NVIDIA G-Sync Pulsar monitor - Asus ROG Strix XG27AQNGV

[edgintheledge]

Lag testing 60hz ULMB pt 3

Went ahead and tested top vs bottom screen lag in different fixed refresh rate modes and wouldn't you know it... 60hz gives me the exact same latency measured at the very top and very bottom. This doesn't happen in higher hz ulmb modes.

I turned vsync on just to make sure there wasn't any possibility of something funky happening with the tearline skewing any runs, that's why the numbers are so high (though that should be irrelevant over 1000+ measurements averaged together). See previous post for more context.

60hz ulmb

• Run 1 - Top 81.41ms, Bottom 81.18ms : No significant difference (~1000 measurements taken)
• Run 2 - Top 80.26ms, Bottom 80.94ms : No significant difference (~1700 measurements taken)
• Run 3 - Top 77.80ms, Bottom 77.76ms : No significant difference (Only ~300 measurements)

120hz ulmb

• Run1 - Top 59.72ms, Bottom 65.38ms : 5.66ms difference (~1200 measurements)
• Run 2 - Top 59.65ms, Bottom 65.46ms : 5.81ms difference (~1100 measurements)

I genuinely can't imagine any way this monitor could possibly be rolling the backlight at 60hz and still have

• Identical lag at the top and bottom of the screen
• No skew when flicking my eyes really fast
• No skew when tracking fast horizontal motion (reappears when strobing is off or ulmb enabled for 120+hz)
• Even repeating pattern of 3f off, 1f on when filmed as close to 240 as my phone would muster (239.97fps) (video here)

I feel fairly confident in saying that the 60hz mode does not roll and the global strobe behavior is why it is so much sharper than the other modes that are rolling. I think this also lends credibility to Discorz saying "The poor diffusion of edge-lights raises the minimum MPRT floor". I suppose the last way to confirm would be a 1000+fps camera, but I don't have one of those unfortunately.

[brownvim]

Lag testing 60hz ULMB pt 3
60hz ulmb

• Run 1 - Top 81.41ms, Bottom 81.18ms : No significant difference (~1000 measurements taken)
• Run 2 - Top 80.26ms, Bottom 80.94ms : No significant difference (~1700 measurements taken)
• Run 3 - Top 77.80ms, Bottom 77.76ms : No significant difference (Only ~300 measurements)

Yeah, that's really strong evidence it's using global strobing. Is it just me, or does it look like it flickers a lot less than OLED BFI?

[brownvim]

What's the recommendation now for anyone using these Pulsar monitors on the latest firmware?

With the latest firmware, this is what I’ve settled on:

180–360 fps range: Regular G-Sync Pulsar is the best. Try Dynamic frame gen x6 if you can, keeps steady high frames.
90-180 fps: Even small frame drops (just 1–2 fps) introduce noticeable ghosting/double imaging in regular Pulsar. So in this range it’s better to lock the frame rate or roll back firmware to keep Pulsar ghosting free.
For lower frame rates / 60 fps content: Switch to 60 Hz ULMB2 and lock the game to exactly 60 fps using Fast Sync + Special K Latent Sync.
This is currently the clearest mode on the monitor — noticeably sharper than regular Pulsar.
Trade-offs are: brightness at lower pulse widths, you get perceivable flicker, and you add roughly 10ms of input lag.

So in short:

High fps games → stay in Pulsar
Lower fps / console-style games → 60 Hz ULMB2 + locked 60 fps is often the best looking option.

[RealNC]

I suppose the last way to confirm would be a 1000+fps camera, but I don't have one of those unfortunately.

Or a photo with a very fast shutter speed so see if only a part of the screen is lit.

