NVIDIA Low Latency Mode - On or Off for less Input lag?

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.
Defox
Posts: 11
Joined: 16 Jul 2020, 11:59

NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by Defox » 29 Jul 2020, 09:56

Greetings!

Based on the tests, when GPU is not loaded at 99%, for some reason the input lag at the "Ultra" value is little increased compared to "Off". Will there be an increase input lag in the same situation when set to "On" compared to Off? I have not found any comparison between "On" and "Off"

What do I need to choose (on or off) for the lowest latency with a constantly not GPU bounded?

UPD:
In a recent video of Battle(non)sense about "NVIDIA Reflex", I noticed that it recommends setting the low latency mode to "Off". Why not "On"? If I am correctly aware, then the "On" value forces the maximum number of prepared frames to 1, which will give us confidence that the game will not prepare more than 1 frame.

Are there any explanations and assumptions why he recommends installing it "Off"?
Last edited by Defox on 15 Oct 2020, 06:35, edited 1 time in total.

Defox
Posts: 11
Joined: 16 Jul 2020, 11:59

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by Defox » 29 Jul 2020, 11:45

in non-G-SYNC scenarios, according to Nvidia, "Ultra" LLM has a "just-in-time" frame delivery mechanism, so in non-GPU-bound situations, it may occasionally time delivery of a frame too late, missing a fixed refresh window, causing the previous frame to repeat once over, instead of the next frame showing. If this even happens every few seconds, each of these repeat frame occurrences could cause a minor average input lag increase over LLM "On" or "Off" in non-GPU-bound situations.
I searched deeper and found this post.
So the "On" value as well as "Off" does not increase the latency in the case of no GPU bounded compared to "Ultra", right?

There was one more question. In the case of a GPU bound, does "On" reduce input lag as well as "Ultra", or does "Ultrа" do it better? What should I use with the GPU bound for the minimum input lag?

TTT
Posts: 156
Joined: 28 Jul 2018, 14:17

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by TTT » 01 Aug 2020, 06:59

Leave it set to 'On' universally and forget about it.

On is the same as max pre-rendered frames '1', off is let the game decide.

Setting it to Ultra is supposed to reduce latency when you have your game uncapped and you are GPU bound and still only achieving 60-100fps.

Battlenonsense did a vid though and showed you can decrease latency even more by not having it on and capping your GPU usage below max.

User avatar
MaxTendency
Posts: 49
Joined: 22 Jun 2020, 01:47

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by MaxTendency » 01 Aug 2020, 09:04

TTT wrote:
01 Aug 2020, 06:59
Battlenonsense did a vid though and showed you can decrease latency even more by not having it on and capping your GPU usage below max.
Somewhat true, but this behavior is game dependent and doesn't apply universally. On games like ow capping fps can reduce input delay but thats not the same with every game. Hardware unboxed did a newer video on this testing a wide variety of games and there seems to be no one stop solution to weather you should cap your fps or not.

Image
Click on the picture to take you to the full video.

User avatar
jorimt
Posts: 1412
Joined: 04 Nov 2016, 10:44
Location: USA

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by jorimt » 01 Aug 2020, 09:35

MaxTendency wrote:
01 Aug 2020, 09:04
but this behavior is game dependent and doesn't apply universally.
This is true, but...
MaxTendency wrote:
01 Aug 2020, 09:04
Hardware unboxed did a newer video on this testing a wide variety of games and there seems to be no one stop solution to weather you should cap your fps or not.
If that's the same video I think it is (saw it when it was released), it's confusing, doesn't lay out its methodology explicitly enough, and is thus wildly inconclusive.

At one point I wasn't even sure they knew what they were testing for any more, and in some instances they may have been conflating it with other factors, including counting in sheer framerate input lag (like in the screenshot you embedded, the differences of which were directly due to an increase in frametime proportionate to the avg FPS drop, not the LLM setting used).

Any setting that manipulates the pre-rendered frames queue has the most effect in reducing input lag when the GPU is saturated, and is preventing the CPU from handing over the newest information to complete render.

In non-GPU-bound situations, LLM is more likely to affect frame pacing than input lag. And in GPU-bound situations, it's more likely to first affect average framerate before it affects input lag.

LLM is unnecessarily obsessed over, as it honestly has the least potential impact on input lag out of GPU-side user configurable settings. You're looking at a frame or so of input lag reduction in specific GPU-bound situations, at best.

If these settings were any more impactful, AMD and Nvidia would continually market the crap out of them, yet they don't.

Basically, if you use G-SYNC, it's safe to leave LLM set to "On." If you use fixed refresh, it's safe to leave LLM to "On," or try "Ultra" per game (may or may not induce more stutter depending on the system/game combo).

