What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Everything about latency. Tips, testing methods, mouse lag, display lag, game engine lag, network lag, whole input lag chain, VSYNC OFF vs VSYNC ON, and more! Input Lag Articles on Blur Busters.
Post Reply
lizardpeter
Posts: 208
Joined: 01 Dec 2020, 14:41

What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by lizardpeter » 31 Jul 2021, 12:24

So I was considering getting a 360 Hz monitor as an upgrade from my 240 Hz TN monitor, but I was concerned about the panel not being fast enough for a true 360 Hz.

I know that lowering the contrast can in some cases help to reduce these problems. Also, I’ve seen that potentially reducing the dynamic range from something like 0-255 to 16-235 can help. Does anyone know if both of these actually help? I assume it’s because pixels don’t have to transition from full white to full black ever.

Any other tips or personal experience are also helpful. I have the AW2518H and was thinking about getting the AW2521H.
i9 9900k | RTX 2080 Ti | 32 GB 4x8GB B-Die 3600 MT/s CL16 | XV252QF 390 Hz 1080p | AW2518H 240 Hz 1080p | PG279Q 144 Hz 1440p

Razer Viper 8K | Artisan Zero Mid XL | Apex Pro TKL | 1 gbps FiOS (Fiber)

deama
Posts: 368
Joined: 07 Aug 2019, 12:00

Re: What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by deama » 31 Jul 2021, 21:07

lizardpeter wrote:
31 Jul 2021, 12:24
dynamic range from something like 0-255 to 16-235 can help. Does anyone know if both of these actually help? I assume it’s because pixels don’t have to transition from full white to full black ever.
That's not how it works.
If you have it at 0-255 then having some background set to 0 will make your monitor display it at full black.
If you have it at 16-235 and you have some background set to 16 then your monitor will have it at full black, the 16 will get translated to 0 on your monitor's end.

All it does is reduce the range of shades you'll get to have less depth, which may help a bit with reducing bandwidth requirements allowing you to overlock your monitor higher, or pick a higher resolution.

lizardpeter
Posts: 208
Joined: 01 Dec 2020, 14:41

Re: What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by lizardpeter » 01 Aug 2021, 03:13

That’s interesting. I thought I’ve read somewhere on here that it avoids the full blacks and whites. I guess if that’s not the case then it wouldn’t really have a benefit in this particular situation.
i9 9900k | RTX 2080 Ti | 32 GB 4x8GB B-Die 3600 MT/s CL16 | XV252QF 390 Hz 1080p | AW2518H 240 Hz 1080p | PG279Q 144 Hz 1440p

Razer Viper 8K | Artisan Zero Mid XL | Apex Pro TKL | 1 gbps FiOS (Fiber)

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by Chief Blur Buster » 02 Aug 2021, 18:13

deama wrote:
31 Jul 2021, 21:07
That's not how it works.
If you have it at 0-255 then having some background set to 0 will make your monitor display it at full black.
If you have it at 16-235 and you have some background set to 16 then your monitor will have it at full black, the 16 will get translated to 0 on your monitor's end.
That means you have to override your monitor's end to ignore this translation step.

You can test this by watching your grey brighten when forcing:
(A) NVIDIA menus to 16-235
(B) Monitor menus to override the translation (change "Color Range" to other manual settings than "Auto")
You know if it works if the greys brighten and whites dim.

But you can manually create your gamut by fiddling with contrast/brightness until you get something like 8-248, as you might prefer a slightly wider gamut with less overdrive-overshoot guard margin.

Alternatively, simply adjust contrast to raise blacks slightly and dim whites slightly. This gives more overdrive overshoot headroom above white and below black, to get visually overshootless saturated colors. You can test that stuff in www.testufo.com/ghosting and watch the ghosting/coronas reduce when reducing the LCD contrast slightly.
lizardpeter wrote:
31 Jul 2021, 12:24
I know that lowering the contrast can in some cases help to reduce these problems. Also, I’ve seen that potentially reducing the dynamic range from something like 0-255 to 16-235 can help. Does anyone know if both of these actually help? I assume it’s because pixels don’t have to transition from full white to full black ever.

Any other tips or personal experience are also helpful. I have the AW2518H and was thinking about getting the AW2521H.
Easiest step you can do today:
1. Upgrade your mouse to at least 2000 Hz. The mouse microstutter of 1000 Hz diminishes the 360 Hz upgrade.

Tomorrow, I would like to see this happen:
2. 256x256 overdrive lookup tables (with 2-or-3-past-refresh-history lookbehind) will be the biggest improvement for 360 Hz and 480 Hz panels without requiring a panel upgrade. In addition to a slight contrast-ratio reduction.

