USB mouse hard overclocking (2000 Hz+)

qsxcv · Feb 1, 2016

what the fushfjwjdjfjwjdndjajjx

W M O
M
O

edit:
mlt04 does 4000hz stably. but since dpi is 400, you need to keep the mouse moving pretty quickly:

edit2

looked at usb data pin on oscilloscope
indeed 4000hz polling works

but whether the mouse responds to every poll depends on the firmware

for instance, g100s:
http://i.imgur.com/Xeoj28u.png
still runs at "500hz" because there's some limit in the firmware to only respond every 2ms.

qsxcv · Feb 1, 2016

my teensy+g100s: limited to 1000hz

((( will try to figure out why later.

first, lets look at my other mice
g302: limited to 1000hz yes i checked process parent, and underclocking to 125hz works. probably (unintentionally) firmware limited
g303: ^
g402: ^
g502brick: ^
logitech's intervals are always quantized to 1ms: http://i.imgur.com/JkQmzOI.png. probably because there's stuff in the firmware that's like hardcoded to run every 1ms

finalmouse @ 4000 polling: http://i.imgur.com/jEUNbI9.png
finalmouse @ 2000 polling: http://i.imgur.com/5FxMCT9.png

torq x5: limited to 1000hz but unlike logitech's newer mice, the interval isn't quantized to milliseconds: http://i.imgur.com/Jjj2tgs.png
god damn my torqx5's clicks are nice.

kinzu v1: soo close. 2000hz: http://i.imgur.com/tMh8MkG.png 4000hz: http://i.imgur.com/my4sF7I.png

Quote:

Originally Posted by HAGGARD

On the flipside, swamping your CPU with 4k+ USB interrupts per second increases CPU load and decreases polling stability/precision (http://www.overclock.net/t/1550666/usb-polling-precision).

not always
for instance, on my computer, the jitter doesn't differ between 1ms or 2ms polling

but in any program that prints out the hz (e.g. dimr, mouserate, mousemovementrecorder, etc...)
higher polling rates will always appear to have more jitter. this is an illusion due to how the function 1/x is not linear.

for instance if you have 15us jitter, at 1ms polling, mousemovementrecorder would show numbers between 1015 and 985. but with the same amount of jitter at 2ms polling, mousemovementrecorder would show between 496 and 504.

also, the sort of jitter does not affect the data coming out of the mouse at all because the timing of the usb communications is not affected by this. i've checked on my oscilloscope and the jitter of the usb timing there is on the order of nanoseconds, even for my laptop, with power saving and other stuff on which makes the mousetester interval plots look quite bad.

uaokkkkkkkk · Feb 1, 2016

It'd be interesting if 4000hz would make the Paten/Fatal1ty 1010 actually hit 999hz(stuck at 333hz at 1000hz normally). I was apparently doing it wrong since 31 made it 31hz instead. Meh, oh well. Might try another time.

qsxcv · Feb 1, 2016

make sure to enable xhci in bios

and devicemanager should look like this
http://i.imgur.com/wbm0WyX.png

blahhh so tempting to install win10 on my main drive

SweetLow · Feb 1, 2016

Quote:

Originally Posted by c0dy

And I've only been able to change the Hz when I checked "Process Parent". Maybe this helps out

You have composite device, not pure hid device. But no difference.
If device have more than 31(62) Hz when you select 31(62) Hz - you make setup as expected and have right driver

Quote:

Originally Posted by qsxcv

mousetester interval plots are much better imo

If you have 2000+ you can see that in any rate test program

Quote:

Originally Posted by qsxcv

looked at usb data pin on oscilloscope
indeed 4000hz polling works

Hardware confirmation - nice!

Quote:

mlt04 does 4000hz stably.

So I must make what i said - wait 8000 Hz test version some time [hours] later

Quote:

Originally Posted by qsxcv

lets look at my [other] mice
...

Wow!

