The advancement of the Personal Computer (PC) over the last two and a half decades has enabled people to work and play in ways that were previously unimaginable.

Today’s PC buyers continue to seek out exciting new applications that require continued increases in performance and capability of the PC. As such, performance and capability remain key purchase criteria, along with the cost of the actual PCs themselves. In addition, as the cost of energy is starting to increase in some areas around the world, the energy cost associated with running the PC is starting to become an item of interest.

As Intel delivers products and platforms that target the needs of a diverse and expanding PC client market segment, shipping client platforms that deliver the intersection of leading performance, expanded capabilities and energy efficiency. Intel calls this approach Energy Efficient Performance, or EEP.

What EEP Means

Although performance and capability correlate to the productivity and the experience that both the desktop and mobile client PC delivers, energy efficiency can mean different things for different platforms.

Put simply, energy-efficient performance (EEP) is the intersection of great performance, expanded capabilities and energy efficiency.

Mobile clients already deliver the lowest PC energy costs, so users with a great focus on lowering energy costs will identify mobile PCs as the preferred client solution. That said, however, an even more important element of energy efficiency on the mobile PC is the battery life achieved while delivering a required level of performance and capability. This can significantly increase productivity and usability when working in areas that do not have readily available electricity access (airplanes, outdoors, conference rooms, etc.).

For the desktop PC, energy efficiency mainly translates into the energy cost required to deliver a required level of performance and capability. EEP shouldn’t be confused with Thermal Design Power (TDP), which is the thermally-relevant power draw of processor used to design PC platforms. TDP plays a role in system design around the cost and form factor of the PC, but does not communicate either performance, or the energy cost at that performance level.

As energy efficiency becomes an interesting element when comparing PCs, it is important that the industry develop a meaningful and broadly adoptable metric that enables a person to make an informed buying decision. There are, however, some challenges with measuring energy efficiency and energy efficient performance. Although, historically, the industry has developed a number of standard practices for gauging performance like SYSmark*, SPEC* CPU as well as battery life for the notebook PC using MobileMark*, there is currently no standard methodology for evaluating client performance coupled with its associated energy cost. This whitepaper outlines a proposed methodology for evaluating both the performance and AC energy efficiency of the client platform that may serve as an interim solution until an industry developed solution becomes available. In addition to this proposed methodology, Intel will fully participate with industry standards organizations to develop a consortium-based approach to this area of platform evaluation.

Modeling Energy-Efficient Performance

There are a number of points that must be considered when examining how to measure and report energy efficient performance

First, for a test approach to be meaningful to users, it must mirror the kinds of activities routinely done during everyday interactions with the PC. It must also report results that have meaning to the buyer/user. In the case of energy efficient performance, it must report both the level of performance that can be achieved on the PC as well the energy cost to run the PC and achieve that performance.

In an office environment, much activity is centered around “office suite” applications, such as word processors, spreadsheets, slide presentation, database, virus scanning, and file compression/decompression. Web content creators use a different set of applications, including HTML editors, image editors, media encoders, as well as video editing packages. An effective methodology will use these types of real-world applications as its workload.

In addition to the right workload, an effective methodology needs to define a usage context for that workload that simulates how the PC will be used over the course of an average day. This should be a reasonable mixture of executing the defined workload, some amount of time the system sits idle, and the system being put into Standby or shut off during non-working/use hours.

The final results of an effective methodology should report out delivered performance at a given level of energy cost, communicating both performance, and the energy cost of delivering that performance. Metrics that attempt to combine these measures are best avoided, as they can obscure both the delivered performance and energy cost. This is especially true of taking ratios, as there is not a guaranteed linear correlation of the value of performance or energy cost. Platform choices are more easily made when both performance and its energy cost are clearly articulated.

Finally, the challenge of choosing the right workload is that one size does not fit all. Different usage models – digital office, digital home, gaming – require different representative workloads. As much of the concern about PC energy use is found in the enterprise however, it follows that a digital office workload is the logical starting point.

