facebook rss twitter

Review: DFI NB80-EA Granite Bay - Overclocking

by David Ross on 11 December 2002, 00:00

Tags: DFI (TPE:2397)

Quick Link: HEXUS.net/qao2

Add to My Vault: x

Test Setup

Examining the bios more closely shows that there is no option to adjust the voltage of either the AGP socket or the RAM. This is a shame as it could limit our ability to overclock the board. However both the AGP and PCI clocks can be locked which should limit the need for additional AGP voltage.

The board is designed to run with DDR2100 and we were delighted to find that Corsair XMS 3200 and 3500 run without problem. As the board accepts higher specification memory the need to increase the voltage should not be required to achieve some high FSB speeds. There are no DDR ratio adjustments provided so the RAM speed will be locked at 2xFSB, lets hope that Dual DDR will be worth the wait.

In the Vapochil the CPU temperature even at 1.90v true and a 160FSB is well under control at 12c. Remember that the bios temperatures normally display the equivalent of mid load windows temperatures, therefore 12c is perfectly within the expected range.

Having sorted the CPU out attention turned to the RAM. 2 sticks of XMS 3500 CAS2 were used in the first bank of two memory slots enabling dual band bandwidth. Clearly DDR3500 RAM is complete overkill for a board that is designed to run with DDR2100 and we don't expect to push the RAM anywhere near its limits. So the most aggressive RAM timings possible were used. It should be perfectly possible to re-create our results using good quality DDR2100 or DDR2700 memory modules.

Experience has shown that the boot up process puts a great deal of strain onto the memory and CPU. We were delighted to find a bundled software FSB adjustment program supplied with the board. This enabled the system to be booted at a low FSB, we chose 160Mhz to give a balance between boot speed and reliability. The FSB was then pushed further once windows was loaded. It is worthy to note that we did not experience any difference in the performance tests between setting a high FSB in the Bios and setting it in windows.

FSB speeds of up to 176Mhz were achieved while in windows, pushing the P42.4b to 3,168Mhz. Although it was possible to run some benchmarks at this speed they were somewhat unreliable. The optimum speed we found was 172Mhz FSB (3,096Mhz CPU) where the system was rock stable. This limitation is more than likely caused by the CPU not the motherboard.

Temperatures at 172Mhz FSB were well controlled by the Vapochil. The CPU temp at idle sat around 8.5c and rose to 24c under prolonged full load. The Vapochil chiller unit showed -13c at idle rising to -5 under sustained load. Both temperatures meant that the computer had temperature well under control.