Intel's (NASDAQ:INTC) competitive positioning in the mobile market will improve significantly as both design and manufacturing process has undergone significant improvement over prior generations. However, benchmarks of the upcoming Qualcomm Snapdragon 810 have emerged, and we won't get further details on the actual performance of Intel's mobile SoCs until the Intel Developer Forum (September 9th- 11th). Intel has addressed all the practical issues (power usage, packaging, performance). However, factors like pricing, GPU performance, additional CPU cores may buy Qualcomm (NASDAQ:QCOM) enough time until it reaches 16nm tech.
The race to develop the most competent mobile solution is on.
The technological leap in Broadwell Core M
When it comes to CPUs, performance improves incrementally. Usually, Intel works on a tick-tock cycle, in which it improves the manufacturing tech, and then later it significantly improves on actual design. But what's really interesting about Broadwell Core M is Intel's commitment to developing a custom mobile solution that improves upon design, while improving upon manufacturing process. A combination of smaller packaging, further power efficiency, improvements to the instructions per cycle (IPC) and lower TDP (thermal design power) drive significant efficiency gains.
Intel addresses practicality by ensuring that Intel's upcoming line-up of CPUs for mobile won't need fans to keep thermal output in check. This will improve Intel's odds of penetrating into the tablet and smartphone segment, as a smaller CPU will allow for thinner form factor devices. Improving upon the thinness of a smartphone/tablet device is a big priority for mobile OEMs, as it helps them to market the improved portability of their device.
Two of the distinct design improvements for mobile come in the form of different burst modes, paired with off-switching of the GPU. The burst mode PL3 (refer to figure above) drains the battery quickly; quick burst will last for milliseconds. The quick burst may help with applications that require a massive burst of CPU in a short period of time (like loading up an application). AnandTech explains GPU off-switching:
While Intel has been able to clamp down on GPU idle power consumption over the years, they are increasingly fighting the laws of physics in extracting more idle power gains. At this point Intel can significantly scale down the frequency and operating voltage of their GPU, but past a point this offers diminishing returns. Transistors require a minimum voltage to operate - the threshold voltage - which means that after a certain point Intel can no longer scale down their voltage (and hence idle power consumption) further. Intel's solution to this problem is both a bit brute force and a bit genius, and is definitely unlike anything else we've seen on PC GPUs thus far. Since Intel can't reduce their idle voltage they are going to start outright turning off the GPU instead; the process of duty cycling.
When you look at the above figure, with DCC (Duty cycle control) Intel shows that it can perform at lower power, by operating the graphics around 12.5% of the time, which generates power savings. Intel's performance specs are much closer to where they need to be, as the anticipated power efficiency gains from this is 20% lower power usage from traditional blends of the 14nm Broadwell family.
Snapdragon 810 performance specs have been leaked
Snapdragon 810 is slated to be released in 2015. The timing coincides with the early spring launch cycle of the Samsung Galaxy series. However, Intel Broadwell Core M will also be released in the same time frame, which puts investors in a bit of a tough pinch when determining how well Qualcomm's high-end solution will perform. Snapdragon 810 will use 20nm tech, which is inferior to Intel's 14nm tech. On the graphics spectrum, Intel's solution may be inferior, but on the CPU side I'm fairly confident that even if Intel's app processor design is somewhat inferior, the added density, paired with improved instructions per cycle, will drive enough performance to keep Intel in a solid position to compete with Qualcomm in 2015.
According to GSM Arena:
The Snapdragon 810 features four Cortex-A57 and four Cortex-A53 processor cores which provide 25-55% increase in performance over the Snapdragon 805 and draws 20% less power. In terms of GPU, the Snapdragon 810 has a brand new Adreno 430, which is advertised as 30% faster than the Adreno 420.
Qualcomm plans to compete with Intel's Broadwell Core M, through additional cores, along with a reduction in power draw of 20% (the reduction in power draw most likely comes from power saving features, paired with the better transistor density that goes with the transition from 28nm to 20nm). The Adreno 430 a step-up from the Adreno 420 (slated to be released with Snapdragon 805), which will most likely get its performance boost from the transition from 28nm to 20nm tech.
Qualcomm can remain competitive as it has the most superior modem on the marketplace. Intel's XMM 7260 is expected to be inferior to Qualcomm's Gobi 9x35. The packaging of better broadband solutions may keep Qualcomm from completely capitulating to Intel's manufacturing tech gains. Furthermore, Qualcomm's short-term solution of adding more cores to a CPU won't be "power efficient," but it will pad benchmark results.
Intel's demonstration of Broadwell Core M improved upon many of Haswell's shortcomings when it came to mobile design. However, the economics of the mobile market are somewhat different. Qualcomm has a fully competent CPU/GPU/modem solution in store, paired with semi-custom designs that can address NFC and wireless charging.
Over the short-term Qualcomm has been able to identify a couple ways to address performance. TSM's 16nm node using FinFET (vertical transistors) will produce in massive volume in 2015, as the majority of tape-outs are expected to come in during the 2015 timeframe. By then, Intel will already be on its second generation of FinFET, and will have more time to develop better GPU/CPU/modem designs. Therefore, Intel is significantly closer to competing with Qualcomm on both price and performance, and this could weigh on Qualcomm's performance further out into the future (which I will address in a future article).
Disclosure: The author has no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
The author wrote this article themselves, and it expresses their own opinions. The author is not receiving compensation for it (other than from Seeking Alpha). The author has no business relationship with any company whose stock is mentioned in this article.