NVIDIA Pixel Clock Patcher

[sephr]

Is the 1.3.3 patcher working well with the new 358.66 drivers?

[Mangix]

any chance of making a linux version available? Most of the work has been done: (https://github.com/CFSworks/nvlinpatch) although not all the patches are supported.

[niciuffo]

(11-03-2015 04:20 PM)ToastyX Wrote:  

ToastyX Wrote:What happens when you try to go over 165 MHz? Is the driver not allowing it ("test failed" message) or is the TV not displaying it?

When going over the 165MHz limit, the TV just displays this message:


The nVidia CP doesn't say anything about the resolution not being valid.
To add to this, a pixel clock of 163.3500MHz works, but a value of 165.8250MHz does not.

ToastyX Wrote:Hardware limits are not exact like that. It's usually possible to go higher than 165 MHz unless there's a new driver limit that I don't know about. If it's a driver limit, 165.00 MHz will work but 165.01 MHz will not. If the TV has a pixel clock limit, it won't be exact like that.

I tried messing with the "Total pixels" values to test at which point the TV refuses to accept the signal (I changed the Horizontal values).
Up to this point, the TV accepts the resolution:

At this point, the TV refuses it ("Mode not supported"):

ToastyX Wrote:Those values are impossible because the total includes the active + front porch + sync width. It's probably reducing the sync width without telling you.

That might be true. One thing I should note, which I found weird. This is the default resolution (1920x1080@60Hz), seen by the TV's OSD:

And this is the custom (1920x1080@75Hz, settings in my last post) resolution:

Could this be a bug on the TV side, or is it really only displaying that amount of pixels? The pixel clock was still below 165MHz, so I don't know why it would not display the full 1080p image.
Is there a test I can do to verify the resolution?

ToastyX Wrote:Technically, 0 front porch values are possible, but the driver might not be handling it correctly. The TV should tell you what refresh rate it's actually receiving in the OSD.

I tried setting the front porch values to 0, and the sync width values to 1 (can't set them to 0). Total pixels: 2020x1089.
I am able to test a resolution with a pixel clock higher than 165MHz. I tired setting the refresh rate to 120Hz, and the TV accepted the signal, although the OSD still said "1440x900@75Hz".
Moving the mouse around, I can confirm that it was not running at 120Hz.

ToastyX Wrote:The only way to reduce the pixel clock without reducing the refresh rate is to reduce the totals. The totals include the front porch and sync width, so reducing those might allow you to reduce the totals further.

I tried reducing those values to 1, to remain within the 165MHz limit, and the max stable refresh rate was 77Hz I think. It was very unstable though, every ~15 seconds the TV would display a black screen for a couple of seconds.

[Lumonaut]

On 358.87 I was able to patch and find all values but cannot add custom resolutions. Even with "Enable resolutions not exposed by the display" checked in Nvidia Control Panel any custom resolution at all gives me "Test failed. Custom resolution 2560x1440 at 96Hz (32-bit) is not supported by your display."

I have a 980 Ti. Any ideas?

[ToastyX]

(11-07-2015 05:59 PM)Lumonaut Wrote:  On 358.87 I was able to patch and find all values but cannot add custom resolutions. Even with "Enable resolutions not exposed by the display" checked in Nvidia Control Panel any custom resolution at all gives me "Test failed. Custom resolution 2560x1440 at 96Hz (32-bit) is not supported by your display."

I have a 980 Ti. Any ideas?

You shouldn't need the patch to add 2560x1440 @ 96 Hz when using the NVIDIA control panel with the 980 Ti.

You didn't mention what monitor you have, but I'm assuming it's the QNIX/X-Star monitor because of the resolution. Did you install the monitor driver?

https://www.monitortests.com/qnix.zip

If you're using Windows 8 or later, you'll have to temporarily disable driver signature enforcement first: http://www.howtogeek.com/167723/how-to-d...d-drivers/

Then install the INF file using the Device Manager: Monitors -> Right-click "Generic PnP Monitor" -> Update Driver Software... -> Browse my computer for driver software -> Let me pick from a list of device drivers on my computer -> Have Disk... -> Browse for the Qnix.inf file -> OK -> Next -> Install this driver software anyway

Then reboot and it should let you add the resolution. If not, you can try using CRU instead, which requires the patch.

[ToastyX]

(11-07-2015 04:00 PM)niciuffo Wrote:  

ToastyX Wrote:What happens when you try to go over 165 MHz? Is the driver not allowing it ("test failed" message) or is the TV not displaying it?

When going over the 165MHz limit, the TV just displays this message:
[img]

The nVidia CP doesn't say anything about the resolution not being valid.
To add to this, a pixel clock of 163.3500MHz works, but a value of 165.8250MHz does not.

ToastyX Wrote:Hardware limits are not exact like that. It's usually possible to go higher than 165 MHz unless there's a new driver limit that I don't know about. If it's a driver limit, 165.00 MHz will work but 165.01 MHz will not. If the TV has a pixel clock limit, it won't be exact like that.

