- LightSolver unveils Laser Processor Unit designed to unravel physics equations at excessive pace
- Laser computing system makes use of optical reminiscence to keep away from processor and reminiscence information bottlenecks
- Firm begins providing early lab entry with roadmap to at least one million variables by 2029
Laser-based computing is rising as a possible challenger to established high-performance computing techniques, with LightSolver revealing what it calls a Laser Processing Unit (LPU).
The Tel Aviv-based startup says its {hardware} can instantly clear up partial differential equations, a elementary class of issues in physics and engineering.
In contrast to quantum techniques or GPU clusters, the method relies on a grid of lasers performing in unison.
Mapping equations onto the LPU
LightSolver says its {hardware} can map equations comparable to the warmth equation (which fashions warmth move) and Schrödinger’s equation (a core equation in quantum mechanics) instantly onto the LPU.
By utilizing the pure properties of lasers as electromagnetic waves, LightSolver’s tech avoids lots of the constraints present in digital techniques.
This contains bottlenecks attributable to information switch between reminiscence and processor.
In response to LightSolver, the LPU operates with embedded optical reminiscence. In follow, this implies laser states are preserved contained in the resonator, permitting every calculation step to construct on the final with no need to maneuver information externally.
The result’s constant-time iteration measured in nanoseconds, impartial of downside measurement.
Whereas LightSolver compares its platform to each high-performance and quantum computer systems, it stresses that it’s not constructed on photonic chips.
Immediately’s photonic processors are two dimensional, whereas its laser-based design is three dimensional, which the startup argues will permit larger scalability.
LightSolver has set a roadmap to achieve 100,000 variables by 2027 and a million by 2029.
The corporate is already starting to make its {hardware} accessible for testing. An Alpha model of its processor and a digital emulator are being provided to researchers by what it calls its LPU Lab.
LightSolver believes this step will assist scientists and engineers experiment with the know-how earlier than it reaches industrial grade.
“Classical computer systems digitize analog nature, and we pay the value in longer run instances and wasted power. By operating large-scale physics simulations on a bodily machine, we are able to clear up them extra effectively than any HPC or quantum system accessible at present,” mentioned Ruti Ben-Shlomi, CEO and co-founder of LightSolver.
Analysis papers have been printed on the topic, and displays have been given at computing conferences, together with ACM Computing Frontiers 2025.
The corporate has additionally shaped partnerships with simulation software program suppliers and is working with HPC facilities and nationwide labs.
