RAM | Reseller Advocate Magazine: Issue 78

By Marco Chiappetta

Many of the top PC equipment manufacturers are touting energy-efficient, eco-friendly computing products they claim will save end users’ money and help the environment. While more energy-efficient products offer significant benefits to mobile and enterprise-class products, their value proposition in the consumer space is questionable at best. Do these energy-efficient products really save consumers money on their electric bills? Do end-users even care? We hope to answer those questions and more and find out if "green" is really golden.

Is Green Really Golden?
Over the past few years, many of the most prominent equipment manufacturers have expanded their product stacks with more eco-friendly, low-power products. Although we suspect Al Gore would like a pat on the back for the increased environmental awareness, the inconvenient truth is he can't take all of the credit. We think some of the seeds were actually planted by AMD way back in early 2004, right after Intel released its first batch of Prescott-based Pentium 4 processors.

Although an increased emphasis has been placed on energy-efficient computing as of late, questions remain as to whether or not these supposed eco-friendly products truly offer any real-world benefits to end users.

As many of you are no doubt aware, Pentium 4 processors based on Intel's Prescott microarchitecture and subsequent derivatives consumed much more power than their Northwood-based predecessors. In and of itself, increased power consumption isn't necessarily a bad thing if higher performance is a side effect, but Prescott-based Pentium 4 processors were outperformed by the older Northwood-based CPUs in many application usage scenarios. At the same time, AMD was riding high with its Athlon 64 lineup, which was not only faster than Intel's most of the time, but also used less power. Soon thereafter, power efficiency and performance-per-watt were the buzzwords du jour, as AMD promoted its desktop and server processors to general consumers and IT professionals.

The Independent Dynamic Core and Dual Dynamic Power Management Technologies incorporated into AMD's native quad-core Barcelona microarchitecture, which is used in current Opteron and Phenom processors, are designed to lower power consumption and increase power efficiency.

Today the tables have turned, and Intel's Core 2 line of processors typically outperform AMD's latest offerings both in terms of overall performance and power consumption at any given price point. However, the buzzwords remain, and it's becoming increasingly more difficult to find any major product announcement that doesn't extol the virtues of some sort of new "power saving" feature or technology.

How'd They Do It?

AMD's 7-series motherboard chipsets are manufactured using a 65nm process node and have a relatively low 10 watt TDP. AMD claims that the 7-series chipset's new design make it roughly twice as power efficient as the company's previous generation.

Virtually every component in a typical PC now incorporates technology designed to increases its power efficiency. Processors from both Intel and AMD feature clock-gating technology that disables unused portions of the chips when not in use. They also have built-in mechanisms that decrease voltages and clock speeds and allow the processors to enter deep sleep states while idle.

Motherboards and video cards have also joined the action. They too boast clock-gating features and, like CPUs, are continually manufactured on more advanced process nodes in an effort to lower power consumption. NVIDIA and AMD have unveiled new hybrid SLI and CrossFire features that allow a discreet graphics card to be completely shut down when paired with a modern IGP to conserve power. NVIDIA has even gone so far as to incorporate graphics processors into its high-end chipsets for the AMD platform, whereas IGPs used to be relegated to bargain-bin motherboards only. And similar IGP-equipped chipsets for the Intel platform will soon follow.

Many modern, energy-efficient power supplies, like the Corsair VX550 seen here, are 80 PLUS certified. To be 80 PLUS certified, the program's performance specification requires that power supplies be 80%, or greater, energy efficient at 20%, 50%, and 100% of their rated load with a power factor correction rating of 0.9 or higher.

Power supply manufacturers have begun focusing on power efficiency as well. In fact, electric utilities from across North America have contributed millions of dollars to a program dubbed 80 PLUS, whose aim is to get desktop and server system builders to integrate more energy-efficient power supplies into their designs. Power supplies that are granted 80 PLUS certification meet strict criteria for power efficiency and are typically of higher quality than off-brand, less efficient products.

One area where we've seemingly taken a step backward, however, is in regard to server-class memory. Fully Buffered DDR2 memory modules, or FB-DIMMs, for example, actually consume more power than standard DDR2 or DDR3 DIMMs. FB-DIMMs use more power because each stick is outfitted with a specialized chip called and Advanced Memory Buffer, or AMB. While the AMB is what allows FB-DIMMs to be packed in ultra high densities in server space, it also consumes power and generates heat. The additional chip and requisite cooling make FB-DIMMs somewhat less attractive to power-conscious admins. Until a better alternative arrives, though, FB-DIMMs are a necessary evil for high-end Intel-based servers.

Why Bother?
Although eco-friendliness and low power consumption are both good things, it's difficult to quantify their benefits with desktop machines. With notebooks and in the enterprise, it is much easier to justify the need for decreased power consumption. For example, low power processors, GPUs, chipsets, and drives offer tangible benefits to notebook manufacturers and consumers alike. With notebooks, low power components result in increased battery life for consumers and minimal thermal concerns for manufacturers. And in the enterprise, where hundreds or even thousands of servers may be operational, shaving a few watts of power consumption off of each machine can result in huge energy cost savings. For home users, though, saving a few watts here and there may not offer any significant benefits, depending on your point of view.

All FB-DIMMs feature an Advanced Memory Buffer, or AMB, chip. The AMB gives FB-DIMMs bidirectional interconnects to the system's memory controller and the next FB-DIMM in the bank. This allows for higher densities, but it also increases power consumption versus standard DDR2 DIMMs.

