OLED BFI Options?

High Hz on OLED produce excellent strobeless motion blur reduction with fast GtG pixel response. It is easier to tell apart 60Hz vs 120Hz vs 240Hz on OLED than LCD, and more visible to mainstream. Includes WOLED and QD-OLED displays.
Ogmilkman
Posts: 20
Joined: 26 Sep 2020, 08:19

Re: OLED BFI Options?

Post by Ogmilkman » 29 May 2023, 20:05

Chief Blur Buster wrote:
10 May 2023, 22:09
I think we will have to wait till 480 Hz OLEDs for 240Hz+BFI.

Currently, only VR OLEDs (e.g. PSVR2) is capable of hardware based sub-refresh BFI.
Is it correct to say that 480Hz OLED displays will have less motion clarity than the current 360Hz IPS with ULMB 2? I assume these OLEDs will only be able to reach 240Hz with BFI and a 720Hz OLED (360Hz with BFI) is necessary to achieve the same motion clarity as a PG27AQN currently.

Or should we expect 480Hz OLED without BFI to have better motion clarity even without BFI?

GammaLyrae
Posts: 116
Joined: 28 Mar 2018, 01:44

Re: OLED BFI Options?

Post by GammaLyrae » 29 May 2023, 21:09

Ogmilkman wrote:
29 May 2023, 20:05
Chief Blur Buster wrote:
10 May 2023, 22:09
I think we will have to wait till 480 Hz OLEDs for 240Hz+BFI.

Currently, only VR OLEDs (e.g. PSVR2) is capable of hardware based sub-refresh BFI.
Is it correct to say that 480Hz OLED displays will have less motion clarity than the current 360Hz IPS with ULMB 2? I assume these OLEDs will only be able to reach 240Hz with BFI and a 720Hz OLED (360Hz with BFI) is necessary to achieve the same motion clarity as a PG27AQN currently.

Or should we expect 480Hz OLED without BFI to have better motion clarity even without BFI?
I'd expect that it'd be down to user preferences / tolerances. In the short term, OLED is going to offer a higher MPRT (the LG C1 is around 4~ms with BFI on at 120hz, a feat I expect to be repeated by "Dough"'s upcoming 240hz monitor sized OLED), but each frame has zero crosstalk and zero overshoot. It's very clean.

The XG2431 can achieve 1ms~ MPRT at any refresh, but the amount of visible crosstalk or overshoot depends entirely on which refresh rate you're running the monitor at. The closer you are to max, the more crosstalk is visible. Technically still 1ms MPRT the entire time, but users seeking a "glass smooth, CRT-like appearance" would favor 120hz or lower on this specific monitor.

Until someone does proper benchmarks for either of the ULMB2 monitors on the market, I would remain cautiously optimistic. The only thing we've been promised in the marketing is that strobe at max refresh (or any refresh without a QFT trick) will be higher quality than competing monitors strobing under similar conditions.
Last edited by GammaLyrae on 29 May 2023, 21:28, edited 1 time in total.

bumbeen
Posts: 84
Joined: 25 Apr 2023, 14:35

Re: OLED BFI Options?

Post by bumbeen » 29 May 2023, 21:25

Ogmilkman wrote:
29 May 2023, 20:05
I assume these OLEDs will only be able to reach 240Hz with BFI and a 720Hz OLED (360Hz with BFI) is necessary to achieve the same motion clarity as a PG27AQN currently.
while maintaining brightness they'd have to chop the refresh rate. But if we're fine with a 50 nit output they could do BFI at full refresh. I think it's a bit disingenuous of nvidia to say "THE MOTION CLARITY OF 1400 FPS!!" It doesn't have that. Not even close. Even in their own video where they show the 480hz vs the 120hz+BFI you can see the background is incredibly choppy by comparison. Just because one element of the screen doesn't have persistence motion blur does not mean the motion clarity overall is equivalent to 4x the number of frames. If that were the case then everyone would be saying the pg27aqn sucked solely because it lacked BFI and the XG27AQM(R) would be the far superior display. We would all still be using CRTs if it were true, 60hz CRTs do not have the motion clarity of a 1000fps display

edit: adding on to this. An OLED at 360hz with BFI will have the motion clarity of the PG27AQN at 360HZ with ULMB2, probably slightly better. The backlight pulse basically replaces the response time of the pixels. So there's slightly more pixel response latency with ULMB2 enabled. The backlight waits until the pixels at the bottom of the display have changed before flashing. So you're having to wait for that to happen before you can see it. On an OLED this will happen much faster since the pixel response time is quicker. It would simply wait until it got all the pixel information for the frame, then flash it and you'll see it, while the LCD is still waiting on the pixels to change.

