In the heat of a "dual-core duel," AMD responds to Intel's Conroe

Sunnyvale (CA) - Now that Intel's next-generation Pentium architecture is confirmed to have fewer look-ahead pipelines instead of more, and will focus on power consumption even at the risk of moderating its performance gains, AMD may have been successful in forcing CPU market leader Intel into a long-term strategic U-turn.

But after having challenged Intel on Tuesday to a "dual-core duel" in the open market, AMD may be finding itself today ironically blindsided by Intel's strategy to reveal less about its future dual-core and multicore architectures than was anticipated. With little information to go on from Intel, AMD found itself this week with little information to shoot down. In what appeared to be a carefully prepared briefing for Tom's Hardware Guide and other journalists yesterday at a suite outside of Intel's Fall Developers' Forum in San Francisco, AMD ended up revealing very little, other than to confirm the news that had been leaked on Tuesday, that the company would be releasing a dual-core Turion 64 mobile processor with DDR2 memory support, sometime during the first half of next year.

Latest Videos FromTom's Hardware

Watch full video here:

"Their legacy front-side bus architecture is probably hampering them more than anything at this point in time," Kirby told Tom's Hardware Guide. "We haven't seen any movement [towards] them getting rid of that front-side bus legacy architecture, that's been around for over 20 years." Kirby argued that by sticking to a single pathway to memory, contentions and bottlenecks only increase as more cores are added to a processor, as more threads simultaneously request access to both I/O ports and memory. "[With] our Direct Connect architecture," he added, "since our CPUs are directly connected to each other, and our memory controllers and I/O [controllers] are directly integrated into our CPUs, it's just a better formula for being able to provide a balanced system bandwidth."

Intel's move - announced Tuesday at IDF - away from its own NetBurst architecture and towards shorter pipelines, is seen by some as another strategic victory for AMD, which was first to the finish line with 64-bit x64 architecture. At the same time, not having Intel's big pipelines to aim at ends up giving AMD one less Intel target to shoot down. So when questioned about Intel's having come around to seeing things AMD's way, Kirby changed the subject quickly, like a presidential debater reorienting a panelist's question. "Even if [Intel is] moving to a shorter pipeline," he said, "they're still in a trap with the front-side bus. That front-side bus is going to haunt them for as long as they keep that. They're going to need to make some efforts to take care of that issue."

In his first keynote address to IDF as CEO of Intel, Paul Otellini announced his company is recalibrating its performance indicators around a "performance-perwatt " scale, which may replace megahertz as Intel's key power factor. Kirby countered by saying his company's Opteron processors already meet Intel's power targets, having originally been launched at 89 watts, and with successive HE and EE units launched for 55 and 30 watts, respectively.

But then Kirby threw some new math into the mix: "The one thing Intel probably didn't mention is the additional power for their memory controller," he told us. "When we say we're at 68 watts or 95 watts, that includes our memory controller as well as the core technology itself. If you were to add up the max power for each of the power rails dedicated on [Intel's] front-side bus or memory controller hub, that would equate to an additional 22 watts. So say you're trying to get down to 80 watts, if you add 22, you're really at 102 watts, if you're going to do an apples-to-apples comparison [against] Opteron-based platforms."

Hyperthreading - Intel's technology for splitting threads within a single core - will apparently be abandoned in its Merom/Conroe/Woodcrest architecture, though maintained to some extent in existing Xeon server designs. Just a few months ago, Intel was touting hyperthreading as a key architectural advantage. With Intel's U-turn having abandoned HT, is it now a dead issue? "We've seen hyperthreading as being a dead issue for quite some time," responded AMD's Kirby. "We've always said we believe that dealing with more threads is better done by actually adding real cores to the processor. Hyperthreading had its good times in the single-core world, as long as the application was highly tuned for it, but there were a lot of applications out there that weren't tuned for it, that actually hindered performance, which no one wants to see. I think Intel finally said they'll throw cores at the problem to deal with those threads, instead of trying to emulate additional cores, especially with all the extra contention that adds to the front-side bus."

"If you want to look at the marketplace," countered Kirby, "we think we've won that duel." He cited figures claiming AMD with over 20% market share in a four-way CPU battle, including with Intel. "In 2005, we're going to have our partners introducing more Opteron-based platforms than have been introduced in the last two years combined...Let's do an actual duel, platform-to-platform, and let's bust through this hype and provide the information our end customers need about which platform is the best dual-core platform."

It may be difficult, going forward, for AMD to be touting greater market share and more partnerships, at the same time it's arguing in court that Intel has attained monopoly status in some markets. AMD may see its battle with Intel as a street duel; but if Intel's lower key this week is indicative of any intentional strategy, it may actually resemble less of a street fight, and more a game of chess.

TOPICS