You are almost certainly CPU-limited, not GPU-limited. At 1080p with every graphics setting on low, the RTX 5070 renders frames faster than the Ryzen 7 9700X can prepare them, so the processor, not the graphics card, sets your frame rate. A 144 Hz target also depends on your monitor, cable, and display settings actually running at 144 Hz. The remaining gap usually comes from memory configuration, background processes, or game-engine overhead you cannot tune away.
Most people assume low settings should unlock high frame rates because rendering gets cheaper. That logic holds only until the GPU stops being the bottleneck. Below that point, lowering settings buys almost nothing because the frame is waiting on the CPU. Understanding where that crossover sits, and what moves it, explains why your specific combination lands near 144 FPS instead of well above it.
The CPU-GPU frame-time handoff
Every frame passes through two stages. The CPU runs game logic, physics, animation, audio, and draw-call submission, producing a frame time in milliseconds. The GPU then shades and presents that frame, producing its own frame time. Your effective frame rate is set by whichever stage takes longer, because they pipeline and the slower one gates throughput.
At 144 FPS, each frame must complete in about 6.94 milliseconds (1000 divided by 144). That budget is tight. On low settings at 1080p, GPU frame times on a card in the RTX 5070 class often fall well under 5 milliseconds in CPU-light scenes. The CPU frame time on a mainstream 8-core chip frequently sits between 6 and 9 milliseconds depending on the title and scene. When CPU time exceeds GPU time, the GPU idles waiting for work, and you see utilization well below 100 percent.
That idle GPU is the diagnostic signature. If your graphics card sits at 60 to 80 percent utilization while frame rate plateaus near 144, the processor is the limiter. A plain-language version: your graphics card finishes its homework early and waits for the processor to hand over the next assignment.
Draw calls matter here. Each object the CPU submits to the GPU costs time on the processor. Low settings reduce shading cost but do not remove objects, so draw-call counts stay high. This is why dropping from high to low can raise frame rate modestly and then flatten out.
What 144 FPS actually costs in milliseconds
Frame rate and frame time are reciprocal, and small time differences look large in FPS. A CPU frame time of 6.5 milliseconds yields about 154 FPS. Push it to 8 milliseconds and you drop to 125 FPS. The 1.5-millisecond difference between those cases is invisible on a stopwatch but decides whether you clear your target.
This explains the "barely hitting 144" feeling. You are not far off; you are a fraction of a millisecond per frame short. Consistency also matters. Average frame rate can read 144 while one-percent-low frame rates dip to 90, producing stutter the average hides. Cap frame rate slightly below your refresh rate to stabilize pacing rather than chasing a raw number that flickers.
Refresh rate itself is a common silent failure. A 144 Hz monitor running at 60 Hz in the operating system caps you near 60 FPS regardless of hardware. Verify the active refresh rate in display settings before diagnosing anything else. Cable and port matter too: some older HDMI paths cap refresh rate at 1080p, while DisplayPort generally carries the full rate.
Configuration bottlenecks beyond the graphics card
Memory is the most underappreciated limiter on a Ryzen platform. The 9700X pairs with DDR5, and its infinity fabric and memory controller reward tight timings and specific speeds. Running memory at default JEDEC speeds instead of enabling the advertised EXPO profile can cost meaningful frame time in CPU-bound games. Two matched sticks in the correct slots outperform four mismatched sticks because of how the memory controller schedules access.
Thermal throttling quietly raises frame times. If the CPU downclocks under sustained load, its frame time grows and frame rate sags after the first minute of play. Check sustained clocks, not just peak boost. Power delivery and motherboard VRM quality also influence how long the chip holds boost.
The following flags usually indicate a non-GPU problem:
GPU utilization below roughly 85 percent while frame rate is capped near your target. Frame rate that starts high and settles lower after several minutes, suggesting thermal or power throttling. One-percent-low frame rates far below the average, pointing to CPU spikes or background load. Refresh rate in the OS set below the monitor's rated rate. Memory running at default speed with the EXPO profile disabled.
