External frame limiter Vs. in-game limiter

[RagingBuddhist]

Whenever I use an external limiter like NVCP I see big variances in performance. I get seemingly the same performance, but when capped using Nvidia the power usage is lower, clocks are lower, and temps are lower. As an example I took screenshots of a test in Hunt Showdown. All the settings are the same for both test besides where the FPS cap was put in place. All the games I’ve tried this in gave the same results, I’m not sure of the cause. Reflex+boost isn’t an option in the games I’ve tried so it’s not that. Power mode set to normal in NVCP. Can anyone with more expertise share their wisdom? I’d prefer to use in-game limiters, but with such a big difference in power usage why would I?

[RealNC]

The Nvidia limiter is taken into account by the driver when making power management decisions. So more aggressive power management is possible. Obviously this comes at a latency cost. Nvidia themselves recommend setting it to "maximum perf" if you care about latency:

https://nvidia.custhelp.com/app/answers ... C-and-more

The driver has no control over in-game limiters, so it doesn't alter the power management strategy.

[RagingBuddhist]

I wonder how much more latency it adds using max framerate in NVCP? My room is much cooler using it I get the same fps so I don’t really see a downside.

[RealNC]

I wonder how much more latency it adds using max framerate in NVCP? My room is much cooler using it I get the same fps so I don’t really see a downside.

Worst case is 1 frame. Usually less than that. I say that because if it was more than 1 frame, then it wouldn't be able to reach the FPS cap if the GPU was clocked too low

[RagingBuddhist]

I’ve played for a couple of days using NVCP to cap FPS and I have to say it’s been nice. Honestly it’s surprising to me nobody else has talked about this much. The only part I don’t really understand is why GPU usage is higher even though power usage and clocks are lower.

[RealNC]

I’ve played for a couple of days using NVCP to cap FPS and I have to say it’s been nice. Honestly it’s surprising to me nobody else has talked about this much. The only part I don’t really understand is why GPU usage is higher even though power usage and clocks are lower.

GPU usage is calculated using the current clocks. If the clocks go down but the same amount of work is to be done, usage goes up. Basically, if at 2GHz usage is 40%, then at 1GHz that would mean 80% usage. (Although there's more factors here, it's not really linear, but that's the gist of it.)

The same goes for the CPU, btw. If you force a power saving profile that locks CPU to low clocks, you will see CPU usage go up. Same if the CPU clocks down with the "balanced" profile when there's not much load.

[RagingBuddhist]

Thank you for your responses you seem very knowledgeable, and I’ve learned a lot from this. Now it’s time to get back to gaming without stressing over performance, hopefully.

[daemonjax]

rtss let's you use a floating point for framecap, instead of just integer values. Comes in handy. Plus with rtss you can choose where it does its framelimiting (at presentation, or wherever).