HAGGARD · Feb 1, 2016

Quote:

Originally Posted by qsxcv

not always
for instance, on my computer, the jitter doesn't differ between 1ms or 2ms polling

In idle they can be more or less the same; try playing music or have some other work-intensive program open though and this should change.
Quote:

also, the sort of jitter does not affect the data coming out of the mouse at all because the timing of the usb communications is not affected by this. i've checked on my oscilloscope and the jitter of the usb timing there is on the order of nanoseconds, even for my laptop, with power saving and other stuff on which makes the mousetester interval plots look quite bad.

That's true but also obvious. The hardware USB communication between mouse and host should be rock solid because the mouse is always doing the same thing regardless and the host controller is made with these timings in mind; has an internal clock of its own (microframes timings). I. e. there's no hardware jitter. The timing disruption comes from the CPU not addressing/processing the inputs timely enough and the CPU's ability to do that properly obviously depends on the amount of stuff it has to do - including amount of inputs.

Also: WMO wins AGAIN? This will not make me install Windows 10. This will not make me install Windows 10. This will not make me... - I bet it even does 8kHz.
Quote:

but since dpi is 400, you need to keep the mouse moving pretty quickly:

Oh yeah, this is something I forgot to mention in that last post as a little benefit of higher polling rates. The lower the CPI and the higher the polling rate, the faster you can move the mouse before you get maximum poll latency on every input. Basically as long as count event rate < polling rate every count is a single report and gets latency anywhere between zero and full poll interval as opposed to full poll interval all the time.
400cpi @ 1kHz: 1000 / 400 = 2.5inch/s or 6.35cm/s
400cpi @ 4kHz: 4000 / 400 = 10inch/s or 25.4cm/s

qsxcv · Feb 1, 2016

sweetlow, is it possible to get this to work with the intel usb3.0 driver for windows7
Quote:

In idle they can be more or less the same; try playing music or have some other work-intensive program open though and this should change.

good point which i always forget

.
will try in ~18hours
Quote:

Basically as long as count event rate < polling rate every count is a single report and gets latency anywhere between zero and full poll interval as opposed to full poll interval all the time.

before someone uses this to argue that "low dpi is better than high dpi", the decrease in latency due to finer quantization at higher dpi is more significant.
well it all depends on how exactly one defines latency, but regardless, above 500hz polling, polling-related latency is imperceptible and insignificant compared to many other things

btw offtopic:
http://www.ebay.com/itm/151969244623
see second picture
so this where the 6000fps mlt04 idea comes from. not sure why people thought that certain wmo/imo's are 6000fps though

SweetLow · Feb 1, 2016

Quote:

Originally Posted by qsxcv

sweetlow, is it possible to get this to work with the intel usb3.0 driver for windows7

You tempt me second time in this thread!

My main system exactly the same (Intel USB from HM77 south bridge + Windows 7), and i already try to find patching points in code of Intel host controller driver (IUSB3XHC.SYS) But this code is NOT from Microsoft and, I assume, even not Microsoft compiler compiled. All reasonable assumptions about how compiled code and data may look not worked (compare to USBXHCI.SYS, for example - the code is differ from good old USBPORT.SYS, but i found needed sequence literally after five minutes). So i can't search result quickly - and postponed this.

HAGGARD · Feb 1, 2016

Quote:

Originally Posted by qsxcv

well it all depends on how exactly one defines latency, but regardless, above 500hz polling, polling-related latency is imperceptible and insignificant compared to many other things

True. Even 500->1k is diminishing returns territory in terms of latency but can noticeably worsen the CPU poll stutter. And then the actual latency difference between 1kHz and 4kHz for any arbitrary count from those 2.5inch/s to 10inch/s is also only 375us on average, right? (*But then that would be the same regardless of movement speed... I guess I need to go in paint for 9000 hours again to wrap my head around that.)
Quote:

good point which i always forget

.

Which should be another reason for people to consider capping their framerates: less stress for the system equals more stable polling.
Quote:

btw offtopic:
http://www.ebay.com/itm/151969244623
see second picture
so this where the 6000fps mlt04 idea comes from. not sure why people thought that certain wmo/imo's are 6000fps though

Wireless peasants. But you mentioned how they probably used the wireless MLT04 in the IME4.0, so maybe that's how?
Quote:

Originally Posted by SweetLow

So i can't search result quickly - and postponed this.

