In general terms, polling rate can be described as the rate at which the data generated by the mouse is transmitted from the mouse to the PC via USB. Polling rate is measured in Hz; i.e., number of times per second. The higher the polling rate—the lower the polling interval—the more frequently are the cursor position and any other input events (button inputs) updated, resulting in improved positional accuracy and generally reduced latency. At 1000 Hz, the polling interval is 1 ms, which means the PC receives a new update every 1 ms. At 2000 Hz, the interval is 0.5 ms, at 4000 Hz, the interval is 0.250 ms, and at 8000 Hz, the interval is 0.125 ms.
8000 Hz: The Technology and How to Use It
In the past, multiple mice claimed to be capable of 2000 or even 3000 Hz polling have been released. However, each and every time, these claims turned out to be inaccurate. All of these mice had one thing in common: They were full-speed devices typically incapable of running at polling rates higher than 1000 Hz without modified USB drivers. The X15, on the other hand, is a high-speed device and thus natively capable of polling rates higher than 1000 Hz. Furthermore, it is equipped with a sensor (PMW3389) capable of running at a, for the most part, sufficiently high framerate, which ensures it is not just identical data being sent more frequently.
To set the X15 to 8000 Hz, one needs to make use of its software as the X15 is set to 1000 Hz polling by default. Upon setting the X15 to a polling rate above 1000 Hz, one is greeted by the message shown above. The X15 can be set to 125, 250, 500, 1000, 2000, 4000, or 8000 Hz. However, it is important to note that those values merely denote the maximum applicable polling rate. If the mouse isn't physically moved enough to generate a sufficient number of motion events (for 8000 Hz at least 8000 pixels worth of motion per second), less updates will be transmitted, resulting in a lower effective polling rate. Accordingly, using a sufficiently high CPI step on the X15 is highly recommended. I would advise using at least 1600 CPI, and possibly even higher steps depending on one's effective in-game sensitivity (turn circumference). The higher the turn circumference, the more physical motion is typically generated, and lower CPI is thus required to saturate the polling rate. Conversely, the lower the turn circumference, the less physical motion is generated, and higher CPI is thus required to saturate the polling rate. However, keep in mind that at and above 1900 CPI, smoothing is introduced on the 3389, which in turn increases motion delay. It should be noted that on the X15, click and sensor motion data aren't detached internally. As such, a lower polling rate will increase the click latency compared to 8000 Hz.
In order to get the full benefit out of 8000 Hz polling, certain conditions need to be met. First, it is recommended to have a sufficiently powerful CPU; i.e., one with six physical cores and appropriately high IPC to match. Second, the OS has to be capable of 125 μs or lower interrupt moderation. This is true of Windows 8 or higher, where interrupt moderation on XHCI will typically be 50 μs, but not of Windows 7 and lower, where interrupt moderation is never below 1 ms unless changed manually, which isn't trivial to do. On EHCI, interrupt moderation can be expected to be 125 μs on Windows 8 or higher, which is sufficient but not optimal. Third, it is therefore recommended to plug the X15 into a USB 3.x port in XHCI mode. Any USB 3.x ports forced into EHCI will be equivalent to a native USB 2.x port. As a general rule of thumb, one should use a USB port native to the CPU and not connect any other high-polling devices to a port of the same hub. Even if all of these conditions are met, actual polling stability during higher workloads will further depend on general system and OS health. As such, it is recommended to use a reasonably optimized OS installation light on bloat in conjunction with the X15.
Performance Testing
I'll be testing general tracking, polling stability, and motion delay for 2000, 4000, and 8000 Hz each. All testing is performed at 1600 CPI. Please note that the X15 is moved first and thus receives a head start of roughly 0.3 ms.
2000 Hz:
A recent firmware reworked SPI timing, resulting in a much tighter count distribution. For whatever reason, the X15 averages 0.550 instead of the targeted 0.5 ms. The X15 is ahead of the G403 by roughly 0.5 ms.
4000 Hz:
Count distribution continues to look good. The X15 averages exactly 0.250 ms. The X15 is now ahead by roughly 0.7 ms.
8000 Hz:
Count distribution loosens up slightly compared to 4000 Hz. Instead of the targeted 0.125 ms, 0.160 ms are averaged. This is due to the 3389 updating registers just every other frame, halving the effective sensor framerate. The differential is roughly 0.8 ms.
8000 Hz (EHCI):
For the sake of comparison, here are the results for running the X15 at 8000 Hz on a USB 2.x port (EHCI). Count distribution is less regular compared to XHCI, no more than 0.190 ms is averaged, and the differential therefore remains around 0.7 ms.
Subjective Evaluation
Of course, the performance metrics obtained through empirical testing are just one side of the coin. The more pressing question is whether 8000 Hz is at all noticeable in games, and if so, to which degree.
To properly answer this question, note that someone being unable to notice something does not mean it isn't there objectively, or does not provide an objective advantage. The latter is most definitely true of 8000 Hz polling on the X15, so the matter shifts towards whether said advantage is meaningful and thus noticeable one way or another. That said, playing on a 165 Hz monitor at typically 200 FPS or more, I indeed struggled to notice a difference in terms of latency. As explained above, saturating the full 8000 Hz polling rate takes quite a bit of mouse movement and thus isn't typically reached all the time anyway, so most of the time, the benefit in terms of latency compared to 1000 Hz is around 0.5 ms, which is well below the sensory capabilities of the average human. The greatest effect of 8000 Hz may indeed not be observed in terms of absolute latency, but rather general positional accuracy and smoother cursor feel, more specifically in games requiring high precision in regards to click timing. Particularly games supporting sub-frame input will benefit to a greater degree from 8000 Hz, such as Overwatch or Diabotical with their respective settings enabled. Generally, in order to get any use out of 8000 Hz, I'd recommend using a strong CPU and a 240 Hz or even 360 Hz display. Slower panels will inevitably struggle to even display the granularity afforded by 8000 Hz polling. Those with weaker CPUs may experience worse input response simply due to the higher CPU cost, which means any advantage gained by 8000 Hz immediately cancels itself out.
When choosing between 2000, 4000, and 8000 Hz, I would advise taking the middle route, which is 4000 Hz, as on the X15 in particular, overall responsiveness is virtually on par with 8000 Hz while being lighter on the CPU, and 4000 Hz is also the only of the three landing on target. The parity in terms of latency is further underlined by the fact that 4000 Hz is much easier to saturate, so most of the time, there is no difference in practice whatsoever, aside from the ever so slightly higher click latency.
Appendix: List of Tested Games
As there is little reason to use the EVGA X15 at polling rates higher than 1000 Hz in non-competitive games, I'll exclusively list games that are typically considered competitive. Please note that a game running fine for me won't necessarily run fine for everyone as it merely means it generally works well with 8000 Hz polling. Conversely, a game not working well at 8000 Hz on a specific system isn't generally incompatible with 8000 Hz polling. Unfortunately, Unleash doesn't support application-specific profiles, so one would have to manually lower the polling rate in games that don't work well with polling rates higher than 1000 Hz.