Big thanks to Nicolas for implementing this and Silvan for getting this into our hands! Also, I didn't make it very clear that this is in the PTU (3.24.3) currently!
So for the little less tech savy people, would you say if the green line is above the blue line then you are GPU bottlenecked? and if vice versa CPU bottlenecked?
Quick question. Is it tested that 3d chips work better in SC or did you just pick cuz it's best overall. I have R7 5800X(Got it before 3d was anounced) and I'm wondering if I should upgrade to 5800x3d.
@@tenpoundfortytwo This might also have a small effect. I can imagine that CIG is collecting more data from the client one the PTU to help them optimize the game. This can also impact the main thread perfomance.
100% CPU usage is not 100% CPU usage The CPU is split into sections and each section can do different things If any 1 of those sections is at its maximum capability then the CPU usage % will be capped to a maximum value less than 100% It is possible to have a CPU usage of less than 10% but still be CPU bottlenecked because a section is oversaturated and no other sections in the CPU are seeing much if any usage
You are basically limited to the speed of the core MT runs on; there are many other threads running that do stuff like audio, RT ed. using the rest of the CPU. Those are also in some cases waiting on each other and handing over work, threading can get crazy complicated.
@ this is very much incorrect as faster ram and a slower core frequency can run at the same rate as slower ram with a higher core frequency Its far more to do with latency on the MT as opposed to just being on a faster core by frequency alone For example a 5600x at 3000 vs 4000 is 21fps vs 29fps them go in and tune the timings now your at 31fps and thats at a fixed 4.75GHz RAM is all that changed Performance in SC is never straight forward
So when CIG finally can enable multithreading, they will have the ability to split up many of the game’s simulation into multiple threads effectively fixing the fps in the game? Does this sound right?
@ unless they find a way to have 2 main threads without all the timings going out of whack then no... The render thread and the main thread are each things that schedule what should be done where while also to some extent doing stuff themselves You can just suddenly have 2 people talking to each other and also telling 6 other people what to do it just doesnt work For an example of talking to the wrong main thread if there where 2... Hey what should i do next, well what did you just do, ok go do this That slows everything down as a best case otherwise game crash
@@echo5827 its not, but generally requires A LOT of optimising different elements of the engine/code/physics/draw calls etc. sometimes a lot of stuff needs 2 threads or more to process because of how complex it is.
The multithreading that the developers are talking about is for the Render thread, not the Main thread. It's for rendering things in screen. The Main thread is for all the game stuff. Physics, AI, interactions etc etc. They already do use multi cores but from my understanding it's just difficult to multithread some things. So idk how far they can push it in a simulation heavy game.
I guess we can take this as an indication that they're actively trying to improve things and giving us something to compare to in a few months, a huge perf pass is definitely needed.
They have a much nicer profiler than this; they showed it in the Vulkan dev videos. However, it would be useless and way too complicated for most of us to read.
Just a note of interest. The MT and RT are "tied" together in a way. MT does all the physics, animation, object updates (and in reality it does some of that work part in parallel on many threads or fibers). The one thing Dx11 does not allow is the render work (so called "draw calls") to be submitted in parallel, so it must be done on the MT after all the data is ready and push that work to the one RT. These kinds of "synchronization points" are inherent bottlenecks. This means, the MT thread is slower than it would be if we had their full Vulkan integration because then they would be able to submit render work from multiple threads. So the full Vulkan integration should decrease the timings on the MT as well (and there likely won't be a need to have a RT at all, since the MT and all it's many fibers could submit render jobs without having to "collect" the work on the RT first). in summary, the full Vulkan work is a MT optimization, really. Thanks!
I thought the RT has to send data to the GPU (via the Render API), not the MT to the RT. And the whole Vulkan multithreading is about having many RT threads so that each of those can send subsets of data to the GPU and speed up that communication. MT and RT likely have access to the same memory, so they dont have to communicate and send data to each other. That would make the Vulkan multithreading a RT optimization not a MT optimization. So basically we are fucked, because the MT was the actual problem all along... fucking great!!