Overdrive lookup table is simply an A(B)=C formula where the previous and next refresh cycle is simply a large math matrix of 1920x3 (subpixels) wide by 1080 tall. Normally done via scaler/TCON that is often FPGA or ASIC. But it is perfect for the parallelism of GPU processing.

A = previous refresh cycle pixel color
B = current refresh cycle pixel color
C = exaggerated color to speed up A into B without visible overshoot

The current OD LUTs are currently suboptimal. 360 Hz monitors with puny 17x7 OD LUTs? Yuck.

17x17 OD LUTs: Adding a lawn mower 2hp motor to a Ferrari, my shiny butt*

*Futurama Bender reference

They currently interpolate 17x17 OD LUT to get the remainder of the OD LUT values.

Long term, we need a SweetFX or Reshade filter that can operate on refresh-cycle granularities (via Windows Virtual Display Driver method). Once that's done, I can create a GPU software-based overdrive filter that can dramatically improve 360 Hz monitors. Like a supercharged version of the old ATI Radeon Overdrive.

An open-source virtual window driver that can do GPU-shader reprocessing at refresh cycle level (independently of frame rate), opens a lot of technology to end users, such as software black frame insertion, simulated VRR, and software-based overdrive algorithms, as well as will help multiple refresh rate race initiatives worldwide.

17x17 OD LUTs had their heydey on 60 Hz LCDs but high quality strobing and ultra-Hz need more help from finer-granularity OD LUTs. Only a few sections of the GtG heatmap needs the finer granularity, but they also happen to be the areas of the GtG heatmap where the ghosting/smearing is the most visible. 256x256 OD LUT will brute-force more of those artifacts out.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

lizardpeter
Posts: 208
Joined: 01 Dec 2020, 14:41

Re: What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by lizardpeter » 04 Aug 2021, 00:30

Thanks for the in-depth response. Just to clarify, you're saying that what I originally suggested with switching to 16-235 in the NVIDIA settings and adjusting the monitor settings manually could improve perceived pixel response?
Chief Blur Buster wrote:
02 Aug 2021, 18:13
Easiest step you can do today:
1. Upgrade your mouse to at least 2000 Hz. The mouse microstutter of 1000 Hz diminishes the 360 Hz upgrade.

Tomorrow, I would like to see this happen:
2. 256x256 overdrive lookup tables (with 2-or-3-past-refresh-history lookbehind) will be the biggest improvement for 360 Hz and 480 Hz panels without requiring a panel upgrade. In addition to a slight contrast-ratio reduction.
I just got the Razer Viper 8k a few days ago. I can definitely tell a difference on my 240 Hz monitor. I'm sure it will be even more noticeable on 360 Hz.
Chief Blur Buster wrote:
02 Aug 2021, 18:13
Long term, we need a SweetFX or Reshade filter that can operate on refresh-cycle granularities (via Windows Virtual Display Driver method). Once that's done, I can create a GPU software-based overdrive filter that can dramatically improve 360 Hz monitors. Like a supercharged version of the old ATI Radeon Overdrive.
Don't these kinds of filters add latency? I think I have seen some LDAT tests on YouTube that have shown increased latency, but I am not certain. It definitely seems like the lookup tables would help though.

I'm not sure if you have personal experience with the AW2518H (240 Hz TN) and the AW2521H (360 Hz IPS), but I am not sure if it would be a step forward or a step backward going from TN to IPS. I have seen many mixed reviews. Some people claim to go back to their old 240 Hz TN panels and others love the 360 Hz IPS.
i9 9900k | RTX 2080 Ti | 32 GB 4x8GB B-Die 3600 MT/s CL16 | XV252QF 390 Hz 1080p | AW2518H 240 Hz 1080p | PG279Q 144 Hz 1440p

Razer Viper 8K | Artisan Zero Mid XL | Apex Pro TKL | 1 gbps FiOS (Fiber)

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by Chief Blur Buster » 04 Aug 2021, 15:20

lizardpeter wrote:
04 Aug 2021, 00:30
Don't these kinds of filters add latency? I think I have seen some LDAT tests on YouTube that have shown increased latency, but I am not certain. It definitely seems like the lookup tables would help though.
Yes, filters add latency but at 240Hz+, it's only a tiny amount (1/240sec or 2/240sec) when optimized heavily in full screen exclusive mode. Some of us want better quality for everyday use, not for esports. So they are still beneficial. It will help drive the industry to do better with overdrive, etc.

I know that several are attempting from-scratch developments (independent efforts I had incidentally heard about) but ideally a windows driver version of an existing project like SweetFX or Reshade is preferred, to utilize the large mass of community knowledge. I'm looking forward to all of this. It may be a couple years for the open source projects, but it won't be long in humankind.

