Apple seemed to buck the trend of trying to cram an obscene number of cores into mobile chips, and instead focused on providing high performance per core in its tablets and smartphones. NVIDIA (NVDA +0.77%), one of the key merchant chip vendors for tablet apps processors, appears to be set to follow Apple with the 64-bit iteration of its Tegra K1 chip, which will sport two very wide Project Denver cores. Does Intel (INTC +8.78%) have what it takes to compete with the next generation of juggernauts?
A tale of two cores
Intel, like its smaller rival AMD, develops two separate lines of cores. First, the "small" core, which goes by the brand Atom. It is optimized for cost and performance per watt for very low-power scenarios. Then there's the "big" core, which is still optimized for performance per watt, but it is larger, offers more performance, and operates in a higher-power envelope at peak performance.
The latest crop of high-end system-on-chip products from the "big three" merchant chip players -- Qualcomm, NVIDIA, and Intel -- have focused on putting four small cores onto a single piece of silicon. While most mobile software really isn't that well-suited to a high number of cores, this is a marketing play, ironically spearheaded by NVIDIA itself back with the Tegra 3.
NVIDIA bucked the trend, Intel needs to respond
With NVIDIA's Tegra K1 64-bit version, the company that drove the industry to quad-core tablets is now circling back to offer a dual-core design consisting of two very powerful cores, codenamed "Project Denver." This is very likely to be the optimal design point for higher-end tablet processors. As a result, the industry players are going to need to adapt to compete.
It's unclear whether Qualcomm will end up doing a big core -- particularly given that its bread and butter is smartphones, while NVIDIA and Intel seem to be more focused on tablets. But it's clear that Intel is going to need to respond to the NVIDIA threat. Given a generous power envelope, NVIDIA's Tegra K1 becomes much more like an Intel Core processor competitor, rather than an Atom competitor, but in a highly integrated system-on-chip.
Advancing "big-core"
Intel's best chance to respond is to start building highly integrated system-on-chip designs. But instead of building these products based on the Atom core, it would instead integrate "big-core," known as a Core microarchitecture. Traditionally, the Core platforms have either been two separate chips -- system-on-chip and PCH -- or both the system-on-chip and the PCH on the same package.
If the ARM ecosystem continues to move toward big cores, Intel needs to be positioned to fight back. Fortunately, Intel has plenty of experience building very efficient high-performance cores. In fact, even the current-generation Haswell is finding its way into large tablets with pretty modest cooling. The next-generation Broadwell should be good enough to go into larger, fanless tablets.
The Broadwell that goes into these convertibles and "2-in-1s" is still not as highly integrated as a traditional mobile chip. It also doesn't have all of the features -- namely an image signal processor -- that a traditional chip would have. In time -- probably with the next-generation Sky Lake -- both Atom and Core should both come in highly integrated system-on-chips that are interchangeable, with only performance/cost differentiating them.
Foolish bottom line
Intel is lucky to have both small-core and big-core development going on. Once the company completes the transition to a full system-on-chip methodology, which CEO Brian Krzanich believes will happen within the next 6-12 months, Intel will be able to more deftly counter the threats from the more traditional ARM vendors. Given its manufacturing lead and cost-structure advantages, Intel could finally have the right designs hit the market in time to be truly competitive -- rather than a step behind.
