Blur Busters Forums

Can you explain how it helps at 240hz? Cougar claims "real 2000 Hz, not overclocked". Perhaps these two options are something you'd be interested in yourself. In that overclock thread I linked to they claim that a proper test is not something people are performing, and explain how to do it. Something else you might be interested in.

Actually, this is where they describe their testing:

https://cougargaming.com/us/products/mi ... ngRate.php

The higher the Hz, the less motion blur there is (lower persistence), and thusly, the more visible the same microstutters become.

As we all already know, ULMB is more microstuttery than non-ULMB, due to the lack of motion blur hiding microstutters.

Also, the smaller the difference between two frequencies (e.g. a display Hz and a mouse Hz), the more pronounced beat-frequency effects can become.

This same "beat-frequency" phenomenon can be witnessed by watching framerates versus refreshrates, too. A single refresh-cycle-repeat per second turning 30fps@60Hz into 29.5fps@60Hz creates a more noticeable microstutter (one 2-repeat becomes a 3-repeat, a 50% stutter-amplitude change for 1 frame) than a single refresh-cycle-repeat per second turning 15fps@60Hz into 14.75fps@60Hz (one 4-repeat becomes a 5-repeat, a 25% stutter-amplitude change for 1 frame)

----

Anyway, I'm, of course, talking about 2000 position reports per second, rather than 1000 position reports (each repeated twice) like for some overclocked mouse. Sending more position reports per second can reduce lag slightly, but even better is unique position reports.

Obviously i want 2khz, but if you're going to enable 2khz, you might as well just go full out and enable 8khz or whatever the max is for USB3. It's one of those 'there's no reason not to do it' things.

lexlazootin wrote:Obviously i want 2khz, but if you're going to enable 2khz, you might as well just go full out and enable 8khz or whatever the max is for USB3. It's one of those 'there's no reason not to do it' things.

8 KHz is even better, and will be useful for future ultra-low-persistence retina displays (e.g. 8K displays with 0.1ms persistence).

Displays with higher dpi low-persistence can make mouse movement imperfections more visible than at lower dpi, especially if you've got good vision on ultra-fine fast-moving text (e.g. window dragging, or Retina-rendered GPU graphics with ultra-def textures). That's where we'll definitely want to head eventually towards true-4KHz and true-8KHz mice by the ~2020s or ~2030s.

But on a practical matter, tracking accuracy (and USB signalling accuracy) needs to keep up with the mouse poll rate, and sometimes lower has been better.

For a long time, when the first 1000Hz mice came out, the best 500Hz mice was still superior in tracking. Until 1000Hz matured.

That said, if you can achieve 8KHz reliably, then why the hell not

So I got the Revenger S and it does feel more responsive to me. Build quality seems good. The software does not seem amateurish and has more options than Roccat. The issue is the shape, which was to be expected. Have the Minos x5 arriving tomorrow which will hopefully be a better fit. Hard to see the Revenger S beating out the Roccat Kone Pure Owl Eye due to the shape, and the Minos is not an ergonomic right handed mouse but let's see.

Chief Blur Buster wrote:

lexlazootin wrote:Obviously i want 2khz, but if you're going to enable 2khz, you might as well just go full out and enable 8khz or whatever the max is for USB3. It's one of those 'there's no reason not to do it' things.

8 KHz is even better, and will be useful for future ultra-low-persistence retina displays (e.g. 8K displays with 0.1ms persistence).

Displays with higher dpi low-persistence can make mouse movement imperfections more visible than at lower dpi, especially if you've got good vision on ultra-fine fast-moving text (e.g. window dragging, or Retina-rendered GPU graphics with ultra-def textures). That's where we'll definitely want to head eventually towards true-4KHz and true-8KHz mice by the ~2020s or ~2030s.

But on a practical matter, tracking accuracy (and USB signalling accuracy) needs to keep up with the mouse poll rate, and sometimes lower has been better.

For a long time, when the first 1000Hz mice came out, the best 500Hz mice was still superior in tracking. Until 1000Hz matured.

That said, if you can achieve 8KHz reliably, then why the hell not

It shouldn't be too hard to make a mouse do 8kHz reliably. It's just a matter of writing your firmware loop with a hard 125µs constraint. The harder part is the driver/OS side of things, because Windows isn't a RTOS.

I think at least some mice only look like they have better tracking at 500hz, because the high frequency noise gets averaged out by the longer time base.

Chief Blur Buster wrote:That said, if you can achieve 8KHz reliably, then why the hell not

I wonder if CPU load is now an issue when reading the hardware at 8kHz, especially after the whole Meltdown/Spectre fixes that are said to have a big impact on context switching.

Sparky wrote:It shouldn't be too hard to make a mouse do 8kHz reliably. It's just a matter of writing your firmware loop with a hard 125µs constraint. The harder part is the driver/OS side of things, because Windows isn't a RTOS.

For "true" 8KHz on an optical mouse, wouldn't it need a camera that can operate at 8000FPS? Otherwise, the mouse would need to synthesize the data using extrapolation (current velocity and direction), which is known to produce weird inaccuracies. Works mostly fine when extrapolating to frame rates close to each other, but all the way up to 8000 is probably not going to work that well.

RealNC wrote:For "true" 8KHz on an optical mouse, wouldn't it need a camera that can operate at 8000FPS? Otherwise, the mouse would need to synthesize the data using extrapolation (current velocity and direction), which is known to produce weird inaccuracies. Works mostly fine when extrapolating to frame rates close to each other, but all the way up to 8000 is probably not going to work that well.

This is correct.

All the mouse overclocking simply create duplicate position reports. Or mathematically-guessed position reports. Typical overclocked 2000Hz mice polling usually gives 2 reports each of 1000 unique mouse-position per second. It can, however, reduce latency of a single poll by, say 0.5ms, but it's simply a potentially earlier-reception of a report.

What we need is true 2000 unique, real, positions per second from a 2000Hz camera sensor.

Mouse cameras still only run at ~1000Hz and I'm hoping to see true-2000Hz cameras for the mouse sensor.

RealNC wrote:

Chief Blur Buster wrote:That said, if you can achieve 8KHz reliably, then why the hell not

I wonder if CPU load is now an issue when reading the hardware at 8kHz, especially after the whole Meltdown/Spectre fixes that are said to have a big impact on context switching.

Maybe for few-year-old-systems.Older CPUs might have an issue, but by the time true-8KHz-camera mice (8000 fully unique positions per second, from an 8000fps mouse camera, no duplicates!) come out in a decade from now.

For current systems, I don't think so -- and you don't need every single of all of the 8000 positions. Just the mouse position closest to the frametime you need. So for 200fps, you only need 200 very-fresh (0.5ms-accurate) mouse positions to be used by your software. That's how you destutter.

With the correct mouse drivers and algorithm, and game engines that handles polling correctly, on a modern fast system, it shouldn't have to an issue.

Mouse cameras are not necessarily running at above 1000 frames per second (low-resolution frames, often 16x16 or 32x32) to a computer-accessible level (actual position reports relayed to computer). The mouse firmware may not be updating position registers at the same rate as the camera updates, instead using the camera updates to improve 1KHz accuracy. We specifically need a mouse specifically designed for 2KHz polling, to get the positions accurately to the computer.

Mouse sensors have pretty high framerates to begin with, PMW 3360 for example can do 12kfps. They kinda have to, because they have to have overlapping images in order to track the surface. Not sure how frequently the sensor can be sampled though.