A Day in the Life…

A typical working day in the United States is nine hours: eight hours of working time with an hour for lunch. Workday length varies by country, and can be adjusted to “localize” this approach. Breaking the work-time into four two-hour blocks, the workday looks like this:

The numbers of hours in a workday will vary somewhat by country, and the model can be adjusted to factor in these variations.

During each of the work sessions, a given amount of work gets done, where faster systems will complete the run more quickly. In this proposed methodology, any remaining time in the two-hour window, the system is considered inactive. The methodology assumes that Power Options for an inactive state are configured to put the system to Standby (S3 ACPI state, suspend to memory) after 15 minutes of idle.

Since a picture is worth a thousand words, here is a diagram of the overall usage model: The inactive time remaining at the end of each two-hour window is denoted as a break.

This approach assumes that a system can complete a SYSmark* run in under two hours, something nearly all current-generations can do.

During the lunch hour, the system is assumed to be asleep during that entire hour.

At the end of the workday, the system is then put to Standby until 8:00 the next morning. In the business environment, a capability such as Intel’s Active Management Technology (Intel® AMT) can enable the system to go into a low-power, Standby, while ensuring that the system is manageable if the IT department needed to perform any type of inventorying or critical patch updates. In addition to maintaining a manageable platform, placing the PC into a low-power managed sleep state can cut annual energy costs by over half.

Ultimately, this methodology will output a performance score, as well as daily and yearly energy cost for that level of performance.

SYSmark* 2004 SE: Real Applications in a Human Usage Model

The office usage model is well understood in the PC industry, and a standardized performance benchmark already exists to evaluate PC platforms: BAPCo*’s SYSmark* 2004 SE. This benchmark is based on extensive research into emerging usage models and computing trends, and its expert-designed workloads represent activities that an office productivity or Internet content creation worker will typically perform during the course of their job. Unlike previous script-driven, application-based benchmarks, SYSmark* 2004 SE includes “think-time,” where the benchmark pauses to factor in the time users spend between operations. In fact, from a system energy-draw perspective, SYSmark* 2004 SE spends about half of its run-time in a state that is at or very near system idle.

Here is a graph that shows the system power draw of a SYSmark* 2004 SE run over time. Note the amount time spent at or very near system idle (as denoted by green line).

SYSmark* 2004 SE run System Power Draw During Benchmark Run. See final page for system configuration details. Performance tests and ratings are measured using specific computer systems and/or components. Actual results may vary.

SYSmark* 2004 SE was built to measure performance, but not necessarily energy efficiency. But because energy efficiency and performance measurement have the same overall goal – to mirror real-world activities so as to predict a platform’s effectiveness at executing them – SYSmark* 2004 SE is a good workload for this proposed energy-efficient performance measurement methodology.

Energy Use During a Working Day

A managed PC can assume several power states during a work day. Common power states include active power, idle power and Standby. Each of the power states must be included in a methodology designed to emulate the way people actually use their computers during the course of a workday. This graphic demonstrates the typical distribution of power draw levels during the workday modeled by our proposed methodology:

System power-draw levels during a 9-hour workday. The percentages shown here are the power level amount above the idle power level as compared to the idle power level. Intel® Core™ 2 Duo processor E4300 on the Intel® 965 chipset. See final page for system configuration details. Performance tests and ratings are measured using specific computer systems and/or components. Actual results may vary.

Each piece of this pie-chart represents the amount of time spent at a given system power-draw level. The percentages are the power level amount above the idle power level as compared to the idle power level. The main observation here is that over 35% of the time during the nine-hour working day—about three hours and 13 minutes—the system is at an idle power draw level or at Standby in this proposed methodology. More important however, is that all system power draw characteristics – active, idle, and Standby – are in the context of a reasonable usage model.

At the end of the day’s activities, the system is put to Standby for 15 hours, until it is awoken to start the next day. Accounting for this Standby time, the breakdown of system power draw levels looks like this:

System power-draw levels during 24 hours. The percentages shown here are the power level amount above the idle power level as compared to the idle power level. Intel® Core™ 2 Duo processor E4300 on the Intel® 965 chipset. Performance tests and ratings are measured using specific computer systems and/or components. Actual results may vary. See final page for system configuration details.