Ogmilkman
Posts: 20
Joined: 26 Sep 2020, 08:19

Re: OLED BFI Options?

Post by Ogmilkman » 30 May 2023, 01:42

bumbeen wrote:
29 May 2023, 21:25
Ogmilkman wrote:
29 May 2023, 20:05
I assume these OLEDs will only be able to reach 240Hz with BFI and a 720Hz OLED (360Hz with BFI) is necessary to achieve the same motion clarity as a PG27AQN currently.
while maintaining brightness they'd have to chop the refresh rate. But if we're fine with a 50 nit output they could do BFI at full refresh. I think it's a bit disingenuous of nvidia to say "THE MOTION CLARITY OF 1400 FPS!!" It doesn't have that. Not even close. Even in their own video where they show the 480hz vs the 120hz+BFI you can see the background is incredibly choppy by comparison. Just because one element of the screen doesn't have persistence motion blur does not mean the motion clarity overall is equivalent to 4x the number of frames. If that were the case then everyone would be saying the pg27aqn sucked solely because it lacked BFI and the XG27AQM(R) would be the far superior display. We would all still be using CRTs if it were true, 60hz CRTs do not have the motion clarity of a 1000fps display

edit: adding on to this. An OLED at 360hz with BFI will have the motion clarity of the PG27AQN at 360HZ with ULMB2, probably slightly better. The backlight pulse basically replaces the response time of the pixels. So there's slightly more pixel response latency with ULMB2 enabled. The backlight waits until the pixels at the bottom of the display have changed before flashing. So you're having to wait for that to happen before you can see it. On an OLED this will happen much faster since the pixel response time is quicker. It would simply wait until it got all the pixel information for the frame, then flash it and you'll see it, while the LCD is still waiting on the pixels to change.
To answer my own question:

It seems Nvidia’s claim of 1000Hz+ motion clarity is referring to the MPRT in ULMB 2. With the proposed default factor of 4, a 360Hz monitor will have the same MPRT as a 1440Hz monitor without strobing.

OLEDs are self emissive and BFI support above half the maximum refresh rate would require a subrefresh rolling scan. I feel it’s unlikely the 480Hz OLED display next year will have the backplane to support this subrefresh so it may be a while until available OLED displays are able to match this MPRT (at the very least two years assuming next year we will not have the proper backplane).

This is unfortunate for me as I wanted a do it all monitor (1440p OLED 360Hz+ with great motion clarity rivalling the XL2566K) that I could buy and use until the hardware and display came out for the move to the same specs in 4K.

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

Re: OLED BFI Options?

Post by Chief Blur Buster » 30 May 2023, 03:37

Keep in mind, there are four caveats to get the "1000Hz look" with any strobe backlight (ULMB2 or DyAc+ or Blur Busters Approved)

(1) Strobe Pulse Width of 1ms or Less
(2) Frame rate = refresh rate = strobe rate
(3) No visible strobe crosstalk caused by slow LCD GtG;
(4) Eye tracking on moving objects

Check out my piece on ULMB2 for more explanation about this.

In other words, you will have fewer stroboscopics (www.blurbusters.com/stroboscopics) with a true 1000fps 1000Hz display, than you would with ULMB2. So you're only getting the 1000Hz "look" with strobing if you meet the (1)(2)(3)(4) criteria.

LightBoost 10% (1ms pulses) had the "1000Hz look" way back in year 2013, see the very old Blur Busters LightBoost 10% vs 50% vs 100% article.

NVIDIA only chose to market it as a 1000Hz-look technology, simply because 1000Hz is finally desirable rather than laughingstock.

Esports players in 2013 were way skeptical of high-Hz displays. Today it's much easier to explain the benefits of a 1000Hz display to a modern esports player (and other scientifically-skilled people), than it was in year 2013.
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!

daviddave1
Posts: 381
Joined: 04 Aug 2017, 17:43

Re: OLED BFI Options?

