DSC Problems & Misinformation [DSC IS NOT A HOAX]

[r0ach]

Saw some random guy asking questions about DSC. He knows absolutely nothing about monitors to the point where he basically doesn't even know what overdrive is yet he instantly noticed having DSC on is bad. There's no way all the review sites and Youtube shills haven't noticed the same thing, and if they don't they have to be totally incompetent.

- FORUM SYSTEM EDIT TO ADD SUPPLEMENTAL INFO -
Blur Busters' Word On The DSC Lag Rabbit Hole

This is a complex rabbit hole.

Short answer: DSC should not add lag, but latency-buggy DSC settings/codecs/gpus/devices do.
That's the bad DSC lag you heard from some influencers/youtubers/reviewers. But there's better DSC.

Here's the long answer:

Keep in mind DSC performance varies from monitor to monitor and GPU to GPU. Ideally, the best implementations adds no extra lag, but it's wholly true that lag-buggy early implementations may add wayyyy more lag. And sometimes it's the specific display model's fault, and sometimes it's the specific GPU or specific GPU driver's fault. Many whac a moles to do.

And some early GPU implementations emit a DSC signal that accidentally adds too much DSC latency when it should not. Even for slice sizes in the VESA DSC protocol that doesn't even warrant processing latency, but that it's easier for an early-programmed DSC implementation to buffer more and just post-process. And it does happen. Sadly.

Yes, in the best low-latency implementations possible, DSC often needs to buffer 1 pixel row so it can process the streaming compression properly, but a 1080p 240Hz signal is 135 kilohertz scan rate (135000 scanlines/sec). So 1/135000th second of latency, which is not feelable by a human. Unfortunately some firmwares royally bØяK3d the DSC implementation lag-wise, and here we are. Yup, #ConclusionJumpers.

In fact, Retrotink 4K (Blur Busters Approved video processor box for retro market) can inject BFI into a video signal that is less laggy than many TV firmwares' own native BFI. Simply because of how many TV firmwares more fully buffers and processes the signal before displaying the refresh cycle. See? I do paid work to reduce lag of various display processing.

#ScalerFirmwareImplementationFails
#DontJudgeAllScalersByOthers
#GPUsettingsFail
#JumpingToConclusions

DSC can be realtime stream decompression, and proper DSC feels exactly like 422 chroma (but better quality).
Worst DSC = more lag like what you witnessed.
Best DSC = no more lag than 422 chroma.

DSC can support slice sizes that affects latency of "lazy DSC codecs" (those not programmed to optimal latency):
VESA DSC Specification, Slice Size Chapters
But you can have a slice size of just a scanline and stream process it with nigh no latency. GPUs should expose more settings, etc, to make this adjustable, in theory.

In theory, slices don't really add latency because it's simply retroactive compression help. However, some actual implementation of codecs add unwanted latency, which is supremely dissapointing. But it is not a #DSChoax thing.

While I can say VRR has more lag than VSYNC OFF for many specific use cases (especially "framerates far exceed Hz" and/or "need mouse cursor" situations), the theoretical minimum DSC latency is far lower than the theoretical minimum VRR latency by at least two orders of magnitude from grade-school mathematics (average half a scanout lag penalty versus average 1 scanline lag penalty). So equating DSC and VRR lag as being law of physics equal, is just trying to claim 2+2=5 when the story is sadly more complciated.

It's a cesspool, yes.
But reviewers & users are jumping to conclusions.
Sadly, DSC needs to be lag-tested on a per-product basis, without wholesale DSC-algorithm conclusion jumping.

Switch GPU? Retest.
Switch Displays? Retest.
Yadda, yadda.

Oh and yeah, I edited the topic title for y'all. You are welcome

There are great points in this thread, Some implementations of DSC can be crappy yes. This part is factually correct for some, but not all DSC implementations.. But it is lost in the fog of "LOL DSC IZ HOAX" equivalent of a very completely faked bait topic subject, so I've removed the bait with #Facts. m'kay?

And hey hey, open source programmers of NVInspector style tools, please, please (CAN I HAZ HALP?)... Let us reprogram slice sizes of the outgoing DSC stream. Some displays will improve in lag if they're using buggy slice processors that accidentally adds lag. Changing slice sizes GPU-side can actually bypass some lag-bugs display-side (but not for all displays, sadly).

