facebook rss twitter

Review: nForce3 250 Chipset Overclocking

by David Ross on 20 April 2004, 00:00

Tags: Gigabyte (TPE:2376), AMD (NYSE:AMD), NVIDIA (NASDAQ:NVDA)

Quick Link: HEXUS.net/qaxp

Add to My Vault: x

Testing Notes

AGP and PCI Locking

Our first task was to answer the big questio: does the nforce 3 250 chipset lock both the AGP and PCI Bus? In short, the answer is yes.

As described earlier the BIOS only enabled us to set the AGP bus speed, there was no mention of PCI. As the SATA RAID is now controlled by the chipset I was unsure whether it had any relationship to the PCI bus. I therefore needed to try and check the PCI bus speed using another method. Two tests came to mind. Firstly Gigabyte's Easytune4 software, a Windows-based application that enables changing front side bus speed, Vcore, DDR bus speed, AGP bus speed and PCI bus speed. Secondly we changed the graphics to an old Matrox PCI card which, from previous experience, hates any increase in bus speed. Once the system was booted with an increased front side bus speed, the Gigabyte software confirmed that both the AGP and PCI bus speeds were locked at 66MHz and 33MHz respectively. The PCI graphics card was also quite happy at all the front side bus speeds we tried, thus proving that the PCI bus is indeed locked.

PCI fix

With the reassurance of a locked PCI bus, the next task was to set about seeing just what the AMD Athlon 64, coupled with the Nforce 3 250 chipset, can do. Overclocking is about taking things one step at a time and gradually building a picture of peak performance. I started by seeing just how far I could take the processor's driven clock, with the processor running at stock voltage. Then the fun really began as I ramped up the voltage and got ready to toast my toes on some hot heatsinks.

Cooling and Temperature

I found the standard AMD retail heatsink to be adequate. The test processor seemed very happy chugging along at temperatures of up to 60°C. As soon as the temperature exceeded 60°C it became very unstable. Reducing the temperature for a given overclock made the processor more stable thus emphasising the need for good cooling. For example when running at 2300MHz with the AMD heatsink, Pifast suffered numerous errors, however simply reducing the temperature using either the Zalman or Vapochill enabled the test to run faultlessly time after time. Despite this, it was impossible to achieve better overclocking results using the Vapochill than was possible when using the air cooled Zalman. Clearly if more voltage was available and I was brave enough to use it, the Vapochill might be needed, but with this particular motherboard Freon cooling is wasted and silent air cooling perfectly satisfactory to achieve very respectable overclocking results.

Voltage

The test processor produced a very respectable overclock at stock voltage; it was perfectly stable at speeds of up to 2250MHz. The test AMD Athlon 64 seems to thrive on voltage. The Gigabyte GA-K8NS motherboard was able to provide a very respectable 1.7V to the overworked processor. At this voltage temperatures proved just too much for the AMD heatsink, but the Zalman and Vapochill were able to take it in their stride. Unfortunately even 1.7V was not enough juice to enable the magic 250MHz front side bus with any stability; it is our educated guess that even more voltage would be needed.