Raw Platform Performance
The same can be said about memory architecture and latency. With the exception of Apple’s solution, today’s MSoCs have pretty miserable performance. Apple’s bus architecture is unique among current MSoCs, and it has talent from P.A. Semi (started by the lead designer of the DEC Alpha and StrongARM CPUs) and Intrinsity (started by Paul Nixon, who headed Exponential Technology’s x86 CPU project; Paul is now with TI, but in the MCU division).
We have yet to see a demonstration of a high-performance bus from ARM, TI, or Qualcomm. The Tegra lineup from Nvidia doesn't have a memory bus that is significantly different from the competition (its shipping Tegra 3 still has less memory bandwidth than last year’s Apple A5). But we're not going count Nvidia out because it has experience with NV2’s ring memory bus, a proven track record with the nForce2 platform, and experience with crossbar memory controllers on its GPUs. It just hasn’t prioritized those things in its Tegra family.
Intel, on the other hand, has always done pretty well with the performance of its platforms (just look at its current Sandy Bridge-E architecture). Again, the challenge for Intel is power consumption, rather than performance.
Raw Wireless Performance
Intel has zero expertise with wireless 3G/LTE. But it has an excellent track record with 802.11-based performance. The company’s Centrino platform was responsible for Intel’s dominance in the PC laptop world.
With the acquisition of Infineon Wireless, Intel gained significant expertise with 3G/LTE. Qualcomm, on the other hand, has always had a strong 3G/LTE offering. But its 802.11 had to be acquired, and that was achieved with the purchase of Atheros. For wireless, Qualcomm has the lead in 3G. The advantage isn’t as clear in LTE, though. Both Intel and Qualcomm also face pressure from Nvidia, which owns Icera, a software baseband company, and Samsung, NTT DoCoMo, Fujitsu, and NEC, collaborating to develop 3G/LTE products.
In the next three years, ARM and Qualcomm need to invest significant resources to advancing CPU performance. Their engineers are navigating uncharted territory, attempting to push technology in the same way that the x86 segment had to struggle through several decades ago. From a pure performance standpoint, and ignoring power consumption, Intel, AMD, and Apple (through its acquisitions, which include Intrinsity and PA Semi) have the most human equity in high-performance mobile computing, particularly in the areas of bus architecture and memory management. Nvidia is a wild card, with a seemingly rich portfolio of technologies—though we’ve yet to see an implementation of its more complex bus and memory designs.
But of course, an emphasis on power efficiency is what makes MSoCs so unique, and that is supposed to be Intel’s unconquered challenge.