@@SETHthegodofchaos I think naming them at this point will soon become arbitrary, it's all functions and calls and they cant run asynchronously thats the problem. But I think the progress on graphics will be a slow and iterative change, not an instant double fps boost people are expecting.
@@XxXnonameAsDXxX What? Arbitrary? No, it won't. All games have had and will continue to have exactly one MT which contains the game loop where the render thread(s) and the simulation thread(s) have to sync up. Most games dont take much time for simulation so usually the RT can trigger many frames to be rendered so FPS is very high (e.g. 144Hz) even thought the game may just be simulated 30 times a second. But if the simulation threads are busy for a long time, then that stalls the RT(s). Which is what we are seeing in the video most of the time. Not sure what you want to tell me with the whole "functions" and "asynchronously" stuff. Do you even understand what you are saying there yourself? I have played many games without and then with Vulkan and it always provided a significant enough boost in performance to be noticable as well as reduce stutters (low 1% fps) which made the games way more smooth to the point where they felt like a completely different game at times. Of course, they have to optimize this. But plenty of games have implemented Vulkan by now and were able to release their games with Vulkan in a reasonable amount of time. I am just getting tired of hearing CIG being slow again at everything as if thats the norm...
I still don't get why cockpit view is cutting the fps at lest in half compared to third person. I thought that was a Flash issue but nope, still the same behaviour with building blocks.
9800X3D is coming next week and will be the new fastest CPU. Sadly its jump will not as big as we all hoped for but since Squadron42 is still 1,5 to 2 years away Zen6 will arrive before Squadron42 will be released and i guess they will improve the multithreading again within the next 2 years. Star Citizen and Squadron42 will also make use of 16 Core CPUs and with Zen6 it is said there will be native 16 Core variants with great latencys between the core, thanks to new advantages in packaging and stacking of chiplets with greatly higher gaming performance of each core.
So, being one who understands only the minimal of this type of stuff, am I right in saying the higher the green (GPU) number is it means that the GPU is bottlenecked and the higher the grey (main thread) number is, it means that the CPU is bottlenecked. Trying to understand from a layman's perspective :D
This "tool" is a nice gift from CIG, but client performance is by no means the main problem with SC. Without the servers running stably, as has been the case for years with missions, elevators, ships not working, server crashes (recently like with the 30k that we thought had been forgotten), desyncs , the tool is nice, but nothing more.
Hello, I need some help with the cpu's. Would the ryzen 9 7900x3d or the ryzen 7 7800x3d be better for sc and why? (Pairing it with a 4070 ti super.) Thanks.
What could they do to remedy the main thread then? Further Vulkan stuff? And how long do you think that’d take them? Performance is my biggest issue with SC, alongside the bugs.
You were already able to see this. GPU bottlenecked can also be that you are waiting on the GPU for no good reason, the same is true for the CPU counterpart.
As far as I understood CIG planned to break things out of the main thread in the long run and either put tasks on other threads or even the GPU through Vulcan. I think there is still room for optimization.
If/When C.I.G. adds CPU optimization to the project, do you think it will likely account for, AND USE, the 16 cores of my 7850X3D? Your input is valuable to me. Thank you for all the great videos!
It's very hard to say at this stage! I would imagine the game will always really like fast single threaded performance, but hopefully it will start to use more and more threads properly as it gets optimised. Short answer: I really hope so!
It woiuld have been super interesting to see this graph but with the GPU/CPU Util % added on. I think a big problem we see (which render thread MT will help) is the utilization crashing. You can get staggeringly low utilization on a 4090 (and even wattage) despite playing at 4K in a bunker and getting around 30-40fps. That'd help to identify where the GPU is getting bottlenecked by renderthread slowdown and just where the GPU is being *hammered* by what's on screen.
This looks good but I want to know what the server is doing right now every 15min lol. Vulkan is the way to go but so far I get better FPS on DX11 and it also supports HDR which Vulkan for some reason doesn't and it makes a big difference in visuals in the game.