Post by daviddave1 » 30 May 2023, 03:47

Ogmilkman wrote:
30 May 2023, 01:42
bumbeen wrote:
29 May 2023, 21:25
Ogmilkman wrote:
29 May 2023, 20:05
I assume these OLEDs will only be able to reach 240Hz with BFI and a 720Hz OLED (360Hz with BFI) is necessary to achieve the same motion clarity as a PG27AQN currently.
while maintaining brightness they'd have to chop the refresh rate. But if we're fine with a 50 nit output they could do BFI at full refresh. I think it's a bit disingenuous of nvidia to say "THE MOTION CLARITY OF 1400 FPS!!" It doesn't have that. Not even close. Even in their own video where they show the 480hz vs the 120hz+BFI you can see the background is incredibly choppy by comparison. Just because one element of the screen doesn't have persistence motion blur does not mean the motion clarity overall is equivalent to 4x the number of frames. If that were the case then everyone would be saying the pg27aqn sucked solely because it lacked BFI and the XG27AQM(R) would be the far superior display. We would all still be using CRTs if it were true, 60hz CRTs do not have the motion clarity of a 1000fps display

edit: adding on to this. An OLED at 360hz with BFI will have the motion clarity of the PG27AQN at 360HZ with ULMB2, probably slightly better. The backlight pulse basically replaces the response time of the pixels. So there's slightly more pixel response latency with ULMB2 enabled. The backlight waits until the pixels at the bottom of the display have changed before flashing. So you're having to wait for that to happen before you can see it. On an OLED this will happen much faster since the pixel response time is quicker. It would simply wait until it got all the pixel information for the frame, then flash it and you'll see it, while the LCD is still waiting on the pixels to change.
To answer my own question:

It seems Nvidia’s claim of 1000Hz+ motion clarity is referring to the MPRT in ULMB 2. With the proposed default factor of 4, a 360Hz monitor will have the same MPRT as a 1440Hz monitor without strobing.

OLEDs are self emissive and BFI support above half the maximum refresh rate would require a subrefresh rolling scan. I feel it’s unlikely the 480Hz OLED display next year will have the backplane to support this subrefresh so it may be a while until available OLED displays are able to match this MPRT (at the very least two years assuming next year we will not have the proper backplane).

This is unfortunate for me as I wanted a do it all monitor (1440p OLED 360Hz+ with great motion clarity rivalling the XL2566K) that I could buy and use until the hardware and display came out for the move to the same specs in 4K.
Ty for the info. May I ask what u gonna do now? I assume u have the XL2566K. U going to upgrade to a ULMB 2 1440P IPS monitor soon?
I have the XL2566K I dont really feel the need to sell it and buy a ULMB2 monitor. Optimumtech says in the ULMB2 video he did not really noticed a difference between the OLED 240 HZ and the AQN. so he keeps maining the OLED.
The only monitor that I see worth upgrading now for sure is a 360hz OLED panel with ULMB2 from what u just wrote I understand chances are slim we gonna see that monitor anytime soon.
Hope to hear from u.
Last edited by daviddave1 on 30 May 2023, 05:49, edited 2 times in total.
| Now: ASUS PG248QP 540Hz. | Past : VG259QM with the Qisda panel/PG27AQN/XL2566K

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

Re: OLED BFI Options?

Post by Chief Blur Buster » 30 May 2023, 03:49

GammaLyrae wrote:
29 May 2023, 21:09
I'd expect that it'd be down to user preferences / tolerances. In the short term, OLED is going to offer a higher MPRT (the LG C1 is around 4~ms with BFI on at 120hz, a feat I expect to be repeated by "Dough"'s upcoming 240hz monitor sized OLED), but each frame has zero crosstalk and zero overshoot. It's very clean.
Yep.

While ULMB2 is a significant improvement in strobe quality, brute framerate based motion blur reduction is fantastic. I love my Corsair Xeneon Flex on my desk. Generally, to my eyes, 240fps 240Hz unstrobed (4ms persistence) looks clearer than old LightBoost 100% (2.4ms persistence).

Persistence is frametime on sample and hold, but OLED does it so shockingly beautifully.

There's a biasing factor with OLED-based brute framerate based strobeless motion blur reduction, where you can tolerate roughly 2x the MPRT number as being clear, because of the other mechanics (e.g. fewer stroboscopic artifacts, due to lack of strobing). 240Hz OLED has less blurring than a 240-390Hz LCD with strobing=OFF and framerate=Hz, assuming you can keep framerates peaking at max Hz to minimize persistence.

In my experience, there's a very rough/approximate biasing factor of roughly 2x, when it comes to strobeless persistence, versus strobed persistence. Because of the other benefits.