Microsoft DDK actually has a source code example (60 Hz display mirroring driver, normally used for USB displays) that can be modified to work at different refresh rates or add shader processing or other neato tricks. The open source projects just have to utilize this stuff.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

diakou
Posts: 83
Joined: 09 Aug 2020, 11:28

Re: What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by diakou » 04 Aug 2021, 18:34

Chief Blur Buster wrote:
04 Aug 2021, 15:20
lizardpeter wrote:
04 Aug 2021, 00:30
Don't these kinds of filters add latency? I think I have seen some LDAT tests on YouTube that have shown increased latency, but I am not certain. It definitely seems like the lookup tables would help though.
Yes, filters add latency but at 240Hz+, it's only a tiny amount (1/240sec or 2/240sec) when optimized heavily in full screen exclusive mode. Some of us want better quality for everyday use, not for esports. So they are still beneficial. It will help drive the industry to do better with overdrive, etc.

I know that several are attempting from-scratch developments (independent efforts I had incidentally heard about) but ideally a windows driver version of an existing project like SweetFX or Reshade is preferred, to utilize the large mass of community knowledge. I'm looking forward to all of this. It may be a couple years for the open source projects, but it won't be long in humankind.

Microsoft DDK actually has a source code example (60 Hz display mirroring driver, normally used for USB displays) that can be modified to work at different refresh rates or add shader processing or other neato tricks. The open source projects just have to utilize this stuff.
I am beyond confident Kaldaien can add exactly what you need to Special K in all honesty. This refresh rate granularity type projects that are completely in accordance with the OS are one of his best specialities. Is there a bigger writeup somewhere on exactly what you would need? I could pass that link around a bit and tinker and see what the possibilities are with a few people. Despite it being in something such as SK, Kal has stated a few times that he would love to bake a lot of SK's features into ReShade due to whitelisting and common familiarity. Though ultimately, I'm sure that if there exists someone for such a thing as of today (i.e if Kal is capable of it.) I'm sure he's probably the only person realistic for something like that lol.

Edit 2: Have linked Kaldaien, hoping it's relevant to him or interesting, would be a cool project.
Last edited by diakou on 11 Aug 2021, 14:27, edited 2 times in total.

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: What can be done to reduce ghosting/motion blur and improve pixel response on 360 Hz IPS?

Post by Chief Blur Buster » 04 Aug 2021, 21:19

diakou wrote:
04 Aug 2021, 18:34
I am beyond confident Kaldaien can add exactly what you need to Special K in all honesty. This refresh rate granularity type projects that are completely in accordance with the OS are one of his best specialities. Is there a bigger writeup somewhere on exactly what you would need? I could link around a bit and tinker and see what the possibilities are. Despite it being in something such as SK, Kal has stated a few times that he would love to bake a lot of SK's features into ReShade due to whitelisting and common familiarity.
Thanks. Being an open-invitation project of such an absurdly wonderful generic idethat catapults the refresh rate race to retina refresh rates, this is something that the Blur Busters universe really badly needs -- and will help me convince manufacturers that the puny 17x17 OD LUTs are not enough anymore...

