Overclock.net › Forums › Components › Mice › Things I've noticed that alter mouse movement that nobody talks about:
New Posts  All Forums:Forum Nav:

Things I've noticed that alter mouse movement that nobody talks about: - Page 3

post #21 of 57
I don't know what to say, it works fine for me.

uQFhBl.jpg

E8plCl.jpg

Try doing a clean driver install maybe?
post #22 of 57
HPET really does influence input lag but there's no well defined best option. On some systems it's better left ON, on others it's better left off. People need to experiment with this themselves.

I personally get the best results when enabling HPET in the BIOS but using bcdedit /set useplatformclock false.
Desktop (Main)
(20 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 2600K @4.5Ghz Gigabyte P67-UD4-B3 MSI Twin Frozr Geforce 770 GTX 2GB Corsair Vengeance DDR3 1600Mhz 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 Pro Seagate Barracuda 1.5 TB Seagate Barracuda 1.5 TB Seagate Barracuda 4TB 
Optical DriveCoolingOSMonitor
Le random dvd writer Corsair Hydro Series™ H100i Windows 8.1 Professional 64 bit iiyama ProLite G2773HS 
KeyboardPowerCaseMouse
Battlefield 3™ Razer BlackWidow Ultimate Gaming... Truepower Quattro 850Watt Corsair 600T Black Zowie FK1 
Mouse PadAudioAudioAudio
Puretrak Talent ASUS Xonar Essence ST Sennheiser HD650 Musical Fidelity X-Can v3 
  hide details  
Reply
Desktop (Main)
(20 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 2600K @4.5Ghz Gigabyte P67-UD4-B3 MSI Twin Frozr Geforce 770 GTX 2GB Corsair Vengeance DDR3 1600Mhz 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 Pro Seagate Barracuda 1.5 TB Seagate Barracuda 1.5 TB Seagate Barracuda 4TB 
Optical DriveCoolingOSMonitor
Le random dvd writer Corsair Hydro Series™ H100i Windows 8.1 Professional 64 bit iiyama ProLite G2773HS 
KeyboardPowerCaseMouse
Battlefield 3™ Razer BlackWidow Ultimate Gaming... Truepower Quattro 850Watt Corsair 600T Black Zowie FK1 
Mouse PadAudioAudioAudio
Puretrak Talent ASUS Xonar Essence ST Sennheiser HD650 Musical Fidelity X-Can v3 
  hide details  
Reply
post #23 of 57
Quote:
Originally Posted by CorruptBE View Post

HPET really does influence input lag but there's no well defined best option.
You are experiencing 8000µs DPC latency with HPET on? Wow, something is really wrong with your system!
Quote:
Originally Posted by CorruptBE View Post

I personally get the best results when enabling HPET in the BIOS but using bcdedit /set useplatformclock false.
This is the exact opposite of what you should do. This doesn't allow Windows to use the HPET, and instead uses TSC+ACPI timers which are low performance and need synchronised. If you use bcdedit /set useplatformclock true instead, it forces HPET to be used for all timers, greatly increasing timer frequency. (from around 3-4 MHz to 14+ MHz)

Use WinTimeTester to check this. You will need to do a full system restart when changing this setting.

If your DPC latency is not approaching 500µs, there is absolutely nothing wrong with your system. Single-digit DPC latency numbers offer no real benefit, and are detrimental when it is achieved at the cost of other areas of system performance.
Edited by NotAgain - 9/30/12 at 10:37am
post #24 of 57
Quote:
Originally Posted by NotAgain View Post

I don't know what to say, it works fine for me.

That's because you're using a HDMI device with native 720p support. The test only works on a monitor.
My System
(17 items)
 
  
CPUGraphicsMonitorKeyboard
3570k @ 4.4ghz Gigabyte GTX670 @ 1300/1700 Sony FW900 + Asus VG236he CM QuickFire Red - PBT white keycaps 
MouseMouse PadAudioOther
g4Adder in progress Zowie P-TF Onkyo SE-90 Marantz PM7003 amplifier 
OtherOther
Wharfedale 10.1 speakers Sennheiser HD595 
  hide details  
Reply
My System
(17 items)
 
  
CPUGraphicsMonitorKeyboard
3570k @ 4.4ghz Gigabyte GTX670 @ 1300/1700 Sony FW900 + Asus VG236he CM QuickFire Red - PBT white keycaps 
MouseMouse PadAudioOther
g4Adder in progress Zowie P-TF Onkyo SE-90 Marantz PM7003 amplifier 
OtherOther
Wharfedale 10.1 speakers Sennheiser HD595 
  hide details  
Reply
post #25 of 57
Quote:
Originally Posted by whybother View Post

That's because you're using a HDMI device with native 720p support. The test only works on a monitor.
If your monitor cannot support a 720p signal, it seems like Nvidia is doing a good thing by giving you an image that you can actually see?
post #26 of 57
If you feed a standard 1080p monitor any non-default resolution it will display a stretched image - with every device including ATI cards. Why do Nvidia feel the need to output the default resolution with a scaled or centered image done via the gpu, even when told not to? Is this being done with a secondary buffer, possibly adding an entire frame of input lag? With such code, I wouldn't be surprised to see the scaling function and buffer used at default resolution too.
My System
(17 items)
 
