FYI: PLEASE keep in mind that turning H.A.G.S off has had some other complications besides just FPS issues, most notably issues with the responsiveness of certain applications and secondary monitors. If you have any issues with those, please do leave it on. Let me know your experiences below and thank you for watching! (And Saint, I don't hate you, I just have a new thing that I need to strive to figure out lol)
@@klemintime1452Is it possible that the Tarkov.exe performance regulation in combination with HAGS - ON may play an indirect role after all? I changed the GPU performance mode setting to the lowest setting - controlled by Windows - and suddenly I get much higher FPS. Before that, I was selected on - Max Performance. In this mode there was no difference whether I play in 1440p or 1080p, the fps was relatively the same. But now I notice a higher performance difference. Could you give it a try?
System: 7800x3d and 4070 super OC edition on windows 11 Method: Streets offline mode with scavs as normal and bosses turned off, look at 1 spot near the intersection at the top right of the pinewood complex. HAGS on REFLEX on (on+ boost no difference) 30ms latency HAGS on REFLEX off (RESTART NEEDED) 6ms latency with better 1% fps lows HAGS off not tested, don't really want to change to much settings from stock, and I believe DLSS and frame gen require HAGS on. Happy to have come accros this video, I have much lower latency now :)
@@klemintime1452 The latency's were very very stable, even when I had 30ms it wouldn't go lower than 27 or higher than 33, now I have latency's of 2ms up to about 8ms, I haven't tested online raids yet though. I'll update this comment.
The main thing is that reflex ON (Boost is only when you dont have prefer maximum perfomance in NVCP to ensure you dont downclock your GPU when there is nothing to do) will only kick ON when you are in GPU bound scenario (thats what it was created for) as its mention in the post you showed. It acts as variable FPS limiter which is a bit lower than your max in game FPS (nvidia says it should be max 3% lower) to ensure that your GPU is not Fully utilized bcs you dont want to backpressure you GPU (so you want your render queue always empty) - thats whats create most input lag. Reflex mode will also takover the LLM so there is no point of testing this together. You have showed that you were GPU bound in Streets - How much? Your GPU were utilized >95% all the time? My theory is that when you have really powerful CPU and older GPU you can run tarkov with lower CPU usage(all cores) and really high GPU usage(98-100%). You can enable reflex - this will lower your overall FPS but you will get better latency so GPU usage will be lets say 95% and in that scenario if you start offloading HAGS to GPU it can propbably create some latency/microshutter/fps issues. However most people in EFT are CPU bound and a lot of people run Reflex ON even when they have really powerfull GPU which is sitting in game about 40-50% usage so there is almost no reason to have it on bcs probably in any map you wouldnt hit that 99% mark so most people when they enable HAGS and REFLEX they dont see any difference just like you. So when you are in GPU bound scenario(>98% GPU USAGE) you probably want to run only REFLEX ON and HAGS off. Sorry for my english.. I am not a native speaker.
My reply never sent, but long story short your theory is likely correct. Some people on higher end systems have also ran HAGS with LLM on for similar results. This will essentially just limit the render queue to one frame (as you probably know) to help reduce latency without as heavy of a synchronization as reflex attempts to do. Thanks for commenting!
Theoretically HAGS should help especially in games like tarkov by offloading some CPU tasks onto the scheduling hardware of the GPU. That being said tho, I doubt reflex and HAGS are directly correlated, it's more about the fact that reflex by itself puts a lot of overhead on the CPU and that's just the way it works since it tries to limit the CPU's fps output right on the brink of what the GPU is also able to output and since we're talking about tarkov where the CPU is usually maxed out especially on just single thread. The lack of a render queue and the dynamic cap reflex puts on the CPU might be detrimental since even a slight further increase on CPU load by the game can cause lag and frame time spikes. In my testing using HAGS on and reflex on, not boost, produces a generally lower render latency, although it's definitely more prominent to stutters since I'm usually CPU bound. A good balance I found between fps smoothness and latency for me personally is enabling HAGS, disabling reflex and setting NV low latency mode for tarkov just to "on" so you can limit the render queue to just 1 pre rendered frame. This ensures a better frame stability less prone to stutters and also produces slightly more frames in certain scenarios due to the CPU being a bit more relaxed without the back pressure from reflex. At the end of the day, honestly it's just the game that is a mess when trying to utilize the CPU during online raids. The load from player and server synchronization is insane and the game's not able to balance it properly onto multiple threads
One detail is that the OP of that post never mentioned whether he was GPU or CPU limited, the post about reflex conflicting with hags mentioned this is only during a gpu bound scenario.
That's true, but I can assume that he is definitely CPU-bottlenecked on streets at least. I saw that too though, which is why I tried to test for both.
Hmm I’ve been struggling with consistent frames between raid to raids. One raid I’ll be getting shitty gpu usage and the next I’ll be getting way higher usage and the max core frequency clock for my gpu. It’s a massive difference. Sometimes on light house I’ll load into a road and be getting less than 50-60 fps and even the 40s while other light house raids it’ll be mid 100s make no sense.
