Why GPUs Clock Stretch Under Low Voltage When Benchmarking

This article explores the phenomenon of clock stretching in modern graphics cards when subjected to low voltage during hardware benchmarking. It explains the mechanics behind why a GPU reduces its actual operating frequency to prevent system crashes when starved of power, the difference between reported and effective clock speeds, why heavy benchmark workloads trigger this protective behavior, and how users can detect and resolve the issue to restore true performance.

What is Clock Stretching?

Clock stretching is a hardware-level protection mechanism found in modern GPUs and CPUs. When a graphics card experiences clock stretching, it temporarily extends the duration of its clock cycles without updating the software-reported frequency. As a result, monitoring software displays a high target clock speed, while the actual, effective clock speed drops significantly. The GPU looks like it is running fast, but its actual work output decreases.

The Physics of Low Voltage

Graphics cards consist of billions of microscopic transistors that require a specific amount of electrical voltage to switch states rapidly and reliably. When enthusiasts undervolt a GPU to reduce temperatures and power consumption, they reduce the safety margin of this voltage. If the voltage is set too low for a given frequency, the electrical signal slows down. The transistors cannot switch fast enough to keep up with the target clock speed, which normally leads to calculation errors and an immediate system crash.

Hardware Protection Against Crashes

In older hardware, insufficient voltage simply caused a driver crash or a black screen. Modern GPU architectures, however, utilize fast-acting adaptive clocking systems. The graphics card constantly monitors its own voltage grid at the microsecond level. When it detects that the voltage is too low to sustain the requested frequency, it instantly stretches the clock signal—essentially skipping clock ticks. This gives the transistors the extra nanoseconds they need to complete their operations safely, prioritizing system stability over raw performance.

Why Benchmarking Exposes the Issue

Synthetic benchmarks are designed to push computer hardware to its absolute limits, creating massive, sustained loads and sudden spikes in power demand (transients). These heavy loads cause brief, microsecond drops in voltage known as "voltage droop" (Vdroop).

If a GPU is already operating on a razor-thin voltage margin due to a user-applied undervolt, the heavy load of a benchmark will cause the voltage to droop below the minimum stable threshold. The GPU's internal sensors detect this dip and aggressively stretch the clock to survive the benchmark. Consequently, the benchmark finishes without crashing, leading the user to believe their undervolt is stable, even though the final benchmark score is anomalously low.

How to Detect and Fix Clock Stretching

Because standard monitoring overlays only read the requested clock state, clock stretching often goes unnoticed until a user sees a drop in benchmark scores or frame rates.

To detect it, you must use advanced hardware monitoring software that tracks both the Core Clock (what the GPU requests) and the Effective Clock (what the GPU actually achieves based on real work done). While running a heavy benchmark at 100% GPU load, compare these two numbers. If the Effective Clock is more than 25 to 50 MHz lower than the Core Clock, the graphics card is clock stretching.

To fix the issue and reclaim lost performance, you must restore the voltage-to-frequency balance. You can do this by either slightly increasing the voltage curve in your tuning software or lowering the target clock frequency until the Effective Clock closely matches the Core Clock under full load.