  
CPUGraphicsMonitorKeyboard
3570k @ 4.4ghz Gigabyte GTX670 @ 1300/1700 Sony FW900 + Asus VG236he CM QuickFire Red - PBT white keycaps 
MouseMouse PadAudioOther
g4Adder in progress Zowie P-TF Onkyo SE-90 Marantz PM7003 amplifier 
OtherOther
Wharfedale 10.1 speakers Sennheiser HD595 
  hide details  
Reply
My System
(17 items)
 
  
CPUGraphicsMonitorKeyboard
3570k @ 4.4ghz Gigabyte GTX670 @ 1300/1700 Sony FW900 + Asus VG236he CM QuickFire Red - PBT white keycaps 
MouseMouse PadAudioOther
g4Adder in progress Zowie P-TF Onkyo SE-90 Marantz PM7003 amplifier 
OtherOther
Wharfedale 10.1 speakers Sennheiser HD595 
  hide details  
Reply
post #27 of 57
Quote:
Originally Posted by whybother View Post

If you feed a standard 1080p monitor any non-default resolution it will display a stretched image - with every device including ATI cards. Why do Nvidia feel the need to output the default resolution with a scaled or centered image done via the gpu, even when told not to? Is this being done with a secondary buffer, possibly adding an entire frame of input lag? With such code, I wouldn't be surprised to see the scaling function and buffer used at default resolution too.
The option exists because it should be easy to scale the image on the GPU in less than 8ms (one frame at 120Hz) whereas most displays take several frames to perform image scaling. (32+ms)

Scaling on the GPU also allows for greater flexibility. Many displays will simply stretch the image to fill the screen without offering aspect ratio controls, and few offer a 1:1 mode for pixel-perfect sharpness when required. It's also possible in theory for Nvidia to add other scaling options that no displays offer - in particular, there are some things they could do that would greatly improve image quality on "retina" class displays.

It shouldn't be doing the scaling on the GPU unless you tell it to, or possibly if it knows the display cannot accept the resolution. (I have never experienced that, however)

Nvidia's image scaling also seems to be much higher quality than my display's own image processing, showing considerably less ringing. (though it could be that my display has per-resolution settings I haven't changed yet - I never send it 720p)
post #28 of 57
Quote:
Originally Posted by NotAgain View Post

You are experiencing 8000µs DPC latency with HPET on? Wow, something is really wrong with your system!
This is the exact opposite of what you should do. This doesn't allow Windows to use the HPET, and instead uses TSC+ACPI timers which are low performance and need synchronised. If you use bcdedit /set useplatformclock true instead, it forces HPET to be used for all timers, greatly increasing timer frequency. (from around 3-4 MHz to 14+ MHz)
Use WinTimeTester to check this. You will need to do a full system restart when changing this setting.
If your DPC latency is not approaching 500µs, there is absolutely nothing wrong with your system. Single-digit DPC latency numbers offer no real benefit, and are detrimental when it is achieved at the cost of other areas of system performance.

Derp didn't mean it this way. For me making changes to HPET in the BIOS and the useplatformclock gives me a noticable difference in feel with my mouse. Thus far the best result was with HPET enabled in the BIOS and not useing the platformclock. If I set useplatformclock to true my mouse feels very sluggish.

Strangely enough enabling HPET in the bios but using useplatformclock false also feels different than disabling HPET al together in the BIOS.

EDIT: Not entirely sure about mouse feel but setting my RAM timings manually to what it should be able to handle (8-9-9-24 currently) halved my DPC latency so it did something right.

And a DPC latency reaching 500 is not OK at all. I had a mate of mine hitting more then 100 in a idle state, no sudden spikes, just a stable 100 to 150 ish. He was complaining about a stuttering/laggy feel in games. After doing some tweaking and dragging it down to 50 ish the laggy/stuttering feel was gone.

DPC latency hitting 25 to 35 ish now after changing my RAM timings.
Edited by CorruptBE - 9/30/12 at 4:04pm
Desktop (Main)
(20 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 2600K @4.5Ghz Gigabyte P67-UD4-B3 MSI Twin Frozr Geforce 770 GTX 2GB Corsair Vengeance DDR3 1600Mhz 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 Pro Seagate Barracuda 1.5 TB Seagate Barracuda 1.5 TB Seagate Barracuda 4TB 
Optical DriveCoolingOSMonitor
Le random dvd writer Corsair Hydro Series™ H100i Windows 8.1 Professional 64 bit iiyama ProLite G2773HS 
KeyboardPowerCaseMouse
Battlefield 3™ Razer BlackWidow Ultimate Gaming... Truepower Quattro 850Watt Corsair 600T Black Zowie FK1 
Mouse PadAudioAudioAudio
Puretrak Talent ASUS Xonar Essence ST Sennheiser HD650 Musical Fidelity X-Can v3 
  hide details  
Reply
Desktop (Main)
(20 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 2600K @4.5Ghz Gigabyte P67-UD4-B3 MSI Twin Frozr Geforce 770 GTX 2GB Corsair Vengeance DDR3 1600Mhz 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 Pro Seagate Barracuda 1.5 TB Seagate Barracuda 1.5 TB Seagate Barracuda 4TB 
Optical DriveCoolingOSMonitor
Le random dvd writer Corsair Hydro Series™ H100i Windows 8.1 Professional 64 bit iiyama ProLite G2773HS 
KeyboardPowerCaseMouse
Battlefield 3™ Razer BlackWidow Ultimate Gaming... Truepower Quattro 850Watt Corsair 600T Black Zowie FK1 
Mouse PadAudioAudioAudio
Puretrak Talent ASUS Xonar Essence ST Sennheiser HD650 Musical Fidelity X-Can v3 
  hide details  
Reply
post #29 of 57
Quote:
Originally Posted by CorruptBE View Post

