Posts tagged with Energy Efficiency

This blog is my voice. It is my view and it is not a ’sotto voce’ comment made as a whisper to a journalist. If you are going to blog I think you need to do so with integrity.

I think we can all agree that energy efficiency is a good thing on more than one level in that it helps to lower IT operating costs while reducing your carbon footprint, making you both "greener" and, in these economic times, less poor.

But moving beyond the goodness, IT managers need to know how to measure what they’re using (and saving) and that’s where the conversation becomes interesting

This comes up because there isn’t complete agreement about how you measure power. This is particularly relevant because just last week we introduced 5 new low-power processors. And some are suggesting that by using one metric rather than another we may somehow be trying to hide something. The truth is actually the opposite.

First, let me define terms.

At AMD we think tools should reflect real-world conditions, so we promote what we call ACP, or Average CPU Power, which models real-world processor power consumption for datacenter customers, such as IT managers responsible for power budget planning. Because it reflects CPU power consumption running typical data center workloads, we believe that the ACP metric is better suited to these customers’ needs.

Some in the industry, our competitor in particular, tends to promote TDP (Thermal Design Power). They believe it offers a good way of comparing processors. While it is a more traditional way of measuring power than ACP, if you read through the last hyperlink, you’ll see even Wikipedia points out that, "TDP values between different manufacturers cannot be accurately compared." For many of our customers that becomes problematic.

So, is ACP better than TDP? Like many things in life, it depends on how you are using it.

For datacenter operators looking to estimate processor power budgets as a way to control costs or aid them in sizing a datacenter, we believe ACP – because it reflects a relevant wattage that reflects processor power consumption while running server-class enterprise workloads – is the useful metric. But we also provide TDPs for system designers who need to know what worst-case thermal limits to use when designing a system, which is of course a practical application of that metric.

The thing to keep in mind when managing a datacenter is that ACP is the best reflection of a CPUs power experience while running typical datacenter environment workloads, while AMD’s TDP specification reflects the worse case condition under which the processor is designed to run (hence its usefulness for system designers.) Re-read that statement and you’ll quickly see how, in real-world conditions, the two measurements can be quite different.

Don’t get me wrong: testing a workload at various utilizations and at idle while measuring at the wall is really the way to go to determine how much power a system will truly draw. But if you are looking to get a picture of the energy your datacenter might consume running typical workloads and taking into account other factors like memory, peripherals and power supplies, then ACP is the way to go.

By the way, there is nothing secret about how we measure our ACP numbers. You can get more on that here.

Nigel Dessau is senior vice president and chief marketing officer at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such links sites and no endorsement is implied.