MPRF settings (and even HAGS, for that matter, which FYI, is ultimately related to the pre-rendered frames queue) will really have the most impact on weaker or non-gaming systems. So if your system is at all capable for gaming, said settings are going to be a case of diminishing returns more than not.

Bottom-line is, you don't even want be in a performance position where your system has to resort to relying on LLM in the first place, and if you want to ensure the lowest possible input lag from a GPU standpoint, prevent it from maxing out, LLM or no LLM.

That's really it.
(jorimt: /jor-uhm-tee/)
Author: Blur Busters "G-SYNC 101" Series

Displays: Acer Predator XB271HU / LG 48CX OS: Windows 10 MB: ASUS ROG Maximus X Hero CPU: i7-8700k GPU: EVGA RTX 3080 FTW3 UG RAM: 32GB G.SKILL TridentZ @3200MHz

TTT
Posts: 156
Joined: 28 Jul 2018, 14:17

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by TTT » 01 Aug 2020, 09:39

MaxTendency wrote:
01 Aug 2020, 09:04
TTT wrote:
01 Aug 2020, 06:59
Battlenonsense did a vid though and showed you can decrease latency even more by not having it on and capping your GPU usage below max.
Somewhat true, but this behavior is game dependent and doesn't apply universally. On games like ow capping fps can reduce input delay but thats not the same with every game. Hardware unboxed did a newer video on this testing a wide variety of games and there seems to be no one stop solution to weather you should cap your fps or not.

Image
Click on the picture to take you to the full video.
That example isn't right though, its not GPU bound anyway. Nvidia themselves said Ultra is made for a GPU bound scenario.

The reason capping comes into it is to stop you from hitting 100% GPU, not just capping to random frames.

User avatar
jorimt
Posts: 1412
Joined: 04 Nov 2016, 10:44
Location: USA

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by jorimt » 01 Aug 2020, 09:44

TTT wrote:
01 Aug 2020, 09:39
That example isn't right though, its not GPU bound anyway. Nvidia themselves said Ultra is made for a GPU bound scenario.

The reason capping comes into it is to stop you from hitting 100% GPU, not just capping to random frames.
Yup, that particular graph shows differences in input lag due to frametime, not LLM. At points, they really seemed to lose track of what they were trying to originally test for.
(jorimt: /jor-uhm-tee/)
Author: Blur Busters "G-SYNC 101" Series

Displays: Acer Predator XB271HU / LG 48CX OS: Windows 10 MB: ASUS ROG Maximus X Hero CPU: i7-8700k GPU: EVGA RTX 3080 FTW3 UG RAM: 32GB G.SKILL TridentZ @3200MHz

TTT
Posts: 156
Joined: 28 Jul 2018, 14:17

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by TTT » 01 Aug 2020, 09:50

Don't get why people keep posting this same question, the answer is probably in half the first page threads! :lol:

User avatar
jorimt
Posts: 1412
Joined: 04 Nov 2016, 10:44
Location: USA

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by jorimt » 01 Aug 2020, 09:55

TTT wrote:
01 Aug 2020, 09:50
Don't get why people keep posting this same question, the answer is probably in half the first page threads! :lol:
:) For future reference, if anyone else comes across this thread and is curious about this subject, I have posted quite a bit on it in this forum already. Just search "LLM" and/or "pre-rendered" in my post history:
search.php?author_id=2920&sr=posts
(jorimt: /jor-uhm-tee/)
Author: Blur Busters "G-SYNC 101" Series

Displays: Acer Predator XB271HU / LG 48CX OS: Windows 10 MB: ASUS ROG Maximus X Hero CPU: i7-8700k GPU: EVGA RTX 3080 FTW3 UG RAM: 32GB G.SKILL TridentZ @3200MHz

Defox
Posts: 11
Joined: 16 Jul 2020, 11:59

Re: NVIDIA Low Latency Mode - On or Off for less Input lag?

Post by Defox » 15 Aug 2020, 10:46

jorimt wrote:
01 Aug 2020, 09:35
Bottom-line is, you don't even want be in a performance position where your system has to resort to relying on LLM in the first place, and if you want to ensure the lowest possible input lag from a GPU standpoint, prevent it from maxing out, LLM or no LLM.

That's really it.
There was one more question.
To get lower input latency in "Low Latency Mode = Ultra" - we need to limit in-game FPS below 138 (in the case of a 144 Hz monitor).
But. If there is no in-game limiter. Does it make sense to limit FPS using the nvidia control panel (as we know it increase input lag)? Will this decrease the input lag or, on the contrary, increase it?

Post Reply