In other words, ~500fps=500Hz framerate=Hz unstrobed (2ms MPRT) can be better than 1ms strobes (1ms MPRT). So you'd only need roughly a ~500Hz OLED to look preferable to 1/1000sec strobes, and you'd only need a ~720Hz OLED to look preferable to 1/1440sec strobes.

There are a combination of factors due to this;
- Phantom array effect
- Persistence motion blur
- Minor remnant crosstalk / GtG blurs
- Lack of brightness loss or color degradation with brute framerate-based blur reduction

And display artifacts are different on strobed/strobeless technologies for:
- Stationary eye, stationary image
- Moving eye, stationary image
- Stationary eye, moving image
- Moving eye, moving image

So sometimes person A miss something a person B sees because they weren't using the display (with their eyes, with the content) in the same way. And the strobed vs strobeless advantages/disadvantages diverges differently for each of the above lineitem.

So, in a combined way, it is tolerable to have a slightly higher MPRT with strobless than with strobed. Thus, 240fps 240Hz OLED looks preferable to many eyes than 1/480sec strobed LCD.

(More info at www.blurbusters.com/area51 and www.blurbusters.com/stroboscopics and www.blurbusters.com/1000hz-journey if you want to read up more about it all)

Anyway, frame generation based motion blur reduction will become more popular -- as a strobless blur reduction tech. That's why I turn on DLSS at performance settings when lag isn't important. DLSS is a fantastic ULMB substitute for solo gaming on OLED. 3x framerate increase = 1/3rd motion blur on OLEDs. The benefits are much more linearly visible on OLEDs (following Blur Busters Law more perfectly), since OLEDs don't have as visible GtG pixel response bottlenecks as LCD do.

In the past, DLSS only did 1.5x-2x framerate increases, minor blur difference. But DLSS 3.0 does up to 4x framerate increases, and OLEDs scale in blur virtually linearly 4x (1/4th blur at 4x framerate) in a darn nearly perfect match to Blur Buster Law (without the LCD GtG bottleneck). Consequently, DLSS+OLED is a match made in heaven for strobeless blur reduction purists who don't mind lag. Obviously, no good for esports, 2ms strobe lag vs tens of milliseconds of interpolation lag. Ouch.

But that's not a dead end, fortunately (for future tech).

Now next generation: lagless frame generation. Yes, it's possible. It's already being done in VR (aka "reprojection"), and already written at Frame Rate Amplification Technologies on Blur Busters. LTT also covers it:

phpBB [video]


Now, reprojection can be done with massive ratios (10:1 frame generation) for 90% motion blur reduction (1/10th original display motion blur), without using strobing. Aka, converting 100fps UE5 -> 1000fps UE5.

Good reprojection (ASW 2.0) has fewer artifacts than other persistence-improving techniques such as strobing or detail-level reduction (to increase frame rates).

Pure strobing and pure frames have their place. Especially for retro games. However, the fake-frame dissmiss-artists forgets to acknowledge the above pragmaticness for modern 3D workflows. Video compression (Netflix, YouTube, whatever) is often 23 fake frames per second and 1 real frame per second, being that said video compression uses interpolation/prediction systems (e.g. the I-Frame, B-Frame and P-Frame system of many MPEG1/2/3/4 codecs, use interpolation mathematics).

It's not black box like laggy TV interpolation, because the video compressor had access to ground truth (original frames) and the game good reprojector logic has access to ground truth (inputread/geometry/zbuffer/etc). So, the fakeness of the frame, is pretty much synthetic/academic sematics, considering that all GPU frames were fake (e.g. artisting a photorealistic UE5 frame using triangles (aka polygons) and bitmaps (that we call textures). Semantically, that's still a synthetic frame. Reprojection is simply a different workflow of doing something with fewer artifacts than doing it directly via GPU triangles and textures -- we have to lower detail level to make a frame even more fake, to get higher frame rates -- so we can make frames less fake by using reprojection! It can go artifactless like Netflix fake frames. And if it can be lagless for esports, then why not? See where I am getting at? Wink, wink. We blur-busting enthusiasts all want 1000fps UE5.2 or UE6 quality graphics eventually.

Eventually I want to see lagless frame generation technologies become more popular, such as reprojection. Unfortunately since reprojection requires inputread/geometry/zbuffer data, it has to be baked into the engine for best results and to successfully rewind frametime lag to inputread time lag. (It's possible to turn 100fps 10ms-rendertime-lag to 1000fps 1ms-lag since reprojection can morph past frames retroactively to sync up to current gametime / inputread time).

