Overclock.net › Forums › Components › Mice › Windows audio stack sucks, anything higher than stereo 16 bit 44.1k has higher cursor lag
New Posts  All Forums:Forum Nav:

Windows audio stack sucks, anything higher than stereo 16 bit 44.1k has higher cursor lag - Page 7  

post #61 of 258
microe's MouseTester software that was recently stickied. Gather results, plot them, look at interval vs. time. Post your results!

Edit: Oh yeah, remember to move your mouse fast enough to where you report at maximum polling rate. And zoom in on the data plot to sufficiently highlight poll time in microseconds. Also make sure to test for both active and inactive audio playback.
Edited by HAGGARD - 1/16/15 at 7:42am
post #62 of 258
Quote:
Originally Posted by HAGGARD View Post

Not spoiler (Click to show)
How can you be "on the fence" about it? I mean, nobody is denying that sound components are a performance factor, right? That in turn affects how well the entire system can handle game and I/O tasks. The question is if there's even a way to reduce the demand of audio processing. We should start to flesh things out and get people to test stuff. I will do further tests, but here's an example:



The first measurement was taken during playback of a 16bit/44.1kHz FLAC music file. The second without any audio going on. While all values average around 1ms according to a 1000Hz polling interval, the polls are issued noticably more timely and consistently without active audio processing. The mouse is a WMO, the sound card is a realtek ALC889 onboard chip feeding an external DAC/AMP optically (via WAS-API plugin). Drivers are Windows default.

Of course, that's just to demonstrate that sound components do affect polling behaviour. Measurements should be taken while in-game, so having a program that logs count reporting (utilizing raw input data) constantly would be neat for further testing. But in terms of optimizing your system using media player playback and the MouseTester to observe polling behaviour with and without audio events is already useful enough if you take enough samples to be positive things get better/worse one way or the other.
Say I now want more stable polling during audio playback. My audio controller does have a unique IRQ slot. The next thing I could try and do is change drivers. Then I could try and assign a specific CPU core to the audio process (AudioSrv service -> go to process, it will give you the svchost.exe that handles audio). Or change process priorities of the program or audio process. I could change the output type (while playing I use the analogue line-out of the sound card, not the optical S/PDIF). Ultimately I could try disabling the onboard sound chip and use other sound cards and see if they disrupt polling less, go through driver sets and settings again and so on. Or I would come to the conclusion that polling cannot be entirely stable with audio going on on my system and decide to reduce the polling interval to 500Hz.
Same can be done for DPC latency, which fortunately can be measured real-time while playing the game you want to enhance the experience of.

All that of course if you are "hardcore" about (mouse) performance. Otherwise you could just P'n'P and not criticise people for wanting to tinker with stuff like this, ha.

That is terribly disturbing, actually, never mind that variance only stays within 0.06 milliseconds, the fact that it degrades 290% is real bad. (There are 122 nodes outside of 0.002 ms variance zone on the top graph and only 42 on the bottom).
Workstation mk II
(16 items)
 
ASUS A73S
(16 items)
 
 
CPUMotherboardGraphicsGraphics
Intel Core i7-4790K Gigabyte G1.Sniper 5 GTX 1080 8GB GB GTX 1080 8GB GB 
RAMHard DriveHard DriveCooling
G.Skill Ares 16GB 2133 dual channel kit Crucial 893 GB SSD 3x HGST 3TB SATA3 H110i v2 
OSMonitorKeyboardPower
Windows 10 Acer XB271HK FILCO Majestouch 2 EVGA Supernova 1300 80 gold 
CaseMouseMouse PadAudio
Corsair 900D Cougar 700M Mionix Sargas 400 Asus Xonar u7 
CPUMotherboardGraphicsRAM
Core i7 2670qm ASUSTeK K73SD GeForce 610M 8 GB Samsung 
Hard DriveOptical DriveCoolingOS
750 GB Hitachi HTS54somenumbers Matshita DVD-RAM UJ8B0AW Laptop cooling — snakey heatpipe, some heatsink... Windows 8 pro 
MonitorMonitorKeyboardPower
Some 17" 1600×900 glossy panel, not too shabby. LG Flatron W2343T Most keys are perfect square. A few elongated r... The AC/DC adapter makes a high-pitched sighing ... 
CaseMouseMouse PadAudio
Not slim. Nexus silent mouse Custom handmade Built in, some realtek crap. 
CPUMotherboardGraphicsRAM
Pentium® III-S 1.13 GHz Compaq ProLiant ML330 G2 ATI Rage XL/XC 2 GB Kingston 
Hard DriveHard DriveOptical DriveCooling
A couple 500GB SCSI II disks in raid 1. 120GB SCSI II disk HL-DT-ST DVD-RAM Stock. It’s noisy. 
OSPowerCaseAudio
Windows XP Pro SP 3 Some PSU I pulled out from under my bed. Compaq. Very genuine case. Some Legacy audio. 
OtherOther
Realtek RTL8169/8110 Gigabit ethernet NIC AMI MegaIDE raid controller 
  hide details  
Workstation mk II
(16 items)
 