There are some additional amazingly clever tricks that can be done with such an open source driver, so at this stage, it's a race to see which developer does it first and under what license (GPL, MIT, Apache 2.0, etc)
  1. VRR simulator for non-VRR displays
    Using my TestUFO algorithm www.testufo.com/vrr ... Windows sees a VRR display but actual display is non-VRR. VRR by my VRR simulation algorithm. The math is so simple that I can do it in Javascript, it probably won't do much overhead.
  2. Software BFI to simulate a low-Hz strobed display on unstrobed high-Hz displays
    Much like www.testufo.com/blackframes except done at driver level. Good for 120Hz and 240Hz displays with no strobe backlights. Windows sees native 60Hz display, but display is secretly running 120Hz or 240Hz BFI as applicable. This is a low-overhead filter too.
  3. Software based overdrive with true GPU-shader-based 256x256 overdrive lookup tables
    These tables would be populated by my photodiode tester device from GtG curves, and possibly 3D OD LUT's (multiple refresh cycle depth). This may have more overhead, but not much for simple 65536-byte LUTs (2D LUT of 2 consecutive refresh cycles) or 16777216-byte LUTs (3D LUT of 3 consecutive refresh cycles). The reason I need multi-refresh-cycle OD LUTs is 360Hz displays badly need 3D OD LUTs to have GtG keep up with the joneses...
  4. Improve VRR on existing generic VRR displays, to possibly better than G-SYNC native
    I need a refresh cycle queue & frametime queue to know the history of pixel colors and frametimes, to compute the most perfect possible overdrive for the next VRR refresh cycle. After simple OD LUT, I could replace OD LUT with advanced floating point overdrive algebra formulas. Possible autocomputed algebra formula overdrive per pixel (quadratic regression for each prev2,prev1,current sequence of pixels) creating 65536 (2D OD LUT) or maybe even 16777216 (3D OD LUT) algebra formulas that execute depending on what the last 2 or 3 consecutive greyscale color values (per subpixel) were, with this formula executed per subpixel per refresh cycle, for vastly superior software-based VRR OD. I need to compute 1.5 billion algebra math formulas per second for this superior improved VRR overdrive algorithm for 1080p240, and the great news is RTX cards can do it. Woohoo, goodbye crappy 17x17 OD LUTs! It will require a lot of pre-computing to generate the source math formulas and compress it into the GPU memory (all 16 million formulas). The formulas may also require frametimes between consecutive refresh cycles, for dynamic overdrive, which will make things a lot more complex (perhaps cubic regression formula ala Y = AX^3 + BX^2 + CX + D ... or more advanced formula) to curve-fit GtG curves to all the pixel color combinations and frametime combinations of high-Hz VRR. Compressing these formulas into GPU memory is a challenge, but it would potentially produce better than native G-SYNC quality even onto generic uncertified VESA Adaptive Sync, or piggyback on top of FreeSync VRR. It would need a photodiode to automatically measure the current display for a PC app to generate the necessary LUT of algebra-formulas to compute VRR dynamic overdrive with -- it's either a 3D, 4D, 5D or 6D lookup table which would be terabytes of memory but compressed to only a few megabytes of algebra formulas in an algebra-formula LUT that can fit in only 1% of GPU memory yet still be better than G-SYNC native. Just simply by using a powerful GPU shader on an RTX card instead of a G-SYNC chip. I suspect NVIDIA wouldn't mind because the computing power pretty much asks for humongous RTX horsepower to pull off better-than-native-G-SYNC overdrive quality. And there will be more input lag than native G-SYNC because of the additional processing overhead required for GPU-based dynamic VRR overdrive improvement. This computing power may consume 10% of an RTX in the background, slowing frame rates by 10%, but the VRR suddenly goes better than native G-SYNC quality. So NVIDIA still profits from this GPU-shader algebra software-based VRR overdrive even if they lose on G-SYNC revenue.
Item 4 is kind of a math-formula megaproject that will require lots of automation with measuring equipment, and will have to be tuned on a per-panel basis (panel variances, etc). Which may require commercializing my photodiode lag tester that I developed in-house -- the great thing it can double as a GtG tool and a OD LUT generator tool -- so users could strap it to their monitor, just press a button in the Windows app, then wait a few hours for the computer to automeasure/autocompute the algebra-based 6D VRR OD LUT and install it into the GPU, and suddenly see crappy uncertified VRR suddenly go better than native G-SYNC quality, thanks to my advanced algebra-based 6-dimensional VRR overdrive LUT (three consecutive refresh cycle pixel colors, and the frametimes between them, equals 6-dimensions). Some predictiveness of future frametime may or may not be required, based on testing. Once somebody else creates the open source shader windows driver package, I've got the skillz to turn this into a one-button process for newbie users if they wish. Anyway, big rabbit hole.

Anyway, the other projects (BFI, VRR simulation) is easy to implement. So I'd begin with that first.

There's a lot of potential spinoff applications I can do with a refresh-cycle-granularity shader that's VRR-compatible and fixed-Hz compatible and can simulate any different Hz display onto a different Hz display (e.g. simulate a 60Hz display onto a 240Hz display, while applying multi-refresh algorithms like software-based BFI). I could even simulate a 240Hz VRR display onto a 144Hz display in theory, by blending multiple frames into the same frame. It won't be clearer than 240 Hz, but it would be a good for game developer debugging.

Generally I've been privately telling dozens of developers about this concept idea for their open source project:
- A refresh cycle granularity version of Reshade or SweetFX by adding it to a virtual display driver
'- The virtual display should be able to be capable of a different Hz than the source display (higher or lower)
- The virtual display should be able to execute any GPU shader I create for it

One technique to make different-Hz algorithms work, is that it needs to be in full screen exclusive mode so that the VSYNC's doesn't throttle each other to lowest-Hz display. I have long known this full screen exclusive tip for years since around 2015-ish. Different-Hz multimonitor has lots of problems (See many old complaints in these forums over the last 5+ years) and full screen exclusive fixes this by making each display run at their own independent Hz without throttling each other's framerates.

I probably have some additional code to donate (a registry configuration app from another project).

TL;DR: It's such a beautifully generic idea to combine two existing projects (the Microsoft Windows virtual display driver source code sample) + (either the existing SweetFX or ReShade projects) to pull off something like this. Lots of prior art that just needs minor glue to combine the projects to produce a mega-explosion. Which provides an explosion of very BlurBusters-worthy algorithms like those I've already put on TestUFO for many years. The shaders themselves could be open source or proprietary, but we need an open source driver hardness brought out to the world.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

Post Reply