So by 2030s, lagless reprojection tech will replace strobing in PC gaming hopefully. Ergonomic flickerless/strobless/lagless reprojection is a better motion blur reduction technology than strobing is. Strobing is PWM, and some people get eyestrain. Strobing is useful for current contemporary frame rates.

I love strobing, but strobing is just a bandaid for our inability to do analog framerates (simulating framerateless real life), at least until we brute-force it via 1000fps 1000Hz. The 2030s future of reprojection and true 1000fps+ 1000Hz+ OLEDs/MicroLEDs beckons.

Now, if you want good strobing, then definitely ULMB2 is very high on the list, as is Blur Busters Approved. That being said, refresh rate headroom tricks can make 100Hz Blur Busters Approved clearer-motion than 360Hz ULMB2, via the large-VT + Strobe Utility tweak at www.blurbusters.com/xg2431 ... However, ULMB2 is now one of the best max-Hz strobe backlight technologies. It does a fantastic job utilizing a large number of LCD overdrive tricks to reduce strobe crosstalk, something extremely difficult to do at max Hz (360Hz).
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!

bumbeen
Posts: 84
Joined: 25 Apr 2023, 14:35

Re: OLED BFI Options?

Post by bumbeen » 30 May 2023, 06:18

Chief Blur Buster wrote:
30 May 2023, 03:49
Now, reprojection can be done with massive ratios (10:1 frame generation) for 90% motion blur reduction (1/10th original display motion blur), without using strobing.
wow I had not seen that video, and was very confused until linus showed the bit of the frame floating in black space and then it instantly made sense to me. That is absolutely genius, and it seems like it would even work for multiplayer games like csgo and valorant, although at a higher minimum FPS threshhold that at least matches the tick rate of the game. So 128fps to 1000fps would be quite easy to do here

have gone back to gsync for valorant over ULMB2 on my pg27aqn. The strobing does look absolutely amazing with vsync ON, but I prefer the lower input latency of gsync since the variability of the frame times seems to preclude perfectly timing the game with vblanks, I can get it close with stutters, or I can deal with frame buffers stacking up, not both sadly. So gsync it remains for the time being!

edit, nope no gsync. I will simply use vsync. it just looks too perfect with strobing. It's literally crt quality at 360fps. I guess I will just deal with the extra 4-7ms of added latency. MAybe that is better than gsync with FPS bobbing up and down constantly anyway

Ogmilkman
Posts: 20
Joined: 26 Sep 2020, 08:19

Re: OLED BFI Options?

Post by Ogmilkman » 30 May 2023, 12:31

daviddave1 wrote:
30 May 2023, 03:47
Ty for the info. May I ask what u gonna do now? I assume u have the XL2566K. U going to upgrade to a ULMB 2 1440P IPS monitor soon?
I have the XL2566K I dont really feel the need to sell it and buy a ULMB2 monitor. Optimumtech says in the ULMB2 video he did not really noticed a difference between the OLED 240 HZ and the AQN. so he keeps maining the OLED.
The only monitor that I see worth upgrading now for sure is a 360hz OLED panel with ULMB2 from what u just wrote I understand chances are slim we gonna see that monitor anytime soon.
Hope to hear from u.
I actually have a VG259QM the difference in clarity with the XL2566K wasn't enough for me to reconcile the terrible vertical viewing angles and I returned the latter. I would consider the PG27AQN but can't justify the price point. The motion clarity + higher resolution undeniably creates a clearer picture but the benefits to me are marginal to the price. At 240Hz+ I don't believe any monitor would be holding me back in terms of tactical FPS gaming. At this point better motion clarity is just a benefit I would like to have for my own quality of life, which is why I'm okay with waiting for an endgame OLED monitor.

Personally, I feel a BFI solution via rolling scan will be available sooner than a 720Hz OLED but either way I think I'll be waiting until one of the two are available (unless the PG27AQN can be found on a substantial sale or a cheaper alternative appears).

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

Re: OLED BFI Options?

Post by Chief Blur Buster » 30 May 2023, 17:54

bumbeen wrote:
30 May 2023, 06:18
wow I had not seen that video, and was very confused until linus showed the bit of the frame floating in black space and then it instantly made sense to me. That is absolutely genius, and it seems like it would even work for multiplayer games like csgo and valorant, although at a higher minimum FPS threshhold that at least matches the tick rate of the game. So 128fps to 1000fps would be quite easy to do here
Even that amazing desktop reprojection video is actually low-quality Wright Brothers stuff. Oculus (I still prefer that name over "Meta") Rift VR does a better job of reprojection with ASW 2.0.