I am a bit disappointed because we were told that the issue was the RT for a long time. Now it sounds like it is the MT afterall. What to make of this? Can the RT create a new frame while the MT is still busy or do they need to sync up?
It's a bit disappointing that we're mainly MainThread bottlenecked when I hoped Vulkan Multithreading would fix things for a while now. Have they even mentioned in the past or recently about how they are going to reduce MainThread frametime
My understanding is they can’t touch main thread without first addressing render thread, and many of the delays in main thread are waiting on the render thread to finish. There was a really good ICS on it. Threads can process in parallel but will pause when there is a cross dependency. Vulcan should reduce the number of these pauses and provide new opportunities for splitting off work in the main thread.
yeah, that was also my understanding. what a shame. That is so disappointing. Essentially, it would seem Vulkan will do nothing for SC. Except quantum travel and clouds I guess...
So what is the benefit of spending all the dev effort on multithreading and introducing it, if it's mainthread optimisation that will deliver performance improvements (if it does)? The dev post on this the other day did not explain that because I think they assumed players know the answer to the question, but many of us don't.
Great vid. Happy to see I'm not the only one having issues with .3 PTU. Running with a 5900x 32GB ram and flooring 22~30 FPS on distribution centres with the new event. Still guess the client FPS are tied to the Server FPS, which in this new patch are specially low.. Let's hope they polish it before throwing it to live..
@@HubstepCamaro I'd also imagine that mate. I don't know if it's server FPS or "Server age" or "server health" or whatever, but with around the same amounts of entities, a server with a higher FPS rate gives you less stuttering, overall better 0.1% and a smoother gaming experience. That's been my experience with SC anyhow and YMMV, but that's also how, usually, on PTU environmente you get better FPS compared to live. Except this build which has Save Stanton dynamic Event ongoing, which spawns a lot of stuff all around.
@@HubstepCamaro correct. Developer Clive stated as such years ago if I remember correctly. But we also know that the load that is on the server still effects the client. If there are many objects on the server around our character, then those are also on our PC. Our PC can only handle so much data and objects. CIG needs to find a way to limit whats going on on our end. I assume they will revisit Client OCS again (after dynamic server meshign is in), to further optimize the data received to only a necessary minimum.
That's a great question and only something CIG will really be able to answer! I'm guessing it will be lots and lots of small optimizations that eventually add up to improved performance.
Thanks for your testing and work done by CIG. I would like to share its content in text form on a Chinese forum, with full credit to you. Is that okay?
Hello i need ur help i have 4k screen but as we know even the best hardware is not sufficient ibhave 5700x3d with 6700xt 32gb@3200mt now when i set the reselotion to 1440p tje game window turn smaller its not full screen mode and im pretty sure it still 4k cus fps stays same as native 4k on while showing full screen how do i make it 1440p(2560x1440) on full screen
The Exclusive fullscren mode was removed in 3.24.2 only leaving borderless which doesn't scale to fit if you're using a resolution lower than native. This has been fixed with the new "Fullscreen" borderless mode in the 3.24.3 PTU which will scale to fit to your screen. If you want to fix it for LIVE currently you will have to change your desktop resolution to 2560x1440p.
There seems to be an issue with DLSS in the last few patches, at least on my machine. No matter what upscaling quality preset is used the r_displayinfo resolution will show the game is at 66% (quality mode). This has been pretty negatively effecting my performance as I used to use DLSS balanced at 1440p which is now not being applied even when selected in game settings! Wonder if anyone else has this issue or if theres a workaround? (Tried both Vulkan and DX11)
The CPU is not the issue it is the code the game is built on. It is an unoptimized mess at the moment. If it were not we would need 10 more years of CPU advancements to see good framerates.
Really they just need to implement PresentMons CPU Wait, CPU Busy, GPU Wait and GPU Busy metrics. Every third-party frame time or FPS tool already uses their API calls / framework.
Dont need a tool to see my frame rate that got cut in half with their absurd graphics update that removed the fullscreen. Rtx 4080 with i9 got down to 29-32 in orison cloudview from 52-65.