Comparison of likely limiters
| Limiter | Typical symptom | Rough frame-time impact | Fixable in software |
|---|---|---|---|
| CPU draw-call limit | GPU below 85% use, flat frame rate | Sets the ceiling near 6-9 ms | Partly, via settings and background load |
| Memory at default speed | Lower average and worse lows | Adds roughly 0.5-1.5 ms | Yes, enable EXPO |
| Thermal throttling | Frame rate sags over time | Grows with sustained load | Yes, cooling and power limits |
| Display at 60 Hz | Hard cap near 60 FPS | Caps entire pipeline | Yes, set refresh rate |
| GPU-bound at higher res | GPU near 100% use | GPU time dominates | Yes, lower resolution or settings |
| Engine overhead | Plateau no setting changes | Fixed per title | No, baked into the game |
The engine-overhead row is the counterintuitive one. Some titles have hard CPU costs per frame that no graphics setting removes, so a "low settings" expectation of unlimited frame rate is simply wrong for that engine.
Timeline of where frame time goes
| Stage | What happens | Approximate share of a 6.94 ms budget |
|---|---|---|
| Input and simulation | Game logic and physics tick | Varies, often 1-3 ms |
| Draw-call submission | CPU hands objects to GPU | Often the largest CPU cost |
| Driver overhead | Command translation and batching | Under 1 ms typically |
| GPU shading | Pixels rendered at low settings | Frequently under 3 ms |
| Presentation | Frame queued and displayed | Small but adds latency |
| Waiting | GPU idle for next CPU frame | The visible gap you feel |
Read the table top to bottom and the pattern appears. The CPU-side stages together can consume most of the budget before the GPU does much work. That is the definition of a CPU limit.
Misconceptions that send people down wrong paths
The biggest myth is that lower settings always raise frame rate. They raise it only while the GPU is the limiter. Once the CPU gates throughput, settings changes move the needle slightly or not at all. Another myth: a newer or stronger GPU will fix a CPU limit. It will not, because the processor still prepares frames at the same rate. Upgrading the graphics card in a CPU-bound scenario can even lower average utilization further while frame rate stays flat.
A subtler error is trusting average FPS alone. Averages hide stutter, and the number that reflects smoothness is the low percentile. Professionals watch frame-time graphs, not headline FPS, because a steady 130 feels better than an erratic 145.
Finally, resolution assumptions mislead. Many people test at 1080p expecting CPU limits to vanish, but lower resolution makes the CPU limit more visible, not less. Resolution shifts load toward the GPU; it does not reduce CPU draw-call work. The professional caveat: a machine that is CPU-limited at 1080p is often perfectly balanced at 1440p, where the GPU catches up and frame rate becomes GPU-determined and more stable.
Practical steps to close the gap
Work from cheapest and most likely to most involved. Verify the display first, then memory, then system load, then in-game and driver settings.
Confirm the monitor runs at 144 Hz in the operating system and that you are using DisplayPort or a cable rated for the resolution and rate. Enable the EXPO memory profile in firmware so DDR5 runs at its rated speed, and confirm two sticks sit in the recommended slots. Close background applications and overlays, then recheck GPU utilization; below 85 percent confirms a CPU limit. Monitor sustained CPU clocks and temperatures during play to rule out throttling. In-game, test a frame-rate cap a few frames below refresh to stabilize pacing, and try enabling features that shift work to the GPU only if they do not add CPU cost. If you remain CPU-bound and want more headroom, prioritize faster memory tuning and a platform with stronger single-thread performance over a graphics card upgrade.
Set expectations with thresholds. If GPU utilization stays under about 85 percent and frame rate plateaus near 144, you are at the CPU ceiling for that title. Clearing that ceiling requires faster CPU-side throughput, not lower graphics settings. A cap at 120 to 141 FPS often produces smoother play than an uncapped target you cannot hold.
Limits of what tuning can achieve
Some frame-time cost is fixed by the game engine and cannot be tuned away on any hardware. Every title has a floor for CPU frame time given a scene's object count and simulation complexity. On a mainstream 8-core chip, that floor sits close to your 144 FPS target in demanding scenes, which is why "barely hitting" is the honest expected result rather than a fault.
Driver and firmware versions also shift results between updates, so a measurement from one build may not hold on another. Treat any single frame-rate figure as a snapshot, not a guarantee. The mechanisms above are stable: CPU-gated throughput, memory sensitivity, thermal behavior, and display configuration. The exact numbers vary by title, scene, and software version, and that variability is the real limit on precise prediction.