Anti-lag versus Ultra-Low latency - unexpected results

[Rroff]

Best if you watch the video really - Battlenonsense does an analysis of both with not the results he was expecting

 

[EastCoastHandle]

Hmm, very interesting. It's a balancing act between max fps and keeping the gpu below 97%.

 

[Rroff]

It will need further verification and testing and I can't see there being an easy way to implement it outside of games with very consistent performance but interesting how things like that can go uncaught sometimes.

 

[EastCoastHandle]

I have to wonder what the results would be at 1440p 100% scaling.
Regardless how high your scale the resolution at 1080p it always looks better at 1440p (at least for my monitor).

 

[Randomface74]

cool video

 

[BlueberryPie]

tbh it doesn't surprise me, i never noticed an improvement at higher FPS that was maxing out my card.

 

[shankly1985]

I can not say on the numbers I got on Squad. But I can 100% say I noticed the game feeling more snappy when moving the mouse around, and being a very demanding game at 1440p my VEGA 64 was pegged 100% all the time.

Edit

Its also worth noting my frame rate while playing is anywhere between 60 and 120 so maybe that has something to do with it.

 

[Nasher]

So basically there needs to be some overhead on the GPU for it to work properly.

 

[Grim5]

So basically there needs to be some overhead on the GPU for it to work properly.

To be expected and is something I've noticed in the past as well. Outside of input latency - chuck up a frametime graph and you'd see the same thing happening - graphics cards don't operate optimally at 100% load, there needs to be some balance between the CPU and GPU to get optimal performance - if either the GPU or CPU is creating a bottleneck you're going to have a bad time - and by using an unlocked framerate and letting the GPU run a 100% load you're creating a GPU bottleneck which is no good.

In most games, I play around with the settings extensively to keep the GPU from running into the 90's for usage - now we know it creates a smoother lower latency experience but also results in a more efficient gpu that draws less power per frame.

 

[Rroff]

I guess this shouldn't really surprise me at all - for a long time I used to go a little overkill on the GPU side but cap at 125 FPS (as I was used to it from Quake 3) and always seemed the best for responsive gameplay.

I have to say though it isn't something I've seen emerge as a consistent result from my own frametime experiments back in the day, or I'm sure I'd have noticed this by now, so might be something else going on here as well.

 

[shankly1985]

It seems they is a lot more to this.

 

[Rroff]

Some interesting results there - though somewhat an aside to what Battlenonsense was focussed on some quite useful data on what happens generally with higher framerates, capping and latency some of which isn't intuitive at all.

Different frame rate capping methods work quite different as well - some will measure over a second and once you've hit enough frames hard limit which can have odd results if you get lots of frames very fast in the first half of a second and then a more even rate in the second half, others will use some tolerance or percentiles to spread it out better with looser adherence to the cap and others will enforce a minimum wait between frames being generated - which can sometimes mean performance is reduced over a whole second but generally feels smoother and so on.