Can you also do for the Etro- nevermind.

GHADthc · Feb 1, 2016

Is the polling rate supposed to jump around like crazy in mouse rate tester once 2000-4000Mhz is active? because if that is normal, I think my MLT04 IME 3.0 is hitting 3000+ Mhz...but its certainly not showing up as consistently.

I got my Rival to occasionally jump to 2000+ but it too exhibited the same behaviour.

Edit: But in dimr.exe there seems to be no change...

Edit 2: I think I'm just doing something wrong somewhere....would be nice if someone could break it down in great detail for a tard like me

bruzanHD · Feb 1, 2016

Quote:

Originally Posted by qsxcv

btw offtopic:
http://www.ebay.com/itm/151969244623
see second picture
so this where the 6000fps mlt04 idea comes from. not sure why people thought that certain wmo/imo's are 6000fps though

Probably because the serial numbers correspond as such. X08 ones and X80 ones that is. I mean to me it would only make sense.

SweetLow · Feb 1, 2016

There is not such thing as "sufficient frequency" for orthodoxal overclocker

4k/8k version:
http://sweetlow.orgfree.com/download/hidusbf3.zip
...
2. take drivers from hidusbf3.zip and install it (copy to %systemroot%\system32\drivers\ directly or to the folder of hidusbf setup and then install service).
3. reboot after 2!
4. run setup.exe, and try to change rate to 31 or 62. Rate=31 actually equal to 4000 Hz, 62 - 8000 Hz. (Don't forget to restart device.)
...

P.S. Do not use USB hubs (for first try), it can restrict rate, attach mouse to controller ports directly. I first test my best overclockable mouse (a4tech bw-35) with hub (this hub is USB2.0 hub - but this is not help) - and has 1000 Hz only

When attached directly to controller - it can reach 1400 Hz average (unstable)

acid_reptile · Feb 1, 2016

This is not for windows 7, right?

SweetLow · Feb 1, 2016

Quote:

Originally Posted by acid_reptile

This is not for windows 7, right?

Not now.

qsxcv · Feb 1, 2016

i'll be home in 6-7hours to test 8khz

i don't understand this well, but shouldnt it be possible to do this with the usb2 driver because hi-speed usb supports microframes?

SweetLow · Feb 1, 2016

Quote:

Originally Posted by qsxcv

shouldnt it be possible to do this with the usb2 driver because hi-speed usb supports microframes?

USB2.0 controllers never directly control USB1.1 devices, only through translating hubs, so now i suppose - it's not possible.

qsxcv · Feb 2, 2016

Quote:

Originally Posted by qsxcv

sweetlow, is it possible to get this to work with the intel usb3.0 driver for windows7
Quote:

In idle they can be more or less the same; try playing music or have some other work-intensive program open though and this should change.

Click to expand...

good point which i always forget

.
will try in ~18hours

so for my system, with 5 twitch streams open:

1000hz: http://imgur.com/a/9fVjW
500hz (ignore title

): http://imgur.com/a/4ekiR

indeed the 500us jumps are much more rare, but the main part of the noise (+-0.1ms) looks identical

bond10 · Feb 2, 2016

My CPU usage spikes to 85-90% on 1000 hz. When I'm using 500 hz, it's around 40-50% spikes. 125 hz, less than 15%.

CPU: i5 2500K
Mouse: Deathadder 2013

1000 HZ

Should I stop using 1000 hz??

qsxcv · Feb 2, 2016

WMO DOES 8000HZ POLLING

also that's conclusive proof that my (new) x08, with last 4 digits=4103 (41th month of 2003?), does not run at 6000fps, because if it did it would not be possible to have consistent 125us report intervals

Derp · Feb 2, 2016

W M O
M
O

USB mouse hard overclocking (2000 Hz+)

qsxcv

qsxcv

uaokkkkkkkk

qsxcv

SweetLow

HAGGARD

qsxcv

SweetLow

HAGGARD

GHADthc

bruzanHD

SweetLow

acid_reptile

SweetLow

qsxcv

SweetLow

qsxcv

bond10

qsxcv

Derp

Top Contributors this Month

Recommended Communities