Page 3 of 4

Re: Mousetester graphs

Posted: 17 Aug 2023, 12:23
by Slender
widow13 wrote:
17 Aug 2023, 12:21
Slender wrote:
17 Aug 2023, 12:19
widow13 wrote:
17 Aug 2023, 12:17
Slender wrote:
17 Aug 2023, 11:21


im tried only via and nec (v1/v2) 3.0 board.
All have jitter, it good controllers to move dac, mic, keyboard, and use motherboard usb for mouse only.
But im use usb 2.0 pci fard on mcs9990 chip. It dont have msi mode and have more isr then usb 3.0 pci card 5-10 vs 0.5-1 but this controller have clean graphs + no jitter.
+ (im not sure) mcs9990 dont have interrupt moderation.
Hmmm thanks I might try that out.
Even removing every single usb and leaving only the mouse doesn't get me stable polling.
Did your problems get resolved when u achieved perfect mouse graph ?
no, my gameplay is very smooth, but i have same weird responce and feel like my mouse is heavy.
Eh, I have the same issues. I guess achieving perfect graphs is not guaranteed to solve my issues, might have to try my pc on different house too.
Is it better to use chipset or cpu usb ports ?
if you have amd try both variants.
Correct your SA and another voltages.
use all devices in chipset / only mice in cpu.
affinity chipset usb, cpu usb affinity / no affinity

Re: Mousetester graphs

Posted: 17 Aug 2023, 12:58
by widow13
Slender wrote:
17 Aug 2023, 12:23
widow13 wrote:
17 Aug 2023, 12:21
Slender wrote:
17 Aug 2023, 12:19
widow13 wrote:
17 Aug 2023, 12:17


Hmmm thanks I might try that out.
Even removing every single usb and leaving only the mouse doesn't get me stable polling.
Did your problems get resolved when u achieved perfect mouse graph ?
no, my gameplay is very smooth, but i have same weird responce and feel like my mouse is heavy.
Eh, I have the same issues. I guess achieving perfect graphs is not guaranteed to solve my issues, might have to try my pc on different house too.
Is it better to use chipset or cpu usb ports ?
if you have amd try both variants.
Correct your SA and another voltages.
use all devices in chipset / only mice in cpu.
affinity chipset usb, cpu usb affinity / no affinity
Thanks I'll keep switching around and see what's best!
I'll also take the pc to a different house that's pretty new and let you know of the results.

Re: Mousetester graphs

Posted: 19 Aug 2023, 00:34
by masneb
F1zus wrote:
15 Aug 2023, 07:08
See how my mouse behaves at 8000Hz. Absolutely unstable frequency with jumps up to 200000 Hz.
I tried to bind different cores, 2, 4, 6, the result is the same.
Disconnected all possible usb ports. Disable interrupt moderation.
My amd 5800x processor is overclocked to a stable 4.6GHz, RAM 3800cl14 according to the formulas.
Here are the results at 8000hz

In theory, the higher the polling rate, the more data you need to store in the buffer and the larger the buffer should be.
I have tried 25/50/100 options. Didn't notice much of a difference.
My experience with my 7950X3D is this seems to be inherent to Ryzen systems. I've own a 3900x, 5900x, 13900k, and now a 7950X3D. With AMD you have to turn off c-states in the bios and then CPPC. CPPC seems to be a doozy as windows moves threads around and seems to make more of a problem then it helps.

I eventually decided to just buy a seperate PCIE USB card. It didn't fix all the fluctuations, but it's more consistent and feels better too.

I have this guy.

https://www.amazon.com/gp/product/B072LS4JH7/

But anything with the RENESAS UPD720201 would probably perform the same. I have another USB-PCIE card that I haven't tried yet, as this one feels pretty good now.

Re: Mousetester graphs

Posted: 19 Aug 2023, 20:11
by Ragerlis
Gonna call here some good behaviors BEFORE doing any tweaks, gonna use my old system, a 9700kf and 2080super inside a z390 motherboard with windows 11 without any kind of tweak whatsoever as example as the difference in this board is HUGE.

