Custom Resolution Utility (CRU)

[Chunky1311]

(10-16-2022 06:32 PM)ToastyX Wrote:  

(10-16-2022 08:16 AM)Chunky1311 Wrote:  I'm encountering a weird issue since upgrading to 22H2. Well, at least, I'm only now noticing it and am certain it wasn't present previously.
I can't think of anything else that has changed.

Making ANY change to the EDID of my LG CX using CRU causes issues with gsync.
It causes gsync to make the screen go black at any fps below 40fps.
This includes flickering black if the fps fluctuates even a moment to what would equal less than 40fps.
EXCEPT for 20fps exactly, this will show an image.
While this issue is present (a.k.a. after any alterations to the EDID using CRU) will also cause the PC to completely freeze and need a hard restart then toggling gsync on or off in the NVCP.

Literally all I did was change the HDR Metadata to suit my display.
What's extra odd, is even reverting the changes using CRU, does not fix the issue of the screen going black.
I need to use CRU's 'reset-all.exe' and NOT edit the EDID again whatsoever for gsync to operate properly.
I tested and this issue continues across the last 3 nvidia driver updates so I doubt it's driver related.

This is 100% an NVIDIA driver bug and should be reported to them.

I can report it to them, no worries.
Why does the issue not occur unless the EDID is edited, though?
Is there any extra info I could find to provide them?

EDIT: Setting the refresh rate to below 40hz using the NVCP also causes the screen to go black and stay black with gsync enabled.

[ToastyX]

(10-18-2022 02:49 AM)Chunky1311 Wrote:  I can report it to them, no worries.
Why does the issue not occur unless the EDID is edited, though?
Is there any extra info I could find to provide them?

The driver has a bug that triggers when any EDID override is present. I started getting reports of driver freezes/crashes with EDID overrides starting around driver version 471.11 or 471.41, so it's not exclusive to 22H2. I don't know what conditions trigger the bug because not everyone has the problem, and I haven't seen it myself.

[Chunky1311]

(10-18-2022 07:55 PM)ToastyX Wrote:  

(10-18-2022 02:49 AM)Chunky1311 Wrote:  I can report it to them, no worries.
Why does the issue not occur unless the EDID is edited, though?
Is there any extra info I could find to provide them?

The driver has a bug that triggers when any EDID override is present. I started getting reports of driver freezes/crashes with EDID overrides starting around driver version 471.11 or 471.41, so it's not exclusive to 22H2. I don't know what conditions trigger the bug because not everyone has the problem, and I haven't seen it myself.

Fair enough, that makes things difficult XD
Is there anything I can do that'd offer any insight?
I'll try a clean DDU install and see if it fixes anything.

[Peppe90]

(10-18-2022 07:55 PM)ToastyX Wrote:  

(10-18-2022 02:49 AM)Chunky1311 Wrote:  I can report it to them, no worries.
Why does the issue not occur unless the EDID is edited, though?
Is there any extra info I could find to provide them?

The driver has a bug that triggers when any EDID override is present. I started getting reports of driver freezes/crashes with EDID overrides starting around driver version 471.11 or 471.41, so it's not exclusive to 22H2. I don't know what conditions trigger the bug because not everyone has the problem, and I haven't seen it myself.

Also happened to me a freeze when trying a customized resolution from Nvidia cp on a LCD Tv (I created a custom res that the tv doesn't supports, normally it goes back after 15 secs. but this time it freezed in a black screen. Unplugging and plugging back the HDMI cable made the image to come back on screen but nvidia cp was freezed, I had to force it to close with Task manager).

Never happened to me before but I'm not sure it could be related to the use of CRU since it doesn't seem to do anything on my PC.

I tested the EDID emulation with the PC monitor (I loaded the HD CRT EDID on DP 2 with the monitor connected) and it correctly sends the signal from the EDID (indeed i got 1080i that's the main res of the HD-CRT, also i had all of his res selectable).

I edited the EDID with CRU again (I replaced 1080i and 720p with 576i and 480i), pressed ok and rebooted the PC but on both nvidia cp and windows mode list I always get the res from the original EDID as if it was not edited at all.

I don't know if it is windows preventing it from work or the gpu drivers.

