Viewsonic XG2703-GS QFT Success [Quick Frame Transport Lag Reduction]

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
xcasxcursex
Posts: 3
Joined: 17 Jan 2021, 20:58

Viewsonic XG2703-GS QFT Success [Quick Frame Transport Lag Reduction]

Post by xcasxcursex » 18 Jan 2021, 07:42

Hi all,

Just checking in with a success report in case someone else wants to try this.

Edit: Improved version with advice from the Chief in the following post. The timings below work, but are suboptimal.

TL;DR:
>30% faster frame transport, yay.
See attachment for CRU screenshot. In text for future reference/search engine/etc:

Code: Select all

CRU parameters:

2560    1440
8       3
32      10
40      553
(80)    566
(2640)  2006

119.999
(240.719)
(635.50)



Story time:

Why:
Recently developed some eye/CNS problems and persistence is a major problem, so blur has gone from "annoying" to "blinding". My endless search for low latency has shifted focus (heh heh) to low blur. I'm gaming on a Viewsonic XG2703-GS which is a 1440p 165Hz Gsync monitor, and the 7ms response time with overdrive enabled (!) isn't helping.

This led me to try ULMB, again - I'd tried ULMB in the past but it just didn't compare to G sync - it wasn't the tearing, it was the appearance of the (normal, tiny) frametime variations, presenting themselves not unlike microstutter. Throwing S sync in the mix gave it the edge, not only removing the tearing but most importantly creating rock solid frametimes (since it's effectively a limiter synced to the vertical refresh clock which is plenty stable). That's removed the blocks and now I can really appreciate ULMB in all its glory. I have been missing out. HIGHLY recommended combination.

Of course, the massive improvement in picture quality is not free, scanline sync needs a beefy GPU to keep the frametimes consistent, but even still, there is a small input lag increase. While I'm reading up on this, I stumble over Mark discussing QFT. Fast forward a few weeks....


What:
Actually came here for help after I'd given up, tried one last thing before posting (it pays to search before you post even if you already searched!).... and I have a winner! I have 120Hz (including ULMB) working at a 165Hz pixel clock. That's 2006/1525=1.3154=over 30% faster, thanks!! It's entirely possible that even bigger VBI will work with this monitor, but honestly, I'm still slightly in shock at the fct that it's working and I'm not game to touch it. This is good enough for me for now, but if you do better, please share :)


How:
I'd attempted to get it working at 144Hz (and a bunch of other stuff, I'll spare the boring details) and had zero luck, but the thing that worked in the end was:

Enable the overclock mode on the monitor, copy the 165Hz overclock parameters in CRU, paste them into the 120Hz profile, set the refresh rate back to 120Hz, and crank up the back porch until the pixel clock is as close as possible to the clock for 165Hz.

Pretty basic methodology I know, but it works. At least, it worked doing this based on the 165Hz overclock profile....not on ANYTHING else I tried. The overclock mode does not need to be enabled on the monitor for these settings to work. I just used it to find the settings, you can leave it off if you want to.


If you have a G sync monitor I can only hope for your sake that you'll give this a shot, it's magic... And if you have this nice old Viewsonic monitor, now you have settings you can use without experiments :)
Attachments
CRU XG monitor 120Hz QFT.png
CRU XG monitor 120Hz QFT.png (50.16 KiB) Viewed 4884 times
Last edited by xcasxcursex on 18 Jan 2021, 20:36, edited 1 time in total.

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

Re: Viewsonic XG2703-GS QFT Success [Quick Frame Transport Lag Reduction]

Post by Chief Blur Buster » 18 Jan 2021, 14:25

Great information!

Information for new readers:

Quick Frame Transport thread
Various Purposes: Reduce strobe lag, reduce VSYNC ON latency, improve RTSS Scanline Sync, reduce strobe crosstalk (on some brands; ULMB is unaffected but can benefit from reduced strobe lag)
xcasxcursex wrote:
18 Jan 2021, 07:42
Enable the overclock mode on the monitor, copy the 165Hz overclock parameters in CRU, paste them into the 120Hz profile, set the refresh rate back to 120Hz, and crank up the back porch until the pixel clock is as close as possible to the clock for 165Hz.
That's one of the correct techniques!

Another technique is to lock the radio button on Pixel Clock (keep 165Hz). When you increase Vertical Porch, the refresh rate goes down. Keep increasing Vertical Total (or a porch value, front porch in theory would be superior to back porch) until refresh rate decreases to your target refresh rate (144Hz or 120Hz). You generally can get faster QFTs for lower refresh rates.

In some cases, a very few monitors can do QFT beyond max Hz. But easiest to overclock as much as you can (166Hz, 170Hz), then derive from that to create your overclocked QFT mode for 120Hz or 144Hz.
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!

disq
Posts: 65
Joined: 10 Oct 2018, 16:05

Re: Viewsonic XG2703-GS QFT Success Re: Viewsonic XG2703-GS QFT Success [Quick Frame Transport Lag Reduction]

Post by disq » 18 Jan 2021, 17:06

So, after doing that, should i use G-Sync or ULMB? *confused*

Edit: OP deleted his comment :|

xcasxcursex
Posts: 3
Joined: 17 Jan 2021, 20:58

Re: Viewsonic XG2703-GS QFT Success Re: Viewsonic XG2703-GS QFT Success [Quick Frame Transport Lag Reduction]

Post by xcasxcursex » 18 Jan 2021, 20:33

TL;DR Improved, all hail the Chief!
CRU screenshot attached.
CRU XG monitor 120Hz QFT v2.png
CRU XG monitor 120Hz QFT v2.png (51.02 KiB) Viewed 4849 times

Code: Select all

CRU parameters:

2560    1440
8       550
32      10
40      6
(80)    (566)
(2640)  (2006)

119.999
(240.719)
(635.50)

Chief Blur Buster wrote:
18 Jan 2021, 14:25
Another technique is to lock the radio button on Pixel Clock (keep 165Hz). When you increase Vertical Porch, the refresh rate goes down.