I for one had hags disabled at the start of this wipe, I’ve always had reflex on, and had a increase in FPS once I turned hags on, I’ll do some more accurate testing and let you know.
System : 7800X3D 4060ti 32gb ram windows 11 Method : Load in to Streets of Tarkov offline raid with scav as online bosses off. Go to Concordia and walk around for two minuites. Results, HAGS on REFLEX on : Avg 107 fps Max 113 fps HAGS on REFLEX off : Avg 106 fps Max 113 fps HAGS off REFLEX on : Avg 91 fps Max 132 fps HAGS off REFLEX off : Avg 78 fps Max 120 fps I felt like there were few micro stutter with on/on, but none once I turn either one off(not sure, maybe scav spawning or something). Frametime is much better with reflex on. Max fps were higher with HAGS off for some reason but I did not notice the difference. Afterall, I'll go with on/on.
Id be more interested in the clash of reflex with the nvcp low latency mode setting. Ive heard if you want to reduce latency, turn on reflex and turn low latency mode off. (Their reasoning was that the two settings are redundant and clashing with both enabled) running with reflex on and LL off has the best performance for me tbh
I did this rabbit hole few years ago. thought I found correlation between reddits performance rants and date that windows update set the hags on by default. Tried to find if there was difference between on and off ; did not have any effect on my system.
If it helps, I've done many tests and experiments, from many registry tweaks(priorities, polling rate, quantum latency, etc.), latency tweaks, MSI mode, BIOS advanced settings, timings, undervolt/overclock, nvcp, extensive windows debloating, process lasso, paging file, etc. I'm running a 5800x3D(PBO2 tuned) with a RTX 4090, 32GB Ram CL14 @3600Mhz(1800FCLK). Always trying to squeeze more performance and reduce (or eliminate) the annoying microstuttering I sometimes get only in Tarkov, I saw the aforementioned post in reddit a few weeks ago and I tried disabling HAGS, which *almost* elminated the microstutter completely. I'll do more testing on the weekend when I have time and report back.
I have a 7700k, 1080 ti, and 16gb playing at 1440p. I tried lighthouse offline and streets online. I turned on/off HAGS for lighthouse and saw no difference. Tried turning on/off reflex during streets and saw no difference.
I had to turn hags off because I use sunshine/moonlight to gamestream tarkov to my phone for strength training amd factory scav runs. Hags has caused crashes and stuttering on that software. I have been running it off with reflex on for over a month. I have no hard numbers, but I will 100% notice if my fps changes when I turn hags back on so I'll test for you on Monday.
System : I7 12700k @4.9 3080 32gb ddr4 I had hags and reflex on. Customs offline raid with bots and 120fps cap I was getting to 120fps when I looked to the outside of the map. Otherwise I was between 90 and 110. Hags off I'm running constantly 120 fps. No drops. And Ive seen in hw monitor that my gpu was at 0%, memory around 50% buffer at 30%. With hags off my gpu finally do something and runs at 50% in hw monitor.
I think that GPU usage % is just a bug, but Im glad to hear it helped you! For once, the reasoning behind this completely evades me, and the theories I do have barely make sense lol. Appreciate the followup :)
Hags works better for me on windows 11. When I enabled it in windows 10 there was quite a bit of instability in certain games. I wanted it enabled because of dlss and frame gen. Also latency is better for me with dlss on and frame gen if available.
I saw basically no difference in frame timing or any metric of FPS with all combinations of NVCP low latency, in-game reflex or windows HAGS on or off. I was running with these all on before and will continue to do so but at least I know. 135fps average, 7.5ms, 1440 UW main, 4k second, gysnc on, streets, low texture streets on, high overall, offline mode.
10 raids in and everything was good but felt like i needed to restart the game, i disabled hags kept reflex on my highs feel lower but my lows higher and more smoother through out all 10 raids until the 10th 1 and i only play streets, been running hags off with reflex now since yesterday and it feels like places i would get high fps smooth gameplay lagg now and places i would lagg before are smoother now, i main streets only and have 1 server selected incase this info helps us clarify this tweak
For one thing, something that helps make my FPS more consistent is by disabling CPU core 0 which was causing some random stutter you guys can try it definitely help with my stutter in this case without physical core checked*
I’m no tech expert, but on my old I7 7700k system I felt the game ran better with HAGS off. Not saying it did as I never tested it fully, definitely didn’t compare FPS. It just felt smother. Always had Low Latency on aswell in game
you have to restart your pc once and open the game twice, and load some offline raids in different maps. State your hardware and fps with reflex on and off, hags on and off.
I tested this with my 13900k and 4090 and only map I saw performance improvements offline was factory from 540 to just over 600. The rest was just latency improvements from like 12ms down to 0.8ms
Hello Klementime, I just came across your channel this wipe as I enjoy your thorough explanations. I do have some questions as I can’t seem to figure out a suitable setting for myself. I did try to join your discord from TH-cam but the invite link is no longer valid. Would you be able to provide an updated one? Thanks!
