Glossary
A working vocabulary for clocks, heat, and overlays
These entries are written for msiafterburner.io readers: connect the word on a monitoring strip to the next honest measurement, not a slogan.
Turn definitions into a reading path
Bookmark this page for vocabulary, then jump to longer explanations when a term shows up in monitoring or in a forum thread without context.
Power limit
A cap on how much electrical power the GPU package is allowed to draw or report against, enforced by the driver/VBIOS.
When you hit the power limit, clocks may drop even if temperatures look fine. Raising the power limit can improve sustained boost at the cost of heat and PSU headroom. On many cards it is the dominant limiter under gaming loads.
Example: You see “PerfCap Reason: PWR” in monitoring while FPS dips. Increasing power headroom slightly stabilizes clocks.
Related: Thermal throttling, VF curve, PSU headroom
Thermal throttling
Clock reduction triggered because a temperature sensor crossed a safe policy threshold.
GPUs protect themselves by lowering voltage/clocks when hotspot or edge temperatures rise. Throttling is not always “bad”; it is a safety mechanism, but frequent throttling means your cooling solution or case airflow is the bottleneck.
Example: After ten minutes of ray tracing, core clock drifts down while fan speed pegs at 100%. You are thermally limited.
Related: Fan curve, hotspot, case airflow
VF curve (voltage-frequency curve)
A mapping that tells the GPU which frequency to target at a given voltage point, subject to limits and load.
Modern GPUs boost within a constrained space. Editing the curve is a precise way to undervolt or reshape boost behavior, but mistakes can cause hard crashes if the curve is unrealistically aggressive.
Example: Lowering voltage at 1950 MHz while keeping stability reduces temperatures at the same effective performance.
Related: Undervolting, OC Scanner, instability
Undervolting
Running the same or similar clocks at lower voltage to reduce heat and fan noise.
Undervolting is popular because it trades margin for efficiency. The risk is instability under short transient loads if the curve is too tight. Always validate with workloads that resemble your real games, not only a quick benchmark.
Example: You move the VF point at 0.95V down slightly and GPU temperature drops 6–10°C with identical FPS in your favorite title.
Related: VF curve, stability testing, power limit
OC Scanner
An automated routine in Afterburner that proposes a curve on supported NVIDIA GPUs.
It estimates headroom using internal heuristics. Results vary by silicon quality (“lottery”), cooling, and case thermals. Treat it as a starting line, not a certificate of stability.
Example: Scanner suggests +120 MHz equivalent curve shape: you validate and find +90 MHz is stable in compile-heavy games.
Related: VF curve, memory overclock, validation
Memory clock offset
An adjustment to effective memory transfer rate, often expressed as an offset from stock.
Memory errors can manifest as texture sparkles, driver resets, or sudden crashes. Memory tuning is frequently more sensitive than small core changes. Increase offsets gradually and test with workloads that stress VRAM bandwidth.
Example: +500 MHz effective looks fine in FurMark but crashes in a shader-heavy game. Reduce until shader workloads are clean.
Related: Stability testing, artifacts, VRAM temperature
Fan curve
A function that maps GPU temperature to fan duty cycle or RPM targets.
Good curves prioritize smooth transitions to avoid hunting behavior. Laptop fans may ignore some requests due to OEM EC policies.
Example: You set a gentle slope between 55°C and 75°C so fans rise predictably under long gaming sessions.
Related: Thermal throttling, hysteresis, noise
Frametime
The time between consecutive frames, typically measured in milliseconds.
FPS is an average; frametime shows stutter. RTSS frametime graphs help distinguish pacing issues from average FPS drops.
Example: FPS reads 120 but gameplay feels choppy: frametime spikes show periodic hitching.
Related: RTSS, stutter, pacing
RivaTuner Statistics Server (RTSS)
A companion component commonly used with Afterburner for on-screen display and frame pacing tools.
RTSS provides global and per-application profiles for overlays and scanline sync tuning. It is powerful and worth learning separately from Afterburner’s core tuning tabs.
Example: You enable OSD items for FPS, frametime, and GPU temperature to compare tuning changes objectively.
Related: Scanline sync, OSD, anti-cheat
PerfCap reason
A monitoring hint explaining what limiter is capping performance at a given moment.
Common reasons include power, thermal, voltage reliability, and utilization. Reading PerfCap helps you choose the next tuning action instead of guessing.
Example: PerfCap toggles between PWR and THERM: you improve case airflow before pushing clocks higher.
Related: Power limit, thermal throttling, utilization
Hotspot temperature
A sensor reporting a hotter junction on the GPU die than the edge or average GPU temperature.
Hotspot can be tens of degrees above “GPU temp” depending on sensor placement and workload. It matters for boost behavior and long-term reliability margins.
Example: GPU temp looks fine but hotspot spikes: you repaste or improve mount pressure on a DIY maintenance plan.
Related: Thermal throttling, boost clock
Profile / skin
A saved configuration and UI presentation layer in Afterburner.
Profiles let you store tuning presets. Skins change layout and readability but not underlying capabilities. Keep a known-good default profile export habit if you experiment often.
Example: You save “Quiet gaming” and “Benchmark” profiles and switch intentionally between them.
Related: Fan curve, VF curve
Reported GPU voltage
A sensor or driver-reported estimate of voltage relevant to the GPU core at a given moment, not always a literal physical measurement at every point.
Voltage readouts help interpret VF tuning and power behavior, but naming and accuracy vary by vendor, BIOS, and workload. Memory voltage is a different domain from core voltage. Treat sudden impossible-looking numbers as a reason to cross-check with other limits (power, thermal, reliability).
Example: You undervolt aggressively and voltage looks low yet the card still draws high power. Power limit and boost tables still govern the package.
Related: VF curve, power limit, undervolting
ASIC quality (silicon lottery)
Manufacturing variation between otherwise identical GPU models that changes how much frequency headroom exists at a given voltage and temperature.
Two cards with the same model name are not identical silicon. Copying someone else’s offsets is a starting guess, not a promise. Validation on your own unit is the only honest proof.
Example: Your friend’s “+150 core” profile crashes on your card at +90. Same model, different leakage and bins.
Related: OC Scanner, stability testing, VF curve
Exclusive fullscreen vs borderless windowed
Different presentation paths for games that change how overlays inject, how vsync behaves, and sometimes which GPU is active on hybrid systems.
RTSS and other overlays may appear in one mode but not another depending on the title and anti-cheat. Borderless often routes through the desktop compositor; exclusive can block or alter injection. Use borderless for quick overlay tests when troubleshooting.
Example: Overlay shows in borderless but not exclusive fullscreen. Check RTSS profiles and the game’s fullscreen mode setting.
Related: Frametime, RTSS, scanline sync