Can we really trust this tool considering performance do change a lot depending on servers? On one server i can get 70fps on Lorville but at another i drop down to 25. Maybe it has more stuff to render in due to player activities, not sure but.
Wow cig released something that worked the first time. Impressive. Just a shame the game is optimized like sh and designed for pc's that will be manufactured in 2032.
It is so disappointing to hear. I was really hoping multithreading to save the game and It’s horrible performance. So TLDR, the game was always designed poorly and only now possibly they will start to optimize. But I have 0 hope of that. These devs don’t seem to care about performance in this tech demo.
Well considering my CPU is the 7800X3D I know for sure it would be my GPU that would be the bottleneck. I have the 6800 XT but am waiting for RX 8000 cards to drop to upgrade. Be it an RX 7000 reduced in price or if the 8000s are good enough for their price I'll snag one of them. Switched from Nvidia during the first gen of RTX and with their prices I just wont go back.
Big thanks to Nicolas for implementing this and Silvan for getting this into our hands! Also, I didn't make it very clear that this is in the PTU (3.24.3) currently!
Glad you like it! Thanks for the video and explaining it so well! Couldn't have done it better myself.
So for the little less tech savy people, would you say if the green line is above the blue line then you are GPU bottlenecked? and if vice versa CPU bottlenecked?
Quick question. Is it tested that 3d chips work better in SC or did you just pick cuz it's best overall. I have R7 5800X(Got it before 3d was anounced) and I'm wondering if I should upgrade to 5800x3d.
do not mention that this engine and how it is used is simply useless
@@akiodonoSilvan is a gangsta
Important Piece of information left out. This is only available 3.24.3 not the current live build.
Good point! I did mention that it was on the ptu, but yep didn't make it very clear!
@@tenpoundfortytwo This might also have a small effect. I can imagine that CIG is collecting more data from the client one the PTU to help them optimize the game. This can also impact the main thread perfomance.
Vulkan is also about moving stuff from the MT to the RT.
The dIsplay info stats had those 4 data points but wasn't intuitive to read in a block of numbers. This is a cool addon none the less.
100% CPU usage is not 100% CPU usage
The CPU is split into sections and each section can do different things
If any 1 of those sections is at its maximum capability then the CPU usage % will be capped to a maximum value less than 100%
It is possible to have a CPU usage of less than 10% but still be CPU bottlenecked because a section is oversaturated and no other sections in the CPU are seeing much if any usage
You are basically limited to the speed of the core MT runs on; there are many other threads running that do stuff like audio, RT ed. using the rest of the CPU. Those are also in some cases waiting on each other and handing over work, threading can get crazy complicated.
@ this is very much incorrect as faster ram and a slower core frequency can run at the same rate as slower ram with a higher core frequency
Its far more to do with latency on the MT as opposed to just being on a faster core by frequency alone
For example a 5600x at 3000 vs 4000 is 21fps vs 29fps them go in and tune the timings now your at 31fps and thats at a fixed 4.75GHz RAM is all that changed
Performance in SC is never straight forward
We only use 10% of our brains! 😅
So when CIG finally can enable multithreading, they will have the ability to split up many of the game’s simulation into multiple threads effectively fixing the fps in the game? Does this sound right?
@ unless they find a way to have 2 main threads without all the timings going out of whack then no...
The render thread and the main thread are each things that schedule what should be done where while also to some extent doing stuff themselves
You can just suddenly have 2 people talking to each other and also telling 6 other people what to do it just doesnt work
For an example of talking to the wrong main thread if there where 2... Hey what should i do next, well what did you just do, ok go do this
That slows everything down as a best case otherwise game crash
This is why at citcon they mentioned that they’re moving some calculations off CPU to GPU.
they mentioned that at last years con too. its been the goal i think for quite a while since sc is notoriously cpu heavy
As far as I remember, multithreading is still not fully implemented on the main thread
My cpu knows this... it hates the new patches.
it isnt, its fully implemented on the dev branch tho
Impossible to fully multithread the mt
@@echo5827 its not, but generally requires A LOT of optimising different elements of the engine/code/physics/draw calls etc. sometimes a lot of stuff needs 2 threads or more to process because of how complex it is.