ASUS A73S
(16 items)
 
 
CPUMotherboardGraphicsGraphics
Intel Core i7-4790K Gigabyte G1.Sniper 5 GTX 1080 8GB GB GTX 1080 8GB GB 
RAMHard DriveHard DriveCooling
G.Skill Ares 16GB 2133 dual channel kit Crucial 893 GB SSD 3x HGST 3TB SATA3 H110i v2 
OSMonitorKeyboardPower
Windows 10 Acer XB271HK FILCO Majestouch 2 EVGA Supernova 1300 80 gold 
CaseMouseMouse PadAudio
Corsair 900D Cougar 700M Mionix Sargas 400 Asus Xonar u7 
CPUMotherboardGraphicsRAM
Core i7 2670qm ASUSTeK K73SD GeForce 610M 8 GB Samsung 
Hard DriveOptical DriveCoolingOS
750 GB Hitachi HTS54somenumbers Matshita DVD-RAM UJ8B0AW Laptop cooling — snakey heatpipe, some heatsink... Windows 8 pro 
MonitorMonitorKeyboardPower
Some 17" 1600×900 glossy panel, not too shabby. LG Flatron W2343T Most keys are perfect square. A few elongated r... The AC/DC adapter makes a high-pitched sighing ... 
CaseMouseMouse PadAudio
Not slim. Nexus silent mouse Custom handmade Built in, some realtek crap. 
CPUMotherboardGraphicsRAM
Pentium® III-S 1.13 GHz Compaq ProLiant ML330 G2 ATI Rage XL/XC 2 GB Kingston 
Hard DriveHard DriveOptical DriveCooling
A couple 500GB SCSI II disks in raid 1. 120GB SCSI II disk HL-DT-ST DVD-RAM Stock. It’s noisy. 
OSPowerCaseAudio
Windows XP Pro SP 3 Some PSU I pulled out from under my bed. Compaq. Very genuine case. Some Legacy audio. 
OtherOther
Realtek RTL8169/8110 Gigabit ethernet NIC AMI MegaIDE raid controller 
  hide details  
post #63 of 258
An overwhelming majority of those deviations are only a fraction of a hundredth of a millisecond. There's no way you can see or feel the difference of 10 microseconds. You plot those same graphs on a scale of 0-1-2 ms instead and it looks like a solid line. You're blowing up the scale in that graph and there's no real-world feeling there.
post #64 of 258
Quote:
Originally Posted by shatterboxd3 View Post

An overwhelming majority of those deviations are only a fraction of a hundredth of a millisecond. There's no way you can see or feel the difference of 10 microseconds. You plot those same graphs on a scale of 0-1-2 ms instead and it looks like a solid line. You're blowing up the scale in that graph and there's no real-world feeling there.

Remember that in this scenario I am only using the foobar media player, while in a game scenario there is a lot more stress on your system.
Quote:
Originally Posted by KYKYLLIKA View Post

That is terribly disturbing, actually, never mind that variance only stays within 0.06 milliseconds, the fact that it degrades 290% is real bad. (There are 122 nodes outside of 0.002 ms variance zone on the top graph and only 42 on the bottom).

