|MSI 990FXA-GD80 AM3+ Motherboard|
|Reviews - Featured Reviews: Motherboards|
|Written by David Ramsey|
|Wednesday, 19 October 2011|
Page 6 of 17
Motherboard Testing Methodology
AMD got their 9-series chipsets out months before the Bulldozer processors that were their reason for existing, so vendors have had some time to tweak things...for Phenom II and lower processors, that is. Like most other vendors, MSI introduced a BIOS ugprade when the Bulldozer chips actually became available, and I updated the board before beginning any tests.
For comparison purposes I used an ASUS Crosshair V Formula motherboard (supplied by AMD with their review FX-8150 Bulldozer CPUs), as well as an ASUS P8Z68-V Pro motherboard running an Intel Core i5 2500K processor. I also threw in a previous-generation AMD Phenom II X6 1100T. All test systems used the same memory, video card, and hard drive. The memory frequency was set to the maximum officially supported by each system: 1333MHz for the Intel system and 1866MHz for the two AMD systems.
For overclocking, I simply increased the FX-8150 base multiplier and tweaked the voltage until I reached the highest setting at which the system could complete the SPECapc Lightwave benchmark. I've found this to be the best predictor of system stability: overclocks that will run under other stress tests for hours will crash in minutes under SPECapc Lightwave. In my test of the FX-8150 CPU, I only overclocked the ASUS motherboard, since that was what AMD had supplied. I was able to reach a stable 4.8GHz, but with the MSI board I could only get to 4.6GHz. 4.7GHz would run for quite a while before crashing, and 4.8 frequently wouldn't complete a boot into Windows 7, even when I raised the CPU voltage to 1.5 volts. This is a little surprising since pure multiplier overclocking isolates most of the other board components (i.e. it doesn't depend on the chipset or RAM or anything else to be able to overclock as well). I suspect the problem might be in the power circuitry: although the MSI board does have an 8-phase power system, it lacks the fine adjustments (load-line calibration and such) that the ASUS board has. The ASUS board also has an extra 4-pin EPS-12V connector in addition to its 8-pin connector, both of which I used in my ASUS overclocking tests and which might have afforded it an advantage.
So here's the test lineup followed in all the charts:
Intel Z68 Test Platform
AMD 990FX Test Platforms