The multithreading that the developers are talking about is for the Render thread, not the Main thread. It's for rendering things in screen. The Main thread is for all the game stuff. Physics, AI, interactions etc etc. They already do use multi cores but from my understanding it's just difficult to multithread some things. So idk how far they can push it in a simulation heavy game.
I guess we can take this as an indication that they're actively trying to improve things and giving us something to compare to in a few months, a huge perf pass is definitely needed.
They have a much nicer profiler than this; they showed it in the Vulkan dev videos. However, it would be useless and way too complicated for most of us to read.
Guarantee this frame graph has been in dev builds for years... Not that its a bad thing! New toys for us are always good
Just a note of interest. The MT and RT are "tied" together in a way. MT does all the physics, animation, object updates (and in reality it does some of that work part in parallel on many threads or fibers). The one thing Dx11 does not allow is the render work (so called "draw calls") to be submitted in parallel, so it must be done on the MT after all the data is ready and push that work to the one RT. These kinds of "synchronization points" are inherent bottlenecks. This means, the MT thread is slower than it would be if we had their full Vulkan integration because then they would be able to submit render work from multiple threads. So the full Vulkan integration should decrease the timings on the MT as well (and there likely won't be a need to have a RT at all, since the MT and all it's many fibers could submit render jobs without having to "collect" the work on the RT first). in summary, the full Vulkan work is a MT optimization, really. Thanks!
I thought the RT has to send data to the GPU (via the Render API), not the MT to the RT.
And the whole Vulkan multithreading is about having many RT threads so that each of those can send subsets of data to the GPU and speed up that communication. MT and RT likely have access to the same memory, so they dont have to communicate and send data to each other.
That would make the Vulkan multithreading a RT optimization not a MT optimization. So basically we are fucked, because the MT was the actual problem all along... fucking great!!
@@SETHthegodofchaos I think naming them at this point will soon become arbitrary, it's all functions and calls and they cant run asynchronously thats the problem. But I think the progress on graphics will be a slow and iterative change, not an instant double fps boost people are expecting.
@@XxXnonameAsDXxX What? Arbitrary? No, it won't. All games have had and will continue to have exactly one MT which contains the game loop where the render thread(s) and the simulation thread(s) have to sync up. Most games dont take much time for simulation so usually the RT can trigger many frames to be rendered so FPS is very high (e.g. 144Hz) even thought the game may just be simulated 30 times a second. But if the simulation threads are busy for a long time, then that stalls the RT(s). Which is what we are seeing in the video most of the time.
Not sure what you want to tell me with the whole "functions" and "asynchronously" stuff. Do you even understand what you are saying there yourself?
I have played many games without and then with Vulkan and it always provided a significant enough boost in performance to be noticable as well as reduce stutters (low 1% fps) which made the games way more smooth to the point where they felt like a completely different game at times. Of course, they have to optimize this. But plenty of games have implemented Vulkan by now and were able to release their games with Vulkan in a reasonable amount of time. I am just getting tired of hearing CIG being slow again at everything as if thats the norm...
Homie is in photo mode with more fps than me at all min settings
Sorry, that was clear as mud for me 😂
Maybe it will make sense when I try ingame myself, but thank you for the video nonetheless!
I still don't get why cockpit view is cutting the fps at lest in half compared to third person. I thought that was a Flash issue but nope, still the same behaviour with building blocks.
9800X3D is coming next week and will be the new fastest CPU. Sadly its jump will not as big as we all hoped for but since Squadron42 is still 1,5 to 2 years away Zen6 will arrive before Squadron42 will be released and i guess they will improve the multithreading again within the next 2 years. Star Citizen and Squadron42 will also make use of 16 Core CPUs and with Zen6 it is said there will be native 16 Core variants with great latencys between the core, thanks to new advantages in packaging and stacking of chiplets with greatly higher gaming performance of each core.
