AMD Executive Declares End Of Core-Count Wars

While chipmakers are cramming as many cores as possible onto microprocessors, one AMD executive says that trend won't last long.

Donald Newell, AMD's chief technology officer for servers, says the end of the "core-count wars" is near, as power efficiency and on-die specialized computing are becoming more important considerations than the number of cores on each processor.

"There will come an end to the core-count wars. I won't put an exact date on it, but I don't myself expect to see 128 cores on a full-sized server die by the end of this decade. It is not unrealistic from a technology road map, but from a deployment road map, the power constraints that people expect [servers] to live in wouldn't be feasible for chips with that many cores," Newell told IDG News Service on Friday.

Before the so-called core wars, there was the clock-speed war. The perception was that CPU performance was determined by clock speed, until Apple CEO Steve Jobs challenged the "megahertz myth" openly while offering PowerPC chips that were markedly slower than Intel's processors. Semiconductor firms, system-builders and their customers value much more than simple clock speed as concerns over power consumption and overheating grow.

Joshua Liberman, president of Albuquerque, N.M.-based solution provider Net Sciences said he agrees with Newell about the limits of adding cores, but for different reasons.

Next: Views From System Builders

"Even quad core CPUs provide more raw processing than many businesses can use. While virtualization of multiple server loads on a single price of server hardware certainly can increase the workload that any given CPU will see, there are realistic limits," Liberman said. "Amongst these are I/O to disk subsystems, I/O for network connectivity and even power utilization. And let’s not overlook the many eggs/one basket issue either. Core count is pretty sexy right now, just like bit-depth was for video in the 90s. But we’ll hit a practical limit sooner than later."

Todd Swank, vice president of marketing at Burnsville, Minn.-based system builder Nor-Tech, says while the number of cores on a single die may no longer be significant, there will likely soon be a new feature in the architecture that will create an arms race in CPU technology.

"It wasn’t long ago that the only thing customers wanted was a faster clock speed when they made their purchase decisions and number of cores wasn’t even part of the decision process because all CPU’s were single core," Swank said, adding "Now, you hear very little from customers regarding clock speed and it’s all about the number of cores."

GPU integration and multi-core processing are among the variables that needed to be considered alongside clock speed. But with Intel focusing on its Sandy Bridge line of integrated GPU and CPU capabilities on one die, and AMD readying its Fusion technology with the same goal in mind, now the importance of the number of cores on each chip may be overvalued as well.

"Just as the importance of speeds and feeds fell by the wayside, so too will the priority on core count," Tim Ulmen, principal at Midwest IT solutions group,a Wichita, Kan.-based system builder. "At the time of its inception, Dual Core was impressive and Quad Core was even more so, but as the core count continues to rise from six to eight to twelve the end-user has come to expect the performance gains as a routine part of technological progress and thus has less appreciation for technology and more demand for performance and bottom-line ROI."

As software developers look to apply multi-core chips as efficiently as possible, the number of tasks at hand for x86-compatible cores is increasing beyond the capacity for those cores to keep up, regardless of how many of them can fit onto a single chip.