The problem with such posts is that the guy could have some weird system configuration where - against normal behaviour - HAGS off helps with performance. Another - simpler - explanation is: that guys test methodology was flawed and he came to the wrong conclusion. Say doing scav raids on Streets, spawning in early with lots of PMCs on the server, then doing the same with HAGS off and getting into a late raid with fewer people. He then concludes "HAGS bad". I can do the tests you suggested at some point, but I doubt I can reproduce it (had HAGS on with Reflex, then switched to HAGS off because of that reddit post. Didn't feel any different but didn't do any methodic testing)
Exactly my thinking too, but I know this guy since he has been in the community for a bit. I'm fairly certain he did some decent testing, and that combined with the mass of people saying stuff about it further added to my confusion. Appreciate you looking at it :D (and yeah, this is partially the pain of doing these tests)
Just eyeballing the numbers as I play, for me there's a massive increase in latency when both HAGS and Reflex are enabled. [RAID 1 STREETS ONLINE HAGS(ON) REFLEX/LOW LATENCY(ON) (Render Latency 22-24ms) GPU Utilization around 60%, CPU around 30%, FPS 80-110] Game restart after disabling reflex [RAID 2 STREETS ONLINE HAGS(ON) REFLEX/LOW LATENCY MODE(OFF) (Render Latency 3-6ms)] No noticeable difference in GPU/CPU utilization or FPS] Game restart after enabling reflex (Enabling Reflex mid raid didn't seem to change anything.) [RAID 3 STREETS ONLINE HAGS(ON) REFLEX/LOW LATENCY(ON) (Render Latency 20-24ms) No noticeable difference in GPU/CPU utilization or FPS] Game restart after disabling reflex [RAID 4 STREETS ONLINE HAGS(ON) REFLEX/LOW LATENCY(OFF) (Render Latency 3-6ms) No noticeable difference in GPU utilization or FPS]
Interesting- at least other people are seeing that too. Also I never thought that skibidi toilet would be in my comment section; I've officially made it
😅@@klemintime1452 Yeah! It's definitely interesting and it seems that the Low Latency Mode in NCP on a driver level is NOT causing an increase in latency. So like previous patches when Reflex was broken, maybe people should go the same route and keep it OFF while having Low Latency Mode enabled on a driver level.
Good question. A chunk of people within the HW community dislike Geforce Experience because of its forced login and telemetry that runs in the background. Its a pain and is outdated compared to other GPU software (hence why Nvidia is coming out with the NVidia APP to replace it soon!)
@@inreisen i know thats the only downside, i have to tick iff vsync after every raid because pc will explode 😂. But def gameplay in raid feels smoother, not just loby. pls let me know your results im curious.
@@juremandic9058 you can set an FPS cap via the nvidia controle panel for the tarkov.exe, put it high enough so ingame you're not limited by it and FPS in menu's will be under control.
I'm sorry for the wall of text I'm about to submit here but everyone is so busy looking at this from the GPU side and not from the CPU side. This is more likely a CPU ''issue'' than a GPU issue, why? Tarkov is known as an extremely CPU bound game and not only Tarkov relies on a single/high priority fast CPU thread. Everyone is so focussed on GPU bound or not GPU bound scenarios that they are completely ignoring CPU Thread scheduling/prioritizing and coordinating. WDDM(Windows Display Driver Model) is responsible for managing multiple applications that is giving more work to the GPU, in order for this to work as smooth as possible it relies on a high priority Thread on the CPU that is scheduling, prioritizing and coordinating the workload submitted by multiple applications. Yes the GPU is repsonsible for the rendering part but the CPU is carrying the load of preparing AND submitting orders/commands to the GPU. The CPU submits commands in clusters, this increases the overall performance in frame rate but this also increases input latency. A technique called frame buffering has become common where the CPU submits these clusters in batches because doing this 1 frame at a time is simply not efficient. So, when a user clicks a button nothing will happen untill the GPU gets to the next cluster of submitted frames, the larger the cluster the longer the wait(potentially). HAGS offloads works from that high priority CPU thread and gives it to the dedicated GPU based scheduling processor. The potential framerate improvements is in CPU bound games since some of the work is offloaded from the CPU, this effect should be most noticeable on lower end CPU's, now this should be most noticeable on CPUs that run at 100% load but even Tarkov won't push CPUs to 100% but it will push that 1 CPU Thread hard. I believe Tarkov and the windows scheduler might be colliding in some form of way but this is just a theory of mine so far. If I'm right than 1 person cannot test and prove this theory as a lot of CPUs are different. Intel P and E-cores might be more efficient at this or it might have the opposite effect..
Good theory. As you said though, HAGS should *help* in a CPU bound scenario (theoretically) but instead the post claimed the opposite. Just really weird.
This might sound dumb but aren't you GPU-bound in your tests? Seeing DLSS increase your FPS significantly and afaik DLSS only improves frames like that when you're GPU-bound. H.A.G.S is more so for CPU-bound situations since it takes the load off from CPU to the GPU.
