A gamer plays a first-person shooter while monitoring FPS and CPU/GPU performance on a widescreen display.
A 240 FPS target is a frame-time budget, and an average of 240 doesn't guarantee you're meeting it. A recent XDA Developers piece by Hamlin Rozario argues that at 200+ FPS in esports shooters, the CPU can become the limit. In that case, lowering graphics settings or buying a faster GPU may do little. Rozario's broad claims come from his own experience. The piece includes no benchmarks, so treat them as a hypothesis to test on your own PC. The method for testing it is below.

A gamer plays a first-person shooter while monitoring FPS and CPU/GPU performance on a widescreen display. The arithmetic​

FPS is the inverse of frame time. Rozario's figures check out:

Frame rateInterval between frames
60 FPSabout 16.7 ms
120 FPSabout 8.3 ms
180 FPSabout 5.6 ms
200 FPS5.0 ms
240 FPSabout 4.17 ms

The article frames 4.17 ms as the time the CPU has to prepare each frame. That is a simplification. The figure is the spacing between delivered frames. CPU and GPU work overlap in a pipeline, so the CPU's share of that window is not fixed.

NVIDIA's developer blog defines "game latency" as the time the CPU takes to process inputs or world changes and submit a frame to the GPU. It describes the span from the start of a frame's simulation to the Present() call as the total time the CPU spends on that frame. CPU-side work, GPU rendering and display timing are related but separate measurements.

Why an average hides the problem​

Rozario's central point is sound: an average can hide slow frames. Intel's guidance on reading CPU benchmarks makes a similar point. It says frame time should be consistent, or the pacing will be uneven and cause stutter. It also says FPS should be read together with frame time.

His example of a 240 FPS average with 1% lows near 100 FPS is illustrative, not a measured result. "1% low" is also not one universal calculation. Tools define and average it differently, so name the tool when you compare numbers. TechSpot's PresentMon guide notes the tool's capture summary reports average, min/max, 1% and 5% lows.

How to tell whether the CPU is the limit​

Don't rely on a generic utilization percentage. Intel's open-source PresentMon is a free tool for Windows that captures frame metrics. PresentMon traces CPU, GPU, and display frame durations and latencies, and works across DirectX, OpenGL, and Vulkan. Two of its metrics matter most here:

  • CPUBusy is how long the CPU spent working on a frame, in milliseconds.
  • GPUBusy is the time the GPU was actively working on the frame, counted whenever at least one engine is executing work from the target process.

A PresentMon GitHub discussion from CapFrameX developers gives a useful rule of thumb:

  • If CPUBusy is close to the average frame time and GPUBusy is lower, you're likely CPU-limited.
  • If GPUBusy is close to the frame time and CPUBusy is lower, you're likely GPU-limited.
  • If both are well under the frame time, a frame limiter or another constraint is probably capping you.

That thread also shows the metric can behave differently across systems. Read it as evidence of a pattern, not proof of cause.

Here is a practical procedure:

  1. Pick a repeatable scene, such as a practice range or in-game benchmark. Capture long enough to include normal variation.
  2. Record average FPS, the tool's stated 1% low, and frame time alongside CPUBusy and GPUBusy.
  3. Change one variable, such as resolution or a GPU-heavy setting, and capture again.
  4. Compare the traces, not a single on-screen FPS reading.
  5. Count how often frame times exceed roughly 4.17 ms. Don't expect every chaotic moment to stay under it.

If GPUBusy drops after a settings cut but frame delivery barely improves, the GPU was probably not the main limit. One source describes a GPU-limited case as one where GPU busy time approaches the frame budget and lowering quality improves frame time.

Where the article overstates​

"Lowering settings won't help" is true only conditionally. Lower resolution, detail or upscaling cuts GPU work, which helps if the GPU was the limiter. It can't speed up a saturated game thread. A faster GPU raises frame rate only when you are GPU-bound.

NVIDIA's Reflex material adds nuance. Reflex aligns engine work to finish just in time for rendering. That removes the GPU render queue and reduces CPU back-pressure in GPU-heavy scenes. So latency and throughput are not the same thing.

Some practical cautions on measurement:

  • PresentMon's README says GPU execution metrics are less accurate with Hardware-Accelerated GPU Scheduling enabled. In a GPU-bound case, GPU active time may read about 0.5 ms too high, depending on the workload and GPU.
  • For OpenGL and Vulkan titles, CPU frame time may be slightly less accurate.
  • PresentMon needs membership in the "Performance Log Users" group or administrator rights. Without that you may see an "access denied" trace error, or incomplete process data.

About the CPUs mentioned​

Rozario names the Ryzen 7 9800X3D and his older 5800X3D, plus an RTX 5090. He offers no head-to-head numbers, so there is no verified upgrade benefit to report. I couldn't retrieve AMD's product pages for this piece, so I'm not restating spec details. His claim that flagship CPUs from a few years ago struggle at 240 FPS is also his own framing.

Takeaway​

  • The 4.17 ms figure is a useful target, but it isn't a hard "CPU deadline."
  • Judge consistency by frame-time percentiles and CPU/GPU timings, not by average FPS.
  • Test before spending. Falling short of 240 FPS doesn't by itself prove a CPU upgrade is needed. Capture a trace, change one variable, and see where the time goes.
 

References

  1. Your CPU has less than 4.2ms to keep up at 240FPS, and that's harder than it sounds XDA 2026-10-30T20:30:21+00:00
  2. README.md github.com
  3. github.com github.com