So, being one who understands only the minimal of this type of stuff, am I right in saying the higher the green (GPU) number is it means that the GPU is bottlenecked and the higher the grey (main thread) number is, it means that the CPU is bottlenecked. Trying to understand from a layman's perspective :D
Pretty much yep! The easiest way to understand it is whichever is highest is the bottleneck!
@@tenpoundfortytwo Thank you, and thank you for doing this video, much appreciated :D
Yes great stuff cig,after 12 years you can see better how unoptimized it is..
Exactly. This guy jizzzzingg over a frametime graph. This shits abhorrent.
You, Mr, are a legend. Thanks for this.
This "tool" is a nice gift from CIG, but client performance is by no means the main problem with SC. Without the servers running stably, as has been the case for years with missions, elevators, ships not working, server crashes (recently like with the 30k that we thought had been forgotten), desyncs , the tool is nice, but nothing more.
whenever anything running via the MT is optimized then both server and client will benefit.
it's good to see the high spike ... they can chase a little more about what doing that in some situation and try to make it a little more smooth.
Brother, if you trust a CIG in-house metric then I got a bridge to sell you. lol
flashbang white at the end, ahh! Thanks for all the work
Hello, I need some help with the cpu's.
Would the ryzen 9 7900x3d or the ryzen 7 7800x3d be better for sc and why? (Pairing it with a 4070 ti super.)
Thanks.
Nice tool, but in the End CIG needs to make the game more GPU focussed...
Is no-one going to talk about the landing at 7:13?
Thank you for making another great video!
It was nearly a disaster! Thanks for your support as always!
@@tenpoundfortytwo- any landing you can walk away from is a good one…it looked slick, that’s the main thing
RTSS does this 100x better and is keybindable.
What could they do to remedy the main thread then? Further Vulkan stuff? And how long do you think that’d take them? Performance is my biggest issue with SC, alongside the bugs.
You were already able to see this.
GPU bottlenecked can also be that you are waiting on the GPU for no good reason, the same is true for the CPU counterpart.
What can tank the RT ? A lot of ships and players on screen ? Like 200 people for exemple ? Or it will still be MT bottlenecked in this situation?
Sound more excited for an in-game framtime graph than all of citizencon lol
Great works thank you !
Thanks for making this.
My GPU bound hard with my 4090… also DisplayPort bound and looking forward to the 5090.
As far as I understood CIG planned to break things out of the main thread in the long run and either put tasks on other threads or even the GPU through Vulcan.
I think there is still room for optimization.
If/When C.I.G. adds CPU optimization to the project, do you think it will likely account for, AND USE, the 16 cores of my 7850X3D? Your input is valuable to me. Thank you for all the great videos!
It's very hard to say at this stage! I would imagine the game will always really like fast single threaded performance, but hopefully it will start to use more and more threads properly as it gets optimised. Short answer: I really hope so!
Yes the cpu is the issue, with a 3090ti and a 13900k I get double the fps on lorvile latest live patch
They should make that so i can see how bad of a state the server is in.
It woiuld have been super interesting to see this graph but with the GPU/CPU Util % added on. I think a big problem we see (which render thread MT will help) is the utilization crashing. You can get staggeringly low utilization on a 4090 (and even wattage) despite playing at 4K in a bunker and getting around 30-40fps.
That'd help to identify where the GPU is getting bottlenecked by renderthread slowdown and just where the GPU is being *hammered* by what's on screen.
Thanks for the video! 🙂
This looks good but I want to know what the server is doing right now every 15min lol. Vulkan is the way to go but so far I get better FPS on DX11 and it also supports HDR which Vulkan for some reason doesn't and it makes a big difference in visuals in the game.
there are also some other things like your drive and ram could be affected on your performance. Would be nice to see if they are affecting things
I am a bit disappointed because we were told that the issue was the RT for a long time. Now it sounds like it is the MT afterall. What to make of this? Can the RT create a new frame while the MT is still busy or do they need to sync up?
Very nice, Thank you very much
Solve this one! Localized keyboard with no "tilte key" how to access console?! (Button cannot be mapped)
Could try the windows on screen keyboard in the accessories menu?
