AMD’s Zen has been a topic of conversation for over a year now, as the chip has made its way from design schematics to functional silicon and into shipping products. Intel’s Kaby Lake refresh has been less a topic, since it’s a minor respin on existing 14nm hardware, but both companies have previously stated they would target a 2016 release date.
According to DigiTimes, however, both manufacturers will delay new product introductions supposedly due to weak market demand and upstream channel inventory management issues. While this could theoretically be true, it’s not a particularly convincing reason for either company. AMD has made it extremely clear that it will launch Zen into the PC enthusiast space first, followed by the enterprise data center. AMD currently competes in neither of these markets; the company’s FX-class processors are midrange CPUs at best (the FX-8350, at $159, is the highest-priced AMD product in the Amazon top 20).
If AMD wants to retake space above the $200 price point, it can launch a handful of Zen SKUs at that price point while retaining the Piledriver-class desktop parts for less expensive systems. This would be a similar strategy to when the company introduced the Athlon 64 and Athlon 64 FX, both of which were significantly more expensive than then-existing chips in the Athlon XP family. Since AMD has no plans to launch a Zen APU this year, there’s no chance that Zen would cut into APU sales, particularly in mobile. Finally, AMD still has roughly half of the second quarter and all of the third to adjust its own production of FX-class processors to minimize the chance of an inventory build-up.
Intel’s Kaby Lake could theoretically run into some inventory issues, but these wouldn’t necessarily delay the refresh. When Intel first launched mobile Haswell, we noted that the chip had been slow to ship in updated systems; especially ultrabooks. OEMs often choose to launch new SKUs around new chips, but they aren’t required to do so, and an Intel refresh that lands outside the OEMs typical refresh cycle probably won’t be adopted en masse until it makes sense for the OEM to do so. Given how little difference Intel delivers with each successive generation these days, there’s less risk of carrying older hardware.
In short: AMD’s Zen is supposed to launch into segments where AMD doesn’t currently compete, while Intel’s Kaby Lake can be easily segmented and introduced over time, much as Intel did with Broadwell, or as OEMs did with mobile Haswell.
The major question from enthusiasts has been when AMD will launch Zen, and the company has been a bit coy on this question. CEO Lisa Su said at Computex that Zen would begin sampling to top-tier customers this quarter and to larger OEMs in Q3. While both of these steps are important, neither of them exactly translates into a firm commitment to the retail channel.
While I have no special knowledge on this topic, I can say that both AMD and Intel typically align their channel launches with at least a handful of OEM products in order to claim that hardware is immediately available and to avoid claims of a paper launch. The other thing to remember is that very few product launches ever take place in December: Staff are typically on vacation or gearing up for CES, not prepping to kick a product out the door. A December launch is also too late to recognize any significant revenue from the new platform in the current fiscal year and it’s generally too late for people who might plan to buy a chip for Christmas.
If Lisa Su plans to sample OEMs starting August 1, AMD might still make an October / November launch window. If Q3 sampling means “September 30,” than AMD will probably hold off on retail availability. It could sample reviewers prior to a wide retail launch, or it could hold the chip until CES and debut it there for maximum impact. If I had to guess, I’d guess that AMD will either roll the chip late in the year (mid-November) for immediate retail availability, or launch it at CES 2017 for both channel customers and OEMs.
There are still rumors swirling around that Zen won’t ship until later in 2017, but these rumors appear to link back to either unclear remarks Devinder Kumar made last December, in which he implied that Zen would tape out over the next few quarters, or to confusion over the company’s data center versus consumer products roadmap. AMD has explicitly said that its data center roadmap wouldn’t launch until later in 2017, but has not stated exactly when consumer products will hit store shelves.
Zen is rumored to have targeted Haswell’s performance, but there’s still a great deal of wiggle room in that statement. The best-case scenario for AMD, of course, would be a one-to-one match against an Intel CPU, where an eight-core AMD was equivalent to an eight-core Intel. Given how badly AMD has slipped compared to its larger rival, however, this may be difficult to achieve. Even rough parity would be a huge leap forward for Sunnyvale, particularly if it manages to make up enough ground that it doesn’t have to sell its eight-core chips as competitors to Intel’s quad-cores. This is doubly true in single-threaded code, where the fastest FX-9590 can be beaten by a Core i3 in certain scenarios.