Good reprojection (ASW 2.0) has fewer artifacts than other persistence-improving techniques such as strobing or detail-level reduction (to increase frame rates).

So for those people coming and not realizing we can reduce display motion blur gigantically in a strobeless way (no flicker like CRT / plasma / phosphor / BFI / strobe). Ergonomic flickerfree method of eliminating display motion blur.

Good, properly done reprojection is not frames that are more fake

Pure strobing and pure frames have their place. Especially for retro games. However, the fake-frame dissmiss-artists forgets to acknowledge the above pragmaticness for modern 3D workflows.

Video compression (Netflix, YouTube, whatever) is often 23 fake frames per second and 1 real frame per second, being that said video compression uses interpolation/prediction systems (e.g. the I-Frame, B-Frame and P-Frame system of many MPEG1/2/3/4 codecs, use interpolation mathematics).

It's not black box like laggy TV interpolation, because the video compressor had access to ground truth (original frames) and the game good reprojector logic has access to ground truth (inputread/geometry/zbuffer/etc).

So, the fakeness of the frame, is pretty much synthetic/academic sematics, considering that all GPU frames can meet the Merriam-Webster dictionary definition of "fake" in some aspects (e.g. artisting a photorealistic UE5 frame using triangles (aka polygons) and bitmaps (that we call textures).

Semantically, rendering a GPU frame is still a synthetic frame.

Reprojection is now potentially a "less fake" way of doing ultra high frame rate, because of what I said above.

Reprojection is simply a different workflow of doing something with fewer artifacts than doing it directly via GPU triangles and textures -- we have to lower detail level to make a frame even more fake, to get higher frame rates -- so we can make frames less fake by using reprojection! It can go artifactless like Netflix fake frames. And if it can be lagless for esports, then why not?

Eliminating reprojection artifacts

The artifacts of reprojection can be diminished further by using a base framerate above flicker fusion threshold (aka 100fps starting frame rate), to eliminate even more reprojection-artifacts.

Sample and hold stutter and blur are exactly the same thing. Low-frequency sample and hold shakes/vibrates like a music string -- And high-frequency sample and hold blurs like a fast music string -- www.testufo.com/eyetracking#speed=-1 (look at bottom of UFO as it changes speed over the course of 30 seconds). By pushing ALL frequencies above flicker fusion threshold (starting frame rate, and all jitter frequencies, etc) well up there, reprojection becomes much more artifactless.

Reprojecting 100fps->1000fps is MUCH fewer artifacts than reprojecting 45fps->90fps.

And you have no double-image effect like Oculus, because it's sample-and-hold (no double-impusling)

Sometimes blur is good. Right tech for right job.

Sometimes we love motion blur. Hollywood Filmmaker Mode. Motion blur has the right artistic tool. But it's bad when trying to simulate a Star Trek Holodeck: More blur than real life. So, there you go. Right Tool For Right Job.

We all love artsy looks like Minecraft or retro games, but we also enjoy real-life-style games (e.g. photorealistic shooters), and that's where displays are throwing bottlenecks at us, or artifacts of attempting higher frame rates at low blur (strobe flicker or low-detail ugliness), which reprojection does fix.

See where I am getting at? Wink, wink.

Large-ratio reprojection is a massive motion blur reduction technology

With large-ratios, reprojection can eliminate as much motion blur as strobing can!

The biggest technological catch is that refresh rates and final frame rates need to be stupendously high.
- 500fps for 2ms MPRT strobelessly
- 1000fps for 1ms MPRT strobelessly
- 2000fps for 0.5ms MPRT stroblessly

This is assuming GtG is practically zeroed out, GtG darn nearly 0ms. OLED is able to do that. With 240Hz OLEDs and 500Hz+ LCDs, and refresh rates still increasing (and 1000Hz tests already done in the lab), we're already very close to 1000Hz displays becoming a reality before the end of this decade.

Now that brings the GPU problem. And reprojection solves the GPU problem.

We blur-busting enthusiasts all want 1000fps UE5.2 or UE6 quality graphics eventually.

