Intel’s only rival in the PC processor market continues to strive, and struggle, to keep up with the advances
Every time you think AMD is about to drop dead, some good news pops up. But you can be just as sure that any time things are looking up, some disaster or other lies just around the corner.
So where does that put AMD today? The context here is Intel’s ceaseless march of technology. While we nitpick over its delayed 14nm technology, there’s no denying Intel has broadly stuck to its Tick-Tock strategy of releasing new CPU architectures and new production processes in alternating years.
AMD, meanwhile, simply doesn’t seem able to react. Its current performance CPUs are based on the Bulldozer architecture that first appeared in 2011 and was supposed to take the fight to Intel’s Core. It’s been slightly revised on a few occasions since. But there’s no getting round it – Bulldozer is a failure. In fact, we’d say it’s so bad that AMD would’ve been better off developing the Hammer CPU core that dates back to 2003 and was launched under the Athlon 64 branding.
AMD currently has little to excite PC enthusiasts or gamers. It’s been pegged back to chips that just about make sense on account of being cheap. That means tight margins and narrow profits. Not good. But is there any hope of a turnaround?
Enter Zen, its all-new x86 CPU architecture. Due out in 2016 – though if AMD’s past form is anything to go by it’ll more likely than not miss that target – few details are currently known. However, the weight of opinion and rumour points to a launch on 14nm production technology, compatibility with AMD’s FM3 socket and support for both DDR4 memory and PCI Express 3.0 interconnects.
It’s thought Zen will offer SMT or simultaneous multi-threading capability similar to Intel’s Hyperthreading tech. It will also likely be a step away from the modular architecture of Bulldozer with its pairing of what you might call two small cores into a module with shared resources back to a more traditional big core approach with more emphasis on IPC or instructions processed per core and per clock cycle.
Another detail is a TDP, or total design power, for the first Zen chips of up to 95W, combined with a greater focus on power efficiency, which has been one of Bulldozer’s most conspicuous failings. Taken together, the combination of 95W TDP and increased efficiency confirms that AMD intends to compete at the performance end of the market.
If all this sounds similar to Intel’s strategy since its switch from the highclocking Netburst Pentium 4 architecture to the IPC-tastic Core processors, even the code-name for the first Zen processors, Summit ridge, sounds distinctly Intel-ish. Arguably, however, Zen will succeed or fail more on execution than broad strategy.
Bulldozer’s modular architecture could have been a winner. But it was very poorly executed from day one and for a company with limited resources like AMD, rectifying that initial mis-step has proven pretty much impossible. The final question is whether Zen is AMD’s final roll of the dice. If Zen fails, does all of AMD go down with it? The answer is possibly. It will depend on AMD’s plans to also launch its own chips based on the ARM instruction set, a new architecture known as K12.
A Zen failure will almost certainly kill off AMD as a vendor of x86 processors. If that happens, AMD will be left trying to make a mark on the viciously competitive market for ARM processors. And for the record, AMD probably can’t survive solely on the strengths of its graphics chips. If the CPU business collapses, AMD is unlikely to survive as an independent entity.