Help me understand those graphs. So they are interpreting the SD as almost like “lag”? So the averages are basically the same (which is very surprising for me 60 vs 360 fps), but the SD is significantly smaller in the higher fps groups. Does that mean the lower fps has higher latency spikes? The averages are basically the same. I’m used to reading medicine literature where they actually give you stats to see if the differences are significant lol
As to the Nvidia graph, I would want to see what the y axis means and how latency (I’m assuming that is what is above the fps number?) was chosen or obtained. I have a strong suspicion that graph is misleading.
And yea I get 60 to 360 will have a “large” reduction in latency. I still would argue 20ms is meaningless, but I was just showing you what Nvidia’s data regarding 60 and 360 hz and latency was. It could be mislabeled, incorrectly analyzed, involve misleading methods, etc but we don’t know that so I think thats a moot point to make. I do, however, appreciate you scrutinizing it
Edit: hmmm ok it appears they artificially adjusted latency for the experiment. Not entirely sure what that means, but it looks like they forced the averages to be the same (ish) and tested the fluctuations to determine how fps affects latency? That’s way over my head so I don’t understand but I still don’t see a meaningful difference in user experience. Even 30ms is nothing.
0.03 seconds is astronomically small that it’s hard to even imagine that time frame. Basically happens 30 times a second? Idk I know I don’t notice anything over low 100s so I just keep it at 140 fps even though my monitor is 240
That specific part of the demo was showing latency normalized between different framerates artificially, I just wanted to show that their internal terminology is poor and they kind of use different terms and different times which is really annoying.
This article is a little older, and they’ve gotten better about this recently.
Here’s another slide that is showing player performance at “low, medium, and high” latency at different frame rates.
The point of this study was to determine whether latency or the perceived smoothness of higher fps independent of latency was more important. It’s kind of a moot point to the end user because in the real world they go hand-in-hand in most circumstances outside of frame generation. Frame generation wasn’t a thing in the consumer space back then.
The GamersNexus engineer interview I linked is still the best explanation of the full end-to-end system chain thus far.
Some parts of the chain can be reduced or eliminated, but you always have peripheral input latency, CPU render, GPU render, display. These cannot be avoided. You have to add all 4 to get the total. So, if CPU render is 3ms, GPU is 6 ms, and display is 2.7, and the mouse is 2ms the total can’t be lower than 13.7ms.
In the future, APU’s could perhaps better parallelize CPU and GPU render so they overlap, but we aren’t there yet.
Hey I love data! Those graphs my brain can work with.
No offense, but posting a video of people talking about it without them actually “teaching” it just doesn’t help me change my view on something. I have too many questions and I need to know the details. Some people might like the Gamers Nexus video, but it just makes me have more questions since they don’t actually explain things on a “data driven, evidence based” level.
It’s been an interesting conversation! I learned about system latency tonight
That last graph titled “refresh matters” seems to not show what the title says though in any meaningful way. Let me know if I’m interpreting it wrong…
So they had users track…something? It went in a straight line, strayed, or jumped around and they had to follow it? While the latency set definitely affected the “time to completion”, it really looks like the fps didn’t do much at all? The biggest difference was with the “jump cohort”, but it looks like it was maybe 0.2 seconds difference between 60 and 360 fps? Sorry it’s hard to zoom in on my phone lol
YES! This was is much more my cup of tea! I’ll watch it tomorrow fully since it’s bedtime now.
I watched a minute of the “more fps = less latency” data though and I still feel like the latency reduction you gain from fps increase is pretty meaningless. 60 -120 seems pretty good, but after that it’s so tiny! I liked how the higher fps results in a tighter SD though. Kinda stabilizes the latency spikes or whatever. Pretty neat! Thanks for sharing that
Edit: ok maybe “meaningless” is the wrong word lol. I don’t think I can notice a difference in that reduction of latency
It’s also worth noting that in my calculations, the final value would represent the “longest” value on that chart. The reason the lowest can be so low and also lower on 144ps at 60hz than 360fps at 60hz is just luck. The gunfire just happened to better sync up both with input capture on the CPU/GPU render while the line the display was rendering also happened to sync up to that position on screen, so display latency and peripheral latency were reduced in that instance to almost nothing.
As you noticed, the results form a much tighter spread with high fps, high refresh rate, so what you experience moment to moment is much more consistently closer to that lowest value.
I’m critical of things when they don’t make sense to me. Criticism isn’t immune to scrutiny
This has to be the 50th topic made regurgitating the same thing over and over again. It’s just boring at this point. Everyone agrees FS is super slow, lied about how the game was going to be, and doesn’t say a dang thing to us. I’m at just boring at this point to have the same conversation over and over again.
Instead, I got taught about system latency yesterday, which is a tangent on DT’s poorly optimized engine.
I heard some doomurs about one of their main auto desk stingray engineers retiring. Could explain the constant revision errors we get with Darktide, not that it wasn’t a thing in VT2 they at least figured out that it was happening and admitted to it at one point. Equally likely the studio environment is more like a game of telephone than a development procedure, with some glowing revelations about the process like ‘we had no plans for our game we said we’d have content for every 3 months other than woo vacay time’.
Yea every week that goes by I’m like “definitely next Thursday will be a comms link”….it is definitely odd. They have to know their behavior is detrimental to the game. There’s gotta be more to it than what we think
My experience was the same as the TC: level several classes to 30 and instantly trip and fall into a bottomless pit of disappointment. The more you engage and learn about the systems, the worse it gets.
In the end, the persistent bugs and lack of polish are the final nails in the coffin.
Fatshark won’t do anything if we don’t constantly and loudly make complaints. If you don’t like a thread, mute it. I’m rightfully upset about the state of the game and I’m voicing this.
Obviously, the best way to get them to do something is to stop playing the game entirely. When they see those numbers drop, if they decide to pull the plug, no big loss IMO - I’ve gotten all I’ll ever get out of this game and I don’t see myself touching it again for the foreseeable future. But if they actually do something, then the community will have succeeded, and we can hopefully move on to greener pastures and a better video game.
People’s reaction time ranges from like 100 (top .1%) to 400ms, with the curve peaking around 200ms. Even at the absolute max of 100ms, it’s a little over five times the refresh rate at 60hz; you’ll see five frames at 60hz before the fastest person should be able to make a reaction. There have been studies which prove people get used to the refresh rate changes and play at statistically equivalent capability. The only thing high refresh rates do is improve the ‘feel’ and general experience, but actual performance impacts are all placebo.
FatShark has done nothing , despite you making complaints loudly. So, I don’t know why you think it is effective.
I agree with you guys in general. However, I’d advocate that the best way to address problems is to offer constructive feedback AND solutions.
I’m not saying don’t voice your opinion or whatever. I’m saying there is a right way and a wrong way to say whatever you want. If this way isn’t working (it clearly isn’t), then try a different one. Posting the same topic 50 times doesn’t add anything to FS. They know what you are describing. Contribute something different
We HAVE offered constructive feedback and solutions, and nothing has happened so far. I linked a thread with both of these things in my OP. Oftentimes the more vitriol an issue creates, the more urgent it is that it needs to be fixed - and crafting hasn’t been fixed for a little over a year running, now.