HAGS offloads some scheduling of GPU tasks to a dedicated part of the GPUs hardware- its not *directly* taking the load off. Also, some of the tests mentioned in the video (like the ones with DLSS performance) made the load on the GPU low enough for the CPU to become the bottlenecking factor.
the truth is it does improve performance, but everyone says it impacts performance so you turn it off and give them the upper hand, or am I the one lying, the internet man.
Sorry for the late reply, but according to a lot of people's experience, yes. The issue was in the video that I did not see something similar. I tend to run both with no issue, but it seems like it devastates some users performance, and I still cannot find out why.
Definitely in a CPU bottleneck. Even though you are less than 100% usage on your CPU, you can still be in a CPU Bottleneck (likely with one or a couple threads). If you were GPU bound, your GPU would be at 100% though. Since it isn't, its very likely a CPU bind.
Kinder have same problem i7 12700kf 30% load rtx 3080 30-40% load. My fps is fine 100 -140 fps except on streets edit was useing task manager and saw in one of you other video that its not accurate
It's all garbage, don't waste your time with these gamer myths. A buddy and I have identical configurations (7800X3D with the same RAM and we're both not even close to maxing out the GPU even though it's a different one). He has done NO optimisations. Meanwhile I've done it all. Buildzoid's primary and secondary RAM timings, HAGS, various other optimisations (at least the ones that don't look like snake oil, such as reducing % reserved for background tasks from 20 to 10% and such). Initially, he had 5% performance more than me CONSISTENTLY. I couldn't figure out why. Every single raid, in the same spot on the map, duo together, looking in the same direction and he had 5% more fps every time. But he was crashing every raid after 25-30 min. Turns out, his page file was really small and a Streets raid would begin with 22GB RAM usage but grow to 31GB+ over a 30 minute game and eventually crash. As soon as the increased his page file, crashes stopped but his performance dropped by... you guessed it... 5%. He now has IDENTICAL performance to me. We're talking within 1% every single raid consistently. And he's done no optimisations or other stuff. Just turned on DOCP or whatver its called these days. None of this shite makes a difference. Get a decent setup, do the main 2-3 fixes to optimise performance and then just enjoy playing and forget about it. Unless you wanna get 64GB of RAM so you can turn off your page file. Maybe that will make a difference, idk.
Do you really need 64 gb of RAM though? I definitely don't have those crashing issues with stock paging file and simply use ARC with Empty standby List running every 5 minutes.
@@klemintime1452 No, you don't need it, I was just thinking you won't need a paging file at all then. Or you can also create a virtual drive from your RAM and put the page file there. Because it seems the game uses it even if you have enough memory left. At least that's the impression I got.
Thanks for this, especially the ram timings thing. I was going to overclock my 3600 CL18 ram but at 30% of the process I kept getting no posts and stuff on secondary timing and I gave up. At the end only thing I changed from XMP was Gear 2 to Gear 1 and it made a noticable difference, like 5% by eyeballing the fps numbers before/after the process.
Though I think the application is fairly well made, the fact that you are FORCED to log in to access most features makes it a cheap way for them to gather data about their users. Secondarily, the telemetry they have running in the background has some overhead as discussed by Hardware Unboxed (if you google "Hardware Unboxed Nvidia Driver Overhead, " you should find it). It's a nice software suite though, don't get me wrong.
![[#2024MAMA] G-DRAGON - 무제(Untitled, 2014)+POWER+HOME SWEET HOME+뱅뱅뱅+FANTASTIC BABY | Mnet 241123 방송](http://i.ytimg.com/vi/Ox29z5Nf1Uk/mqdefault.jpg)
![[#2024MAMA] BIGBANG (빅뱅) - 뱅뱅뱅 (BANG BANG BANG) + FANTASTIC BABY | Mnet 241123 방송](http://i.ytimg.com/vi/EQsYfiPR9uM/mqdefault.jpg)
FYI: PLEASE keep in mind that turning H.A.G.S off has had some other complications besides just FPS issues, most notably issues with the responsiveness of certain applications and secondary monitors. If you have any issues with those, please do leave it on. Let me know your experiences below and thank you for watching!
(And Saint, I don't hate you, I just have a new thing that I need to strive to figure out lol)
idk if this matters, but you didn't have the Tarkov.exe in the HAGs settings
@@snowmandev That's a seperate setting; HAGS is a global setting and that setting is just for power plans as far as Im aware.
@@klemintime1452 thanks for clarifying!
@@klemintime1452Is it possible that the Tarkov.exe performance regulation in combination with HAGS - ON may play an indirect role after all? I changed the GPU performance mode setting to the lowest setting - controlled by Windows - and suddenly I get much higher FPS. Before that, I was selected on - Max Performance. In this mode there was no difference whether I play in 1440p or 1080p, the fps was relatively the same. But now I notice a higher performance difference. Could you give it a try?
Tarkov is going to run at the fps it wants to run at.
System: 7800x3d and 4070 super OC edition on windows 11
Method: Streets offline mode with scavs as normal and bosses turned off, look at 1 spot near the intersection at the top right of the pinewood complex.