Oh my, if only stream codecs did not add unnecessary lag, I wish! Sometimes it's sender-side lag, and sometimes it's receiver-side lag, oh my, I wish I could find the correct tree in the lag forest, sometimes! But DSC is not a hoax. DSC is great for the refresh rate ....if implemented properly. 4K 1000fps 1000Hz OLED with lagless esports framegen FTW.

And please copy-paste the permalink to this post to your favorite DSC latency newbies:

Code: Select all

https://forums.blurbusters.com/viewtopic.php?f=2&t=12730&p=99795#p99795

I insist. Please freely throw this link to anybody who talks about DSC latency.

You're welcome.

</>
Yes, my famous micdrop, for descendants of "Humans Cant See 30-vs-60" debators.<- Tiny Easter Egg

[Artluxo]

what makes dsc feel bad? input lag? or picture quality/stuttering?

[jorimt]

DSC does not necessarily increase input latency, but it typically does not play nice with multi-monitor configurations, and restricts things such as custom resolution creation and DSR/DLDSR usage.

DSC is necessary to allow refresh rate/resolution combos that would otherwise be impossible with the bandwidth limitations of the given display standard, be it Display Port 1.4, HDMI 2.0, etc.

Not sure what those Discord commenters are on about though...

[r0ach]

what makes dsc feel bad? input lag? or picture quality/stuttering?

Just like the picture says, it's one of those "heavy cursor" settings. I'm guessing it doesn't actually show up in end to end latency (or will be something like 1ms) so this is going to be one of those things where casuals and tech reviewers like Gamer's Nexus pretend the setting isn't detrimental because they're unable to quantify it. Just like freesync.

Do any good players actually play with freesync on? Probably not because it's easier to aim with it off. Can you actually quantify freesync? Sort of. It does actually seem to show up in latency tests with on vs off. The latency test doesn't actually quantify the extent of it's detriment to aim, though.

[jorimt]

Can you actually quantify freesync? Sort of. It does actually seem to show up in latency tests with on vs off. The latency test doesn't actually quantify the extent of it's detriment to aim, though.

VRR dynamically steers the tearline off-screen when the framerate is within the refresh rate, and...that's effectively all it does, fundamentally.

Anything else people attribute their issues to where VRR is concerned don't understand it enough to realize they are mis-attributing the cause of said issues more often than not, which gets old.

It now appears some people are mis-attributing the cause of their issues similarly with DSC, which, again, is just a way to compress the signal over the cable in a visually lossless way as to achieve refresh rate/resolution ratios that otherwise couldn't be reached without it.

That said, it doesn't mean they aren't experiencing genuine issues that may be related to VRR and/or DSC implementation on some models at the firmware and/or driver-level, but it wouldn't be down to the fundamental VRR or DSC behavior themselves.

[r0ach]

Can you actually quantify freesync? Sort of. It does actually seem to show up in latency tests with on vs off. The latency test doesn't actually quantify the extent of it's detriment to aim, though.

VRR dynamically steers the tearline off-screen when the framerate is within the refresh rate, and...that's effectively all it does, fundamentally.

Anything else people attribute their issues to where VRR is concerned don't understand it enough to realize they are mis-attributing the cause of said issues more often than not, which gets old.

It now appears some people are mis-attributing the cause of their issues similarly with DSC, which, again, is just a way to compress the signal over the cable in a visually lossless way as to achieve refresh rate/resolution ratios that otherwise couldn't be reached without it.

That said, it doesn't mean they aren't experiencing genuine issues that may be related to VRR and/or DSC implementation on some models at the firmware and/or driver-level, but it wouldn't be down to the fundamental VRR or DSC behavior themselves.

There is theory and then there is reality. Mouse movement is different with different refresh rates. Altering refresh rate on the fly was never going to be conducive to aiming well even in theory. While DSC is just some new annoying setting that feels like it bogs down your cursor movement aka "heavy cursor" from whatever it's doing in the background.