One thing that works is the reset-all.exe on CRU folder. I can say it works because I got the aspect ratio setting on nvidia cp resetted, after running the exe and rebooting the PC. This make me think that the position where CRU points (where EDIDs are stored on registry) is correct. I wonder why custom res aren't applied.

[ToastyX]

(10-20-2022 08:34 PM)Peppe90 Wrote:  I edited the EDID with CRU again (I replaced 1080i and 720p with 576i and 480i), pressed ok and rebooted the PC but on both nvidia cp and windows mode list I always get the res from the original EDID as if it was not edited at all.

I don't know if it is windows preventing it from work or the gpu drivers.

I don't understand. I said to load the modified EDID using the NVIDIA control panel. Don't save any changes with CRU.

[mdrejhon]

(09-07-2012 09:06 PM)ToastyX Wrote:  Vertical total calculator - Calculates the vertical total required for the specified refresh rate and pixel clock. This can be used to implement Quick Frame Transport (QFT), which can help reduce crosstalk with backlight strobing at lower refresh rates.

Fantastic, and thank you so much for adding this. I do have suggestions (at bottom)

This makes it much easier to do the crosstalk-free modes on the Blur Busters Approved monitors, especially ViewSonic XG2431. We've created a new QFT FAQ for ViewSonic XG2431 on Blur Busters Forums:

Full Panel Zero Crosstalk: Top/Center/Bottom Equally Clear: No Visible Double Images with Strobing!







However, I want to add that Large Vertical Totals incidentally also reduce DLP and OLED lag. They often run in modes where they have to framebuffer the whole refresh cycle in order to do processing.

QFT reduces the latency of those displays quite fantastically, sometimes by tens of milliseconds!

Also, I discovered some displays even support QFT at the full bitrate, e.g. 1/500sec transmission of 60Hz refresh cycles, which turns 16.7ms of lag into 1/500sec of latency.

They don't scanout any faster, but it allows dumping frames from GPU-to-monitor (framebuffer) faster, allowing full-frame monitor processing algorithms to begin sooner (like DLP temporal dithering subfield processing)

Your mileage may vary.

So you may want to also allow people to increase pixel clock to increase vertical total without reducing refresh rate, until the cable's maximum bandwidth is hit. The pixel-receiving circuitry is sometimes 2-5x faster than its max Hz.

We've encontered 60Hz-only displays capable of accepting refresh cycles in 1/240sec, which is useful for framebuffer-processing algorithms (including HDR peak pixel detection post-processing algorithms).

This does not help beamraced-scanout displays like esports LCDs, as trying to QFT 1/500sec a 240Hz LCD while staying at 240Hz, does not seem to provide further latency benefits, and often goes OUT OF SYNC (on panels that synchronizes panel scanrate to signal scanrate).

Not all displays do this.
- Scanrate-matching algorithms goes OUT OF RANGE (panel scanrate matching calbe scanrate)
- Framebuffering-first displays often can accept refresh cycles over the cable sometimes faster than the max Hz.

DLP and auto-relevelling HDR displays necessarily has to framebuffer the whole refresh cycle before processing it (whether for subfields or for detecting brightest pixel and adjusting average picture level, or other algorithms that requires full framebuffering). Sometimes their HDMI/DP transceivers are able to accept pixels faster than max Hz into the display's scaler framebuffer for earlier begin of full framebuffer processing. Sometimes at max bandwidth of the cable too, since some of newer 60-120Hz displays use newer chips that can fill memory faster than the max-Hz refresh cycle, despite not supporting higher Hz.

Remember many 60Hz displays could accept a 120Hz signal that frameskipped. Those displays work perfect with QFT 60Hz, with a non-frameskipped 60Hz that has 8.3ms less latency!

So displays that does unsuccessful frameskipped overclocking, are also candidates for successful lag-reducing QFT, because they were accepting at least specific refresh cycles successfully at faster cable transmissions. Sometimes QFT doesn't increase lag because they key on the beginning versus end of VBI, and if they key on beginning of new refresh cycle before processing previous refresh cycle, there's no lag improvement.

But if their processing begins immediately upon delivery of last pixel, giant latency reductions occur -- several OLED's and DLP's successfully has lower latency with QFT. Not always beyond max Hz, but at least you can lower 60Hz latency massively.