HAGS on REFLEX on (on+ boost no difference) 30ms latency
HAGS on REFLEX off (RESTART NEEDED) 6ms latency with better 1% fps lows
HAGS off not tested, don't really want to change to much settings from stock, and I believe DLSS and frame gen require HAGS on.
Happy to have come accros this video, I have much lower latency now :)
Keep in mind it could have been run-to-run variance as well, so just keep an eye on it :)
@@klemintime1452 The latency's were very very stable, even when I had 30ms it wouldn't go lower than 27 or higher than 33, now I have latency's of 2ms up to about 8ms, I haven't tested online raids yet though. I'll update this comment.
In online raids I also get consistent 5ms latency
The main thing is that reflex ON (Boost is only when you dont have prefer maximum perfomance in NVCP to ensure you dont downclock your GPU when there is nothing to do) will only kick ON when you are in GPU bound scenario (thats what it was created for) as its mention in the post you showed. It acts as variable FPS limiter which is a bit lower than your max in game FPS (nvidia says it should be max 3% lower) to ensure that your GPU is not Fully utilized bcs you dont want to backpressure you GPU (so you want your render queue always empty) - thats whats create most input lag.
Reflex mode will also takover the LLM so there is no point of testing this together.
You have showed that you were GPU bound in Streets - How much? Your GPU were utilized >95% all the time?
My theory is that when you have really powerful CPU and older GPU you can run tarkov with lower CPU usage(all cores) and really high GPU usage(98-100%). You can enable reflex - this will lower your overall FPS but you will get better latency so GPU usage will be lets say 95% and in that scenario if you start offloading HAGS to GPU it can propbably create some latency/microshutter/fps issues. However most people in EFT are CPU bound and a lot of people run Reflex ON even when they have really powerfull GPU which is sitting in game about 40-50% usage so there is almost no reason to have it on bcs probably in any map you wouldnt hit that 99% mark so most people when they enable HAGS and REFLEX they dont see any difference just like you.
So when you are in GPU bound scenario(>98% GPU USAGE) you probably want to run only REFLEX ON and HAGS off.
Sorry for my english.. I am not a native speaker.
My reply never sent, but long story short your theory is likely correct. Some people on higher end systems have also ran HAGS with LLM on for similar results. This will essentially just limit the render queue to one frame (as you probably know) to help reduce latency without as heavy of a synchronization as reflex attempts to do. Thanks for commenting!
Theoretically HAGS should help especially in games like tarkov by offloading some CPU tasks onto the scheduling hardware of the GPU.
That being said tho, I doubt reflex and HAGS are directly correlated, it's more about the fact that reflex by itself puts a lot of overhead on the CPU and that's just the way it works since it tries to limit the CPU's fps output right on the brink of what the GPU is also able to output and since we're talking about tarkov where the CPU is usually maxed out especially on just single thread.
The lack of a render queue and the dynamic cap reflex puts on the CPU might be detrimental since even a slight further increase on CPU load by the game can cause lag and frame time spikes.
In my testing using HAGS on and reflex on, not boost, produces a generally lower render latency, although it's definitely more prominent to stutters since I'm usually CPU bound. A good balance I found between fps smoothness and latency for me personally is enabling HAGS, disabling reflex and setting NV low latency mode for tarkov just to "on" so you can limit the render queue to just 1 pre rendered frame.
This ensures a better frame stability less prone to stutters and also produces slightly more frames in certain scenarios due to the CPU being a bit more relaxed without the back pressure from reflex.
At the end of the day, honestly it's just the game that is a mess when trying to utilize the CPU during online raids. The load from player and server synchronization is insane and the game's not able to balance it properly onto multiple threads
thanks bro, its really working, get much less stutters with NV reflex on.
One detail is that the OP of that post never mentioned whether he was GPU or CPU limited, the post about reflex conflicting with hags mentioned this is only during a gpu bound scenario.
That's true, but I can assume that he is definitely CPU-bottlenecked on streets at least. I saw that too though, which is why I tried to test for both.
Hmm I’ve been struggling with consistent frames between raid to raids. One raid I’ll be getting shitty gpu usage and the next I’ll be getting way higher usage and the max core frequency clock for my gpu. It’s a massive difference. Sometimes on light house I’ll load into a road and be getting less than 50-60 fps and even the 40s while other light house raids it’ll be mid 100s make no sense.
I for one had hags disabled at the start of this wipe, I’ve always had reflex on, and had a increase in FPS once I turned hags on, I’ll do some more accurate testing and let you know.
what settings you using now and have you done it vice versa hags on reflex off
System : 7800X3D 4060ti 32gb ram windows 11
Method : Load in to Streets of Tarkov offline raid with scav as online bosses off. Go to Concordia and walk around for two minuites.
Results,
HAGS on REFLEX on : Avg 107 fps Max 113 fps
HAGS on REFLEX off : Avg 106 fps Max 113 fps
HAGS off REFLEX on : Avg 91 fps Max 132 fps
HAGS off REFLEX off : Avg 78 fps Max 120 fps
I felt like there were few micro stutter with on/on, but none once I turn either one off(not sure, maybe scav spawning or something). Frametime is much better with reflex on. Max fps were higher with HAGS off for some reason but I did not notice the difference. Afterall, I'll go with on/on.
will you put on vsync to unlimit the fps while turning reflex off?
