In June and August of last year, I wrote a pair of columns in which I extolled the value of virtualization as a solution to excessive energy consumption. The primary benefit, as I described it, is that virtualization makes it possible to consolidate multiple applications onto a single server. That is, apps that currently run on dedicated systems can be moved en masse to a single server that consumes less power -- generally, far less power -- than that required by the dedicated servers.
This economy derives from two principal factors:
1) Modern servers are much more energy-efficient than their forbears. This is true in absolute terms and relative terms. In fact, in relative terms, such as watts per mips, today's systems are orders of magnitude more efficient.
2)
Until very recently, the need for IT to really include eco-concerns as part of overall strategy did not have universal appeal. Surely, sites located in areas such as southern Manhattan where power distribution is already running at maximum capacity have a grave problem. And likewise sites that need more room but have tight expansion constraints. For them, green has been a key preoccupation for a while.
For most other IT sites, however, the main driver for green has been cost reduction -- and until the last few months, the cost of energy was tolerable even if somewhat higher than budgeted. So, pressure existed to reduce unnecessary consumption, but not place the issue at the center of IT concerns. However, with oil now regularly surpassing $130 per barrel, there is no longer any
Early this month, I attended the Technical Forum of the Green Grid vendor consortium. The Green Grid is a recently formed group that brings together major businesses to establish useful tools and policies for eco-responsibility in IT shops. Its activities include defining metrics for the IT industry, establishing best practices, and encouraging adoption of both.
The two-day forum was narrowly focused on the quest for useful, usable metrics that measure energy efficiency in data centers. While many members of the technical committee have been working on this problem long before Green Grid existed, I was surprised by how little consensus there was on how to measure energy efficiency and how crude the proposed measures currently are. This observation does not in any way denigrate
In this column, I have previously examined energy-saving options on processors and hard disks. This time around, I'd like to examine one of the other principal energy sinks on the standard PC: graphics cards. Graphics cards are a confusing area of technology because almost all the attention and press the cards receive is dedicated to the high-end, super-expensive cards favored by gamers and hardware aficionados. Those users live and die by the next release of whiz-bang features and the number of anti-aliased triangles that can be displayed.
But if you're choosing graphics capabilities for a business system, the likelihood that anti-aliased triangles are important to your choice is close to nil. And that means that you'll be able to save energy, because generally, the more powerful the graphics card, the more electricity it uses.
How people use their computers determines what kind of cards they need, and the power that the cards will use. Most knowledge workers have very modest graphical needs. As long as they can see their Excel charts, enjoy basic Web 2.0 features, and view the occasional video clip on a single monitor, they are content.
For many users, this capability is delivered by the video capabilities built directly into the motherboard. They don't need a separate graphics card. The motherboard graphics system has modest but sufficient capabilities. This inexpensive option results in systems that consume less electricity.
The second tier of users consists of those who have somewhat more demanding needs. In the business world, these users tend to be software developers and "power users." They don't necessarily need faster graphics. Rather, they require a graphics card that can drive more than large LCD monitors. (By large, I mean a monitor that requires screen resolution greater than 1280 x 768.) The graphics cards required to drive multiple large monitors vary from entry-level to mid-level products. In most cases, the entry-level products are sufficient.
For example, software developers, who are a breed apart when it comes to system needs, rarely have to do more than display code, Web pages, and some basic animations. What they do need is multiple large monitors. So, for them video cards such as the Nvidia Quadro series are sufficient. For example, the Quadro NVS 290 supports two monitors at 1920 x 1200 resolution and consumes a mere 21 watts (with a street price of currently $160 or so). For users with somewhat higher requirements or those needing basic 3D graphics, the Nvidia Quadro FX 570 will almost surely suffice. (It supports two monitors at 2560 x 1600 resolution, has a street price of around $200, and requires 38W of power.)
One question that will surely arise in keeping graphics-card prowess and power consumption in line with user requirements is how to measure the graphics performance as experienced by the user. The most effective ways is with the ViewPerf benchmark developed by the Standard Performance Evaluation Corp. (SPEC).
SPEC is a vendor-neutral, hardware benchmarking organization. While most SPEC benchmarks must be purchased, ViewPerf is available at no cost at SPEC.org. Release 10 of the benchmark ships as a Windows program (while version 9 runs on Windows and Linux). It's a large executable and takes a good 10 minutes to run. It finishes up by providing a table of six performance results for the graphics subsystem. Take the average of these and you have a single-number rating for the graphics capabilities. You can then test other cards to see whether they provide comparable performance. Some manufacturers, such as Nvidia provide their cards' ratings on each of the tests, so you can perform highly detailed comparisons.
These comparisons are particularly important when it comes to the small group of users who require (or at least think they require) very high-end hardware. Ultra-high end graphics cards are monsters of consumption. For example, the Nvidia Quadro FX 5600 requires 171 watts of power (and sell for more than $3000). Typically, such cards are the domain of specialists who use high-end workstations. These include engineers doing very complex 3D modeling and some financial analysis.
Often, though, users' requirements can be satisfied at lower prices and significantly reduced power consumption. To prove the point, the Nvidia 4600 has nearly the same performance profile as the 5600, but it costs half the price and consumes nearly 25 percent less power. Using SPEC tests will help determine performance metrics for cards; vendor spec sheets will tell you how much power each card consumes. In that way, you'll be able to choose less power-hungry cards, without sacrificing performance.
In sum, the graphics card is a PC component that can easily consume the majority of a system's power. As a result, it's important to carefully match the user's needs with the performance of a given card. By taking measurements of needs and capabilities and matching them carefully, you can significantly lower power consumption, heat generation, and, of course, the price tag.
See ClimateBiz.com
See GreenBiz.com