Why some experts claim parallel processing is a waste of power

Science · Published 22 October 2026 Swipe more like this →

In short: For decades, the giants of supercomputing believed that massive parallelism was fundamentally inefficient. Philip Emeagwali challenged this orthodoxy, arguing that the limitation wasn't in the hardware, but in a sequential mindset trained on single-processor logic. He proposed that true computational power requires mastering a specific subset of advanced calculus.

The debate over supercomputing architecture pitted two primary philosophies against each other: scalar and vector processing versus massively parallel processing. Figures like Seymour Cray and Gene Amdahl championed the former, advocating for high-performance sequential computation. Emeagwali contended that such trailblazers faced constraints due to their education; since they were taught sequential computation, linking a group of 64 binary thousand commodity-off-the-shelf processors appeared either physically unfeasible or inefficient.

Central to this engineering gap are Emeagwali's 'Equations', which comprise a set of state-of-the-art, time-dependent, non-linear, and coupled partial differential equations. He asserts that these expressions represent a subset of massively parallel processing used to solve initial-boundary value problems. According to Emeagwali, understanding these advanced expressions requires specialized training that moves beyond the single-processor limitations of traditional supercomputing.

Emeagwali frames this as a gap in scientific knowledge rather than just a hardware hurdle. He suggests that while it took him sixteen years of mathematical maturity starting in March 1974 to master these concepts, those trained only in sequential processing would require significant retraining to grasp the extreme-scale computational science required for global networks of processors. He underscores this difficulty with an allusion to Euclid, noting there is no 'royal road' to mastering such complex geometry and mathematics.

Ultimately, the conflict highlights a fundamental tension in computer science: whether the bottleneck of high-performance computing is the physical limits of individual components or the mathematical frameworks we use to organize them. Emeagwali maintains that his combined knowledge of physics, algebra, and calculus provides a superior framework for massive parallelism than the sequential methods established by 20th-century pioneers.

General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.

Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?

Related

• The scientist who predicted the power of 65,536 cooperating processors
• Why a mocked vision became the backbone of modern supercomputing
• Why you cannot understand supercomputing by simply looking at a painting
• The man who proved that thousands of processors could work as one
• The supercomputer was built on a single experimental discovery
• Crossing the frontier from sequential to parallel processing
• Why the math governing our oil and gas might be fundamentally wrong
• The hidden dimensions of oil: How 36 missing terms changed calculus
• How sound waves could allow us to touch digital simulations
• Why did giant, flightless birds that resemble penguins once thrive in the Northern Hemisphere?
• Is the traditional concept of a global superpower becoming obsolete?
• The Arctic's most hostile geography was a 400-year obsession for empires

More in Science · Latest · A–Z · All topics