@@samson1897 I put g-sync on but no vsync. I'm not an expert on this subject but I'm guessing my frame rate is not capped.
Id be more interested in the clash of reflex with the nvcp low latency mode setting. Ive heard if you want to reduce latency, turn on reflex and turn low latency mode off. (Their reasoning was that the two settings are redundant and clashing with both enabled) running with reflex on and LL off has the best performance for me tbh
IIRC LLM gets overrun by Nvidia Reflex in a given game. If Reflex is detected as enabled, LLM "gives way" to reflex.
I did this rabbit hole few years ago. thought I found correlation between reddits performance rants and date that windows update set the hags on by default. Tried to find if there was difference between on and off ; did not have any effect on my system.
If it helps, I've done many tests and experiments, from many registry tweaks(priorities, polling rate, quantum latency, etc.), latency tweaks, MSI mode, BIOS advanced settings, timings, undervolt/overclock, nvcp, extensive windows debloating, process lasso, paging file, etc. I'm running a 5800x3D(PBO2 tuned) with a RTX 4090, 32GB Ram CL14 @3600Mhz(1800FCLK). Always trying to squeeze more performance and reduce (or eliminate) the annoying microstuttering I sometimes get only in Tarkov, I saw the aforementioned post in reddit a few weeks ago and I tried disabling HAGS, which *almost* elminated the microstutter completely. I'll do more testing on the weekend when I have time and report back.
Appreciated; looking forward to hear :)
brother any update?
I have a 7700k, 1080 ti, and 16gb playing at 1440p. I tried lighthouse offline and streets online. I turned on/off HAGS for lighthouse and saw no difference. Tried turning on/off reflex during streets and saw no difference.
Thank you for the flashbang warning 🤝
🤜🤛 NP bro
I had to turn hags off because I use sunshine/moonlight to gamestream tarkov to my phone for strength training amd factory scav runs. Hags has caused crashes and stuttering on that software. I have been running it off with reflex on for over a month. I have no hard numbers, but I will 100% notice if my fps changes when I turn hags back on so I'll test for you on Monday.
System :
I7 12700k @4.9
3080
32gb ddr4
I had hags and reflex on. Customs offline raid with bots and 120fps cap I was getting to 120fps when I looked to the outside of the map. Otherwise I was between 90 and 110.
Hags off I'm running constantly 120 fps. No drops.
And Ive seen in hw monitor that my gpu was at 0%, memory around 50% buffer at 30%.
With hags off my gpu finally do something and runs at 50% in hw monitor.
I think that GPU usage % is just a bug, but Im glad to hear it helped you! For once, the reasoning behind this completely evades me, and the theories I do have barely make sense lol. Appreciate the followup :)
@@klemintime1452 maybe it is, but in other games, it showed normal gpu utilization. Anyways, runs smoother then before so glad for the tip
Hags works better for me on windows 11. When I enabled it in windows 10 there was quite a bit of instability in certain games. I wanted it enabled because of dlss and frame gen. Also latency is better for me with dlss on and frame gen if available.
I saw basically no difference in frame timing or any metric of FPS with all combinations of NVCP low latency, in-game reflex or windows HAGS on or off. I was running with these all on before and will continue to do so but at least I know. 135fps average, 7.5ms, 1440 UW main, 4k second, gysnc on, streets, low texture streets on, high overall, offline mode.
10 raids in and everything was good but felt like i needed to restart the game, i disabled hags kept reflex on my highs feel lower but my lows higher and more smoother through out all 10 raids until the 10th 1 and i only play streets, been running hags off with reflex now since yesterday and it feels like places i would get high fps smooth gameplay lagg now and places i would lagg before are smoother now, i main streets only and have 1 server selected incase this info helps us clarify this tweak
For one thing, something that helps make my FPS more consistent is by disabling CPU core 0 which was causing some random stutter you guys can try it definitely help with my stutter in this case without physical core checked*
I've had other viewers discuss this too- could help especially if there are a lot of tasks in the background that are hogging that thread.
@@klemintime1452 I believe thats the reason or maybe the core too slow it helps me with stutter for sure despite no FPS gain
I’m no tech expert, but on my old I7 7700k system I felt the game ran better with HAGS off. Not saying it did as I never tested it fully, definitely didn’t compare FPS. It just felt smother. Always had Low Latency on aswell in game
hmmm, good to note. Might be good to test on weaker systems.
thats seems like a lot of work, count me out!
you have to restart your pc once and open the game twice, and load some offline raids in different maps. State your hardware and fps with reflex on and off, hags on and off.
I tested this with my 13900k and 4090 and only map I saw performance improvements offline was factory from 540 to just over 600. The rest was just latency improvements from like 12ms down to 0.8ms
Are u still getting those frames ? Have the same setup but struggle with settings lately
Hello Klementime, I just came across your channel this wipe as I enjoy your thorough explanations. I do have some questions as I can’t seem to figure out a suitable setting for myself. I did try to join your discord from TH-cam but the invite link is no longer valid. Would you be able to provide an updated one? Thanks!