I tried messing with the "Total pixels" values to test at which point the TV refuses to accept the signal (I changed the Horizontal values).
Up to this point, the TV accepts the resolution:
[img]
At this point, the TV refuses it ("Mode not supported"):
[img]

This looks like a TV limitation. 164.9959 MHz should still work if it were a driver limitation.

Adjust the refresh rate to fine tune the pixel clock. If you can get it to fail at around 164.98 MHz, then the TV has its own pixel clock limit.

(11-07-2015 04:00 PM)niciuffo Wrote:  

ToastyX Wrote:Those values are impossible because the total includes the active + front porch + sync width. It's probably reducing the sync width without telling you.

That might be true. One thing I should note, which I found weird. This is the default resolution (1920x1080@60Hz), seen by the TV's OSD:
[img]
And this is the custom (1920x1080@75Hz, settings in my last post) resolution:
[img]
Could this be a bug on the TV side, or is it really only displaying that amount of pixels? The pixel clock was still below 165MHz, so I don't know why it would not display the full 1080p image.
Is there a test I can do to verify the resolution?

It's probably a bug in the TV's firmware. The video card controls the resolution, so it would be obvious if the TV was not displaying the full resolution, especially with text.

(11-07-2015 04:00 PM)niciuffo Wrote:  

ToastyX Wrote:The only way to reduce the pixel clock without reducing the refresh rate is to reduce the totals. The totals include the front porch and sync width, so reducing those might allow you to reduce the totals further.

I tried reducing those values to 1, to remain within the 165MHz limit, and the max stable refresh rate was 77Hz I think. It was very unstable though, every ~15 seconds the TV would display a black screen for a couple of seconds.

Don't set the values too low or the TV won't be able to handle them.

77 Hz is possible with these values:
H: 1920 8 16 1960
V: 1080 3 5 1093

75 Hz is possible with these values:
H: 1920 24 32 2008
V: 1080 3 5 1093

[ToastyX]

(11-06-2015 09:58 PM)Mangix Wrote:  any chance of making a linux version available? Most of the work has been done: (https://github.com/CFSworks/nvlinpatch) although not all the patches are supported.

I've been wanting to make a Linux version, but I haven't gotten around to it. I'd have to install Linux and find all the different limits and test all the limits with different driver versions. Then I'd have to do the same for AMD.

[ToastyX]

(11-04-2015 11:12 PM)sephr Wrote:  Is the 1.3.3 patcher working well with the new 358.66 drivers?

It works for 64-bit versions of 358.xx. The patch has to be updated for 32-bit. Is anyone using the 32-bit driver?

[Lumonaut]

(11-07-2015 09:04 PM)ToastyX Wrote:  

(11-07-2015 05:59 PM)Lumonaut Wrote:  On 358.87 I was able to patch and find all values but cannot add custom resolutions. Even with "Enable resolutions not exposed by the display" checked in Nvidia Control Panel any custom resolution at all gives me "Test failed. Custom resolution 2560x1440 at 96Hz (32-bit) is not supported by your display."

I have a 980 Ti. Any ideas?

You shouldn't need the patch to add 2560x1440 @ 96 Hz when using the NVIDIA control panel with the 980 Ti.

You didn't mention what monitor you have, but I'm assuming it's the QNIX/X-Star monitor because of the resolution. Did you install the monitor driver?

https://www.monitortests.com/qnix.zip

If you're using Windows 8 or later, you'll have to temporarily disable driver signature enforcement first: http://www.howtogeek.com/167723/how-to-d...d-drivers/

Then install the INF file using the Device Manager: Monitors -> Right-click "Generic PnP Monitor" -> Update Driver Software... -> Browse my computer for driver software -> Let me pick from a list of device drivers on my computer -> Have Disk... -> Browse for the Qnix.inf file -> OK -> Next -> Install this driver software anyway

Then reboot and it should let you add the resolution. If not, you can try using CRU instead, which requires the patch.

Already disabled driver enforcement and installed that. Yes it's the QNIX QX2710. Already rebooted and that's the error that Nvidia's control panel gives me. It gave me that before I installed Toasty's patch as well but I thought that might fix it.

[ToastyX]

(11-07-2015 10:02 PM)Lumonaut Wrote:  Already disabled driver enforcement and installed that. Yes it's the QNIX QX2710. Already rebooted and that's the error that Nvidia's control panel gives me. It gave me that before I installed Toasty's patch as well but I thought that might fix it.

That sounds like a driver bug, or they added a new limit that I haven't heard about. I know some people had trouble adding any refresh rate with some of the earlier Windows 10 drivers, but that should have been fixed by now. I don't know why some people are still affected by that. Have you tried wiping the driver with DDU? http://www.wagnardmobile.com/DDU/

If that doesn't work, the CRU method should still work.