This was what I tried originally - I figured the dotclock would be the limiting factor, so I used it as the basis for the required VT for the target refresh rate... The result was a refresh rate of 120.001Hz which everything just called 120Hz and worked fine, other than that the monitor wouldn't engage ULMB. It was okay with several rates in the high 119.9** range, but beyond 120, ULMB was greyed out in the monitor menus. This spot was the one closest to the 165Hz pixel clock, within the constraint of being under 120Hz.

Chief Blur Buster wrote:
18 Jan 2021, 14:25
Keep increasing Vertical Total (or a porch value, front porch in theory would be superior to back porch)


I had a "duh" moment here when I realised the 'back porch' is the back porch of the blanking interval, not the back of the image. How I saw that diagram so many times and didn't notice I was naming the porches backwards.... :D
Front porch makes sense now that I don't think it's the back.... Notes for the uninitiated, and perhaps a sanity check:

EDID limits the front porch size to 63. This means we can't just edit the existing resolution. Instead, we need to use a DisplayID extension to add the new timings. Since the timings seem to collide between the extensions, I removed the resolutions from the existing CEA-861 extension, and duplicated them into the DisplayID extension. There is a limit of 5 resolutions per extension, so you will either need to choose 5, or add another extension. (For this monitor and my needs, I just left out the 24Hz res). Now that's done, we can make a large front porch to get the same total. The result is in the attached image, and in text in the TL;DR at the top of this post.

disq wrote:
18 Jan 2021, 17:06
So, after doing that, should i use G-Sync or ULMB? *confused*

ULMB i guess, therefore the need for scanline sync?


That's right :) I actually recommend trying this before scanline sync - changing the monitor timings like this means re-tuning the sync reference in RTSS, so if you get this working first, then enable ULMB, and then do the scanline sync tuning

xcasxcursex
Posts: 3
Joined: 17 Jan 2021, 20:58

Re: Viewsonic XG2703-GS QFT Success [Quick Frame Transport Lag Reduction]

Post by xcasxcursex » 20 Jan 2021, 01:44

disq wrote:
18 Jan 2021, 17:06
Edit: OP deleted his comment :|
I edited it, but it's awaiting moderation. Anyway if I knew then what I knew now, I might have deleted it :(

The good news is that I was eventually able to get the QFT working, including with the large front porch, so that's nice. The very bad news is that when the system is running at 120Hz over a 165hz pixel clock, the monitor thinks it's at 165Hz, resulting in it disabling ULMB. Darn.

The reason I had not noticed this previously is that somehow (???) the system is magically ignoring the custom resolution as well as any specified in the monitor EDID, and using the default settings for that refresh rate. This results in a system where bad resolutions cause a black screen, but good resolutions while they *appear* to be working, are actually using a nonexistent (in EDID or overrides) timing. I only noticed it after everything seemed to be working and I tried to tune my scanline to the new blanking interval... only to notice that it was showing 1525 lines - the default 'native' scan rows count, when it should have been showing 2006. I fiddled with it until I was able to get the 2006 line 120Hz override to actually function, and everything seemed fine, other than that the monitor info panel showed that it was running at 165Hz, and had disabled ULMB as a result.

ULMB with scanline sync is still preferable to Gsync.... But it was almost so much better. It's super frustrating to know the hardware is capable of doing it but the firmware crippled it. Shame it's not open source...........

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

Re: Viewsonic XG2703-GS QFT Success [Quick Frame Transport Lag Reduction]

Post by Chief Blur Buster » 27 Jan 2021, 22:54

disq wrote:
18 Jan 2021, 17:06
Edit: OP deleted his comment :|
xcasxcursex wrote:
20 Jan 2021, 01:44
disq wrote:
18 Jan 2021, 17:06
Edit: OP deleted his comment :|
I edited it, but it's awaiting moderation. Anyway if I knew then what I knew now, I might have deleted it :(
<Forum System Quirk>

Correct. It wasn't deleted. It was temporarily hidden by the automatic system.

Everytime a post is edited by a 1-post or 2-post forum member, it goes back into Moderation Queue. This makes newbie posts temporarily disappear when edited. It reappears when manually approved.

It is a weird quirk of forum approve queue when a new-OP tries to edit posts.

We had to do this because hundreds of spammers often later edited their posts with advertisements after posting a fakely legit-sounding post. So all edits of 1-postcount and 2-postcount users go into Moderator Approval Queue (for human eyes to approve first).

A Moderator Approval Queue is a manual spam filter system, to fight against next-generation advertising spammers, maybe from GPT-2 or GPT-3 style natural text AI algorithms. Regardless, it is a cat-and-mouse game.

Once a forum newbie here has a few posts (threshold varies, but it's usually approximately 5).... Members can edit/post instantly, and post permitted images/links/etc instantly.

(as long as it passes the automated spam filter -- we use 2 automated spamfiltering packages but combined isn't even 100% success rate. So, we have to have a Moderator Approve Queue for 1-postcount and 2-postcount forum newbies).

</Forum System Quirk>
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