Computing aspires to an exascale level
Simultaneously supporting AMD x86 and ARM computing platforms, hardware vendors are aiming to enable practically identical user experiences across almost all computing and communications devices. Features and applications should be virtually indistinguishable for all supported devices, whether accessed using a smartphone, tablet, mobile device, 2-in-1 PC, ultra-thin notebook, desktop gaming computer, or home server.
Unifying collaboration between diverse computing hardware and software ecosystems is vital to the success of ambidextrous computing. However, a new computing system architecture specifically designed to harness the capabilities of diverse computing platforms, microprocessors, devices, and software is needed to make this possible. Fortunately, this is already happening...
Heterogeneous System Architecture
A new open computing system architecture called Heterogeneous System Architecture (HSA) is emerging as the next-generation industry standard that integrates different types of microprocessors and compute elements. HSA's architecture and specialised programming tools enable these different compute elements to work together in harmony and eliminates inefficiencies with sharing data and routing computing tasks. This could be a key innovation for exascale performance.
Heterogeneous computing using HSA optimises processing by using the best processing element for the workload, delivering better efficiency through low power consumption, smaller form factors, and minimised technology complexity. HSA's advantages may also boost supercomputing efforts by harnessing the "contributed" compute processing power of hundreds of thousands of computing devices.
Imagine how much processing power would be available if every desktop, notebook, tablet, and smartphone in the world could share its processing capabilities among each other. This may sound impossible, but there are already many web-based applications doing exactly that. With ambidextrous designs and heterogeneous systems architecture, they could do so far more effectively.
Embracing "IP diversity"
"Diversity is the leaven of creative effort," according to famed scientist Theodosius Dobzhansky—and harnessing technology diversity may be the key to unlocking the compute power needed to achieve exascale supercomputing.
The growing momentum of ambidextrous designs and heterogeneous architectures may compel the entire computing industry to embrace a new era of IP diversity, enabling competitors to collaborate towards a shared goal, while differentiating intellectual property through innovation. Working together like the different parts of the human brain, open software architectures, frameworks, and programming tools could enable harmonious collaboration among proprietary technologies from a wide array of different manufacturers.
The goal is to create a whole that is greater than the sum of its parts. A new industry mindset fostering and embracing interoperability among diverse processing technologies—while maintaining their differentiating value—could hold the key to achieving exascale computing near the end of the decade. Until then, the human brain remains the undisputed processing champion.
- Sasa Marinkovic
Head of Technical Marketing, AMD
|Related Articles||Editor's Choice|