Dr Saswat Kumar Ram, Associate Professor, Department of Electronics and Communication Engineering, SRM University AP
Almost every digital device we use is governed by a clock. Inside a smartphone processor, a laptop CPU or a microcontroller, a precisely timed electrical pulse ticks billions of times per second, telling every part of the chip when to move to its next step. This synchronous design has been the foundation of computing for more than half a century. But as chips grow larger, faster and more power-hungry, the clock is becoming both a bottleneck and a burden. A long-studied alternative, clockless or asynchronous computing, proposes a radical idea: remove the global clock altogether and let circuits coordinate themselves.
The Role of the Clock
In a synchronous circuit, data moves between storage elements called registers or flip-flops. The clock signal tells all of them when to capture their inputs at the same instant. Between ticks, logic gates perform calculations, and the designer must guarantee that even the slowest path finishes before the next tick arrives. This arrangement makes design manageable and predictable, and it underpins the automated tools that allow engineers to build chips containing billions of transistors.The clock, however, has a cost. It must reach every corner of the chip at almost exactly the same moment, which requires a large and carefully tuned distribution network that consumes a significant share of the chip’s power. The clock ticks whether or not there is useful work to do. And because it must be set for the slowest possible operation, faster parts of the circuit spend time waiting.
What Is Clockless Computing?
Clockless computing, also called asynchronous or self-timed design, replaces the global clock with local coordination. Instead of every component waiting for a universal tick, each stage signals to its neighbours when it has finished and when it is ready to accept new data. This is usually done through handshake protocols: a sender indicates that data is available, the receiver acknowledges it, and the process moves forward. Work flows through the circuit like a relay race rather than a marching band.Because timing is determined by actual completion rather than a fixed schedule, an asynchronous circuit can run as fast as the data and the physics of the circuit allow at any given moment. If an addition happens to be quick, the result moves on immediately instead of waiting for the next clock edge.The Potential Advantages are Lower power consumption: with no global clock distribution and no switching in idle parts of the circuit, asynchronous designs can use energy only when work is actually happening.Average-case performance: circuits can finish as soon as the computation is done, rather than always taking the worst-case time.Reduced electromagnetic noise: synchronous chips produce sharp spikes of current and radio-frequency emissions at each clock edge. Asynchronous circuits spread activity over time, producing a smoother, lower-peak emission profile.Resilience to variation: because circuits wait for completion signals, they can adapt to changes in temperature, supply voltage and manufacturing variation, which are growing concerns at advanced process nodes.Modularity: asynchronous blocks communicate through well-defined handshakes, which can make it easier to combine components without worrying about aligning clocks.Graceful operation at low voltage: some self-timed designs continue to function reliably when the supply voltage drops, which suits energy-harvesting and ultra-low-power devices.
How It Works in Practice
Asynchronous design uses several techniques. In bundled-data approaches, ordinary data wires are paired with a request signal that includes a deliberate delay matched to the logic, so the receiver knows when the data is valid. In more robust delay-insensitive schemes, the data itself is encoded so that the circuit can detect when each value has arrived, often using dual-rail encoding, where each bit is carried on two wires. Completion detection circuits then signal when a stage has finished.Many practical designs are hybrid. The popular globally asynchronous, locally synchronous (GALS) approach divides a chip into several synchronous islands, each with its own clock, and connects them through asynchronous interfaces. This preserves the convenience of conventional design tools within each block while avoiding the need for one global clock across the whole chip.
Where It Has Been Used
Clockless ideas are not purely theoretical. Asynchronous techniques have appeared in research processors, specialised chips and commercial products over several decades. Academic and industrial groups have built asynchronous microprocessors, and the approach has been explored in low-power microcontrollers, secure smart-card style chips where the lack of a clock can make side-channel attacks harder, and communication circuits. Asynchronous interfaces are routinely used in mainstream chips to move data safely between blocks running on different clocks. More recently, the approach has attracted interest in neuromorphic and event-driven computing, where chips process sparse events as they arrive, much as biological neurons do, rather than on a fixed schedule.
Why It Matters Now
Interest in clockless design has been rekindled by several pressures. Power and heat limit how fast and dense conventional chips can become. At the smallest transistor sizes, variation between devices makes it harder to guarantee a single global timing target. Artificial intelligence workloads, especially at the edge, reward designs that process sparse, bursty data efficiently. And billions of battery-powered or energy-harvesting devices need to sip energy and sleep deeply, waking only when something happens. Event-driven, clockless approaches align naturally with these demands.
Links to Embedded Systems and Edge AI
For embedded devices, the appeal is especially strong. A sensor node that does nothing for minutes and then must react instantly to an event is a good match for event-driven hardware that consumes almost no power while idle and wakes up in nanoseconds. Always-on listening for a wake word, anomaly detection in a vibration sensor or spike-based processing of signals from event cameras are examples where asynchronous or event-driven techniques are being explored. Some neuromorphic chips, designed to run spiking neural networks, use asynchronous circuitry to process information only when spikes occur, aiming for dramatic energy savings on certain tasks.
As a result, the benefits of asynchronous design often have to be large enough to justify leaving a familiar and well-supported path. In practice, the most successful adoption has come from targeted uses and hybrid approaches rather than wholesale replacement.
ecosystem may help build the expertise needed.
What Comes Next
Most experts expect clockless techniques to grow gradually rather than overthrow synchronous design. Hybrid GALS architectures are likely to become more common as chips are divided into many domains. Better design tools and open frameworks could reduce the cost of asynchronous design. Event-driven and neuromorphic processors may find commercial niches in always-on sensing, robotics and edge AI. Ultra-low-power devices running on harvested energy could make particular use of self-timed circuits. And as the limits of conventional scaling grow more pressing, designers will be increasingly willing to consider architectural alternatives that were once seen as exotic.
The clock has served digital electronics extraordinarily well. Clockless computing asks whether chips might do better by working at their own natural pace, using energy only when there is something to compute and finishing tasks as soon as they are done. The idea is not new, and it faces real hurdles in tools, training and tradition. Yet with power efficiency, variability and event-driven intelligence now central concerns, it deserves fresh attention. The future of high-speed digital systems may not be a faster clock but, in some places, no clock at all.




