Custom Resolution Utility (CRU)

[mdrejhon]

(11-19-2022 04:59 AM)ToastyX Wrote:  

(11-19-2022 03:09 AM)mdrejhon Wrote:  There are many mitigation measures that people do to compensate, and it's important for that person to adjust.

My point is increasing the refresh rate wouldn't mitigate any of that because the frame rate would still be low. In fact, changing the refresh rate can actually increase judder if it's not a multiple of the frame rate.

That's definitely true, increasing refresh rate can cause problems with certain content -- not all content, software,

(11-19-2022 04:59 AM)ToastyX Wrote:  Increasing the refresh rate for desktop use only has minor improvements like smoother mouse movement and smoother scrolling. If someone gets headaches from general computer use, the refresh rate is not likely the cause.

I know, but it definitely exists.

There are literally over 100 niche ergonomic causes of display discomfort of all kinds (eyestrain and/or nausea and/or motionsickness and/or headaches) other than PWM, blue light, etc.

We're already experts at diagnosing many peoples' headaches. I also now fly in-person to teach classrooms (e.g. photo at bottom of services.blurbusters.com ...)

Also, you need geometrics (60Hz -> 144Hz -> 360Hz -> 1000Hz) to see major improvements in things like scrolling. For average users, a 4x geometric is recommended. Doubling Hz halves motion blur, assuming GtG=0 and that's why Blur Busters is a huge fan of geometric refresh rate upgrades (~2x-4x), especially on OLED which more linearly follows Blur Busters Law.

(11-19-2022 04:59 AM)ToastyX Wrote:  I'm aware that global refresh displays exist, but they are such a small minority, and they tend to be TVs with additional processing lag, so anyone that cares about lag shouldn't be using them. The vast majority of monitors and TVs don't do global refresh.

Don't forget that the panel does not have to be global refresh, it can simply be one of the thousands of LCD and OLED displays that needs to pre-buffer the full refresh cycle first before processing. All of those thousands of LCD and OLED displays gets lag reductions with QFT, because of pre-buffering, even if the panel is not global refresh.

To clarify, definitions:

Real-time scanout: A display that can accept a pixel row from the input, and refresh that pixel row directly to the panel in real time. Often a rolling-window of a few scanlines is used, for things like image scaling (bilinear/bicubic/etc) due to algorithmic need to know the pixel row(s) above/below.

Full refresh cycle buffering before scanout: A display that, for some purpose, needs to buffer the whole refresh cycle from the signal input, BEFORE it scans out to the panel (even if the panel is not global refresh). Some algorithms are much easier to do with global buffering, such as many HDR algorithms, and many kinds of OLED-processing algorithms.

(11-19-2022 04:59 AM)ToastyX Wrote:  Strobed LCD is a different issue. The whole point of using QFT with strobing is to finish the scanout faster so there's more time for the pixels to change before the backlight strobes, which reduces crosstalk. QFT wouldn't reduce lag with strobing unless you make the backlight strobe sooner, but then that would negate the crosstalk reduction.

No, it does not negate the crosstalk reduction.

You still get crosstalk reductions AND ability to strobe sooner, because of laws of physics. In a 1ms VBI between a 15.7ms scanout, you may have 0ms before/after a 1ms pulse. In a 12.9ms VBI between a 4.2ms scanout, you have 4.35ms before AND after the 1ms pulse. So 4.35ms lag reduction AND crosstalk-free 60Hz strobing. If the GtG is sufficiently fast, you can even push the strobe pulse slightly sooner, e.g. 5.35ms or 6.35ms lag reduction, and still stay zero crosstalk, if the GtG curve is still below human visibility threshold and the curve is still completely inside the VBI time.*
*Note: There's a phase offset depending on GtG lag behavior, however once GtG transition is centered in the VBI (so that the leading/trailing edge is invisibilized equally at both ends), you now have room to slide the curve forward/back. For custom QFT-calibrated Pure XP Custom modes (at the 60-100Hz territory), XG2431 is the first zero-crosstalk IPS LCD's on the market that's not a VR LCD.

If you are unaware, ViewSonic paid me to do one of the world's most perfect strobe tunings (beating ULMB, beating LightBoost), on the XG2431.



If you're currently unaware, I have already done paid consulting services for multiple manufacturers so I've gained some immense knowledge about these matters...

(11-19-2022 04:59 AM)ToastyX Wrote:  I'd like to know more about which displays do global refresh. The problem is most sites don't test lag properly or mention the scanout method, so I don't have enough information. For instance, RTINGS lag measurements include vsync lag measured in the center of the screen, so it's usually higher than the actual lag, and I don't see them talking about the scanout method.

