Overclocking Results
For this roundup I decided to test maximum stable bclk for Super Pi 1M, Super Pi 32M, and then a maximum stable frequency through 5 runs of Intel Burn Test. I chose these settings with the idea that SPi 1M stability indicates maximum suicide base-clock frequencies, Super Pi 32M indicates a rough measure of strenuous base-clock stability, and the Intel Burn Test indicates a rough measure of what the board is capable of for 24/7 operation. Below are the settings used for the various tests.
For all the boards I would set the voltage in the BIOS, check it with a multi-meter while running Super Pi, and then adjust accordingly so each board was feeding the CPU the same core and vtt voltages. Voltages used were 1.35vCPU, 1.30VTT, 1.8vPLL, 1.65vDIMM with the CPU running the 9x multiplier, 16x QPI multiplier, and the 2:6 memory divider. For the Intel Burn Test voltages were left the same while the CPU multiplier was bumped to 22x.
The ECS P55H-A was the first board on the bench table. By far the cheapest of the four boards, I had low goals for this board but it proved to be more than capable of hanging with the crowd.
Locking in at 207MHz base-clock was more than acceptable and would be sufficient to push any LGA1156 CPU to their thermal limitations on an air heatsink.
Super Pi 32M is a bit more strenuous and as such the CPU had to drop down to a base clock of 204MHz. This is nothing to scoff at; the P55H-A definitely held its own for the price.
Intel Burn Test is one harsh test and looping it 5 times plays hell with the boards' thermal tolerances and CPU stability. I had to compromise at 3950MHz with the highest core temperature experienced at 83 Celsius in order to pass all 5 loops but the board could pass 3 of 5 loops at 4070MHz.
With a premium price you would expect premium performance from the EVGA P55 FTW SLI and it delivers without a hitch. This CPU has a documented maximum base clock on air cooling of around 217MHz and the P55 FTW SLI gets close at 214MHz.
Strangely enough Super Pi 32M stability was identical to the Super Pi 1M stability. However, even the slight jump to 214.5MHz would cause instant instability. I'm chalking that up to the weak CPU. 214MHz through 32M means 210MHz should be solid for 24/7 stability.
I said Intel Burn Test is a harsh test and it really shows with the EVGA P55 FTW. While the board was able to keep the CPU stable through 5 loops at 4095MHz it came at a cost; 88 Celsius was the highest temperature recorded through the run. Strangely enough, the next jump to 4116MHz wouldn't pass a single loop while core temps hit 91 Celsius. Looks like we are thermally limited without adding more voltage to stabilize the chip.
The Gigabyte GA-P55A-UD6 hits the same base clock wall as the eVGA P55 FTW at 214MHz. Just like the other board, 214.5MHz resulted in an instant crash indicating the CPU was not willing to go a MHz faster.
No surprise here, the Gigaybte GA-P55A-UD6 holds 214MHz base clock through Super Pi 32M also. The board is definitely capable of more, perhaps the upcoming 32nm LGA1156 CPUs will unleash the speed within.
For 24/7 stability the Gigabyte trails the eVGA by 20MHz while also hitting 88 Celsius through Intel Burn Test. The next jump to 195MHz base clock would fail 1 of the 5 loops through ITB while hitting 90 Celsius so we are once again thermally limited.
The Gigabyte GA-P55A-UD6 is the first board of the roundup that features an auto overclocking feature. Located within the BIOS is an option labeled CIA2. With CIA2 enabled to "Full Thrust" it yielded a modest 19% overclock. This automated overclock is designed to be stable regardless of CPU quality, and you can see it is pumping 1.36v through the CPU where it only needs roughly 1.22v to pass Intel Burn Test at those speeds. I would suggest manually overclocking to save yourself some electricity and lower the stress on your components.
Surprisingly, the MSI P55-GD65 was the only board to break the 214MHz base clock wall with this CPU. Sure, it was only 214.5MHz, but 214.5MHz is faster than 214.0MHz. This great break, however, didn't hold over so well for Super Pi 32M.
Given how it cracked 214MHz through Super Pi 1M, I was expecting the P55-GD65 to hold out nicely in Super Pi 32M also. Just like the ECS P55H-A though, the MSI P55-GD65 fell a few MHz short for Super Pi 32M, dropping to 211.5MHz.
Even though the P55-GD65 dropped a bit with the Super Pi 32M test it led the pack with the Intel Burn Test at 4105MHz while hitting a record 89 Celsius. While only 10MHz higher than the eVGA P55 SLI FTW, the important fact is the P55-GD65 was stable through all 5 loops at 4.1GHz while the P55 FTW couldn't pass a single loop when trying to make the jump to 4.1GHz.
I mentioned the OC Genie prior when discussing the P55-GD65, and here are the results. Leaving all settings at default in BIOS and then hitting the OC Genie button caused the system to boot at the above settings. It definitely feeds more volts to the CPU than necessary but hitting 189.5MHz base clock on an "auto" setting is pretty impressive.
So all that is quite a lot to handle. To make things easier there is a handy chart quickly showing you the max base clock and 24/7 frequencies attained by each board. Going strictly by the chart you would think the MSI P55-GD65 was the outright winner but that would be far from the truth. The MSI P55-GD65 may have the highest 1M base clock and 24/7 frequency, however the EVGA P55 FTW was the easiest to overclock. For these tests I was trying to duplicate the same voltage setting across all the boards and the EVGA P55 FTW was the closest in voltages set in BIOS to voltages experienced. The MSI P55-GD65 would over-volt by .04v to .06v under heavy load forcing me to compensate. The EVGA and Gigabyte boards were spot on through the Super Pi and ITB loading while the ECS P55H-A was a nightmare to work with. Voltage regulation was rough on the ECS P55H-A with the "stable" voltage jumping 0.02 to 0.03v in either direction through the ITB testing which might be responsible for its lower than average performance.