This graph is a mousetester graph with apex legends running, edge with 20+ tabs and a DAC, usb mic, my smartphone, 2 1000hz mice and a 1000hz keyboard plugged into the motherboard, no pcie usb, ZERO tweaks and a 6 seconds long graph, no cuts, no cherry picked.

Image

Now the important part, how the to achieve that? In my case, after a fresh windows install, my graphs are a clusterfuck like someone posted here, 20, 30 40k spikes, unstable, sometimes it even misses a few milliseconds for whatever reason, installing drivers, hardcore tweaking, nothing works, nothing does shit.

To fix that, install the chipset drivers, but not the windows optional drivers, the actual right drivers from intel, as everysingle shit connected to the intel motherboard pass through the controller chipset.

DO NOT install the windows update optional drivers, windows do not know what chipset you have, it can just guess, it need the inf files from intel. (The optional or even the microsoft generic ones will work but this shit will be VERY unstable)

In my case, this can be fixed downloading https://www.intel.com/content/www/us/en ... ility.html

This is NOT the drivers, its just files to tell the OS exactly what you have inside your board, the OS will auto install everything up after, I don't know which download you need to your board, you will have to search for it, this one covers everything up 10th generation, 11, 12 i have no idea how it works, and its INTEL ONLY.

After doing that, polling rate should be stable, after that start doing the hardcore tweaks spreading in this topic and whatever guide you guys find.

As AMD you can dodge this clusterfuck by just using the usb ports linked straight to the processor, but its a good idea to do the same for obvious reasons.

DO NOT KEEP OR INSTALL THE WINDOWS UPDATE OPTIONAL DRIVERS, they are just a close enough drivers, will work, but can be unstable.

And yes, hardcore tweaking can make this graph 0.001+-, i dont like to do that because its just too niche and you cant do that kind of shit on lan.

Edit: all this isnt just for mouse stability, my ethernet card without the right chipset drivers have near 25% of TDP retransmission, 0.01% with it, nvme ssd cant hold max speeds and many other microproblems, its an overall system stability.

Re: Mousetester graphs

Posted: 20 Aug 2023, 06:08
by widow13
Ragerlis wrote:
19 Aug 2023, 20:11
Gonna call here some good behaviors BEFORE doing any tweaks, gonna use my old system, a 9700kf and 2080super inside a z390 motherboard with windows 11 without any kind of tweak whatsoever as example as the difference in this board is HUGE.

This graph is a mousetester graph with apex legends running, edge with 20+ tabs and a DAC, usb mic, my smartphone, 2 1000hz mice and a 1000hz keyboard plugged into the motherboard, no pcie usb, ZERO tweaks and a 6 seconds long graph, no cuts, no cherry picked bullshit.

Image

Now the important part, how the fuck to achieve that? In my case, after a fresh windows install, my graphs are a clusterfuck like someone posted here, 20, 30 40k spikes, unstable, sometimes it even misses a few milliseconds for whatever reason, installing drivers, hardcore tweaking, nothing works, nothing does shit.

To fix that, install the goddamn chipset drivers, but not the windows recommended drivers, the actual right drivers from intel, as everysingle shit connected to the intel motherboard pass through the controller chipset.

DO NOT install the windows update optional drivers, windows do not know what chipset you have, it can just guess, it need the inf files from intel. (The optional or even the microsoft generic ones will work but this shit will be VERY unstable)

In my case, this can be fixed downloading https://www.intel.com/content/www/us/en ... ility.html

This is NOT the drivers, its just files to tell the OS exactly what you have inside your board, the OS will auto install everything up after, I don't know which download you need to your board, you will have to search for it, this one covers everything up 10th generation, 11, 12 i have no idea how it works, and its INTEL ONLY.

