Energy-efficiency battle is key in war between AMD and Intel

In the ongoing chip wars between AMD and Intel, the battle over power efficiency is heating up. With datacenter operators fixated on electric bills as well as their facilities' environmental impact, and with the EPA pondering Energy Star standards for server hardware, being able to claim the title of Most Energy-Efficient has become far more important.

Even being able to claim 10 percent superior energy efficiency can be a tremendous selling point for one of these chipmakers. If you have a datacenter with 5,000 machines and can save $20 per system per month on electricity, you're looking at pocketing $1.2 million per year. (I suspect there's savings on cooling, too, though I'm not sure how much that would be.)

Some independent groups have pitted AMD and Intel engines in power-efficiency tests, the results of which have cropped up in the past couple of weeks. If you've been reading my Sustainable IT blog, you'll have already read about one such study conducted by Neal Nelson and Associates. In it, Nelson pitted a 3GHz server running AMD Opteron against a 3GHz server running Intel Xeon.

Nelson determined that the AMD-based server used 7.3 to 15.2 percent less power at five different user load levels and 44.1 percent less power while the systems were idle and waiting for work.

That, he said, translates to annual electricity savings between $20.29 and $36.04 per server, depending on the workload, the study concluded. At idle speeds, it amounts to a $99.76 per-server, per-year saving.

The test elicited some heated discussion the likes of which you might expect during an election year. Some suggested that Nelson must have been paid off by AMD to have reached his findings; others aimed a bit higher than just below the belt, at least keeping comments and criticism in the context of Nelson's methodology instead of his morality.

Nelson still stands by his testing, though he plans to run more benchmarks using a new Intel processor. He expects to have results in around a week, and I'll share them.

AMD, not surprisingly, was quite pleased with Neal's results. Intel issued this statement: "We stand by all of our energy-efficient claims, period. We also recommend that IT managers who don't do their own in-house testing turn to the dozens to hundreds of independent and certifiable benchmark organizations for the best, most credible perspective."

As I said, Nelson's tests aren't the only ones to emerge in the past couple of weeks suggesting that AMD has developed more energy-efficient processors than Intel.

A couple of weeks ago over at Tom's Hardware, Bert Töpelt and Daniel Schuhmann posted the results of their own power-efficiency tests between several Intel and AMD processors. In terms of raw energy efficiency, AMD consistently had a better showing. "Whenever low acquisition costs, low follow-up costs, as well as low power consumption are important, AMD's processors are still first choice. AMD also currently offers the cheapest dual-core processor. Finally, AMD processors are very suitable for use in quiet systems," they concluded.

However, unlike Nelson, the guys over at Tom's considered other factors when rating the CPUs. In addition to energy efficiency, they equally weighted in chip price and performance. In that context, Intel's Core 2 Duo E6700 topped anything AMD had -- though AMD's offerings are nearly all superior to the rest of Intel's.

Meanwhile, over at AnandTech, Jason Clark and Ross Whitehead recently had an AMD Opteron Socket-F machine square off against an Intel Xeon Woodcrest machine in what they dubbed a "Low-Power Server CPU Shoot-out." Their conclusion: "AMD is clearly the leader when it comes to performance per watt using the workloads in this article."

Interestingly, they attributed AMD's advantage not to the processor but to the fact that, unlike Intel, AMD doesn't employ fully buffered DIMMs. (In my Sustainable IT blog about Neal's test, one commenter suggested that "As for FB-DIMMS. It was Intel's decision, and they went the wrong path. And from the Wild Wild Web the information seems to be that future chipsets will not support FB-DIMMS due to lack of performance versus the power it consumes.")

My takeaway in all this is that, for the time being, AMD does appear to have an advantage over Intel, at least from a raw energy-efficiency standpoint, which is becoming increasingly important as datacenter operators grapple with high power and cooling bills. But the challenge remains of devising a useful and reliable metric for measuring energy consumption against overall performance.

That doesn't appear to be an easy task, given just how many factors there are to consider. Performance is, after all, a relative term, depending on what kind of tasks your servers are performing.

But you can expect that AMD and Intel will continue to weigh in on the discussion as groups such as the EPA and The Green Grid, as well as independent groups such as we here at InfoWorld, devise benchmarks for gauging energy efficiency. The stakes are simply too high to ignore.

Posted by Ted Samson