[Discorz]

I genuinely can't imagine any way this monitor could possibly be rolling the backlight at 60hz and still have

• Identical lag at the top and bottom of the screen
• No skew when flicking my eyes really fast
• No skew when tracking fast horizontal motion (reappears when strobing is off or ulmb enabled for 120+hz)
• Even repeating pattern of 3f off, 1f on when filmed as close to 240 as my phone would muster (239.97fps) (video here)

I feel fairly confident in saying that the 60hz mode does not roll and the global strobe behavior is why it is so much sharper than the other modes that are rolling. I think this also lends credibility to Discorz saying "The poor diffusion of edge-lights raises the minimum MPRT floor". I suppose the last way to confirm would be a 1000+fps camera, but I don't have one of those unfortunately.

Good point! Basically slower scan = larger skew, faster scan = smaller skew, ultra fast/no scan = no skew. If skew is truly non-existent, it's probably a global flash. If 100 has more skew compared to 10, than it's a form of QFT. https://testufo.com/scanskew#speed=1.5

I guess there is a possibility for global flash. Maybe they didn't like how rolling looked at 60 so they switched back to global flash. Dispersion limitations without QFT would've been too much at such low frequencies. It's already becoming apparent at 120 Hz.

Those oscillographs don't really tell if something is moving or not. If that's the case than the pulse fading in/out comes from the backlight (all zones) intensity gradually rising-holding-falling, and squashing comes from holding the intensities for shorter periods. I guess the fading is there to soften the impact on the eyes a bit. QFT at 60 Hz can mean huge GtG headroom after all.

Interestingly, what are the chances that fading of the global 60Hz-PW100 pulse resembles a rolling 120Hz-PW100 pulse so closely? Rising and falling is pretty much identical on those two graphs.

High speed footage would still be of great help. What PW did u record in your slowmo video? Maybe try PW 100 if not.

[edgintheledge]

Or a photo with a very fast shutter speed so see if only a part of the screen is lit.

Correct me if I’m wrong as I’m not the most knowledgeable on cameras, but can’t the lack of global shutter combined with lower frame rate cameras cause partial frames to show even if you use faster shutter?

This site has a test that claims my phone has ~5ms of rolling shutter which could cause partial frames. I have definitely observed this effect with strobe lights (the old xenon tube style, not the led type) splitting up between two different frames as a result even though a strobe light is physically incapable of rolling.

Good point! Basically slower scan = larger skew, faster scan = smaller skew, ultra fast/no scan = no skew. If skew is truly non-existent, it's probably a global flash. If 100 has more skew compared to 10, than it's a form of QFT. https://testufo.com/scanskew#speed=1.5

I guess there is a possibility for global flash. Maybe they didn't like how rolling looked at 60 so they switched back to global flash. Dispersion limitations without QFT would've been too much at such low frequencies. It's already becoming apparent at 120 Hz.

Those oscillographs don't really tell if something is moving or not. If that's the case than the pulse fading in/out comes from the backlight (all zones) intensity gradually rising-holding-falling, and squashing comes from holding the intensities for shorter periods. I guess the fading is there to soften the impact on the eyes a bit. QFT at 60 Hz can mean huge GtG headroom after all.

Interestingly, what are the chances that fading of the global 60Hz-PW100 pulse resembles a rolling 120Hz-PW100 pulse so closely? Rising and falling is pretty much identical on those two graphs.

High speed footage would still be of great help. What PW did u record in your slowmo video? Maybe try PW 100 if not.

This makes me realize… I have completely neglected testing any of this stuff at PW100 The top vs bottom lag testing was done at PW14 and the high speed was at PW10. I’ve always been keeping the monitor at very low pw. When I get a chance later, I’ll check the skew at PW100 and if it does skew more, that kinda completely changes my understanding of this mode’s behavior (and possibly means more GtG headroom on the table?)

[RealNC]

Or a photo with a very fast shutter speed so see if only a part of the screen is lit.

Correct me if I’m wrong as I’m not the most knowledgeable on cameras, but can’t the lack of global shutter combined with lower frame rate cameras cause partial frames to show even if you use faster shutter?

