Overclocking With Gigabyte Z77X Motherboards
Overclocking
It’s one thing to compare scores in our standard spate of benchmarks, but it’s quite another to see how these mainboards can overclock; after all, overclocking is really where the rubber meets the road.
We tested all three boards using the same components and chassis and at the same ambient room temperature. We let all the case fans run, but we also used a mainstream liquid cooler on our CPU. We adjusted all settings using the UEFI BIOS instead of relying on any Windows-based overclocking software.
Corsair H55 Hydro liquid cooler
To gain the most control over what our systems were doing, we turned Turbo mode off, switched all CPU voltages from auto to manual (even when we didn’t alter the default voltages), and changed the Vcore loadline calibration to Extreme to accommodate for vDroop.
After making our adjustments, we booted into Windows, let the system idle for a few minutes, and then used OCCT to both stress the CPU and monitor temperatures and voltages. Also note that we set OCCT to abort once our CPU reached 80 degrees C to prevent any potential damage. The goal was to see how well the boards handled the overclocked workload with mainstream cooling, to set a realistic expectation for any of you reading this that may want to try overclocking yourself, without using exotic cooling.
In regard to changing various settings, we generally left the base clock alone. After doing some tweaking, we found that raising it even a little bit, while already pushing to CPU, was generally a waste of time; our systems remained more stable if we left the base clock at 100MHz than if we tried to find a balance between upping the base clock and changing the multiplier. For example, the system could handle our stress test for several minutes at 4.7GHz (100MHz base x 47 multiplier, 1.23V) before hitting our preset heat ceiling and cancelling the test; at 4.64GHz (101MHz base x 46 multiplier, 1.23V), the system actually BSODed just seconds into the test. Thus, even though the overall clock speed was lower, the higher base clock made the system less stable. This is common with Sandy and Ivy Bridge chips, by the way.
Ultimately, we focused only on the CPU multiplier and voltage. From there, it was a simple matter of finding the sweet spot between feeding the CPU the necessary amount of voltage for the clock without pushing the temps too high.
So how did our boards fare?
The UP5 TH was nice and stable at 4.6GHz, but it just couldn’t punch past the 4.7GHz mark without overheating. In the end, this board liked a base clock of 100MHz and a multiplier of 46 with a core voltage of 1.23V.
We found the overclocking headroom of the UD4H to be nearly identical to that of the UP5 TH; 4.6GHz was the magic number, and we reached it by upping the multiplier to x46 and leaving the base clock at 100MHz. The one difference is that we reached a stable 4.6GHz on the UP5 TH with 1.2V instead of 1.23V. It looked like we were going to have some success with a slightly raised base clock for a few minutes there, but the system just couldn’t handle the heat and coughed up a hairball.
Although on paper the UP7 looked like it would be the toughest board in the bunch, it couldn’t quite push past the 4.6GHz mark, either; however, it did manage to achieve stability at that clock speed at just 1.18V, which was a bit batter than both the UP5 TH and UD4H. Keep in mind, to maintain stability while overclocked, it's not just a higher voltage that helps, but clean, smooth power delivery as well. And the UP7's beefier power array seems to have paid off here.
At the end of the day, then, all three mainboards achieved the same 4.6GHz overclock, but at three different voltages. The UP7 was the winner by a hair.
There is one additional bit of information to note: Knowing that our 80-degree ceiling wasn’t exactly an accurate measurement of how far we could push these systems--just how far we could do it safely--we ditched OCCT once or twice and let each of our systems fly on a load of Prime95 for several minutes (using Core Temp to monitor temperatures). Under those circumstances, we managed 4.8GHz on our boards; the systems appeared stable, but we didn’t let the test run long to prevent running the CPU for an extended period with higher than stock voltages, frequencies, and temps.
We tested all three boards using the same components and chassis and at the same ambient room temperature. We let all the case fans run, but we also used a mainstream liquid cooler on our CPU. We adjusted all settings using the UEFI BIOS instead of relying on any Windows-based overclocking software.
;)
Corsair H55 Hydro liquid cooler
To gain the most control over what our systems were doing, we turned Turbo mode off, switched all CPU voltages from auto to manual (even when we didn’t alter the default voltages), and changed the Vcore loadline calibration to Extreme to accommodate for vDroop.
After making our adjustments, we booted into Windows, let the system idle for a few minutes, and then used OCCT to both stress the CPU and monitor temperatures and voltages. Also note that we set OCCT to abort once our CPU reached 80 degrees C to prevent any potential damage. The goal was to see how well the boards handled the overclocked workload with mainstream cooling, to set a realistic expectation for any of you reading this that may want to try overclocking yourself, without using exotic cooling.
In regard to changing various settings, we generally left the base clock alone. After doing some tweaking, we found that raising it even a little bit, while already pushing to CPU, was generally a waste of time; our systems remained more stable if we left the base clock at 100MHz than if we tried to find a balance between upping the base clock and changing the multiplier. For example, the system could handle our stress test for several minutes at 4.7GHz (100MHz base x 47 multiplier, 1.23V) before hitting our preset heat ceiling and cancelling the test; at 4.64GHz (101MHz base x 46 multiplier, 1.23V), the system actually BSODed just seconds into the test. Thus, even though the overall clock speed was lower, the higher base clock made the system less stable. This is common with Sandy and Ivy Bridge chips, by the way.
Ultimately, we focused only on the CPU multiplier and voltage. From there, it was a simple matter of finding the sweet spot between feeding the CPU the necessary amount of voltage for the clock without pushing the temps too high.
So how did our boards fare?
;)
The UP5 TH was nice and stable at 4.6GHz, but it just couldn’t punch past the 4.7GHz mark without overheating. In the end, this board liked a base clock of 100MHz and a multiplier of 46 with a core voltage of 1.23V.
We found the overclocking headroom of the UD4H to be nearly identical to that of the UP5 TH; 4.6GHz was the magic number, and we reached it by upping the multiplier to x46 and leaving the base clock at 100MHz. The one difference is that we reached a stable 4.6GHz on the UP5 TH with 1.2V instead of 1.23V. It looked like we were going to have some success with a slightly raised base clock for a few minutes there, but the system just couldn’t handle the heat and coughed up a hairball.
Although on paper the UP7 looked like it would be the toughest board in the bunch, it couldn’t quite push past the 4.6GHz mark, either; however, it did manage to achieve stability at that clock speed at just 1.18V, which was a bit batter than both the UP5 TH and UD4H. Keep in mind, to maintain stability while overclocked, it's not just a higher voltage that helps, but clean, smooth power delivery as well. And the UP7's beefier power array seems to have paid off here.
At the end of the day, then, all three mainboards achieved the same 4.6GHz overclock, but at three different voltages. The UP7 was the winner by a hair.
There is one additional bit of information to note: Knowing that our 80-degree ceiling wasn’t exactly an accurate measurement of how far we could push these systems--just how far we could do it safely--we ditched OCCT once or twice and let each of our systems fly on a load of Prime95 for several minutes (using Core Temp to monitor temperatures). Under those circumstances, we managed 4.8GHz on our boards; the systems appeared stable, but we didn’t let the test run long to prevent running the CPU for an extended period with higher than stock voltages, frequencies, and temps.