The link on the channel page was indeed expired- try this one too: discord.gg/zrdxtvKS6t
The problem with such posts is that the guy could have some weird system configuration where - against normal behaviour - HAGS off helps with performance. Another - simpler - explanation is: that guys test methodology was flawed and he came to the wrong conclusion. Say doing scav raids on Streets, spawning in early with lots of PMCs on the server, then doing the same with HAGS off and getting into a late raid with fewer people. He then concludes "HAGS bad". I can do the tests you suggested at some point, but I doubt I can reproduce it (had HAGS on with Reflex, then switched to HAGS off because of that reddit post. Didn't feel any different but didn't do any methodic testing)
Exactly my thinking too, but I know this guy since he has been in the community for a bit. I'm fairly certain he did some decent testing, and that combined with the mass of people saying stuff about it further added to my confusion. Appreciate you looking at it :D (and yeah, this is partially the pain of doing these tests)
Just eyeballing the numbers as I play, for me there's a massive increase in latency when both HAGS and Reflex are enabled.
[RAID 1 STREETS ONLINE
HAGS(ON) REFLEX/LOW LATENCY(ON) (Render Latency 22-24ms)
GPU Utilization around 60%, CPU around 30%, FPS 80-110]
Game restart after disabling reflex
[RAID 2 STREETS ONLINE
HAGS(ON) REFLEX/LOW LATENCY MODE(OFF) (Render Latency 3-6ms)]
No noticeable difference in GPU/CPU utilization or FPS]
Game restart after enabling reflex (Enabling Reflex mid raid didn't seem to change anything.)
[RAID 3 STREETS ONLINE
HAGS(ON) REFLEX/LOW LATENCY(ON) (Render Latency 20-24ms)
No noticeable difference in GPU/CPU utilization or FPS]
Game restart after disabling reflex
[RAID 4 STREETS ONLINE
HAGS(ON) REFLEX/LOW LATENCY(OFF) (Render Latency 3-6ms)
No noticeable difference in GPU utilization or FPS]
Interesting- at least other people are seeing that too.
Also I never thought that skibidi toilet would be in my comment section; I've officially made it
😅@@klemintime1452 Yeah! It's definitely interesting and it seems that the Low Latency Mode in NCP on a driver level is NOT causing an increase in latency. So like previous patches when Reflex was broken, maybe people should go the same route and keep it OFF while having Low Latency Mode enabled on a driver level.
why is it that downloading geforce experience as unthinkable, actual question because I dont understand why its seen as taboo
Good question. A chunk of people within the HW community dislike Geforce Experience because of its forced login and telemetry that runs in the background. Its a pain and is outdated compared to other GPU software (hence why Nvidia is coming out with the NVidia APP to replace it soon!)
Thank you!@@klemintime1452
HAGS ON, REFLEX OFF, VSYNC OFF IN CONTROL PANEL, ON IN GAME, this is the smoothest experience i get in game. 5800x3d, 3060, 32 gb 3600 mhz.
If you do that your lobby fps limit will be uncapped that's why it feels so smooth. But if you mean your raids are also smooth then I shall try it.
@@inreisen i know thats the only downside, i have to tick iff vsync after every raid because pc will explode 😂. But def gameplay in raid feels smoother, not just loby. pls let me know your results im curious.
@@juremandic9058 you can set an FPS cap via the nvidia controle panel for the tarkov.exe, put it high enough so ingame you're not limited by it and FPS in menu's will be under control.
Vsync on 🤦🤔 that's a first??
I know HAGS has issues with recording/streaming software, simple answer it was his recording/streaming software causing the issue but who know
I'm sorry for the wall of text I'm about to submit here but everyone is so busy looking at this from the GPU side and not from the CPU side. This is more likely a CPU ''issue'' than a GPU issue, why? Tarkov is known as an extremely CPU bound game and not only Tarkov relies on a single/high priority fast CPU thread.
Everyone is so focussed on GPU bound or not GPU bound scenarios that they are completely ignoring CPU Thread scheduling/prioritizing and coordinating.
WDDM(Windows Display Driver Model) is responsible for managing multiple applications that is giving more work to the GPU, in order for this to work as smooth as possible it relies on a high priority Thread on the CPU that is scheduling, prioritizing and coordinating the workload submitted by multiple applications.
Yes the GPU is repsonsible for the rendering part but the CPU is carrying the load of preparing AND submitting orders/commands to the GPU.
The CPU submits commands in clusters, this increases the overall performance in frame rate but this also increases input latency. A technique called frame buffering has become common where the CPU submits these clusters in batches because doing this 1 frame at a time is simply not efficient. So, when a user clicks a button nothing will happen untill the GPU gets to the next cluster of submitted frames, the larger the cluster the longer the wait(potentially).
HAGS offloads works from that high priority CPU thread and gives it to the dedicated GPU based scheduling processor. The potential framerate improvements is in CPU bound games since some of the work is offloaded from the CPU, this effect should be most noticeable on lower end CPU's, now this should be most noticeable on CPUs that run at 100% load but even Tarkov won't push CPUs to 100% but it will push that 1 CPU Thread hard.
