Anti-Lag Thread

What do you mean?
I posted the bench screenshot.
It doesn't show?

Here is a direct link.
Sometimes PrntScr gets blocked by adblockers when using it directly on forums.
https://prntscr.com/oopucf

 

oh I see now. thanks

---------------
Just an update now that I have proper voltage on ram.

 

Hey, hows it going?
If I were to enable HPET in win10 would it have any kind of effect on RAL?

 

No - I use QueryPerformanceCounter and it is not dependent on HPET AFAIK...

 

Good to know thanks. I understand that if you don't enable or disable it in win10 it's default is more like a default "standby mode". It won't turn on until an app, driver, etc request it's use. I was just wondering if RAL actually did.

I also noticed that Nvidia got on board with it as well. Congrads, you made them angry, LOL.
I can tell to because even though they offer it they down play it as just a "reduction pre-rendered frames".

 

Yup - NV implementing it is a personal win for me Of course their implementation is lacking at the moment - they need to go through the same iterations I did to make it better...

About prerendered frames - this is a self-sustaining urban legend, which refuses to die. NV are not saying that ullm is that - it’s all some stubborn users...

 

Sigh, one day they will get it. One day...
----
I'm a bit confused to the "different options" that Nv offers though. From my understanding either you want it on or off. IE: latency reduced to the best it can work. Or, disable it.

 

It’s not that simple. The regular NV low latency mode is indeed “force 1 prerendered frame”. On AMD this is the default mode, but on NV having more prerendered frames might improve FPS in some cases. As for ULLM (or RAL) - this incurs extra CPU power consumption, as well as sometimes reduced FPS (although still improved latency). For RAL I believe we mostly fixed the FPS reduction issues (at least on DX11) - but NV still has that road ahead of them...

 

Ah, I see. Thanks for the clarification.
BTW, I left a message for you under "conversations" not sure if you can see it on your end.
Click/look at your user name in the top right corner of the website. You should see a red icon indicating a message.
I'm note sure how it looks if you aren't using "desktop view".

 

who knows how to switch off anti-lag mode via registry? it seems by default anti-lag mode is ON at boot. I have A4Tech mouse with Bloody software and i can see weird behavior of my mouse. I think Adrenalin software overrides Bloody software settings and it looks like it changes mouse CPI rate. If i press Alt+L either close Bloody software and then re-start it is fine. Sometimes it happend in games(espacially Far Cry Dawn). So i would prefer to switch off this function for beginning. is it possible?

 

Question:
Are you using MarkC or other registry tweak to remove acceleration?

 

Nope, but i think i should play with that later. Thanks.

 

@HiAlgoBoost Does "Anti-Lag" use some sort of "Dynamically" sized flip queue? I'm wondering why it's not enabled by default and how it's different from the old options of simply changing the flip queue size.

We know it reduces input lag over default, but historically reducing the flip queue size could either reduce overall fps (slight to meaningful depending on game) and/or cause more discernible hitching/stutter so it's important to know if the new Anti-Lag setting retains these historical drawbacks.

Generally speaking there's a reason why the default is what it is for this sorta thing was my understanding. I did see on hardware unboxed that antilag does reduce overall fps in the same way more or less that reducing the flip queue size does, but they didn't mention or really even seem to take a look at hitching/stutter/smoothness with antilag on vs off.

I currently have an nvidia card myself and their own Ultra low lag mode causes hitching/stutter out the whazzoo in many games I test (with uncapped fps at least).

 

Limiting flip queue size to 1 hardly ever reduces performance, you can test this also on AMD e.g. with DXVK API wrapper. It also reduces stutter instead of increasing it, this is very noticeable e.g. in Crysis 3 in GPU limit.

 

I haven't experienced any hitching or stutter when using anti-lag. I believe it works by matching CPU to the GPU to make sure the GPU isn't being fed too many frames at once, therefore reducing the GPU bottleneck and making them work more closely to each other.

 

@HiAlgoBoost Stuff like this is a massive UX problem that stops you from getting the important message across. It creates confusion instead of impressing people. I'm falling in love with Unicode misc characters (not emojis ), why not make a really clear indicator of whats going on?
Like this: "FPS→AL-FPS" which will look like: "42→47"
Now the reliability of the reading is on you, but I feel it's the polish that gets nVidia so much further than AMD/RTG.

On-topic: I will update the drivers and post my findings tomorrow (mainly CSGO and couple others)

 

Has anyone seen this yet?



He claims he not only gets less input lag without anti-lag, but even less when capping the frame rate far lower than just leaving it uncapped and getting as many frames as possible.

How is this possible? I always thought lower frame rate meant MORE input lag, not less.

 

It's called "backpressure". You get it when using vsync, and also when the GPU is fully saturated. In both of those cases, the GPU can't render anymore, so there's a buffer pileup. With FPS capping however buffers are kept empty and there's no bottleneck anywhere.

Lower FPS makes little input lag difference. The difference between 144FPS and 100FPS for example is just 3ms. The lag reduction from FPS capping is much larger than that. This makes gsync/freesync very useful, since if you have that, you can cap your FPS to whatever you want and get extremely low lag.

Capping your FPS to get rid of lag has been known for years now. Me and others around here have been advocating it for a long time. I'm very surprised battlenonsense didn't know about it already. It really is old news.

 

I don't think you understand. He was saying he was getting less lag when capping his frames to a lower frame rate, rather than letting his frame rate be unlimited with vsync off, so there was no buffer build up either way. He was completely uncapped, vs capping in game.

See this screenshot.


Lower input lag with both lower frame rate and vsync off across the board and it's funny that the lag is even less than with anti-lag on. When he specifically caps the frame rate, the GPU usage goes down and the lag goes down as well.

I don't think this has to do with multiple frames being stored in the buffer. I've always been told that more frames=less lag no matter what, but what he's proving here is 60fps cap (not 60hz vsync) is giving him less input lag.

 

Full GPU saturation causes backpressure When you use vsync, the GPU is blocked "artificially" from processing fast enough, which causes backpressure. When full GPU saturation happens without vsync, the GPU is blocked naturally because it just can't do any more processing. In both cases, there's a GPU bottleneck and that causes a pile-up because the GPU can't process the pile-up fast enough.

Well, yeah, but the lag difference is much smaller than the lag caused by backpressure. For example, going from 85FPS to 60FPS causes an average lag increase of 2.5ms (5ms worst case)*. But if at 85FPS you get full GPU saturation with a resulting lag increase of over 30ms that you can eliminate by capping to 60FPS, then it's an overall win (a big one.) You increase lag by 2.5ms average due to lower FPS, but reduce it by 30ms at the same time due to removing backpressure.

* 2.5ms average and 5ms worst case comes from the fact that worst case is the bottom of the screen (5ms), best case is the top of the screen (0ms), so the middle of the screen is the average (1/2 that of the bottom.)