How the Methodology Works

This methodology specifies a workday modeled around four SYSmark* 2004 SE runs. For cost of benchmarking purposes, one run of SYSmark can be measured and used as a representative sample for the rest of the methodology.

That said, general best practice with any benchmark is to run it three times and take the median value of the three runs. That practice applies here.

In addition to obtaining a performance score, the system’s power draw level from the wall (AC socket) will need to be recorded as well. There are several ways to obtain this data. This can be done using a simple, inexpensive watt-meter, or using a more sophisticated meter that supports data-logging. The latter piece of equipment will make the task of gathering system power data easier, and will yield the more accurate results.

The next consideration is the watt-meter’s sampling rate. Low-cost AC watt-meters do not include the ability to log power draw over time. More costly tools may support logging, but if an insufficient sampling rate is used, many transients (spikes and troughs) may not be captured and could skew average power measurements. Intel’s measurements presented here used a logging sample-rate of 2Hz (2 samples per second).

More expensive watt-meters enable automatic logging of data, where a tester would only have to start the meter before beginning the SYSmark* run, and then stop it after the Internet Content Creation run completes at which time the PC will automatically reboot. After rebooting, a tester would restart the meter at the beginning of the Office Productivity run and stop it after the run completes. This reboot time should not be counted as part of the power measurement. This also helps ensure that BAPCo* rules are followed. It is not part of the performance measurement period and not part of what BAPCo* intends to model with the workload. The information that will need to be captured includes:

• The time it takes SYSmark Internet Content Creation to run (may be captured by stop watch or the watt-meter).
• The average power consumed during this time (captured by the watt-meter automatically or something that will have to be calculated manually from the captured sample).
• The time it takes SYSmark Office Productivity to run (may be captured by stop watch or the watt-meter).

The average power consumed during this time (captured by the watt-meter automatically or something that will have to be calculated manually from the captured sample).

To ensure stability of measurements, testers may want to perform more than one measurement. One recommendation is to take three runs and use the data from the run that the median performance.

Next, measure the idle and Standby system power draw levels.

For idle, just let the system sit with no active applications running on it, and take samples over a three minute period, and then calculate average idle power from those samples.

To measure Standby system power draw, go to the Windows* Start Menu, and select Shut Down. Then select Standby and the system will go to Standby mode. Take samples over a three minute period, and then calculate average idle power from those samples.

This methodology can be applied to platform evaluation irrespective of the platform’s energy savings settings. Note however that for best energy efficiency, users should enable features like Enhanced Intel® SpeedStep™ Technology (EIST), whose benefits include achieving optimal idle power energy-efficiency. EIST exists to help with energy efficiency, and if this is a primary concern, then this technology should be enabled. With that in mind, tested platforms presented here had EIST enabled.

Mobile Platform Considerations

If the system being evaluated is a mobile platform, an industry-based benchmark is already available called MobileMark* 2005. This methodology is not meant to replace MobileMark* 2005, as that benchmark already conveys the relevant energy efficiency of a mobile platform, which is its battery life.

However, this methodology could be used to evaluate a mobile platform’s on-AC (plugged into the wall) energy-efficient performance. The system should be connected to an external monitor, and the system’s graphic sub-system configured to display only on the external monitor. The backlight in mobile platforms is the biggest consumer of power, and by removing it from the evaluation “equation,” a tester can get a clearer picture of the power draw characteristics of the rest of the platform (CPU, chipset, RAM, disks, etc.). This approach will provide a better comparison between desktop and mobile platforms because monitor power is not included in the desktop calculations and removing the LCD monitor avoids burdening the notebook with energy used by a display device where the desktop measurement excludes the energy used by a display device. Alternatively, a tester could first evaluate the platform configured as described here, and then retest the system using the integrated display to understand the difference between the two configurations.

Finally, the laptop’s battery should be left in the system and should be fully charged prior to beginning the SYSmark* run, as some system’s AC power adapters cannot handle the transient power demand spikes of the system, and the battery effectively augments the available power level to cover these spikes. Such a system will not run unless the battery is present.