Derp didn't mean it this way. For me making changes to HPET in the BIOS and the useplatformclock gives me a noticable difference in feel with my mouse. Thus far the best result was with HPET enabled in the BIOS and not useing the platformclock. If I set useplatformclock to true my mouse feels very sluggish.
Strangely enough enabling HPET in the bios but using useplatformclock false also feels different than disabling HPET al together in the BIOS.
EDIT: Not entirely sure about mouse feel but setting my RAM timings manually to what it should be able to handle (8-9-9-24 currently) halved my DPC latency so it did something right.
And a DPC latency reaching 500 is not OK at all. I had a mate of mine hitting more then 100 in a idle state, no sudden spikes, just a stable 100 to 150 ish. He was complaining about a stuttering/laggy feel in games. After doing some tweaking and dragging it down to 50 ish the laggy/stuttering feel was gone.
DPC latency hitting 25 to 35 ish now after changing my RAM timings.
Firstly, I did not say 500µs was OK, I said it's only if you are approaching 500µs that you will start running into problems. (primarily audio-related) Anything in the 100–200µs range is fine and will have no impact on system performance, certainly not input lag. Unless you have a particularly bad driver or piece of software running, most systems should be well under 250µs without doing any tweaks.

Again, DPC latency is measured in microseconds, not milliseconds. If your mouse is being polled at 1000Hz, then it is updating at 1000µs. So if your DPC latency is below that, it cannot be impacting mouse input.

At best, your display is updating at 120Hz (8000µs) and if it's LCD-based there is at least another 2000µs latency on top of that.

Anything you see by reducing DPC latency from 100–200µs to 0–50µs as far as mouse input is concerned, is entirely placebo.



If you are disabling HPET in the BIOS, and setting useplatformclock to false, you are forcing Windows to use the slowest, and least accurate timers—TSC+LAPICs which are slow and need synchronised.
If you have HPET disabled in the BIOS and have useplatformclock enabled, Windows should just be using LAPICs—slow, but no synchronisation needed as there is only one clock running.
If you have HPET enabled with useplatformclock disabled, Windows will now be using a combination of TSC+HPET which need synchronised.
If you have HPET enabled and enable useplatformclock, Windows will solely be using HPET, which is a significantly faster clock (about 14MHz compared to less than 4MHz) that is more accurate, and does not require synchronisation.


Using only HPET likely will increase DPC latency. On my system I went from 80–140 DPC latency (average around 95) to 100–180 DPC latency. (average 110) Running TimerResolution then brought this down to 60–120, average 80. This is a full OS install with peripherals plugged in and lots of software running in the background, not a system "optimized" for gaming by disabling everything that is not the game itself.

I can get my system down to 3–20µs DPC latency, averaging around 12, but this requires disabling features that I use, power saving, and using non-optimal drivers.


There is more to system performance than just DPC latency, and as previously mentioned, if you aren't even close to approaching 500µs, reducing it to sub-50µs numbers results in sub-optimal performance in other important areas.

Again, try running WinTimeTester on your system and see what clock QueryPerformanceFrequency is running at, and whether or not your clocks are synchronised. (Ratio should settle on 1.0000 after a few seconds) With HPET enabled mine is running at 14.31818MHz.
Quote:
Originally Posted by h8m3zilla View Post

i find win8 being more responsive and somewhat smoother than win7 even with 300 series drivers, though comparing it to a year old win7 installation might be the reason ;o
It's probably mostly down to it being a fresh install, though I believe they have removed some of the legacy timer options.
Edited by NotAgain - 9/30/12 at 5:22pm
post #30 of 57
You guys are still arguing about nil effect with hpet and legacy pic interupts. The issue is thread scheduling and many others. Windows doesn't multitask very well, it never has, in fact linux is rather ****ty at it to.You can have a super accurate timer, but if the kernel doesn't respond to interupts or just flat ignores them, then it doesn't matter how accurately you measure the occurence. Ifs a OS issue. Also as I told you, if you want to test drive a low input latency OS, theres a link up above.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Mice
Overclock.net › Forums › Components › Mice › Things I've noticed that alter mouse movement that nobody talks about: