Why high polling taxes the CPU
Each mouse report is a packet the CPU has to take in and process through the USB/HID driver. At 1000 Hz that's a thousand handlers per second; at 8000 Hz it's eight thousand, and the load scales roughly linearly — sometimes worse at the very top end because of driver overhead. In CPU-bound games (many competitive shooters, sims, strategy titles) those extra cycles come straight out of your game thread. The result isn't smoother input — it's microstutter and uneven frametimes, especially on 4-6 core and older chips.
Telling help from harm
Don't judge by average FPS — watch frame consistency instead: 1% and 0.1% lows plus the frametime graph in MSI Afterburner or your game's built-in overlay. Run the same scene at 1000 Hz, then at 8000 Hz, and compare how flat the frametime line stays and whether CPU usage climbed. If your lows dropped or the line started spiking in a sawtooth pattern, the higher poll rate is costing you, not helping.
What to actually do
Start at 1000 Hz as your baseline — it's a safe ceiling for almost any hardware. If your CPU has real headroom (a recent 8+ core part), try 2000 or 4000 Hz and compare frametimes before reaching for the max. The latency itself drops at any refresh rate, but you'll most easily notice the gain from 8000 Hz on a fast 240 Hz+ monitor; at 60-144 Hz it's usually lost behind the panel's own response time. And for wireless mice, factor in the extra battery drain and heat — that's part of the cost too.
Verdict
8000 Hz is a tool for a specific setup: a strong CPU, a 240 Hz+ display, and competitive play where every millisecond counts. On mid-range or weaker hardware it's more likely to add stutter than give you a felt edge. Test it on your own rig with the frametime graph and keep whatever rate runs cleanest — for most people that's 1000-2000 Hz.