Page 2 of 10

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 02 Oct 2020, 15:06
by Avean
If i have understood this correctly, its still adviced to use V-Sync + fps limiter globally on with G-Sync? And if a game supports Reflex you can enable it without touching the global settings?

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 02 Oct 2020, 15:45
by jorimt
Avean wrote:
02 Oct 2020, 15:06
Yes, that should be generally fine.

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 02 Oct 2020, 18:34
by diakou
jorimt wrote:
02 Oct 2020, 15:45
Avean wrote:
02 Oct 2020, 15:06
If i have understood this correctly, its still adviced to use V-Sync + fps limiter globally on with G-Sync? And if a game supports Reflex you can enable it without touching the global settings?
Yes, that should be generally fine.
Seth from NVIDIA actually just recently added some info from themselves on it (still, they're not the true authority since there's many rules for what info can be shared "truthfully", but this is still adding info!)

Image

From: https://www.nvidia.com/en-us/geforce/gu ... tion-guide

Side note: Funny how when money is involved and marketing is a thing that the truth comes forward, a lot of latency optimizers and people who've been "tweaking" have been constantly told nothing they touch will have any benefit. (referring to real ones who want everyone to have lower latency, not the people who sell and optimize for money)
I enjoy NVIDIA's wording on this article regarding "advanced latency optimization" they capture the concept well :) Naysayer culture has been very lame and it can be found anywhere, so far very little in blurbusters, but still a ton (even from mods at times :roll: )
Image

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 02 Oct 2020, 21:44
by howiec
diakou wrote:
02 Oct 2020, 18:34
From: https://www.nvidia.com/en-us/geforce/gu ... tion-guide
Side note: Funny how when money is involved and marketing is a thing that the truth comes forward, a lot of latency optimizers and people who've been "tweaking" have been constantly told nothing they touch will have any benefit. (referring to real ones who want everyone to have lower latency, not the people who sell and optimize for money)
I enjoy NVIDIA's wording on this article regarding "advanced latency optimization" they capture the concept well :) Naysayer culture has been very lame and it can be found anywhere, so far very little in blurbusters, but still a ton (even from mods at times :roll: )
Image
Yeah, I'm glad that Nvidia and others in the industry (e.g. mouse designers) are paying more attention to optimizing performance with respect to latency AND timing, and not just "throughput" from an engineering sense. Both are important from an interactive games/app perspective.

I'll probably post a Windows PC optimization guide of things that I've found to actually make a difference after all the testing I've already done. Of course, all of this can already be found in various places all over the internet.

MSI mode, write combining, NT scheduler settings, interrupt affinity, and CPU power settings are only some of the things you can optimize.

Off the top of my head some other things that actually do make a difference incl (there's more than just this):
  • MMCSS priorities
  • Various driver tweaks via Nvidia profile inspector
  • Various game/engine-specific tweaks biasing low latency
  • Disabling various event logs
  • Disabling various services, devices, & drivers
  • adjusting xHCI interrupt moderation
  • FSE & disabling FS optimizations
  • Windows timer resolution => 0.5ms
  • bcdedit useplatformtick
  • Disabling HT/SMT
  • Disabling HPET

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 02 Oct 2020, 22:05
by jorimt
diakou wrote:
02 Oct 2020, 18:34
Seth from NVIDIA actually just recently added some info from themselves on it (still, they're not the true authority since there's many rules for what info can be shared "truthfully", but this is still adding info!)

Image

From: https://www.nvidia.com/en-us/geforce/gu ... tion-guide
Good to see they're making an effort to share more specific information more frequently now. Certainly a recent (and welcome) development.

Beyond everything I already know there (and regularly recommend myself), that message suggests that Reflex does indeed auto cap the FPS with G-SYNC + NVCP V-SYNC, just like LLM "Ultra" did before it (in supported games). This is one point Battle(non)sense wasn't too clear on in his videos, so I initially assumed it didn't, as I haven't had time to test Reflex myself yet (I'll make a note of this in the OP).

But if that's the case, then if you want to use an in-game (or another external) limiter with Reflex, you'll have to limit it slightly below Reflex's limit for it to take effect (when the framerate can reach it).

