A study published in December claimed that a new approach to computing could make all of our devices a heck of a lot faster without having to replace a single component.
Modern technological devices use different chips to handle processing needs, such as the central processing unit (CPU,) graphics processing units (GPUs), as well as hardware accelerators for artificial intelligence (AI) workloads and digital signal processing units, according to Live Science. Right now, most of these components process data independently and in sequence, which is why processing times can sometimes feel slow.
Think of your device like a baton race, where each process has to happen in order but cannot occur simultaneously. Sometimes this creates a bottleneck that scientists now claim they’ve solved, allowing for “simultaneous and heterogeneous multithreading (SHMT), according to the paper’s preprint on the arXiv server.
‘End Of Human Race’: Stephen Hawking Gave Blunt Warning About AI Before His Death | @DailyCaller https://t.co/LBMC7bC8Dq
— KAY SMYTHE (@KaySmythe) May 1, 2023
SHMT “lets different processing units work on the same code region at the same time, while letting them also take on new workloads once they’ve done their bit,” Live Science explained. (RELATED: China Finally Decides To Review Research Misconduct After More Than 17,000 Retractions. This Is A Very Big Deal)
The study’s lead author Hung-Wei Tseng argued that devices don’t need new processors to reach the speeds uncovered in the research. In tests performed throughout the study, tasks were completed 1.95 times faster with 51% less energy used on conventional systems.
In layman’s terms, the processing power on all of our devices could have been significantly faster this whole time if this new methodology actually work in a macro-application. More research is needed to determine exactly how these principles can be implemented in practical settings, but it could be a revolution in technology unseen since we gave up on dial-up.