Reprojection will successfully get us there, and possibly even on some of today's GPUs (e.g. RTX 4090 can do 1000fps reprojection today; confirmed via tests). 75% of an RTX 4090 can be directed to render original UE-quality frames, and 25% of an RTX 4090 can be directed to convert 100fps to 1000fps in a nigh-lagless manner. NVIDIA theoretically could call this DLSS 4, but it also has to be baked into the game engine (due to need for access to inputread/geometry/zbuffer ground truth, for accurate artifact-free reprojection).

Reprojection can rewind local input lag too! Turning 100fps 10ms rendertime lag, to 1000fps inputreadtime lag. This is done by making sure reprojection has access to current ground truth (inputreads) and past ground truth (fully rendered frames with zbuffer and geometry). The current ground truth (mouselook/strafes/etc) morphs the most recent historical frame to current freshest 6dof inputreads.

With a two simultaneous-workflow GPU pipeline (the original-frames pipeline, and reprojection-pipeline), you can convert a varying-framerate game into a permanent constant framerate=Hz experience that is synchronized to freshest 6dof inputreads (mouselook/strafes/etc).

So a fluctuating 80-120fps gets converted to a perfect 500fps@500Hz via a variable-framerate-compatible reprojection engine that has permanent access to continual 1000Hz 6dof data (even between original-render frames). You get the lagless 500fps VSYNC ON look, from an original fluctuating-framerate stream.

The most important two gods of good blur-reducing stutter-reprojecting reprojection
(A) game timestamp is god (microsecond monotonic counters)
(B) inputread timestamp is god (calculating and executing local player 6dof position to sub-millisecond relative accuracy)

As long as (A) and (B) is met, it successfully becomes a DLSS+ULMB+GSYNC sequel.

This is "DLSS+ULMB+GSYNC+FlickerFree+EyeCare" sequel rolled into one.

- DLSS because reprojection is a frame generation tech (even if lagless)
- ULMB because reprojection also reduces motion blur
- GSYNC because reprojection also can de-stutter a variable input framerate to a constant output framerate
- FlickerFree EyeCare because this is flickerless, strobeless, BFI-less, PWM-less.

So it is simultaneously DLSS 4, ULMB 3, and GSYNC 3 rolled into one!

The reprojection processor on the GPU would run at a higher GPU priority than the original frames pipeline, to ensure the framerate=Hz lagless reprojection experience.

Enemy movements may still stutter, but remember the stutter-to-blur continuum -- 100-200fps is just slightly extra persistence blur, not stutter. So enemy movements might look slightly blurrier than the rest of the screen. Reprojection fixes the most important motion blur (global motion blur of panning / scrolling / turning / strafing), which is why reprojection is the prioritized motion blur reduction technology.

So, reprojection (on ultra-high-Hz nigh-instant-GtG displays) will easily make BFI obsolete for some people, since strobless (Flickerfree! Ergonomic! PWM-free!) motion blur reduction is the Blur Busters Holy Grail of sorts for many of us.

We love strobing, but it's still a (good) bandaid for the fact that computer frame rates is finite/synthetic version of real life, and real life is framerateless.

Since large-ratio reprojection technology is a fantastic motion blur reduction technology;

Next Steps for desktop reprojection

Now, the next steps for reprojection I think should be:

NVIDIA and Epic should collaborate to add lagless frame generation to DLSS 4/5 or ULMB3 (via reprojection) to a future Unreal Engine 6.

(or Intel+XeSS, or AMD+FSR, if they have the skill to beat the fabulous NVIDIA gorilla to this punch)

I even told this suggestion NVIDIA directly (to Seth Schnieder during past emails / past meeting), that they need to consider this. I could tell they knew what I was talking about -- there was no confusion. I have no doubt the engineers at NVIDIA are smart. Let's hope.

There is enough computing power in an RTX 4090 to do 1000fps reprojection, and NVIDIA could use the extra sales that reprojection can cause. Even the budget gamer who can only afford a $400 GPU is more likely to buy an RTX 4060 if it can do DLSS4 reprojection to turn 20fps games to 60fps (with some artifacts), while the enthusiasts like me will use an RTX 4090 to turn 100fps RTX ON to 240fps-540fps.

If it's even more stomachable to the beancounters at GPU manufacturers, DLSS4 with reprojection could debut during the 5000-series instead. Okay. I can certainly understand the business case (I'm a business too), even if the users will scream holy hell about not getting easy reprojection on older GPUs, like that famous LTT reprojection video is extolling about.

