OnQ Blog

True Heterogeneous Computing

Sep 3, 2013

Qualcomm products mentioned within this post are offered by Qualcomm Technologies, Inc. and/or its subsidiaries.

Some processor manufacturers are discussing heterogeneous computing within the limited domain of the CPU, where tasks are switched between CPU cores. We don’t accept this limitation, as we believe the future of true heterogeneous computing goes well beyond the CPU, utilizing the full spectrum of diverse and specialized processors within the modern mobile SoC including the GPU, DSP and multimedia cores. The unique capabilities and features of each of these processors can be used together to ensure the right task is assigned to the most efficient processor for the job, maximizing the performance and power efficiency beyond what could be accomplished on a CPU’s general computing core alone.

For example, audio and object detection are more efficiently executed on the low power DSP, while image composition and processing are better suited for the massively parallel GPU. Because different algorithms will align to different processor architectures, we don’t agree that adding another CPU core will increase the performance or power efficiency of the overall system. Rather, we have achieved increases in both performance and power efficiency by enabling the right tasks to shift from the CPU to other more efficient processing cores.

Qualcomm solves this balancing act in mobile devices by engineering the CPU to take full advantage of Asynchronous Multiprocessing or aSMP. This power efficiency feature—found only in  Qualcomm Snapdragon mobile processors, products of Qualcomm Technologies, Inc.—is designed to allow each CPU core to operate independently, dynamically scaling the voltage and frequency, to meet specific performance needs. This is designed to eliminate the need to have different sized CPU’s or clusters of CPU cores working at different times to save power or increase performance.

Other processor manufacturers’ designs lack this precision control, so they are forced to add small cores to compensate for the large, power guzzling cores needed for peak performance. In addition, even though the competitors shifts tasks between the high performance/high power consuming cores and lower performance / lower power consuming cores, all the cores process the tasks in exactly the same way, similar to hammering nails with a small and a big hammer. As long as you are only dealing with nails, that approach is adequate. However, today’s mobile processors have to deal with much more dynamic tasks and that’s why Qualcomm Snapdragon processors are built around a true heterogeneous computing model, literally reaching into a full toolbox of specialized processors to ensure the most efficient operation possible.

A true heterogeneous architecture allows the CPU, GPU, DSP and multimedia subsystem as well as other specialized cores such as camera ISP to work more effectively and efficiently together to deliver higher performance as well as lower power consumption compared to using the CPU alone. This leads to an overall better user experience for devices powered by Snapdragon processors. 

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries ("Qualcomm"). Qualcomm products mentioned within this post are offered by Qualcomm Technologies, Inc. and/or its subsidiaries. The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.

Adam Kerin

Sr Manager, Marketing