at all. It is very likely that a higher FSB would have run equally stable but, because of time constraints and the fact that the CPU and RIMMs were loaned from [H]ardOCP, I didn't want to push things further. CPUID showed the correct clock speed as 1680 MHz but failed to disclose any other parameters.
|