After doing that, polling rate should be stable, after that start doing the hardcore tweaks spreading in this topic and whatever guide you guys find.

As AMD you can dodge this clusterfuck by just using the usb ports linked straight to the processor, but its a good idea to do the same for obvious reasons.

DO NOT KEEP OR INSTALL THE WINDOWS UPDATE OPTIONAL DRIVERS, they are just a close enough drivers, will work, but can be unstable.

And yes, hardcore tweaking can make this graph 0.001+-, i dont like to do that because its just too niche and you cant do that kind of shit on lan.

Edit: all this isnt just for mouse stability, my ethernet card without the right chipset drivers have near 25% of TDP retransmission, 0.01% with it, nvme ssd cant hold max speeds and many other microproblems, its an overall system stability.
So basically install chipset/lan/gpu drivers before updating windows so you skip the auto install or skip windows update completely ?
Do you get ur chipset drivers from the mobo website or from intel/amd ?

By the way interval graphs are not that bad for me either, frequency graphs are bad.

Re: Mousetester graphs

Posted: 20 Aug 2023, 07:01
by ayukreizi
since the topic of mouse polling rates is being discussed again i want to take the opportunity to ask about an "anomaly" that i experience and want some input on my interpretation of results.
at least it would be an anomaly when taking into the account that it is generally recommended to plug your mouse into the USB port connected to the CPU instead of the Chipset one.

i'm on the 7800X3D, X670E Asus Hero MB with a VAXEE XE Wireless but tested with several mice in order to exclude HW related errors. Wired as well as wireless.

Here are what my results boil down to. I have tested this "anomaly" across several installations of Windows. Windows 10 as well as 11.

CHIPSET:
Image
Image

CPU:
Image
Image

as you can see i can get the results of my mouse on chipset to stabilize around .001~.002
compared to the CPU results that stabilize around 0.008~0.010

now, with my CPU results i've gone through all of the known advice to increase my polling stability and i cannot get it lower than those values outside of increasing its stability around 0.008.
i'd like to ask people who have extensive knowledge with testing mice and maybe Chief (can't tag him for some reason): are those results to be seen in the traditional way?
do they indicate that the CPU port is inheritently worse than my Chipset port? and should others that measure such results on their CPU port always use their Chipset port?
i've tested and retested it in several configurations and am lost for any other explanations.

thanks for your insight.

Re: Mousetester graphs

Posted: 20 Aug 2023, 09:53
by Slender
ayukreizi wrote:
20 Aug 2023, 07:01
since the topic of mouse polling rates is being discussed again i want to take the opportunity to ask about an "anomaly" that i experience and want some input on my interpretation of results.
at least it would be an anomaly when taking into the account that it is generally recommended to plug your mouse into the USB port connected to the CPU instead of the Chipset one.

i'm on the 7800X3D, X670E Asus Hero MB with a VAXEE XE Wireless but tested with several mice in order to exclude HW related errors. Wired as well as wireless.

Here are what my results boil down to. I have tested this "anomaly" across several installations of Windows. Windows 10 as well as 11.

CHIPSET:
Image
Image

CPU:
Image
Image

as you can see i can get the results of my mouse on chipset to stabilize around .001~.002
compared to the CPU results that stabilize around 0.008~0.010

now, with my CPU results i've gone through all of the known advice to increase my polling stability and i cannot get it lower than those values outside of increasing its stability around 0.008.
i'd like to ask people who have extensive knowledge with testing mice and maybe Chief (can't tag him for some reason): are those results to be seen in the traditional way?
do they indicate that the CPU port is inheritently worse than my Chipset port? and should others that measure such results on their CPU port always use their Chipset port?
i've tested and retested it in several configurations and am lost for any other explanations.

thanks for your insight.
your graphs looks like stable but anyway you have jitter 960-1040, what system you use, amd?
if you use amd try to disable GPU driver in Dev.man and retest mousetester

Re: Mousetester graphs

Posted: 20 Aug 2023, 11:07
by widow13
ayukreizi wrote:
20 Aug 2023, 07:01
since the topic of mouse polling rates is being discussed again i want to take the opportunity to ask about an "anomaly" that i experience and want some input on my interpretation of results.
at least it would be an anomaly when taking into the account that it is generally recommended to plug your mouse into the USB port connected to the CPU instead of the Chipset one.

i'm on the 7800X3D, X670E Asus Hero MB with a VAXEE XE Wireless but tested with several mice in order to exclude HW related errors. Wired as well as wireless.

