Conceptual illustration: Mais FPS, created with AI.
Frame generation sends the counter soaring, yet aiming still feels heavy. Or a new mouse advertises faster reports, but the game on your TV feels unchanged. Both upgrades can deliver their promised feature while leaving another part of the journey slow.
The question is what happens between your hand moving and the screen showing the result. Frame generation can make that result look smoother. Whether it arrives sooner depends on the whole chain.
Follow a command to the screen
A mouse or controller detects input and reports it. The game reads that information, runs its CPU-side work and submits rendering. Work may wait in a queue before the GPU completes it. Frame generation and presentation scheduling can add their own processing. The display receives the signal, processes it, scans out the image and changes its pixels.
NVIDIA’s latency overview separates peripheral, PC and display stages. Some work overlaps, so this is a map, not permission to add unrelated review measurements. Network ping measures the server round trip. It can delay online outcomes, but it is a different quantity from local click-to-screen latency.
Polling is one stop on that journey
A 1,000 Hz reporting rate has a nominal interval of 1 millisecond; 8,000 Hz means 0.125 milliseconds. That arithmetic describes report spacing, not total mouse latency, let alone the entire game’s response. Sensor sampling, button processing and firmware still matter.
Higher polling can deliver fresher input, but Razer acknowledges CPU and game-engine limitations. More reports require processing; some setups become less consistent. Its wireless specifications also show shorter battery runtime at higher rates. Test your game’s frametimes while moving the mouse before assuming 8 kHz is a free upgrade.
Controllers need model-, connection- and platform-specific comparisons. A cable or wireless label alone cannot establish which configuration responds faster. Microsoft’s Dynamic Latency Input documentation describes aligning supported controller reports with game input reads, but explicitly limits DLI to Xbox consoles. Do not transfer that behavior to a PC merely because the controller looks identical.
The engine and GPU still have to answer
A high reporting rate cannot force a busy game thread to read the next input sooner. Likewise, frames waiting in a render queue can contain older input by the time you see them. This is where latency-focused scheduling matters.
NVIDIA Reflex Low-Latency Mode aligns engine work with rendering to reduce queuing. AMD Anti-Lag 2 integrates into supported games to reduce latency in GPU-bound situations. These are latency tools, with hardware and game requirements; enabling frame generation is a separate decision.
Upscaling can reduce rendering workload by reconstructing from a lower-resolution image. Frame generation inserts additional images between rendered output. A generated image does not mean another complete game-simulation update. Consider 60 rendered frames feeding a nominal 2X output of 120 images: that is multiplier arithmetic, not measured performance or a promise of native-120 responsiveness. Interpolation can add delay while constructing and scheduling extra images. Generation also consumes resources, so measure the base again after activation.
Know what today’s multiplier actually counts
NVIDIA’s DLSS 4.5 documentation defines 6X as one rendered frame plus five generated on compatible RTX 50-series configurations. Dynamic mode adjusts the multiplier toward a target. Intel XeSS 3 supports up to three generated frames on Intel hardware.
AMD’s SDK 2.3 requirements distinguish RX 7000 support for Upscaling 4.1.1 from ML Frame Generation 4.0.1’s RX 9000-and-newer requirement; analytical fallback support is broader. A shared brand name does not make every feature interchangeable.
AMD recommends at least 60 FPS before interpolation for best results. Improve an unpleasant base before multiplying it. This is implementation guidance, not a universal cliff between 59 and 60 FPS.
Your display gets the final word
Pixel response describes color transitions. Display input lag concerns when incoming video becomes visible; neither figure alone measures your whole system. RTINGS tests these separately. A fast panel response cannot cancel processing earlier in the chain.
Check the active refresh rate and the relevant picture mode. Sony’s support guidance recommends Game mode and model-dependent processing changes to reduce delay. TV motion smoothing is a separate stage from the game’s frame generation.
FPS above refresh rate can still reduce latency by providing newer rendered information; Blur Busters demonstrated this with V-Sync off. Tearing, GPU saturation, queuing, VRR and frame caps change the tradeoff. “Above my monitor’s Hz is always wasted” and “uncapped is always best” are both poor universal rules. Generated output also should not be mistaken for that same increase in rendered input updates.
Compare the experience fairly
Keep resolution, upscaling, graphics settings and synchronization consistent. Where supported, compare frame generation off and on with the same Reflex, Anti-Lag 2 or XeLL setting. Otherwise, the latency helper can take credit for a misleading result.
PC Gamer’s March 31 test reported lower PC latency with generation in Dragon Age: The Veilguard, but Reflex was off in its no-generation baseline. That external test does not isolate generation’s contribution.
Record rendered and displayed output separately where supported. PresentMon’s definitions also distinguish PC latency from click-to-photon measurements. Check what the number includes before comparing it with a monitor review.
Then repeat familiar aiming, camera turns and interface movement. Change polling, display mode and generation separately; inspect frame pacing and visual artifacts as well as response. This is a suggested workflow, not a MaisFPS hands-on test.
Keep the configuration that feels responsive and looks good. In a demanding adventure, generated frames may be a welcome improvement. In a precision-heavy match, lower latency may matter more. The counter should help explain the experience, not overrule your hands.