Heck, I can do Oculus reprojection on an old GTX 1080 GPU, converting Oculus Rift 45fps to 90fps. But, I would still implore NVIDIA to introduce DLSS4 (at least forward reprojection) retroactively to RTX 4000 series.

I feel that NVIDIA does not need to fear sales and revenue reductions -- reprojection will help increase GPU sales by giving more 60-120fps experiences to more mainstream people (including casual gamers), while giving the high end gamers the 500+fps strobeless blur reduction experience.

It can be basic for now, e.g. slightly laggy reprojection that doesn't have rendertime-latency rewind capability at first. The complete ULMB+GSYNC+DLSS combined sequel that is lagless-ready for esports, can come later (e.g. DLSS5 or DLSS6). But the starting pistol of reprojection needs to begin soon -- like 2023. Definitely 2024 at the latest for one of the flagship game engines.

There is little demand now only because nobody knows this as well as Blur Busters does.

Remember our LightBoost HOWTO? Remember our G-SYNC 101? We helped the sales years before many marketing people at NVIDIA realized this. This is BlurBusters' "Big Hit #3" -- we are even about to crowdfund a third party reprojection processor for a desktop game engine. But I'd rather see NVIDIA beat me to the punch. Will be glad to see NVIDIA lift all refresh rate boats, win win even if my project becomes a throwaway project. So, NVIDIA, you heard me. I consider this potential Hit #3 of the mid-2020s after LightBoost HOWTO and G-SYNC 101.

Why am I loud? Big boom of OLED

I've seen the Wright Brothers version of this in action. It's that impressive.

I am actually, in real life, witnessing with my eyes, the tantalizing simultaneous ULMB+GSYNC+DLSS+EyeCare+FlickerFree rolled into one, when framerate=Hz reprojection is executed on a high-Hz OLED.

So, NVIDIA needs to address the boom of OLED, increase GPU sales to OLED purchasers. With gamers kind of frustrated at NVIDIA, this technology would be a major appeasement. Framerate progress should no longer be held back -- RTX ON pathtracing and triple-digit frame rates is currently mostly mutually exclusive -- but it does not need to be.
Today, 240fps reprojection for pathtraced content on today's 4000 series!

Tomorrow, 1000fps ULMB3 replacement.

This is the motion blur reduction and stutter-killing technology of the 2030s.

Strobing will remain important for a long time (especially today's LCD, plus also on retro content that is ruined by frame generation). That being said, framerate=Hz on OLED means that 4ms MPRT on OLED looks better than 2ms MPRT strobed, since the extra biasing factor (The simultaneous "ULMB+GSYNC+DLSS+FlickerFree+EyeCare" effect) puts a gigantic elephant-weight on the motion quality of strobeless motion blur reduction.

The Business Case For Reprojection on OLED

Why is reprojection vastly more effective on OLED?
Why is it potentially very profitable for GPU vendors?
  • OLEDs have massive upgradefeel appearance with reprojection algorithms
    120-vs-240fps is more visible on OLED than LCD, more mainstream people will notice.
    ---
  • OLEDs on desktop usually dont have BFI
    Reprojection is a successful alternative
    ---
  • OLEDs stutters are more visible at 60fps, requiring 90fps+ to look good
    The ultra fast GtG of OLED makes 60fps more stuttery on OLED than LCD. Reprojection fixes this by going to higher framerate. Some people get motionsick OLED stutters, or from from stutters of turning RTX ON, so having RTX ON at framerate=Hz, in a lagless way via rendertime-rewind-to-inputtime, I believe, visually creates a lagless motion nirvana that leads to more sales all across the board.
    ---
  • OLEDs does not have much brightness headroom
    Reprojection does not degrade brightness
    ---
  • OLEDs has fantastically fast pixel response and looks great at framerate=Hz
    240Hz OLED has clearer motion than most unstrobed 240Hz-500Hz LCD
And hey, eventually, even VR games will be able to get all that RTX ON goodness, with no framerate compromise! With the new reprojection algorithms that convert a varying framerate input, to a constant framerate=Hz output. So these innovations can be ported back to VR markets too, lifting all Hz boats, and expanding beyond the sales-limiting esports silo, too.

Note: This forum post is about to be converted to a Blur Busters article shortly; I often use my forums as a drafting table for future articles. More than half of www.blurbusters.com/area51 are rewritten/polished/curated versions of my past forum posts here. I intend to fire off a race between all the GPU vendors, even though I am rooting for NVIDIA's brilliant engineers to do it first, helped by their brilliant robot arm pursuit camera
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