I agree. Do test yourself.
Remember that general settings affect polling times as well though. If I change to high performance energy mode, disable speedstep/C1E and change clock resolution to .5ms, the variance is noticably smaller (polls are issued mostly double-digit microsecond precise in that case). But again, in-game things change, and while we would need test software to check just how much worse it gets, I think making sure idle variance is as low as possible is worthwhile. Like, I'll go as far and letting my inner r0ach hang out now and say that I feel the difference in mousing smoothness between 1ms clock resolution (csgo.exe default) and .5ms set with TimerTool.exe.
post #65 of 258
Quote:
Originally Posted by HAGGARD View Post

Remember that in this scenario I am only using the foobar media player, while in a game scenario there is a lot more stress on your system.
I agree. Do test yourself.
Remember that general settings affect polling times as well though. If I change to high performance energy mode, disable speedstep/C1E and change clock resolution to .5ms, the variance is noticably smaller (polls are issued mostly double-digit microsecond precise in that case). But again, in-game things change, and while we would need test software to check just how much worse it gets, I think making sure idle variance is as low as possible is worthwhile. Like, I'll go as far and letting my inner r0ach hang out now and say that I feel the difference in mousing smoothness between 1ms clock resolution (csgo.exe default) and .5ms set with TimerTool.exe.

I realize that foobar has a very light footprint, so lets see some testing in an average persons usage. Lets say I'm listening to Pandora while I'm DMing in CSGO. Can you do that and replicate this test for a period of say 5 seconds during a high action time in game? I'd like to see what the results are.
post #66 of 258
Quote:
Originally Posted by shatterboxd3 View Post

An overwhelming majority of those deviations are only a fraction of a hundredth of a millisecond. There's no way you can see or feel the difference of 10 microseconds. You plot those same graphs on a scale of 0-1-2 ms instead and it looks like a solid line. You're blowing up the scale in that graph and there's no real-world feeling there.

That is true, but facts lie in actual numbers, not feelings.

Imagine this as a statistical dataset.
Workstation mk II
(16 items)
 
ASUS A73S
(16 items)
 
 
CPUMotherboardGraphicsGraphics
Intel Core i7-4790K Gigabyte G1.Sniper 5 GTX 1080 8GB GB GTX 1080 8GB GB 
RAMHard DriveHard DriveCooling
G.Skill Ares 16GB 2133 dual channel kit Crucial 893 GB SSD 3x HGST 3TB SATA3 H110i v2 
OSMonitorKeyboardPower
Windows 10 Acer XB271HK FILCO Majestouch 2 EVGA Supernova 1300 80 gold 
CaseMouseMouse PadAudio
Corsair 900D Cougar 700M Mionix Sargas 400 Asus Xonar u7 
CPUMotherboardGraphicsRAM
Core i7 2670qm ASUSTeK K73SD GeForce 610M 8 GB Samsung 
Hard DriveOptical DriveCoolingOS
750 GB Hitachi HTS54somenumbers Matshita DVD-RAM UJ8B0AW Laptop cooling — snakey heatpipe, some heatsink... Windows 8 pro 
MonitorMonitorKeyboardPower
Some 17" 1600×900 glossy panel, not too shabby. LG Flatron W2343T Most keys are perfect square. A few elongated r... The AC/DC adapter makes a high-pitched sighing ... 
CaseMouseMouse PadAudio
Not slim. Nexus silent mouse Custom handmade Built in, some realtek crap. 
CPUMotherboardGraphicsRAM
Pentium® III-S 1.13 GHz Compaq ProLiant ML330 G2 ATI Rage XL/XC 2 GB Kingston 
Hard DriveHard DriveOptical DriveCooling
A couple 500GB SCSI II disks in raid 1. 120GB SCSI II disk HL-DT-ST DVD-RAM Stock. It’s noisy. 
OSPowerCaseAudio
Windows XP Pro SP 3 Some PSU I pulled out from under my bed. Compaq. Very genuine case. Some Legacy audio. 
OtherOther
Realtek RTL8169/8110 Gigabit ethernet NIC AMI MegaIDE raid controller 
  hide details  
Workstation mk II
(16 items)
 
ASUS A73S
(16 items)
 
