Copyright by William Lloyd Bircher 2010 - The Laboratory for ...

Power (Watts)
30
25
20
15
10
5
0
Compute
Intensive
Graphics
Intensive
Compute
Intensive
Productivity
0.9
2.9
0.9
2.7 3.3 3.3 3.3 3.2 2.9 2.9 2.6 2.9 2.8 2.6 2.8
1.4 1.4 1.4 1.4 0.9 0.9
1.3 1.3 1.4 1.3 1.7 1.4 1.6
4.0 3.8 3.8 3.7 10.2 10.6 10.2
14.8
10.5
14.1
14.3 14.2
11.9 11.1 11.0
3.5
10.6 10.3
3.0 3.3 2.9
3.4
2.5
3.0
2.4
2.5 2.4 2.4 2.3
1.6
2.0
1.3
1.6
2.0
1.3
1.3
1.9
1.1
1.7
2.0
1.1
1.1
1.9
1.1
1.4
2.0
1.1
1.2
1.9
1.1
Figure 4.1 Average Power Consumption (Watts)
The remaining subsystems have little variation from workload to workload. For the disk
subsystem this can be explained by two factors. First, most workloads used in this study
contain little disk access with the exception of DBT-2. For most others, the initial
loading of the working set is the majority of the disk access. Using synthetic workloads
targeted at increasing disk utilization, less than a 3% increase in average disk power can
be achieved compared to idle. This is due to the second factor which is a lack of disk
power management. Modern hard disks often have the ability to save power during low
utilization through low power states and variable speed spindles. However, the server
disks used in this study do not make use of these power saving modes. Therefore, disk
power consumption is dominated by the power required for platter rotation which can
account for almost 80% of max power [ZeSo03]. For I/O and chipset subsystems, little
workload to workload variation is observable. In the case of the chipset, offset errors due
45
CPU
Memory
Memory
Controller
GPU
Disk
0.6
0.2
0.5
0.9
0.8
0.5