Latency Pipeline, "Input Lag," Reflex, & Engineering Interview

One way to mitigate lag from missed inputs is to have extra frames rendered and ready, and when a missed input is detected, it can drop the frame that's about to be rendered, and pick the frame which matches the input best and use the input with that frame.

But that requires really smart rendering and physics and a gpu capable of rendering that fast.
 
Sponsored by NVIDIA, but this is really good lesson on gaming input latency. Have to say, I learned a couple things. Really good down to earth talk.
...
Maybe because it isn't sponsored by nvidia at all. Gamersnexus doesn't do sponsored content pieces (of the kind where the entire video's existence is paid for and bought by the sponsor) and doesn't even sell in-video ad space to nvidia (or amd, or intel).

Source: https://gamers.nexus/ethics-statements/advertising
 
?

You think that they're going to risk ftc scrutiny by lying about this or you're referring to them interviewing an nvidia engineer suddenly, somehow, making the video sponsored?
LoL
 
Why, lol to you, too. I am happy, though befuddled, to have amused you.

Back on topic:

One way to mitigate lag from missed inputs is to have extra frames rendered and ready, and when a missed input is detected, it can drop the frame that's about to be rendered, and pick the frame which matches the input best and use the input with that frame.

But that requires really smart rendering and physics and a gpu capable of rendering that fast.
Quite a few classic emulators can do this nowadays. It's even been integrated into retroarch, if I remember correctly.
 
Why, lol to you, too. I am happy, though befuddled, to have amused you.

Back on topic:


Quite a few classic emulators can do this nowadays. It's even been integrated into retroarch, if I remember correctly.
That’s just Kyle. You get used to him after a while. :D
 
Maybe because it isn't sponsored by nvidia at all. Gamersnexus doesn't do sponsored content pieces (of the kind where the entire video's existence is paid for and bought by the sponsor) and doesn't even sell in-video ad space to nvidia (or amd, or intel).

Source: https://gamers.nexus/ethics-statements/advertising

It does say it's sponsored - just not by nVidia:

1707695923871.png
 
Why, lol to you, too. I am happy, though befuddled,
The nuance between being paid by money or by access to content for free can become extremely thin.

The implication is not that there was a secret exchange of money.
 
GamerNexus like VideoCardz has integrity and does NOT accept under-the-tables from NVIDIA, AMD or Intel.

Anyone who claims otherwise will be banned, period.
 
Was the claim really under the table money ? Which indeed sound like pushing it.

Companies does not have to pay those people with secret money, how much does Intel paying for the flight and giving a tour of their fabs to LTT worth ? How much does Intel engineer giving hours of interview on Intel Arc launch worth for someone that make money with content ? How much free sample of their product ahead of the public launch so you can have a review before others worth ? You can pay them in content, content is money for them.

You can have a full video about input lag without any mention of AMD solution for it, without any spoke word-agreement about it, if it is Nvidia engineer doing a presentation for you. Back in the days a big newspaper product review section would have never talked with the people doing the products in any way, never accepted free sample, always bought them in a regular store, does not work like this in that world, the bar moved.

Military, Movie, hardware and game studio have bought positive coverage (or polite-understood-in context when it need to be negative because it is just too bad of a product) for years without under-the-tables deals, but with very open to be seen by everyone deals.
 
GamersNexus has sponsored content all the time. Sometimes it's sponsored by them (i.e. their store sales), other times its various brands. Just got to look at the video and video description.

That's just sort of how the YouTuber tech channel thing works....for better or worse.
 
A video about Latency and an Nvidia technology to reduce latency interviewing Nvidia team members.

Nvidia certainly had a hand in organising this. That's not a bad thing and I think it's kind-of obvious.

Like imagine going to a military base, being shown around by uniformed men and interviewing officers and someone gets upset about "YOU DIDN'T DISCLOSE THAT THE MILITARY ENDORSED THIS VIDEO" like, duh. They obviously endorsed the video, it isn't a shakycam video on liveleak linked from a news article. The military allowed it to happen.

