SAE / Dividiti - Efficient, reliable and cheap computing -- everywhere
Problem and solution
Everywhere -- from tiny computers embedded in “things” to enormous supercomputers -- increasing the efficiency and decreasing the cost of computing is critical to innovation and wellbeing. Unfortunately, unlocking the optimisation potential is hard due to the sheer complexity and poor understanding of trade-offs (e.g. speed vs. energy efficiency vs. accuracy). This typically results in overly-expensive, under-performing and energy-hungry computer systems. Combined with the cost and time-to-market pressures, this in turn leads to only incremental improvements in the efficiency and cost over product generations.
The cTuning foundation have developed Collective Knowledge (CK), a disruptive methodology for designing and optimising computer systems. Similar to Wikipedia, CK invites a broad community to share representative programs, data sets, tools and predictive models as reusable components, crowdsource and reproduce experiments, and apply predictive analytics to continuously grow knowledge about optimising computer systems.
How did SAE help?
The SAE project TETRACOM helped in two significant ways. First, it provided know-how and €50K funding to mature the CK solution, and release it under a permissive licence. Second, CK was validated with ARM, the world-leading supplier of microprocessor technology. Using CK, ARM was able to obtain valuable insights into performance of its products in a fraction of the time required by conventional analysis. This validation was essential to showcase the potential of CK to spur the design of next generation, high performance and energy efficient computer systems.
The experiment demonstrated that adopters of the CK solution will dramatically increase the performance of their products (making them cheaper, smaller, faster, more energy efficient and more reliable), and thus save millions of euros within 2 years and tens to hundreds of millions of euros within 5 years. To exploit the CK solution commercially, a start-up called dividiti was founded in 2015. One year later, dividiti’s clients already include a cloud computing company and an automotive company from the Fortune 50 list. Based on the strong demand for their services, dividiti are projecting the revenue in the region of €300K in 2016. From 2 full-time co-founders, the headcount will increase to 4 full-time staff and several part-time staff by the end of 2016. It is expected that the headcount will double and the revenue will quadruple every year from 2017.