Does "Nvidia Reflex" Limiting Mode enable Pseudo-Reflex Always or Only when Reaching the FPS Cap?

Reading over the helper note for the new "Nvidia Reflex" option/mode it states that for DX11 and DX12 games using this mode will also enable a kind of Pseudo Reflex even if the game in question does not support Reflex natively (assuming I'm reading/understanding the note correctly that is).

My question is, does this only occur when the Limit is reached? Or even when framerate drifts below the target does this "Pseudo-Reflex" mode still activate? E.g. Suppose I enable Reflex in a game that natively supports it -- it will auto-cap with the driver limiter a few frames beneath monitor refresh and also enable the Reflex mode/low latency mode when framerate is drifting below that target. So in that case Reflex is ON regardless of whether the limit is actually reached or not. Does the RTSS "Nvidia Reflex" limiter option/Pseudo Reflex that's active for DX11/DX12 titles also function the same way where Reflex is constantly active even when framerates drift below the FPS target?

I expect whatever RTSS does in this mode is also what Nvidia's Driver level Limiter does (they're identical as I understand it where the effect of enabling one should be the same as enabling it with the other?). I had noticed in some titles the Driver limiter "feels" anecdotally a bit more direct/less input delay VS RTSS Async (of course maybe it's placebo, can't rule it out entirely as I've not tested this with equipment) and now I'm wondering if this is because the Nvidia driver limiter also turns on pseudo reflex when it's in use for those DX11 and DX12 games.

Thanks! Just wanted to better understand that helper note. As always, love RTSS, my favorite PC gaming tool

 

I don't want to spam the other thread with questions like this, so I'm asking here

I use a hotkey to toggle the limiter on/off. When toggling it off, is Reflex also disabled?

 

I meant in games that have neither FG nor Relfex. I use the Reflex limiter as well in some of those games just to get the latency benefit.

 

Thank you! Apologies, I didn't see that response over yonder. Wow, that's really interesting it works like that huh. So the driver limiter basically force enables ULLM / Reflex in conjunction with wherever the cap is set.

And for @RealNC 's question, that's good to know. So if I play a game that does not natively support Reflex but is DX11/DX12 and do the following:
1) Launch DX11/DX12 game that does not natively support Reflex/FG where RTSS is set to Default Async behavior and limiter is engaged/on
2) Once in gameplay/game is loaded I switch the limiter to Nvidia Reflex mode
3) This also enables "pseudo reflex"
4) I play the game for a bit then switch the limiter back to Async mode from the reflex limiting option
5) Even though the limiter mode was swapped back to Async pseudo reflex stays enabled til the game is restarted.
Hope I got that right, thanks for the new features.

 

Side note, I wish Nvidia would talk more about what the drawbacks to Reflex can be -- they rarely do from what I've seen yet it definitely can impact FPS more than their claimed 3% target (stated in the recent Gamer's Nexus video) in CPU bound scenes/games.

For example in Cyberpunk with lots of RT features ON + high LOD / crowd density / cascaded shadows range with Reflex on in certain scenes my GPU utilization drops to ~80% whereas with Reflex OFF it sits at 98% on my 5800X / 3070 Rig and the FPS dip is meaningful in such scenes I'd say. I always thought queuing CPU command buffers, though it cost a bit of latency, was done to ensure consistent smooth gameplay and full utilization of the GPU/to prevent users from becoming CPU bound no?

 

I think you can just set the limiter to Reflex mode to begin with, no need to only enable it after game start.

Not sure why you call it "pseudo," it's the usual Reflex system. It's gonna work just as well as if it was supported by the game directly, but how well it reduces latency depends on how the game presents frames. So it should either fully reduce latency (best case) or have no impact at all (worst case,) but never actually hurt latency.

Unless you have some other reason to use async instead of Reflex.

 

Oh I just called it "pseudo" to differentiate the Reflex enabled by the RTSS limiter mode VS native Reflex implementations. Probably calling it "limiter-induced Reflex" would've been better

And thanks, I understand you don't need to wait til after the game is launched to swap limiter modes, I framed the steps that way because my focus was on when exactly the limiter-induced reflex turns on and then whether or not a restart is then required to switch it off if you mid-game change the limiter mode back to Async.

 

You can also try this in conjunction with reflex limiter to see if you get further latency reduction:

GeForce 551.23 WHQL driver download & discussion

 

Also I just tested that step 1-5 scenario in Cyberpunk with the in-game reflex OFF and in that game at least turning the RTSS "Reflex Mode" limiter ON clearly did enable Reflex since I saw the same Reflex/ULLM imposed framerate cap of 138 on my 144 hz panel even when attempting to cap to 141. When I turned the RTSS "reflex mode" limiter off entirely (set fps cap value to 0) reflex did appear to then turn off in game -- that 138 FPS imposed value was gone (tested in the menu since the menu runs up to very very high framerates) and I could feel a latency difference again between in-game reflex on vs in-game reflex off.

EDIT: Also there does appear to be a difference bewteen using the nvidia control panel limiter and the nvidia reflex limiter after all -- control panel one you can cap to 141 on 144 hz panel with ULLM OFF and it'll actually reach it whereas doing the same with the rtss option caps to 138 automatically like Reflex does so maybe the driver limiter isn't forcing ULLM/reflex after all? Or perhaps it is and the cap just isn't enforced the same way, curious.

 

It never did, and it never claimed to. NULL and Reflex are entirely separate things.

 

NV framerate limiter does not force ULLM and never did.

If you enable ULLM with Gsync, that auto caps your fps to value below refresh rate max, such as 138 for 144 hz.
Obviously 138 is lower than 141 so it takes precedence. If you would manually cap to say 130, then that would take precedence.

Reflex isn't enforced by anything in NV CPL, that is an RTSS feature, possible since NVAPI exposes the feature and like Unwinder said can be called by any process in the context of the 3D app (or along those lines).
So even if a game does not have implemented Reflex, RTSS or any other injection app can call and enable it.

The practical benefit to doing so versus ULLM depends on game. For example I saw no difference in sim-to-render latency in Division DX12 between Reflex and ULLM, regardless of setting ReflexSleep to 1 or 2 (so input and simulation are processed in different threads then).
Division also has internal reduced latency mode which I had enabled in both cases, though I'm betting it is just pre-render limit.

This behavior was consistent with capped and uncapped framerate as well. Although capped framerate had actually lower sim-to-render latency overall than uncapped.
Uncapped could reach temporary lower latency but also had lots of peaks of higher latency whereas capped had a consistent latency which was equal to frametime.

Which to me is kinda the best you can have, do not think sim-to-render latency can be lower than frametime as that is the time it takes to render the frame in any case.
I could be wrong though on that.

But using that as a practical metric can be handy anyway and to me it makes sense, if your latency is no more than your frametime is, then you essentially have "immediate" response to input.

 

Thanks, yeah I just (wrongly) assume that the driver limiter would work the same as using RTSS switched to the "Nvidia Reflex" mode which was my fault, should've read up on the feature here more closely.

That's really cool then that using said mode in RTSS can provide Reflex support even games without Frame Gen though, very cool.

 

Thank you that's helpful to know