by Rajshree Chabukswar
Learn the optimal strategy for reducing power consumption with Intel SpeedStep® Technology.
As the industry moves toward mobile computing, resources such as battery life become significantly more important for system usability. This paper analyzes the power consumption of different DVD playback software, and provides recommendations for reducing power consumption. For this study, various DVD playback software was analyzed and run with the multiple configurations available. We provide our recommendation on an optimal strategy for reducing power consumption while playing content from a DVD-ROM.
Three pieces of DVD playback software are studied here, in order to understand the effects of power consumption; in one case, with the maximum power-saving mode, and in the other, with the no power-saving mode. The workload used for the analysis is a standard definition of DVD content, shipped with the MobileMark* 2005 benchmarking tool.
DVD Power Profile
Before looking into various case studies, let us look at the DVD power profile while idle, during read time, and during spin-up.
As indicated in the figure above, DVD spin-up causes the most power consumption. Therefore, reducing spin-up instances will save power. Also, continuous read from the DVD consumes almost 2.5W of power. During an idle state, power consumption is the lowest.
DVD Playback Software: Tool #1
This particular software offers three modes for mobile platforms. These modes save more power (compared to running the application with the no power-saving mode active).
Here are the different modes offered when running the DVD playback:
- No power-saving mode: Data is read as it is processed. The DVD is accessed 100% of the time during the playback.
- Power-saving mode: Data is buffered to reduce DVD spin-ups and spin-downs. The buffered data is processed over certain durations of time. The power-saving modes are divided into three modes, based on the amount of total data being buffered with each DVD spin-up. For this study, we focus on the highest-level power-saving mode: aggressive/max power-saving.
The instantaneous DVD power-consumption curves for these modes are as shown below. The data in the first chart has been collected while using the no power-saving mode. Since the data is not buffered, the DVD data is continuously read throughout the length of the video, causing high power-consumption during the entire run time.
In the figure below, data buffering occurs. This causes a reduction in DVD read accesses, which in turn saves power. ‘Aggressive mode’ buffers the most data, as compared to the other two modes.
To verify that the playback software is indeed buffering data by doing read-aheads, an experiment was conducted to catch all ‘ReadFile()’ calls from the DVD Playback Software Tool #1. The log indicated that there are no calls to ReadFile() when DVD activity is low. This time is equal to the processing time (with DVD power close to 0W), as shown on the previous graphs.
The overall profile of DVD and platform power consumption with various modes for the MobileMark workload is as shown below:
|DVD Energy||CPU Energy||Platform Energy|
|No Power Saving||869.84||663.92||6618.76|
|Aggresive/Max Power Saving||263.41||762.99||6039.41|
* Energy in mWHr
As indicated, buffering the data (aggressive, average and conservative) to reduce DVD spin-ups and spin-downs saves a significant amount of DVD power, as compared to not buffering the data (no power-saving). E.g., ~70% of DVD power is saved in the aggressive (most power-saving) mode, as compared to the no power-saving mode. This translates into ~10% platform-level power-saving.
DVD Playback Software: Tool #2
The second case study discusses another piece of industry leading DVD playback software. It offers special mobile features to enhance battery life for DVD playback on mobile platforms. Of the three modes offered by this software, the following two are of the most interest to this study.
- Maximum Performance/No Power-Saving Mode: The processor runs at maximum available speed, irrespective of the system power scheme.
- Maximum Power-saving Mode: The processor runs at optimal speed, based on the power policy selected.
Analysis done with this software indicated that unlike DVD Playback Software Tool #1, it doesn’t buffer/read data during media playback; hence, DVD power consumption is almost the same for all three modes on a 20-minute workload. Furthermore, the root-cause indicated that the operating frequency of the processor changes with the maximum performance/no power-saving mode. The software accomplishes this by changing the system power scheme to run with the maximum frequency available, and restores the original power scheme after the run is completed.
Since the data is not buffered, the DVD power-consumption profile for both modes looks essentially the same. Both modes read data from the DVD drive continuously, causing high power consumption. The figure below shows instantaneous DVD power consumption for the maximum performance mode (no power-saving) and the maximum power-saving mode.