 
CPUMotherboardGraphicsGraphics
Intel Core i7-4790K Gigabyte G1.Sniper 5 GTX 1080 8GB GB GTX 1080 8GB GB 
RAMHard DriveHard DriveCooling
G.Skill Ares 16GB 2133 dual channel kit Crucial 893 GB SSD 3x HGST 3TB SATA3 H110i v2 
OSMonitorKeyboardPower
Windows 10 Acer XB271HK FILCO Majestouch 2 EVGA Supernova 1300 80 gold 
CaseMouseMouse PadAudio
Corsair 900D Cougar 700M Mionix Sargas 400 Asus Xonar u7 
CPUMotherboardGraphicsRAM
Core i7 2670qm ASUSTeK K73SD GeForce 610M 8 GB Samsung 
Hard DriveOptical DriveCoolingOS
750 GB Hitachi HTS54somenumbers Matshita DVD-RAM UJ8B0AW Laptop cooling — snakey heatpipe, some heatsink... Windows 8 pro 
MonitorMonitorKeyboardPower
Some 17" 1600×900 glossy panel, not too shabby. LG Flatron W2343T Most keys are perfect square. A few elongated r... The AC/DC adapter makes a high-pitched sighing ... 
CaseMouseMouse PadAudio
Not slim. Nexus silent mouse Custom handmade Built in, some realtek crap. 
CPUMotherboardGraphicsRAM
Pentium® III-S 1.13 GHz Compaq ProLiant ML330 G2 ATI Rage XL/XC 2 GB Kingston 
Hard DriveHard DriveOptical DriveCooling
A couple 500GB SCSI II disks in raid 1. 120GB SCSI II disk HL-DT-ST DVD-RAM Stock. It’s noisy. 
OSPowerCaseAudio
Windows XP Pro SP 3 Some PSU I pulled out from under my bed. Compaq. Very genuine case. Some Legacy audio. 
OtherOther
Realtek RTL8169/8110 Gigabit ethernet NIC AMI MegaIDE raid controller 
  hide details  
post #67 of 258
Quote:
Originally Posted by shatterboxd3 View Post

I realize that foobar has a very light footprint, so lets see some testing in an average persons usage. Lets say I'm listening to Pandora while I'm DMing in CSGO. Can you do that and replicate this test for a period of say 5 seconds during a high action time in game? I'd like to see what the results are.

Currently, in-game mouse capture is not possible with the MouseTester AFAIK. If it is possible, or if there was a program that logged count reports/poll activity constantly (MouseTester needs a M1 trigger), I am positive there would be quite a lot more disruption. And that is only a busy CPU having a hard time issuing polls in a timely consistent fashion. Imagine what (micro)freezes could do to that, apart from affecting the game itself naturally. Using a CPU/GPU stresstest while doing this test could tell us something right now. I will try that later and also try whether MouseTester can somehow be triggered while in-game.

I don't want to blow this up either. Maybe there isn't even a sound component that is less ressource-heavy in Windows. Since mouse/usb processes are kernel-handled you can't really assign different cores to them either.
Maybe in the end the only thing we can really do is get better CPUs, disable halt states, clock stepping and core parking, use high performance energy modes and maybe use TimerTool as well or go back to 500Hz. In short, general performance tweaks to increase the likeliness of the processor being able to maintain poll intervals during audio/video rendering strain.
Same for DPC latency. Maybe there are no audio drivers that are not a main culprit there. We can only disable all sound devices we don't use and in general reduce DPC latency.

That's what I am interested in now though. What are the least disruptive means of audio playback? Of course, the component and settings have to be tested on the same system, otherwise the results are worthless.

So we need someone with as many different means of audio playback as possible. 8 sound devices of NotAgain sounds good. Waiting on initial and maybe further results.
post #68 of 258
I recently installed an Asus Xonar DG PCI card (M5A99X R22.0 Evo mobo). I am using a Logitech G700 mouse and have noticed no problems but my Rosewill RK-9000RE, plugged into tthe PS/2 plug, IS acting up. Related?

And, yes, I have cleaned the keyboard safely and thoroughly and even resoldered the offending keys.

Is it possible? After reading through all of these posts,, I may remove the Xonar just to see if it has any effect.
post #69 of 258
Turns out, MouseTester can capture/intercept in-game mouse data as well (even in raw input mode). Here is one sample:



Variance increased to up to a whole milisecond (1 missed complete interval) and ~500 microseconds at the extremes. This is without sound events present (except for ambient noise). To test for the impact of sound events you would need to have people firing weapons while you do the swipe. I guess I can set up BOTs to do that, will do further testing some other time. Energy setting was on balanced with speedstep enabled, default 1ms clock resolution, CS in high priority mode. Just so I can keep things consistent for further tests.
post #70 of 258
A lot of the complaints in this thread are from people without high sense mice or don't report their computer specs. a high polling rate mouse WILL be different than standard mice.