But that said, if Nvidia had ANY say over what could or could not be said in the video outside of their IP, then that needs to be disclosed. If GN wasn't allowed to say anything negative or mention competing products, that needs to be disclosed.
 
But that said, if Nvidia had ANY say over what could or could not be said in the video outside of their IP, then that needs to be disclosed. If GN wasn't allowed to say anything negative or mention competing products, that needs to be disclosed.
Yes, it can be worst, but even in the best of best scenario, absolutely no condition was said, no looking at the video between it being published nothing at all, it is still something that can corrupt (the military being a good one, keeping access as a giant value, you do not need anything to be said for the coverage to be positive).

Here it is low value (it is not LTT people being flied oversea to visit factory that rarely have camera in them), but it is still nice content and just to avoid awkwardness, be polite, grateful, you will naturally tend to not ask or talk about competitor solution in a way you would not if a Unreal or Call of Duty game engine guy that knew a lot about it was interviewed and doing a similar presentation.

He cannot disclose it more than having being clear those a Nvidia people and in the current reviewer environment it will be considered not even ok but even good practice, but that something consumer report would never engage in
 
Steve is definitely more direct in his approach but to say this is not sponsored by nvidia because he was not paid doesn’t really say anything.

It’s effectively free marketing which will indeed have its perks in the future. Paid or not. It could just be access for future content.

I don’t see anything wrong with it frankly, it was an informative video and I agree with it since I mainly look at 1% and .1% lows with all of my hardware reviews. I could care less about the average frame rate. Latency/input lag and an overall smooth experience are of far more importance to me.
 
Sponsored by NVIDIA, but this is really good lesson on gaming input latency. Have to say, I learned a couple things. Really good down to earth talk.


View: https://youtu.be/Fj-wZ_KGcsg

My only question is, after they implement the change to the pipeline, such that the 10ms is removed, doesn't the process also add the 7ms that it was waiting in the front, to the overall total system latency? Net improvement 3ms?

Or, since it waits at the beginning, user input can still be accepted so that the latency the game experiences really is reduced by the full 10ms?

I get that the tweaking of the pipeline timing is likely an overall good thing, just wonder if they were overstating the improvement. I suspect it varies based on multiple factors. I like that they said being CPU limited is preferable to being GPU limited. I'm definitely CPU limited.
 
My only question is, after they implement the change to the pipeline, such that the 10ms is removed, doesn't the process also add the 7ms that it was waiting in the front, to the overall total system latency? Net improvement 3ms?

Or, since it waits at the beginning, user input can still be accepted so that the latency the game experiences really is reduced by the full 10ms?

I get that the tweaking of the pipeline timing is likely an overall good thing, just wonder if they were overstating the improvement. I suspect it varies based on multiple factors. I like that they said being CPU limited is preferable to being GPU limited. I'm definitely CPU limited.
In the example provided by Nvidia (reflex is dynamic and adjust according to framerate, so they used 100fps/10ms only for illustration), the time spent in render queue is also 10ms, so holding frame for 7ms, then spending 3ms for CPU to ready image for render queue, means that next frame can enter render queue ready for the GPU to render it while it processes the previous frame. So the 7ms wait time is not added to the total latency. The frame is just waiting before CPU starts working on it, instead of waiting in render queue. As an added bonus, it gives more time for user input to be carried over with the image through the pipeline.

The issues they described in this video are about render queue buffering images because GPU dont have enough overhead to render frames as soon as they are prepared by CPU. They didnt say its preferable to be CPU limited, but to be NEAR GPU limited. You want high FPS if you are concerned about input lag, because in their example they use 100fps as basis. That means each frame is worth 10ms of time on different stages in the pipeline. The latency is reduced if lets say 200fps and each frame is 5ms. Being near GPU limited means that CPU can setup many frames and GPU have enough headroom to render them as they arrive instead of them being buffered.
 
Back
Top