- Sometimes you can only QFT lower than max Hz
- Sometimes you can QFT the max Hz
- Sometimes QFT signal won't lower latency (if processing is keyed to VBI-start)
- Sometimes QFT triggers gigantic latency reductions (on some DLP/OLED).

QFT is not just for strobing; it reduces input lag of several full-framebuffered-processing displays even without strobing, and even certain beyond max Hz (aka displays that don't do submillisecond GPU pixel-to-pixel latency like CRTs and esports LCDs).

Now My Suggestions

1. I recommend adjusting the release notes to indicate there are latency benefits for certain laggy-buffering displays like certain OLED and DLPs. It also even reduces non-strobed LCD latency too with certain sync technologies, especially if you're using Scanline Sync with the tearline adjusted to near bottom of VBI so it's a "end-of-VBI Present()" replacing VSYNC ON's "beginning-of-VBI Present()".

2. I recommend renaming "Vertical Total Calculator" to "Vertical Total Calculator (QFT)" to make sure QFT is part of it. Ideally it should have said "Quick Frame Transport" because it is universal, not just for strobing, but I'd settle for just appending "(QFT)" to make sure users get that assocation. Instead of something obscure like "Vertical Total Calculator".

Searching Google, "Quick Frame Transport" is widely the dominant name for this, especially outside engineer circles, with the most Google search results, and it would be good to keep it in sync with existing Internet information.

[Peppe90]

(10-20-2022 11:34 PM)ToastyX Wrote:  

(10-20-2022 08:34 PM)Peppe90 Wrote:  I edited the EDID with CRU again (I replaced 1080i and 720p with 576i and 480i), pressed ok and rebooted the PC but on both nvidia cp and windows mode list I always get the res from the original EDID as if it was not edited at all.

I don't know if it is windows preventing it from work or the gpu drivers.

I don't understand. I said to load the modified EDID using the NVIDIA control panel. Don't save any changes with CRU.

That's what I did, at first I never pressed the OK button on CRU, just exported the modified EDID and loaded it on Nvidia CP.
But as I said, Nvidia CP and Windows keep seeing the EDID as it was originally.

It is so: once I load the modified EDID on the DVI port, I can open CRU and it sees the EDID (named "Philips 9551 P&P monitor (active)" ) with all the modifications I made. However on Nvidia CP and windows there isn't even one of the added/modified resolutions, I just get all the default ones as if the EDID was un-edited.

Then I also tried pressing the OK button on CRU while the custom EDID was loaded on Nvidia CP but nothing changed. Only CRU can see the modifications.

I also cleaned-up things using the reset-all.exe for being sure windows wasn't using some previous/cached version of the EDID and done the work again but the result is always the same. Can't make the PC to recongnize CRU modifications in any way.

[dexhdSSS]

hi
Can someone help
I used to calmly change from 60 to 100
But now CRU does not change resolution
intel
Notebook hp omen

[Chunky1311]

(10-19-2022 02:04 AM)Chunky1311 Wrote:  

(10-18-2022 07:55 PM)ToastyX Wrote:  

(10-18-2022 02:49 AM)Chunky1311 Wrote:  I can report it to them, no worries.
Why does the issue not occur unless the EDID is edited, though?
Is there any extra info I could find to provide them?

The driver has a bug that triggers when any EDID override is present. I started getting reports of driver freezes/crashes with EDID overrides starting around driver version 471.11 or 471.41, so it's not exclusive to 22H2. I don't know what conditions trigger the bug because not everyone has the problem, and I haven't seen it myself.

Fair enough, that makes things difficult XD
Is there anything I can do that'd offer any insight?
I'll try a clean DDU install and see if it fixes anything.

Just for information purposes, a clean reinstall with DDU did in fact make the issue go away. Driver no longer bugs out after EDID edits.
Still no idea what might have caused the issue to manifest in the first place though.

[Mr4ssault]

Hey, i wanted to try a custom resolution on my lg 27gp850 it has an build in OC mode from 165hz to 180hz. When ever i try to make a custom resolution @180hz the letters turn red, or i get a weird line in the middle of the screen, screen goes black, screen flickers. I have tried every timing (i think). What can i do? I really want a custom res with display scaling, excuse me for my bad english.