[jorimt]

There is theory and then there is reality. Mouse movement is different with different refresh rates. Altering refresh rate on the fly was never going to be conducive to aiming well even in theory.

With VRR, the scanout time, aka the frame scan-in speed doesn’t change; it remains at the maximum refresh rate, regardless of framerate within said refresh rate. Only the amount of times the scanout repeats per second changes with the framerate during VRR operation.

You and those who claim what you do have zero competing data, and don't seem to know enough about the ins-and-outs of fundamental VRR and scanout operation to even begin creating viable alternate explanations to what you feel you are experiencing.

You appear to have already made up your mind, so I think there's little left for us to discuss on this point. Think what you will.

DSC is just some new annoying setting that feels like it bogs down your cursor movement aka "heavy cursor" from whatever it's doing in the background.

If there is any truth to that at all, I haven't personally experienced it, and I'm using a DSC monitor right now.

Further, DSC is not a “hoax;” without it, my monitor couldn’t reach 360Hz at 1440p via DP 1.4, period.

That said, again, implementation is everything, and as I previously mentioned, some monitors may implement these features more poorly than others.

Even my PG27AQN has what seems to be a driver interaction bug with G-SYNC and DSC that causes it to lose signal in certain games unless I deactivate and reactivate G-SYNC, or toggle my scaling settings in the NVCP each time.

Said behavior popped up about three driver releases ago (without me changing anything else with my configuration, and I have reported it to Nvidia), but again, that's not technically the fault of VRR or DSC themselves, but implementation at the firmware and/or driver-level, which will vary by model and vendor.

[r0ach]

You and those who claim what you do have zero competing data, and don't seem to know enough about the ins-and-outs of fundamental VRR and scanout operation to even begin creating viable alternate explanations to what you feel you are experiencing.

I don't think you understand. This is not a complex subject. It's not necessary to read the white paper for how VRR works. All you need to do is click the monitor OSD and turn it on and off then notice how cursor movement is afterwards. It's like claiming only the inventor of the AK47 can possibly know if it's good or not. Anyone can pick the thing up and notice after a few minutes.

As far as I know things like freesync are not popular among high tier players. Do you think there's no reason for this? If it was useful, people would actually use it, but it's not. I haven't even experienced visible tearing in a video game in the last 20 years with vsync off. What do I get by turning freesync on? Nothing besides possible detrimental effects. It was a solution looking for a problem. But it is detrimental to aim on top of not really even providing any benefit.

DSC is just another flop setting to add on top of the freesync gimmick that brings heavy cursor movement. Even overdrive gives you stiffer cursor movement than overdrive off, but at least it does something semi-useful. In reality, most people would probably play virtually identical with the overdrive off response times for whatever monitor they use, so even here in a setting that people make believe is only beneficial with no downside, the negatives can even outweigh the positive there.

[jorimt]

I don't think you understand. This is not a complex subject. It's not necessary to read the white paper for how VRR works. All you need to do is click the monitor OSD and turn it on and off then notice how cursor movement is afterwards. It's like claiming only the inventor of the AK47 can possibly know if it's good or not. Anyone can pick the thing up and notice after a few minutes.

I understand you think you feel an actual difference in mouse response when DSC and/or VRR is on/off, and while there is some limited truth to reduced mouse responsiveness with VRR at extremely low framerates (that most comp players aren't competing at anyway), there's no technical explanation for this in regards to DSC, nor any data to back it up, so self-reporting and anecdote is all anyone has to go off of.

That, and even if DSC was as crappy as you're claiming, it's required for certain models to reach their advertised maximum refresh rate/resolution ratio, so your complaints really have no resolution.

As far as I know things like freesync are not popular among high tier players.

Newsflash, VRR is not popular among comp players. Agreed. I even recommend comp players that want the lowest possible input latency not to use it. VRR is the lowest latency "tear-free" scenario, not the lowest latency scenario, and that will always be the case.

Also, the higher the refresh rate, the less noticeable tearing artifacts are, regardless of the framerate, so VRR becomes less and less necessary the higher the max refresh rate of monitors become.

VRR is and always was intended as a stop-gap to tearing prevention, similar to how strobing is to motion clarity.

