When Intel released its seventh-generation Core processor range in January, many were left unimpressed by the performance of its top-end desktop processors.
Kaby Lake, Intel’s codename for the range, provided a significant boost for mobile devices, but the desktop risks being left to fester, despite Intel’s assurances to the contrary and the re-emergence of AMD as a force in the CPU market.
“We know performance matters,” said Intel’s client computing vice president, Navin Shenoy, talking about the latest generation of chips. “It’s why we continue to focus on improved performance, delivering double-digit productivity performance increases – up to 20% for performance and gaming notebooks and 25% for desktops.”
Yet most third-party benchmarks published to date have found that the high-end processors are not significantly faster than existing CPUs, especially if those sixth-generation parts were overclocked. While Intel has increased the clock frequency of processors with Kaby Lake, there are minimal architectural changes and many of the performance improvements on chips using the company’s 14nm technology rely on optimising specific tasks, such as video decoding and memory management.
“Looking at the core CPU benchmarks, Kaby Lake doesn’t really provide much of a noticeable performance enhancement over [its predecessor] Skylake,” said Mark Hung, an Intel specialist at research firm Gartner. “Given the slowdown in Moore’s law, Intel decided to make a third generation [of 14nm chips] by making further architectural refinement. A lot of the improvements were made with Skylake, so there wasn’t much room to make further optimisation.”
When PC & Tech Authority queried the discrepancy between Intel’s 25% performance increase claim and external benchmarks, the company contested reports that performance improvements were minimal, but appeared to backtrack on the scale of the gains on offer. “Intel’s 7th Gen Core family is expected to provide high single-digit gains on productivity performance,” the company said in a statement, while pointing out that there were bigger improvements in specific areas such as 4K video.
“Seventh-gen Intel Core processor graphics deliver additional 3D graphics performance, driven by process and architecture improvements, when compared to sixth-gen Intel Core processor graphics. These improvements range into double-digit increases,” the company added.
Has Intel lost interest in the niche desktop, or is it simply biding its time until it can perfect the Cannonlake 10nm process due at the end of the year? At least part of this bottleneck in desktop performance is due to the increasing difficulty of “obeying” Moore’s Law, and, according to analysts, Intel has tried to squeeze more from the 14nm production process and left itself little headroom for improvement.
Historically, the company has worked on a “tick-tock” cycle, where the firm moves to a new process one year and then refines that process the year after. For example, Ivy Bridge in 2012 saw chip production move to a 22nm process, which was the tick, and a year later the company improved the performance with the Haswell architecture, the tock. Then Intel moved on to the 14nm process.
“Haswell was a great desktop product in many respects, and then things started to change when Intel moved to 14nm,” said David Kanter, an analyst with the Linley Group research company. “The process was more geared to improving low power operation instead of maxing out peak performance.
“The 14nm process that brought us Broadwell really did great things for mobile, and it was pretty good for servers, but the high-end desktop wasn’t super impressive.”
Skylake followed in the traditional “tock” cycle, but the benefits were less pronounced than earlier revisions, and Kanter believes the company ran out of headroom when it needed to release another upgrade on the same production process. “It’s the third generation, and the problem is that Intel’s planning cycles are long and it didn’t have the option to redesign the CPU and make it faster because of the development cycles,” he said.
While technical issues have played a part in the arrested development of top-end desktop chips, there are good reasons why Intel has chosen to focus on other areas of its business.
Research figures for 2016 released by research companies IDC and Gartner both highlighted that PC sales had shrunk for the fifth year in a row, and within this dwindling market, laptops continue to far outsell desktops. That’s reflected in the improvements made to the processors. “If you look at client PCs, really the vast majority of them are laptops, and there the design considerations are very different,” said Hung. “Power efficiency and video-encoding performance are much more important than raw CPU performance.”
Kanter agrees that Intel is merely following the money trail. “For consumers, the high-end desktop chip is still between $400 and $1,000 and that’s probably going to stay the same,” said Kanter. “Gamers will pay that, but the issue is how often do they upgrade and how much are they going to spend on their CPU?
“Intel is a tech powerhouse, but it’s also focused on the bottom line. The data centre market is more lucrative than the high-end gaming market.”
Another reason Intel hasn’t been pushing the performance barriers is that it doesn’t have to: there’s no credible competitor forcing the company to up its game. Intel’s biggest rival, AMD, has lately switched its focus to GPUs – although it is set to challenge Intel on the desktop once more with its Ryzen chips later this year.
Experts say that Intel has effectively been competing with itself for a number of years. “If Intel can produce something that’s really amazing and all of a sudden PC buyers want to upgrade every three years instead of every five, that moves the whole market, and that’s more important to Intel than what AMD is doing,” said Kanter.
And despite promising early benchmark results from Ryzen, AMD is not expected to challenge Intel on high-end performance. “The world is waiting to see what frequencies the Ryzen gets, because if AMD can come in with a noticeably better frequency and better per-core performance that would be really surprising,” said Kanter.
“My expectation is that it will come in with something that looks roughly competitive with Ivy Bridge or Haswell, so maybe 20% behind, and at a competitive price – although AMD has been keeping its cards close to its chest.”
Either way, analysts believe that Intel will remain the dominant presence on the desktop for some time. “The desktop market is still significant and it’s important to Intel because it’s a very high-margin business and there will be some high-performance features,” said Hung. “I don’t see Intel stepping away from desktops in the near future.”