Special K can drastically reduce latency
Posted: 26 Sep 2020, 16:49
Who you gonna call? The Blur Busters! For Everything Better Than 60Hz™
https://forums.blurbusters.com/
- Frame time consistency-wise, the SpecialK's frame rate limiter was significantly better than both RTSS and NV v3 limiters in all DX11 scenarios, while RTSS and NV v3 were overall on par in terms of frametime stability.
- Approximate latency-wise, the NV v3 limiter was significantly better than both the SpecialK and RTSS limiter in all DX11 scenarios, being the SpecialK limiter better than the RTSS one too.
- However, if we also consider a usability / performance approach between the different FPS limiters, the most balanced and overall recommended option, would be the RTSS limiter instead, which is easier to get working in all games and to be set up vs. the SpecialK limiting method, that is a bit difficult to get working on certain games/clients and it will require to tweak its "Tolerance" setting on per game-basis in some cases to prevent certain instability issues.
- Therefore, based on all the prior data and notes, and when the SpecialK mod is injected succesfully and its limiter is set up properly on a per game-basis, we can clearly state that the SpecialK's frame rate limiter is by far the most consistent and accurate FPS limiting method on DX11-based scenarios. That is, under certain scenarios, the SpecialK limiter can be considered an excellent and recommended alternative to cap your FPS and guarantee the stadiest and smoothest framerate, and specially as a recommended alternative to any suboptimal in-game limiter.
That is based off an older version.deama wrote: ↑26 Sep 2020, 20:43From reading it, seems like it's a framerate limiter tool, similiar to RTSS, that works as well as the nvidia one, however it does slightly worse with input lag, but has better pacing, so you should get less 1% and 0.1% dips, seems more finnicky to get working though.
https://www.reddit.com/r/pcgaming/comme ... x11_based/- Frame time consistency-wise, the SpecialK's frame rate limiter was significantly better than both RTSS and NV v3 limiters in all DX11 scenarios, while RTSS and NV v3 were overall on par in terms of frametime stability.
- Approximate latency-wise, the NV v3 limiter was significantly better than both the SpecialK and RTSS limiter in all DX11 scenarios, being the SpecialK limiter better than the RTSS one too.
- However, if we also consider a usability / performance approach between the different FPS limiters, the most balanced and overall recommended option, would be the RTSS limiter instead, which is easier to get working in all games and to be set up vs. the SpecialK limiting method, that is a bit difficult to get working on certain games/clients and it will require to tweak its "Tolerance" setting on per game-basis in some cases to prevent certain instability issues.
- Therefore, based on all the prior data and notes, and when the SpecialK mod is injected succesfully and its limiter is set up properly on a per game-basis, we can clearly state that the SpecialK's frame rate limiter is by far the most consistent and accurate FPS limiting method on DX11-based scenarios. That is, under certain scenarios, the SpecialK limiter can be considered an excellent and recommended alternative to cap your FPS and guarantee the stadiest and smoothest framerate, and specially as a recommended alternative to any suboptimal in-game limiter.
Any tests to back that up though? Something with a high-speed camera comparing RTSS vs special k?
Unfortunately, none currently, I asked a few friends if they would have the time to help out, but schedules are packed. I'll see if I can get something done.
If you have the time, here are the two direct links to "low latency mode" and NVIDIA Driver being compared with Special K (and if you use special k, but then use the NVIDIA Driver framerate limiter, you'll have even less latency) (it's just linking to a discourse forum with the pictures + explanation, it was posted as opening post of this thread)Chief Blur Buster wrote: ↑23 Sep 2020, 14:07I've added a shadow redirect link to the Programming Forum -- because this is of fantastic interest to software developers who are also connoiseurs of latency.
What NVIDIA is doing about latency is really truly fantastic stuff, from Reflex to FrameView.
A link to forum thread The Amazing Human Visible Feats Of The Millisecond was shared inside one of NVIDIA's team, which probably helped inspired to do some of this stuff, amongst a lot of other stuff.
Aemony:
What’s up with the weird 0.1% percentile FPS of Unlimited and RTSS ? Is Special K just pacing the frames better to prevent such weird drops?
Kaldaien:
That’s the magic of aligning your framerate limiter’s clock to VBLANKAnd some pretty nifty clock drift compensation code I designed that sneaks in occasional clock re-syncs only when re-syncing the clock can be done without inducing a hitch.
I basically always stop delaying frames at precisely the same time that the DWM compositor has swapped the front-buffer out and is ready to accept a new frame. Thus when I tell the engine to continue, it can do so with zero back-pressure from pending VSYNC.
We’re not even playing the same game here, RTSS may try to count the frames and limit them but it’s not taking into account the traffic cop that is the DWM. RTSS releases the GPU into oncoming traffic and causes accidents frequently because its clock is not sync’d up with the DWM
You can basically look at any framerate limiting done by RTSS and you will always find this problem. It can count frames, just probably shouldn’t be directing traffic since it’s not paying attention to all the collisions caused by its bad clock.