Here are what my results boil down to. I have tested this "anomaly" across several installations of Windows. Windows 10 as well as 11.

CHIPSET:
Image
Image

CPU:
Image
Image

as you can see i can get the results of my mouse on chipset to stabilize around .001~.002
compared to the CPU results that stabilize around 0.008~0.010

now, with my CPU results i've gone through all of the known advice to increase my polling stability and i cannot get it lower than those values outside of increasing its stability around 0.008.
i'd like to ask people who have extensive knowledge with testing mice and maybe Chief (can't tag him for some reason): are those results to be seen in the traditional way?
do they indicate that the CPU port is inheritently worse than my Chipset port? and should others that measure such results on their CPU port always use their Chipset port?
i've tested and retested it in several configurations and am lost for any other explanations.

thanks for your insight.
Do you use any affinity on chipset/cpu usbs or running default ?

Re: Mousetester graphs

Posted: 20 Aug 2023, 11:49
by ayukreizi
Slender wrote:
20 Aug 2023, 09:53
ayukreizi wrote:
20 Aug 2023, 07:01
/snip
your graphs looks like stable but anyway you have jitter 960-1040, what system you use, amd?
if you use amd try to disable GPU driver in Dev.man and retest mousetester
as mentioned, 7800X3D with a 4090.
by GPU driver you mean oem0.inf?
Image
because that would make it a synthetic result as you need it to run for gaming if i'm not mistaken? but you're correct that the residual jitters are mostly from nvidia drivers.

here are "synthetic results" in safemode with gpu driver obviously disabled.

Chipset results:
Image

CPU results:
Image

widow13 wrote:
20 Aug 2023, 11:07
Do you use any affinity on chipset/cpu usbs or running default ?
doesn't matter that much in my case and for my specific question. it slightly reduces jitter but the results remain the same mostly. my results are without any core affinity.

so yeah, do those results indicate that the CPU port is inheritently worse than my Chipset port?
or am i mistakenly ignoring something?

Re: Mousetester graphs

Posted: 20 Aug 2023, 11:51
by widow13
ayukreizi wrote:
20 Aug 2023, 11:49
Slender wrote:
20 Aug 2023, 09:53
ayukreizi wrote:
20 Aug 2023, 07:01
/snip
your graphs looks like stable but anyway you have jitter 960-1040, what system you use, amd?
if you use amd try to disable GPU driver in Dev.man and retest mousetester
as mentioned, 7800X3D with a 4090.
by GPU driver you mean oem0.inf?
Image
because that would make it a synthetic result as you need it to run for gaming if i'm not mistaken? but you're correct that the residual jitters are mostly from nvidia drivers.

here are "synthetic results" in safemode with gpu driver obviously disabled.

Chipset results:
Image

CPU results:
Image

widow13 wrote:
20 Aug 2023, 11:07
Do you use any affinity on chipset/cpu usbs or running default ?
doesn't matter that much in my case and for my specific question. it slightly reduces jitter but the results remain the same mostly.

so yeah, do those results indicate that the CPU port is inheritently worse than my Chipset port?
or am i mistakenly ignoring something?
Yeah it seems like it. CPU ports are usually better but that's not 100% of the time.
In my case it makes no difference. How is your frequency vs time looking ?
Btw have u found any particular settings that helped you reach these graphs ? I can't really get good graphs on my system.