Calculating Consumption
According to a report released by the Energy Information Administration in March of 2008, the average cost per kilowatt-hour (kWh) of electricity in the United States was about 10.3 cents as of December 2007, with obvious fluctuations based on location. To see exactly what customers in your area are paying, simply check a recent electric bill or reference the EIA study on its Web site

(http://www.eia.doe.gov/cneaf/electricity/epm/table5_6_a.html).

This is an important piece of data because most electric utilities in the U.S. bill consumers in kilowatt-hours. The problem is, most consumers don't know, or simply don't care, how to ascertain how much a particular piece of equipment really costs in terms of power consumption. We often see power consumption measured in watts, or see products that carry a maximum wattage rating, but this tells only part of the story. Knowing the wattage of a device at any given moment doesn't tell you how much it costs to run each month.

Let's use a quad-core AMD Phenom X4 9850-based desktop system with 2GB of RAM, a single hard drive and optical drive, and a discreet graphics card as a point of reference. According to AMD, the Phenom X4 9850 has a maximum thermal design power (TDP) of 125W. According to our own in-house, real-world testing, a system built around this processor will consume about 230 watts while idling and approximately 340 watts under load. In the absolute worst case scenario, where this system is powered on 24/7 under load 100% of the time, it consumes .34 kWh each hour (340 / 1000 = .34), which equates to 3.5 cents an hour (.34 x .103 = .035), 85.2 cents a day (24 x .035 = 85.2), $26.41 a month (85.2 x 31 = 2641.2 / 100), or $316.92 a year (26.41 x 12 - 316.92).

Now, let's shift gears and use a machine at the opposite end of the spectrum as an example. If we swap out the 125W Phenom X4 9850 for a 45W Athlon X2 4850e and lose the discreet graphics card in favor of a low-power IGP, while using the same storage components, power consumption drops significantly to about 80 watts at idle and 130 watts under load. Using the same math, this new low-power machine will use .13 kWh and cost 1.3 cents to run each hour, 31.2 cents each day, $9.67 per month, and $116.06 per year. In this worst-case scenario, that's a significant $200.86 savings annually. Realistically, though, how many users keep a machine on 24/7 while also running under load?

We have simplified the math here for the sake of clarity, but the thought process should be helpful to those of you trying to decide whether eco-friendly products offer tangible benefits to consumers, strictly in terms of cost. There are a handful of other factors to consider, however. Although the low-power machine in our example does offer significant cost saving in dollars, it is also a much lower performing system. That lower performance may equate to lower productivity for some. Lower performance also means it will take the low-power system longer to complete certain tasks, so while the high-power machine may use more power, it also completes processor-intensive tasks more quickly. For example, the Phenom X4 9850 system can complete some 3D rendering tests in roughly half the time of the Athlon X2 4850e.

There are many tools consumers can use to check the power consumption of an item from the electrical outlet. Devices like the Seasonic Power Angel or Kill-A-Watt meters can be had for roughly $20 and provide a host of interesting data.

Gathering The Data
Figuring out exactly how much a particular system configuration may consume is prohibitively difficult using specifications alone. The total wattage rating listed on a label or in a list of specifications is usually the maximum amount that component will consume under lab conditions. In real-world usage scenarios, the actual consumption numbers will likely be much lower.

If you need to calculate real-world consumption, a multitude of tools are available, from simple meters that plug into an electrical outlet to complex data gathering devices that can monitor and log usage over time. Basic devices like the Seasonic Power Angel or Kill-a-Watt meter are available for less than $20, while high-end analyzers like the Extech True RMS Power Analyzer Datalogger cost about $700. Any meter capable of measuring current can be used, however. To calculate wattage, simply multiply the measured current by the supplied voltage: 115v in the U.S. and 220v overseas.

Is It Worth It?
It’s up to each individual PC user to decide whether investing in "green" components or systems is worthwhile. A casual PC user on a budget who doesn't require a high-power processor or graphics card would likely be well-served by a low power system. Power users or gamers with larger budgets who crave high performance, however, probably won't care about the long-term energy costs of their systems, so eco-friendliness is lost on this audience.

While we have discussed the cost and performance impacts of green computing, there are other things to consider. Using less power has many other desirable side effects, such as the fact that low-power components consume less energy, so they tend to generate less heat. The thermal benefits of low-power components are twofold. For one, less heat means cooling fans don't have to work as hard, which also lowers power consumption somewhat and minimizes the heat expelled from a system into its environment. Even better, some low-power components can be passively cooled, meaning cooling fans may not be required at all.

An inherent benefit to low voltage processors is lower power consumption versus their standard voltage equivalents. They have been a boon to notebooks and server clusters, where they increase battery life and lower power bills, but there are few reasons to use one in a desktop system.

Systems with fewer cooling fans, or fans that aren't working as hard, also offer some acoustic benefits. Quieter systems are desirable in all segments of the market, from HTPCs to high-end desktops. The acoustic impact of a system may be extremely important to some consumers, especially if said machine will reside in a bedroom or living room, where the constant report from loud cooling fans will be an annoying distraction. Finally, there is the environmental impact to consider. According to the Savings Potential Calculator on the 80 PLUS program's Web site (http://www.80plus.org/80sav.htm) , a system that consumes 6,120 kilowatt-hours over its lifetime has consumed an amount of energy equivalent to 175 five-gallon propane tanks, 10 barrels of oil, 479 gallons of gasoline, or 10,913 miles driven in a car. If everyone cut their PC power consumption in half, that would equal a significant savings in energy worldwide. That's an unrealistic expectation, though, and when overall system performance and actual energy cost savings in dollars are factored into the equation, we suspect consumers won't care all that much if their processor is using more power than the next guy's.