What do I get by turning freesync on? Nothing besides possible detrimental effects. It was a solution looking for a problem. But it is detrimental to aim on top of not really even providing any benefit.

That's you're opinion, and you're free to use or not use VRR. That doesn't make it useless.

DSC is just another flop setting to add on top of the freesync gimmick that brings heavy cursor movement.

Either you're not paying attention to what I previously wrote, or you're just trolling (which is not as welcome in our forums as it is on other parts of the internet), but as I've already noted, DSC is REQUIRED to reach certain refresh rate/resolution ratios, else we're limited to 240Hz at 1440p, for instance.

This will be the case until DisplayPort 2.0 and HDMI 2.1 become more prevalent in newer monitor models, and even then, DSC will still be required where bandwidth is exceeded on those.

But yeah, we get it, you hate DSC and VRR. Easy fix; don't buy a monitor with DSC, and then disable VRR on it.

[Chief Blur Buster]

That doesn't make it useless.

Agreed. Right Tool For Right Job.

VRR is one of the world's lowest lag "NOT VSYNC OFF" sync technology, when you're not dealing with mouse cursors (e.g. focusing on aimtrainer performance versus VSYNC ON technologies or other sync technologies that are not VRR). You can't use VSYNC OFF for all use cases, such as virtual reality.

That's you're opinion, and you're free to use or not use VRR.

One quirk of the mouse cursor is it's not always fully representative of VRR latency of aimtrainer.
- Fixed-Hz: The magic is hardware mouse cursors always ran at full refresh rate, even at low game frame rates
- VRR Hz: The mouse cursor only updates once per frametime. So cursors feel unusually handicapped sometimes.

However, when aimtrainering VRR vs VSYNC ON vs VSYNC OFF, you notice you perform much better with VRR than naturally Hz-throttled VSYNC ON, as long as your refresh rate range is bigger than your frame rate range. In fact, your performance between VRR and VSYNC OFF almost equalizes in your favourite aimtrainer but only if you purchased a refresh rate range bigger than the aimtrainers' uncapped frame rate range. This scientific fact is lost because most esports games like to spray framerates far beyond MaxHz of VRR, which is a big problem for esports VRR. And only few esports players have 360Hz+ yet.

The famous old Blur Busters capping trick was to solve a worse latency problem. But why cap VRR at all? Just buy more Hz range than framerate range. We gotz 540Hz models now, after all.

That's why better esports-quality VRR (for those who really hate VSYNC OFF) is a 360Hz+ monitor with a game that only runs 100-300 frames per second. You don't want to have to pick-poison of a framerate cap just to avoid even worse latency. But why cap at all? Just buy a new VRR display with a refresh rate range much bigger than your games' framerate ranges, and watch most of latency issues disappear.

You don't want to use VRR when you're dealing with low-framerate popup menus in games (e.g. RTS), so you might as well turn off VRR for RTS games. Latency testing doesn't cover mouse cursor lag of the "one cursor update per VRR refresh cycle" problem.

Now, one of the beefs I have is that FreeSync has more lag than native G-SYNC, because the graphics driver is responsible for initiating the begin of a new refresh cycle, and that can lag. On native hardware G-SYNC, the latency of starting a new refresh cycle can be even lower. But I have seen some great low-lag FreeSync implementation on RTINGs that has almost the same lag as FreeSync OFF, and as long as your framerate range is entirely inside the venn diagram of VRR range.... problem mostly solved.

VSYNC OFF is still superior for many things obviously, but there are some use cases where VRR really helps (if you don't need to worry about mouse cursor).

VSYNC OFF is still king. But understanding HOW TO OPTIMIZE VRR is the hallmark of an intelligent reputable optimizer, who says "...You should do VSYNC OFF, but if you use VRR, here's some sage knowledge..." point of view.

TL;DR: Buy more refresh rate range than your game's uncapped framerate range, if you want VRR to be as esports-quality as possible (zero out as much as you feasibly can, performance differential relative to VSYNC OFF). It may not be quite VSYNC OFF, but 95%+ of VRR esports disadvantages miraculously disappear when you actually purchase more VRR range than your games' maximum peak frame rate.

And now click:
Blur Busters Official Word On The DSC Lag Rabbit Hole