Lots of shots needed to verify. In ULMB 60Hz, ULMB 120Hz and Pulsar VRR 90FPS@120Hz and fixed refresh rate no ULMB and no Pulsar at 120Hz. Then compare them against each other. Too much effort though

[edgintheledge]

If skew is truly non-existent, it's probably a global flash. If 100 has more skew compared to 10, than it's a form of QFT. https://testufo.com/scanskew#speed=1.5

You're never gonna guess what just happened when I looked at this at PW100...
It skews LOL. I completely neglected to check any of this at max pw but hey, hindsight is 20/20 right?

I would say the amount of skew 60hz PW100 has looks more like unstrobed 120hz skew, not as diagonal as unstrobed 60hz. As you go farther down in pw the skew progressively becomes less and less until I can't notice it anymore/can't measure it through lag testing.

To my eye PW100 does actually wobble a little like a CRT when I flick my eyes, but you REALLY have to try and whip your eyes as fast as you can or turn your head left to right really fast to even notice. It's not nearly as noticeable and wobbly as 60hz/16ms top to bottom difference on a CRT.

Lag testing 60hz ULMB pt 4 (hopefully the final one... for now)

same deal as the previous. Don't worry about the high ms numbers we only care about the difference at the top and bottom.

60hz mode PW100

• Run 1 - Top 75.99ms, Bottom 80.33ms : 4.34ms difference (~1100 measurements)
• Run 2 - Top 75.05, Bottom 79.21 : 4.16ms difference (~350 measurements)

This difference would suggest that the QFT is scanning at 240hz speed since the top to bottom difference is almost exactly the 4.16ms frametime of 240hz, but this is not corroborated by 240fps video. I can see a clear and repeating rolling pattern. A 240hz speed roll wouldn't show up this way on 240fps video. Likely what is happening here is the backlight diffusion is triggering the lag tester to register the bottom measurement as "on" earlier than the backlight actually gets there.

Here is a video of PW10 and 100 side by side. My phone spit out 239.97 and 239.98fps presumably because of something happening with variable frame rate video, but the clip is short enough that there isn't any desync issues.

phpBB [video]

It clearly takes 4f to repeat the cycle on video as expected, but the bottom is basically lit up (due to backlight diffusion) by frame 2 of each cycle, and the backlight appears to get to the bottom by the 3rd frame. This would mean the backlight is taking somewhere between 8.33-12.5ms making the scanout speed/QFT "ratio" roughly somewhere between 80-120hz. This agrees with what my eyes see.

Seems like all roads actually point toward the 60hz mode using a sort of QFT that dynamically adjusts to a higher ratio/scanout speed as you lower the pw. At pw100 the speed is somewhere between 80-120hz and it gets progressively faster (less skew) until it either becomes a global strobe at the lowest pw settings, or is scanning top to bottom so fast that it is effectively a global strobe under the scrutiny of ldat style lag testing.

[kitabatake_radeka]

Hi all, is there any way to limit the monitor to 180 Hz refresh rate, as I set the FPS cap to 180 anyway, so DSC is completely needlessly used. While preserving Pulsar of course. I did create a custom refresh rate in the Nvidia control panel, QHD with 180 Hz, all other options on auto, it worked but Pulsar was disabled, ULMB2 there was instead. I am on the original firmware.

[brownvim]

Seems like all roads actually point toward the 60hz mode using a sort of QFT that dynamically adjusts to a higher ratio/scanout speed as you lower the pw. At pw100 the speed is somewhere between 80-120hz and it gets progressively faster (less skew) until it either becomes a global strobe at the lowest pw settings, or is scanning top to bottom so fast that it is effectively a global strobe under the scrutiny of ldat style lag testing.

Thanks for the continued testing and the new video, appreciate the effort as it clears up me thinking why it looked so good.

So they’re using a more advanced rolling/QFT-style approach that gives better brightness retention.

NVIDIA should really consider bringing this same improved strobing method to the 120 / 240 / 360 Hz ULMB2 modes — maybe even call the next version ULMB3. It would be a huge upgrade if they could make all the fixed modes perform as well as 60 Hz.

Great work again!