Asus Xonar DS PCI for instance caused all sorts of weird behaviors on my HTPC running Windows 7. PRIME 95 reports CPU errors ONLY with that card installed @ STOCK! Plus loads of random
bluescreens and whatnot, all went away after removing that card.

So @r0ach's post can be credible. That said, i do prefer quantitative measurements over qualitative such as yours. Numbers!
Boat Anchor
(11 items)
 
OC'in beast
(21 items)
 
 
CPUMotherboardGraphicsRAM
AMD Threadripper 1950x Alienware Area-51 TR edition Vega Frontier (flashed to RX 64) Micron 16GB DDR4-2666 
Hard DriveOSMonitorKeyboard
WD Blue SSD Windows 10 RS3 Dell P4317Q Cooler Master QuickFire Cherry Brown 
PowerCaseMouse
Dell 1500W Alienware Area-51 Logitech MX5500 
CPUMotherboardGraphicsGraphics
AMD FX-8350 (Vishera) Gigabyte GA-990XA AMD 285X (unreleased) AMD 285X (unreleased) 
RAMHard DriveHard DriveHard Drive
Mushkin Enhanced Blackline 8GB (996679) Western Digital WD10EALS 1TB Blue Corsair Force Series GT CSSD-F180GBGT-BK 180GB Intel 530 Series SSDSC2BW240A401 240GB 
Optical DriveCoolingCoolingCooling
Sony 20X Sata DVD-RW Scythe Mugen 2 rev B (3) Scythe Slip Stream SY1225SL12H-P Antec Tri-Cool 140mm 
CoolingCoolingOSMonitor
Antec Tri-Cool 120mm Scythe Ultra Kaze DFS123812-3000 Windows 7 Pro 64bit ASUS PA248Q 24.1" IPS 
KeyboardPowerCaseMouse
Coolermaster Storm QuickFire Stealth Cherry MX ... Antec Earthwatts EA-750 Antec 300 Logitech G5 
Mouse Pad
Steel Series Starcraft 2 Marine 
  hide details  
Boat Anchor
(11 items)
 
OC'in beast
(21 items)
 
 
CPUMotherboardGraphicsRAM
AMD Threadripper 1950x Alienware Area-51 TR edition Vega Frontier (flashed to RX 64) Micron 16GB DDR4-2666 
Hard DriveOSMonitorKeyboard
WD Blue SSD Windows 10 RS3 Dell P4317Q Cooler Master QuickFire Cherry Brown 
PowerCaseMouse
Dell 1500W Alienware Area-51 Logitech MX5500 
CPUMotherboardGraphicsGraphics
AMD FX-8350 (Vishera) Gigabyte GA-990XA AMD 285X (unreleased) AMD 285X (unreleased) 
RAMHard DriveHard DriveHard Drive
Mushkin Enhanced Blackline 8GB (996679) Western Digital WD10EALS 1TB Blue Corsair Force Series GT CSSD-F180GBGT-BK 180GB Intel 530 Series SSDSC2BW240A401 240GB 
Optical DriveCoolingCoolingCooling
Sony 20X Sata DVD-RW Scythe Mugen 2 rev B (3) Scythe Slip Stream SY1225SL12H-P Antec Tri-Cool 140mm 
CoolingCoolingOSMonitor
Antec Tri-Cool 120mm Scythe Ultra Kaze DFS123812-3000 Windows 7 Pro 64bit ASUS PA248Q 24.1" IPS 
KeyboardPowerCaseMouse
Coolermaster Storm QuickFire Stealth Cherry MX ... Antec Earthwatts EA-750 Antec 300 Logitech G5 
Mouse Pad
Steel Series Starcraft 2 Marine 
  hide details  
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Mice
  • Windows audio stack sucks, anything higher than stereo 16 bit 44.1k has higher cursor lag
This thread is locked  
Overclock.net › Forums › Components › Mice › Windows audio stack sucks, anything higher than stereo 16 bit 44.1k has higher cursor lag