E.g. with G-SYNC, this suggests:

- Where framerate can exceed the refresh rate, Reflex = auto FPS limit slightly below the refresh rate to keep G-SYNC active/in-range.
- Where average framerate generation begins to be limited by GPU capability, Reflex = auto fps limit that sets itself dynamically below the achievable (fluctuating) framerate as to prevent max GPU usage, effectively eliminating the render queue.
diakou wrote:
02 Oct 2020, 18:34
Side note: Funny how when money is involved and marketing is a thing that the truth comes forward, a lot of latency optimizers and people who've been "tweaking" have been constantly told nothing they touch will have any benefit.
Yeah, I'm honestly not sure what exactly triggered this and when (though no doubt Blur Busters and similar sources had a hand in it), but it certainly has seemed to force them (and the industry as a whole) to more fully recognize and prioritize it in recent years.

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 03 Oct 2020, 12:20
by jorimt
FYI all, I've now tested base Reflex functionality myself, and have updated the OP with the below:
viewtopic.php?f=10&t=7522#p57333
G-SYNC on + NVCP V-SYNC on + in-game V-SYNC on OR off + Reflex ("On" or "On + Boost")
  • Reflex applies an automatic FPS limit slightly below the refresh rate. At 144Hz, it limits to ~138 FPS, just as LLM "Ultra" (in supported games) did before it. This auto FPS limit is relative to the current maximum refresh rate.
  • If an in-game or external FPS limiter is to be used with Reflex enabled, the in-game or external limiter must set slightly below Reflex's automatic limit for it to override Reflex's. This will not override Reflex's auto FPS limiting function (and thus, render queue elimination benefits) when the game is GPU-bound and frames are fluctuating, and will instead only override the initial static Reflex FPS limit where the game is not GPU-bound and the system can maintain a framerate that could otherwise exceed the limiters if they were disabled.
G-SYNC on OR off + NVCP V-SYNC off OR "Use 3D application setting" + in-game V-SYNC on OR off + Reflex ("On" or "On + Boost")
  • Reflex does not apply an automatic FPS limit slightly below the refresh rate, but retains all other functionality.
Once I receive an LDAT from Nvidia, I will be performing input lag testing in at least one of those three games to determine and verify how Reflex behaves in both GPU-bound and non-GPU-bound scenarios.

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 04 Oct 2020, 09:21
by speancer
jorimt wrote:
03 Oct 2020, 12:20
Once I receive an LDAT from Nvidia, I will be performing input lag testing in at least one of those three games to determine and verify how Reflex behaves in both GPU-bound and non-GPU-bound scenarios.
Just to let you know, I am especially curious how Reflex behaves in non-GPU-bound scenarios, so I'm looking forward to seeing your test results. I guess them results would give me the ultimate answer whether Reflex is completely useless for me or not, as I am never GPU-bound in my competitive play, my average framerate is 320-340 fps in competitive match with estimated average GPU usage between 35-40% (RTSS measurements).

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 04 Oct 2020, 09:57
by jorimt
speancer wrote:
04 Oct 2020, 09:21
Just to let you know, I am especially curious how Reflex behaves in non-GPU-bound scenarios
I plan on ultimately testing LLM (both "On" and "Ultra") in CS:GO for thoroughness sake as well. I'm honestly curious if that setting does anything in that game. My guess, where input lag reduction is concerned, is no, but we'll see.

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 04 Oct 2020, 10:25
by speancer
jorimt wrote:
04 Oct 2020, 09:57
speancer wrote:
04 Oct 2020, 09:21
Just to let you know, I am especially curious how Reflex behaves in non-GPU-bound scenarios
I plan on ultimately testing LLM (both "On" and "Ultra") in CS:GO for thoroughness sake as well. I'm honestly curious if that setting does anything in that game. My guess, where input lag reduction is concerned, is no, but we'll see.
That's great to know, I was even about to ask if you'd be willing to test it (as you told me in other thread you had not) :D I'm glad we share the same curiosity. Are you able to estimate when would your Reflex and LLM-for-CS:GO test results come up?

Re: NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

Posted: 04 Oct 2020, 10:31
by 1000WATT
speancer wrote:
04 Oct 2020, 09:21
Just to let you know, I am especially curious how Reflex behaves in non-GPU-bound scenarios, so I'm looking forward to seeing your test results. I guess them results would give me the ultimate answer whether Reflex is completely useless for me or not, as I am never GPU-bound in my competitive play, my average framerate is 320-340 fps in competitive match with estimated average GPU usage between 35-40% (RTSS measurements).
Yes, it is interesting.
average GPU usage between 35-40%. Not an indication that the game is not tied to the GPU. With a GPU load of 30% and a heavy load on the memory of the video card, you can get delays above 300ms.