Assembly And Overclocking
Our build had a number of little quirks we can talk about, although we never encountered any show stoppers.
The first thing we noticed is that our DFI X38-T2R motherboard came with a hefty chipset heatsink that would interfere with the large Xigmatek HDT-S1283 cooler if it were situated to blow the hot air out of the rear of the case.
Luckily, the Antec Three Hundred chassis has an exhaust fan on top, allowing us to place the cooler in a manner that doesn’t interfere with the heatsink. (On a side note, we tested DFI’s X38-T2RB version of this board in our last System Builder Marathon. While it seems to be based on the same PCB, it has a different cooling setup that links the different parts of the motherboard with a heat pipe. The T2R version we tested has separate coolers for the chipsets. Both the T2R and T2RB are the same price on NewEgg.)
Our second issue was a worrisome failure to display video on boot. The DFI LanParty X38 curiously displayed the "FF" code—which stands for "fully functional." Some quick experimentation showed that the system booted fine with another graphics card, suggesting an incompatibility with the Radeon HD 4870 X2. However, after a flash to the newest BIOS, the 4870 X2 worked just fine.
Speaking of the BIOS, the DFI X38-T2R frustratingly refused to acknowledge our E8500’s multiplier of 9.5, instead defaulting to the lowest multiplier of 6. After a little poking around in the BIOS, we found we had to manually enable the CPU N/2 Ratio setting. We thought it was a little troubling that the newest BIOS didn’t automatically choose the right settings for the E8500 when left to its own devices.
Our final concern was the temperature of the PowerColor Radeon 4870 X2, which the Catalyst Control Center reported as peaking very close to 100 degrees Celsius under load. Happily, AMD recently enabled fan speed controls in the driver, and we chose to crank it up a notch and live with a little more fan noise rather than allow the GPU temperatures to raise over 60 degrees Celsius when stressed.
After dealing with these issues, we proceeded to run the benchmarks with total stability. Then, with a 45 nm CPU to play with, we proceeded to the overclocking options.
Overclocking
It turns out we didn’t have to worry about the 650 W power supply at all when it came to CPU overclocking, but we did end up having a limiting factor.
After raising all of the usual voltage suspects, increasing memory latency to 5, and setting the CPU voltage to 1.4 V, we brought the front side bus (FSB) up to 453 MHz for a CPU clock speed of 4,303 MHz. The system booted into Windows just fine, but before pushing it harder we decided to run a test to see how stable it was at this clock.
While the Orthos stress tester didn’t demonstrate any instability, we were surprised to see the temperature skyrocket from a reasonable 45 degree Celsius in idle to a hot 85 degrees Celsius ! We quickly shut the test down with the realization that this CPU might happily make higher clocks with liquid cooling, but with an air setup we’d have to do some more tweaking and settle for something a little less.
By reducing the core voltage to 1.3 V and lowering the clock speed to 4.25 GHz, we managed to get an incredibly stable 35% overclock that would max out temperatures at about 65 degrees Celsius in an Orthos run. Satisfied with both the speed and temperature of the CPU, we settled on the 1+ GHz CPU overclock and moved on to the graphics card.
The 4870 X2 card wasn’t as willing as our E8500 CPU was to push the overclocking envelope—the maximum clocks in AMD’s Overdrive utility passed the utility’s stability test at 800 MHz core and 1,000 MHz memory speeds, but actual game play would result in artifacts. We settled for turning the 4870 X2’s fan speed a little higher while pulling back the clock speeds to 777/950 from the stock clocks of 750/900.
With our overclock stable and good to go, we proceeded to run the benchmarks !
