Power consumption and overclocking
Power consumption - idle | |
---|---|
Intel DH55TC (H55) | ASUS P7H57D-V EVO (H57) |
32 | 49 |
We'd expect to see the biggest difference in power draw when the system is idle, as that’s when the various power saving technologies such as ASUS EPU have a chance to shine.
Indeed there is a big difference, but surprisingly one in favour of Intel's DH55TC. In our tests the ASUS board when idle consumes a whopping 53 per cent more power than the Intel.
This is most likely explained not only by the board size difference, but also by the several additional ICs which ASUS has onboard. Each of these will consume some power, even when idle, and by our count there are the following six additional ICs on the ASUS board which the Intel board doesn't have: Marvell 2-port SATA 6Gbps controller, Marvell eSATA+PATA controller, NEC USB 3.0 controller, VIA FireWire controller, ASUS TurboV, and ASUS EPU.
Power consumption - load (2D) | |
---|---|
Intel DH55TC (H55) | ASUS P7H57D-V EVO (H57) |
73 | 82 |
Contrary to expectations once more, the P7H57D-V EVO is able to claw back a large percentage of the power burden it suffers over the DH55TC.
The ASUS's 2D load power consumption whilst running wPrime 1,024M, is just circa 12 per cent higher than the Intel. Like many PSUs, it seems the ASUS EPU technology, in conjunction with T.Probe and Hybrid Phase, work at their most efficient level when closer to 50 per cent of their total power capacity.
Power consumption - load (3D) | |
---|---|
Intel DH55TC (H55) | ASUS P7H57D-V EVO (H57) |
133 | 149 |
When it comes to 3D power consumption, the tiny amount of power drawn by an IGP is unlikely to demonstrate any large gains in power efficiency due to VRM design improvements on the motherboard.
Therefore we now run the 3D load power consumption figures with the Radeon HD 5770 installed as used in some of the 3D Benchmarks. Even so the ASUS P7H57D-V EVO is unable to improve its power efficiency when compared to the Intel DH55TC and the difference remains at circa 12 per cent higher.
Overclocked performance
We know from our Westmere performance analysis that the limit of the Core i5 661 used during testing this motherboard is 4.26GHz.
Therefore in order to avoid introducing any CPU-related instability, any overclocked performance numbers we run should use settings whereby the CPU can be kept at around 10 per cent below this frequency.
The motherboard claims to support memory speeds of up to 2,133MHz when using a Lynnfield CPU and provides built-in overclocked memory profiles for this. However, when using a Clarkdale CPU such as the Core i5 661 used during our testing, the highest memory speed supported is just 1,600MHz, hence our decision to run a BCLK of 160MHz, and the maximum supported memory multiplier of 12x (listed as DDR3-1333 in the BIOS).
In order to avoid any artificial performance constraints being introduced, we keep the QPI multiplier at its maximum. This results in around a 20 per cent overclock and an effective QPI frequency of 7.71GHz.
When overclocking the Core i5 661, and utilising the IGP instead of a discrete GPU, consideration needs to be made for the IGP frequency as this is directly related to the BCLK frequency. In order to avoid any instability the IGP's frequency multiplier was reduced to 5.75x (listed as 767MHz in the BIOS), which resulted in an actual IGP frequency of circa 920MHz.
The voltages were left on auto, but due to utilising the X.M.P. profile in the BIOS, the IMC and DRAM voltages get modified to 1.15V and 1.65V, respectively. By default the X.M.P. profile doesn't overclock the CPU; therefore the CPU ratio was manually adjusted to 24x resulting in a CPU frequency of 3.85GHz.
Therefore to summarise:
CPU Clock: Up from 3.47GHz to 3.85GHz
(+10.9 per cent increase)
QPI Clock: Up from 6.42GHz to 7.71GHz
(+20.1 per cent increase)
MEM Clock: Up from 1,338MHz to
1,607MHz (+20.1 per cent
increase)
BCLK: Up from 133.8MHz to 160.7MHz (+20.1 per cent
increase)
IGP: Up from 900MHz to 920MHz (+2.2 per cent
increase)
As mentioned on page 5, all of the Intel boards
overclocking/over-voltage options with the exception of the BCLK
adjustment were inaccessible.
As such we were unable to run overclocked performance numbers.
Pifast is the only single-threaded benchmark we ran with overclocked
settings. The time taken to run decreased by 3 seconds, down to 23
seconds, an 11.5 per cent improvement.
Staxrip is the only heavily multi-threaded benchmark we ran with
overclocked settings. The time taken to run Pass 1 decreased by an
impressive 32 seconds, down to 254 seconds, an 11.1 per cent improvement.
The time taken to run Pass 2 decreased by 27
seconds, down to 217 seconds, a 12.4 per cent improvement.
Quake Wars: Enemy Territory is the only game we ran with overclocked
settings, and is usually very sensitive to memory bandwidth
improvements.
In order to avoid artificially constraining any performance increase
from the overclock, the Radeon HD 5770 was installed and utilised when
running these numbers.
The framerate increased by 25.87 fps, to 167 fps, an 18 per cent improvement.
Overclocking
Sticking with our tried-and-tested formula of manually increasing the
voltages by 10 per cent first, we proceeded to probe the BCLK limits of the
P7H57D-V EVO.
At first we left the QPI frequency on Auto however the BCLK quickly
topped-out at 188MHz.
It was discovered that the QPI multiplier wasn't automatically
decreasing in order to keep the QPI frequency close to nominal.
Therefore the QPI multiplier was subsequently reduced to the minimum of
24x (listed as 3200MHz in the BIOS).
After this, the maximum BCLK we were then able to achieve was 214MHz.