AMD’s new Ryzen will launch in early March, and Intel is reportedly prepping changes to its own product lineup to counter any AMD threat. It’s a smart move from Intel if true — the rumors we’ve heard about Ryzen suggest the chip will be a potent competitor, and Intel won’t want to be caught flat-footed. Back in 2003, Chipzilla was forced to rush the Pentium 4 Extreme Edition to market to counter the surprise threat of the Athlon 64, which still wound up winning most games and a number of consumer benchmarks. Intel will want to avoid any repeat of that situation, or the drubbing it would take if AMD’s Ryzen debuts at price points that would challenge Intel’s existing product stack.
Canard PC has details on the rumored updates. First up, there’s the Core i7-7740K — a Kaby Lake chip with a 4.3GHz base frequency and a supposed 4.6GHz Turbo frequency. Next, there’s the Core i5-7640K with a similar modest frequency bump (4GHz base clock compared with 3.8GHz). Canard PC isn’t sure whether this chip would have Hyper-Threading or not. This isn’t a crazy proposal, given that Intel just added Hyper-Threading to the Pentium family, but I think a degree of skepticism is warranted given the timing of these announcements.
It is unusual, though scarcely unheard of, for a company to push out a faster version of a core within just two months. Typically this is done when a competitor shoves its faster version of a chip out the door. Given that AMD is prepping to launch a major architectural revision, it therefore makes sense to think Intel might be prepping new launches of its own.
The flip side to this is that preemptively rearranging all your product lines to battle a competitor that hasn’t even launched yet isn’t a great business plan, either. For almost a decade, Intel has had a stable desktop strategy. Pentium chips are still based on modern Core architectures, but run at low clock speeds and without Hyper-Threading (until recently). Core i3 chips have higher clock speeds, Turbo frequencies, and Hyper-Threading. Core i5’s are typically about the same speed as Core i3’s but have four cores without HT, while Core i7’s are Intel’s highest-clocked cores that add HT back again. For more information you can check our guide to choosing between a Core i5 and Core i7 for both mobile and desktop processors.
Intel has already disrupted its own market at the low end by flipping on Hyper-Threadingon its newest Pentium processors, but it isn’t going to be crazy about differentiating the same way at the high end. If Core i5 chips add HT, the distinction between them and the Core i7 gets blurry, fast. None of the solutions to this problem, like shifting to a six-core Core i7 part, are attractive, and all of them will cost Intel some money. Even if it makes these changes, Intel would prefer to make them on its own timeline, not AMD’s.
The final point to consider is that this isn’t the Good Ol’ Days, when CPU frequency flowed like water. Consider the following chart, which shows every Intel chip released from 2011 (Sandy Bridge) to the present day.
Ivy Bridge didn’t increase the CPU frequency, it just lowered its overall TDP. The Core i7-4770K delivered some performance improvements from architectural enhancements, but it didn’t touch the core clock (the integrated FIVR on Haswell makes direct TDP comparisons impossible, since that chip was previously mounted on the motherboard and Intel only used this approach for the two Haswell cores). The 4790K pushed clock speeds up, Skylake pulled them down, and now Kaby Lake has nudged them upwards once again. But Intel’s second-generation 14nm process can’t fix the fact that frequency now scales extremely poorly against TDP — or that it’s running out of frequency to play with.
I have no doubt Intel can manufacture a CPU with a 4.3GHz base clock and a 4.6GHz Turbo. But the higher Chipzilla pushes the clock, the more aggressively it will have to bin its CPUs to ensure they can hit it within a given power envelope. Expanding its power envelope would give Intel more headroom, but not as much as one might hope. All evidence points to the same underlying problem — as you raise clock speeds, power consumption skyrockets. If you can build a 4.5GHz chip in a 91W power envelope, do you really want to try and sell customers on a 5GHz chip in a 140W power envelope? There’s little profit in trading an 11% clock speed boost for a 1.54x power consumption increase. (5GHz picked solely for representative purposes, but this kind of ratio tracks what we’ve seen from overclocking). Even if Intel has a 7740K in the works, it would doubtless prefer to wait and position that chip as a Kaby Lake refresh as opposed to an immediate panicked response to AMD.
I think Intel absolutely has some contingency plans for how it will adjust its own product lines if AMD launches an extremely competitive chip. But I don’t expect to see these cores in the immediate future. Intel can afford to wait and see what AMD debuts and how that debut effects its own hardware sales before it makes a decision on how to respond to its competition.