System Setup

JNav89GT's Test Setup
CPU	Intel P4 530 Retail-3.0GHz LGA 775 (Retail Cooling used for all testing)
Mainboard	MSI 915G Combo
Memory	Crucial Ballistix DDR2 PC2-5300 1GB Dual Channel Kit(2x512mb)-Timings of Cas4 4:4:10 used throught all testing
Video	ATI Radeon X700Pro
Power Supply	Zippy 460W (HP2-6460P)
Operating System	Windows XP SP2

I would like to take a few minutes to describe a few problems I had along the way while testing my system. This might prove useful for some readers to either avoid, or at least possibly allow them to circumvent obstacles to system overclocking.

When I first began testing with the MSI Motherboard, the P4 530 CPU, and the Crucial Ballistix, I was using a Gigabyte PCX5900 video card. The 5900 GPU was originally designed for AGP bus use, and for transition over to PCI Express, NVIDIA designed a HSI bridge chip that allows the GPU to interface with PCI Express bus. While the HSI bridge chip works wonderfully, and without any noticeable overhead or parasitic performance loss, there is a problem with running higher FSB (front side bus) speeds on newer 915P/G and 925X Intel Chipset socket 775 motherboards. As bus speeds approach 240-250(variable due to manufacturer and model) the PCI Express bus either experiences a failure in the frequency locks, or there are no PCI Express locks to begin with, and the bus gets too far out of specification for the HSI chip to cope with these speeds.

I experienced this first hand as the Gigabyte 5900PCX card would become unstable in 3D applications past 240MHz FSB. While I could run the computer in 2D at speeds up to 245MHz FSB, 3D applications would either lock or artifact severely. Attempting to boot computer at speeds above 246MHz FSB resulted in a "no post" situation that required me to clear cmos (bios) and reapply all bios settings at previous max stable. I was able to purchase a Radeon X700Pro locally, and after doing so I was awarded with the ability to run the computer up above the previous limits. Aside from being able to run the computer at higher bus speeds, I was rewarded with much higher 3D scores, although requiring me to scrap all data collected to that point.

Feeling somewhat relieved I was over that hurdle I was soon met with another issue in attaining a stable overclock. Up to 250MHz front side bus was attainable, however somewhat unstable. I had upped CPU vcore in various steps with some improvement, but full stability could not be attained. So far in testing I was using my venerable Enermax Model 465P 460W ATX power supply. This power supply has always treated me well in the past with stable voltage rails. In an effort to rule out power supply issues with this specific model I tried 2 other 500+W power supplies I had in my possession with similar instabilities at higher fsb levels.

Newer 915/925 chipset motherboards come with 24 Pin ATX headers versus the previous standard of 20 pin ATX headers. You can use a 20 Pin ATX plug power supply, but during extreme duress (especially when overclocking), voltages can begin to fluctuate quite a bit. Considering my CPU had reached the limits of air-cooling was a very real possibility, yet luck intervened. On my doorstep arrived a solution for attaining stability at this bus speed and even a bit more.

I was fortunate enough to come in contact with Edward Chang from Zippy Power Supply, and during our conversation at an area conference, he agreed to send a couple of PCI Express ready power supplies for testing. One of these units was the HP2-6460P 460W unit that is quite a beast if I may say so, with quite a hefty weight, one of my quick and dirty tests of power supply quality. This unit also supports a 24 pin ATX header plug, in addition to a plethora of other connectors.

After installation of this new power supply I was able to back the CPU vcore to default voltage and run the system stable up to 255 fsb stable with stock Retail cooler. Past this I was unable to gain stability on the system even with added vcore. I am suspecting the motherboard is now the limiting factor as I had progressively needed to bump the Northbridge voltage up in bios from a default level of 1.55v up to a somewhat scary 2.1v, which in turn generated ALOT of heat and warranted a fan placed on the normally passive Northbridge heatsink.

For the purposes of testing I used default rated voltage for the memory modules of 1.9v per Crucial, which is slightly above normal DDR2 specification of 1.8v. CPU was left at default vcore during all testing.

So after the major impediments to my overclocking efforts were resolved, I was able to start my data collection Again.