This is true, we need more accurate lag testing for the whole screen surface.

(11-19-2022 04:59 AM)ToastyX Wrote:  IThe part about all display (including LCD) HDR needing the full frame is new information to me.

Fixed that for you.

There's many reasons:
- Remember local dimming backlights often need refresh-cycle prebuffering logic too.
- Some LCDs switch between realtime scanout mode and buffered-mode depending on the monitor settings configured into the monitor menus.

In addition, some LCDs are scanrate-multisync and other LCDs are fixed-scanrate (which is why they have high 60Hz latency).
You still have VSYNC OFF latency reductions at QFT 60Hz on many 240Hz LCD panels, with strobing=OFF!

Note: Alternating grey bars = relative frameslice height example (not exact count of frameslices).

As a very small example, all BenQ 240Hz (XL254*), Acer XB252, ViewSonic XG270, and many dozens others, are fixed-scanrate panels and buffers sub-240Hz refresh cycles, and they all have lag reductions during QFT with lower numbers than RTINGS numbers, when tested with QFT.






Until recently, most 240Hz panels were fixed-horizontal-scanrate. Fixed scanrate panels require buffering most of the slow-scanning refresh cycle signal before scanning-out the panel. So they have lag reductions with QFT.

And we haven't entered the territory, that until recently, there was not a single high-Hz big-brand-name 1440p panel with horizontal scanrate multisync capability. Your early 1440p monitor had to buffer the whole refresh cycle before fast-scanout. In fact, almost all NVIDIA native GSYNC panels were fixed-horizontal-scanrate (at first), but unfortunately didn't support custom refresh rates until a few generations ahead, and then when they did, many legacy panels are apparently already QFT-compatible (even lag reductions during VSYNC OFF, strobing OFF, and without needing RTSS)

BTW, we may be launching our hardware-based lag tester, which also lets you test multiple parts of the screen. This is the very same device used to tune XG2431, and the very same device that discovered QFT benefits that YouTubers haven't discovered yet.



In my software, you are forced to measure top/center/bottom before validated lag results are displayed, as well as some calculated scanout information.
(BTW, my tester is VRR compatible too).

While VRR is already a natural QFT-included, since VRR uses a resizeable blanking interval, so if you can use capped VRR, it's an easier lag reduction than using QFT (e.g. 60fps cap at 240Hz VRR, for emulator use). However, QFT is quite handy when you can't do VRR (because of the source side like Intel GPU, or because of display side, or because of a display mode you want to use that doesn't support VRR). QFT is essentially a fixed-Hz equivalent of doing a perfectly perma-capped VRR, since the signal structure is quite similar, and thus has many use cases.

Also, I want to add more use cases I forgot to add:
-- Some signal sources can only do 60Hz or 120Hz, and some of those honor EDID information (interestingly), which leads to some "interesting" EDID hacking opportunities.
-- People who still use 3D mode, benefit hugely from 120Hz QFT too (thousands of us use VorpX today as a sequel to NVIDIA 3D Vision), since 240Hz+ panels are pretty 3D-friendly. With software-based BFI, 240Hz OLEDs will probably be fantastic 3D-glasses panels for those current open source VorpX fans too.
-- People who want to run fixed-Hz 60Hz at the lowest possible latency on a GPU that does not support VRR. 240Hz panels with QFT 60Hz works perfectly fine on Intel GPUs, and creates lag reductions on the majority of 240Hz panels, and stutters less in many emulators than 120Hz-240Hz
-- Etc. (this is all in addition to all the stuff I already wrote earlier). Some niche, some mass market.

While LCD-based QFT non-strobed VSYNC OFF may not benefit the max-Hz player as much as anybody who needs to use a refresh rate below max Hz, this simply shows how a gigantically large number of displays are not horizontal scanrate multisync, as evidenced from their sudden buffering behavior that enables when running at refresh rates below max Hz for any reason.

TL;DR: QFT has latency reductions on many thousands more LCD displays than you think, even for VSYNC OFF. It just needs to be made much more user friendly (plug-and-play automations), and your Vertical Total Calculator made this experimentation a hell lot easier, creating a boom of mini QFT experimentations unbeknownst to you...

People like you, reviewers, professionals, YouTubers, don't know this well because this is territory we are exploring more aggressively than anyone else at this time. So Blur Busters is leaping increasingly into the QFT-education role now;

So I reiterate: I think you underestimate / underrate QFT severely... Because so little of this has been discovered by youtubers/reviewers (yet).