Calculating Energy Cost

When a run is complete, we are interested in several data points:

• The SYSmark* 2004 SE overall score

• Completion time of the test run

• Average system power draw during the Internet Content Creation test run

• Average system power draw during the Office Productivity test run

• Idle system power draw

• Standby system power draw

For example, Intel has measured the following performance and energy cost for the Intel® Core™ 2 Duo E4300 processor with integrated graphics on the Intel® Express Intel® DG965SS motherboard. For full system configuration details, please see final page of this document. Note that both Enhanced Intel® SpeedStep™ Technology and Minimal Power Management are enabled.

Energy Cost per KWh $0.07765

Avg. System Power Idle (watts) 49.892

Avg System Power Standby (watts) 2.898

Avg. System Power OP (watts) 56.934

Avg. System Power ICC (watts) 62.821

Time to Complete - OP (minutes) 51.4

Time to Complete - ICC (minutes) 45.25

Time to Complete - Total (minutes) 96.64

Idle Time After Run (minutes) 15

Standby Time After Run (minutes) 8.36

Performance tests and ratings are measured using specific computer systems and/or components. Actual results may vary.

Note that the kilowatt-hour figures will be small decimal numbers and will be eventually multiplied by much larger numbers. As such, it is important to avoid rounding off until after the final energy cost calculation has been made.

For this example, we are using an energy cost of $0.10/Kilowatt-Hour (KWh). This figure will vary by country, and a localized energy cost value can easily be substituted.

So, to calculate a daily energy cost, we first need to understand how much energy is consumed during the course of a single SYSmark* 2004 SE run.

We first take the average power during the Office Productivity (OP) and Internet Content Creation (ICC) runs:

OP avg. power draw x completion time

56.934W x 51.4 minutes = 2926.408 watt-minutes

Then convert to watt-hours:

2926.408 watt-minutes / 60,000 = 0.048773 KWh

ICC avg. power draw x completion time

62.821W x 45.25 minutes = 2842.65 watt-minutes

Then convert to kilowatt-hours:

2842.65 watt-minutes / 60,000 = 0.047378 KWh

Now calculate energy use during single two-hour window:

OP energy + ICC energy = total SYSmark*

0.048773 + 0. 047378 = 0.096151 KWh

Then comes 15 minutes of the system at idle:

49.9W x 15 minutes = 748.5 watt-minutes

Then convert to kilowatt-hours:

748.5 watt-minutes / 60,000 = 0.012473

Now the system goes to sleep for the remainder of the two-hour work window.

18.35 minutes of Standby x 2.898 W = 24.215 watt-minutes

Then convert to kilowatt-hours:

24.215 watt-minutes / 60,000 = 0.000404 KWh

Next, calculate energy use during the entire workday:

Four Two-Hour Windows

(0.096151 + .012473 + 0.000404) * 4 = 0.436110 KWh

Lunch

60 minutes of standby x 2.898W = 174 watt-min

174 watt-minutes / 60,000 = 0.0029 KWh

After-Hours

900 minutes of Standby x 2.898 W = 2610 watt-minutes

3060 watt-minutes / 60,000 = 0.0435 KWh

Adding the three figures, we get:

0.436110 KWh (four two-hour windows)

0.0029 KWh (lunch)

0.0435 KWh (after-hours Standby)

0.48251 KWh over 24 hours

Getting down to dollars and cents, here’s what we see:

Workday Daily Energy Cost:

0.48251 KWh x $0.07765/KWh = $0.0374669

On non-Working days, the system is asleep, so:

2.898 W x 1440 minutes (which is 24 hours) = 4,176 watt-minutes

Then convert to kilowatt-hours:

4,176 watt-minutes / 60,000 = 0.0696KWh

Converting to dollars:

0.0696KWh x $0.07765/KWh = $0.005405

Yearly Energy Cost (assumes 240 workdays + 125 non-workdays where system is at standby)

($0.0374669 x 240) + ($0.005405 x 125)