I believe Tarkov and the windows scheduler might be colliding in some form of way but this is just a theory of mine so far. If I'm right than 1 person cannot test and prove this theory as a lot of CPUs are different. Intel P and E-cores might be more efficient at this or it might have the opposite effect..
Good theory. As you said though, HAGS should *help* in a CPU bound scenario (theoretically) but instead the post claimed the opposite. Just really weird.
This might sound dumb but aren't you GPU-bound in your tests? Seeing DLSS increase your FPS significantly and afaik DLSS only improves frames like that when you're GPU-bound. H.A.G.S is more so for CPU-bound situations since it takes the load off from CPU to the GPU.
HAGS offloads some scheduling of GPU tasks to a dedicated part of the GPUs hardware- its not *directly* taking the load off. Also, some of the tests mentioned in the video (like the ones with DLSS performance) made the load on the GPU low enough for the CPU to become the bottlenecking factor.
the truth is it does improve performance, but everyone says it impacts performance so you turn it off and give them the upper hand, or am I the one lying, the internet man.
Do hags and Nvidia reflex interfere with each other?
Sorry for the late reply, but according to a lot of people's experience, yes. The issue was in the video that I did not see something similar. I tend to run both with no issue, but it seems like it devastates some users performance, and I still cannot find out why.
Idk why but in GeForce it hardly ever shows my fps or fps%
What preset is the overlay set to in the performance settings? Mine is "advanced" for reference.
@@klemintime1452 advanced
Always says NA for the first 2 things. Think it's FPS and 99% FPS
my cpu and gpu are always at 50% and never above that, any help?
Definitely in a CPU bottleneck. Even though you are less than 100% usage on your CPU, you can still be in a CPU Bottleneck (likely with one or a couple threads). If you were GPU bound, your GPU would be at 100% though. Since it isn't, its very likely a CPU bind.
Kinder have same problem i7 12700kf 30% load rtx 3080 30-40% load. My fps is fine 100 -140 fps except on streets edit was useing task manager and saw in one of you other video that its not accurate
Final Results Keep on on windows 11
DAMN
I have never actually seen hags to anything to anyone's performance seems sus
That's why I dug so far into this- if I could see it as claimed, I *definitely* would, yknow?
It's all garbage, don't waste your time with these gamer myths. A buddy and I have identical configurations (7800X3D with the same RAM and we're both not even close to maxing out the GPU even though it's a different one). He has done NO optimisations. Meanwhile I've done it all. Buildzoid's primary and secondary RAM timings, HAGS, various other optimisations (at least the ones that don't look like snake oil, such as reducing % reserved for background tasks from 20 to 10% and such).
Initially, he had 5% performance more than me CONSISTENTLY. I couldn't figure out why. Every single raid, in the same spot on the map, duo together, looking in the same direction and he had 5% more fps every time. But he was crashing every raid after 25-30 min. Turns out, his page file was really small and a Streets raid would begin with 22GB RAM usage but grow to 31GB+ over a 30 minute game and eventually crash. As soon as the increased his page file, crashes stopped but his performance dropped by... you guessed it... 5%. He now has IDENTICAL performance to me. We're talking within 1% every single raid consistently. And he's done no optimisations or other stuff. Just turned on DOCP or whatver its called these days.
None of this shite makes a difference. Get a decent setup, do the main 2-3 fixes to optimise performance and then just enjoy playing and forget about it. Unless you wanna get 64GB of RAM so you can turn off your page file. Maybe that will make a difference, idk.
Do you really need 64 gb of RAM though? I definitely don't have those crashing issues with stock paging file and simply use ARC with Empty standby List running every 5 minutes.
the game wants to use page files no matter what. it will likely crash without enough allocation. I noticed this same behavior in DCS world as well.
@@klemintime1452 No, you don't need it, I was just thinking you won't need a paging file at all then. Or you can also create a virtual drive from your RAM and put the page file there. Because it seems the game uses it even if you have enough memory left. At least that's the impression I got.
@@Dyils I have 64 Gig ram and can confirm game crashes without page file. Games has to have a large page file.
Thanks for this, especially the ram timings thing. I was going to overclock my 3600 CL18 ram but at 30% of the process I kept getting no posts and stuff on secondary timing and I gave up. At the end only thing I changed from XMP was Gear 2 to Gear 1 and it made a noticable difference, like 5% by eyeballing the fps numbers before/after the process.
hey Klem why do u hate gforce experience, i`ve allways used it to save quick replays, is it bad for performance??? @Klemintime
Though I think the application is fairly well made, the fact that you are FORCED to log in to access most features makes it a cheap way for them to gather data about their users. Secondarily, the telemetry they have running in the background has some overhead as discussed by Hardware Unboxed (if you google "Hardware Unboxed Nvidia Driver Overhead, " you should find it). It's a nice software suite though, don't get me wrong.
@@klemintime1452 Is it better to use nvcleaninstall?