Since the analysis indicated that the operating frequency of the processor changes when running in the maximum performance mode, as compared to the maximum power-saving mode, the instantaneous CPU power profile changes in these modes.
The figure below indicates instantaneous CPU power consumption in the maximum performance (no power-saving) mode, which peaks up to ~24W in a few instances.
Whereas the figure below indicates instantaneous CPU power consumption on maximum power-saving mode, which peaks up to ~7W:
Since overall CPU utilization for DVD playback using these modes is around 10-15%, running at full frequency may not be needed even in the case of maximum performance (no power-saving) mode. If maximum operating frequency is needed, Intel SpeedStep® Technology will automatically change it as demand increases. This also ensures that the user experience is not affected when a high-CPU frequency is needed. By having the software change the operating frequency, CPU power consumption increases by ~25% in the maximum performance (no power-saving) mode.
The overall profile of DVD, Platform and CPU power consumption is as shown below.
|DVD Energy||Platform Energy||CPU Energy|
|Max Perf/No Power Saving||897.82||10143.56||3329.53|
* Energy in mWHr
DVD Playback Software: Tool #3
The third study describes the analysis done on one of the most commonly used applications for DVD playback. It offers two modes: battery optimized performance-enabled (maximum power-saving) or disabled (no power-saving).
This software does not buffer the data in the maximum power-saving mode, but the operating frequency does not change, either, as in the case of DVD Playback Software Tool #2.
The instantaneous power consumption for the DVD and CPU for battery optimized performance-enabled (maximum power-saving) vs. disabled (no power-saving) mode is as shown below.
Since data is not buffered in a battery optimized performance-enabled (maximum power-saving) mode, the instantaneous DVD power consumption profile for both modes demonstrates similar characteristics, as shown in the figure below.
The instantaneous CPU power-consumption profile for these modes demonstrates a slight difference. The figure below indicates CPU power with the battery optimized performance disabled (no power-saving) mode. Note how most spikes are near the 5W line:
The figure below indicates the results with battery optimized performance-enabled (maximum power-saving mode). Note how most spikes are near 2-3W line:
This difference results in lower CPU and platform power consumption with battery optimized performance enabled vs. disabled.
The overall CPU power consumption is slightly higher with battery optimized performance disabled (no power-saving), vs. with enabled (maximum power-saving), causing ~22% CPU power-saving with power-optimized performance enabled. This results into ~5% of platform-level power saving.
|DVD Energy||CPU Energy||Platform Energy|
|Power Optimized Perf Enabled||780.93||703.25||6202.04|
|Power Optimized Perf Disabled||781.04||554.87||6023.57|
* Energy in mWHr
This paper demonstrates DVD, CPU, and platform power consumption comparisons with the three titles of DVD playback software discussed in the previous pages.
The figure below indicates the comparison of DVD energy. Since DVD Playback Software Tool #1 performs data buffering in the maximum power-saving mode, it shows significantly lower power consumption. DVD Playback Software Tool #2 and DVD Playback Software Tool #3 show no differences.
The figure below indicates the comparison of CPU energy. Since DVD Playback Software Tool #2 changes its operating frequency to maximum in maximum-performance (no power-saving) mode, it consumes significantly more energy:
The studies shown above indicate that the technique of buffering implemented by DVD Playback Software Tool #1 reduces DVD power consumption by 70% and overall platform power consumption by 10%, as compared to other techniques.
It is always recommended to reduce DVD spin-up, spin-downs, and read accesses in order to save power.
Also, it is not recommended to change the power scheme to run the processor at the highest available frequency in the software. Since Intel SpeedStep Technology will automatically change the operating frequency as the processing demand increases, it will bump up the frequency as needed.
About the Author
Rajshree Chabukswar is an Application Engineer in Intel's Mobile Application Enabling group, working on client enabling. Prior to working at Intel, she obtained an MS in Computer Engineering at Syracuse University.