Thanks CIG for this nice tool, when can we get a working game tho
We need more performance man...
Its kinda dumb having to buy ridiculously powerful and expensive pcs to reach 60 fps..
I think CIG created this tool is partially due to your influence mate.
I wonder if the MT issues are an engine issue.
I thought the whole point of switching to Vulkan was that it allows pushing non-graphical tasks on to the GPU thus helping with CPU bottlenecks?
It's a bit disappointing that we're mainly MainThread bottlenecked when I hoped Vulkan Multithreading would fix things for a while now. Have they even mentioned in the past or recently about how they are going to reduce MainThread frametime
Vulkan Multithreading is not implemented yet, so it cant resolve anything at the moment ;)
@@Kylliian Vulkan Multithreading only fixes RenderThread time which is a non issue ;)
My understanding is they can’t touch main thread without first addressing render thread, and many of the delays in main thread are waiting on the render thread to finish. There was a really good ICS on it.
Threads can process in parallel but will pause when there is a cross dependency. Vulcan should reduce the number of these pauses and provide new opportunities for splitting off work in the main thread.
yeah, that was also my understanding. what a shame. That is so disappointing. Essentially, it would seem Vulkan will do nothing for SC. Except quantum travel and clouds I guess...
@@jjrickards Very interesting thank you
So what is the benefit of spending all the dev effort on multithreading and introducing it, if it's mainthread optimisation that will deliver performance improvements (if it does)? The dev post on this the other day did not explain that because I think they assumed players know the answer to the question, but many of us don't.
Great stuff £10.42 😁
Only available on PTU at the moment?
Think so!
just bought a better CPU, up until now Ive been using a Ryzen 5 4500. going to a 5800X. *should* make quite a difference.
5800x is one of the best cpu ever made good choice.
Cool, now we have a graph to tell us how bad performance is, instead of performance just feeling bad 😄
Would playing at 1440 instead of 4k be better for all round performance? Native resolution of course
2x better
@@TheKrumstuff cpu usage more or less the same gpu usage will be 2x more at 4k in theory you should have 2x more fps. I use a 3090oc at 1440p
@@StevieCEmpireofUnitedGaming nice, I've never been drawn to going for 4k gaming yet. I prefer more FPS over anything.
@TheKrumstuff correct choice 4k is over-rated. I've Experimented 1k 2k 4k. 2k is the true perfect blend of fps and fidelity
Great vid. Happy to see I'm not the only one having issues with .3 PTU. Running with a 5900x 32GB ram and flooring 22~30 FPS on distribution centres with the new event.
Still guess the client FPS are tied to the Server FPS, which in this new patch are specially low.. Let's hope they polish it before throwing it to live..
Stop. Client FPS has been fully async from “sever FPS” since the release of 3.0
@@HubstepCamaro I'd also imagine that mate. I don't know if it's server FPS or "Server age" or "server health" or whatever, but with around the same amounts of entities, a server with a higher FPS rate gives you less stuttering, overall better 0.1% and a smoother gaming experience.
That's been my experience with SC anyhow and YMMV, but that's also how, usually, on PTU environmente you get better FPS compared to live. Except this build which has Save Stanton dynamic Event ongoing, which spawns a lot of stuff all around.
@@HubstepCamaro correct. Developer Clive stated as such years ago if I remember correctly.
But we also know that the load that is on the server still effects the client. If there are many objects on the server around our character, then those are also on our PC. Our PC can only handle so much data and objects. CIG needs to find a way to limit whats going on on our end. I assume they will revisit Client OCS again (after dynamic server meshign is in), to further optimize the data received to only a necessary minimum.
It would be helpful if they had input from a significantly colorblind individual on their UI choices.
It's a shame I can't move those "displays" and "chat" to a separate monitor, so they don't block any of the screen?
And how exactly can they optimize the main thread ?
That's a great question and only something CIG will really be able to answer! I'm guessing it will be lots and lots of small optimizations that eventually add up to improved performance.
many many ways, but they need to fix the engine first. Their main problem with performance comes from the custom engine.
By actually doing it and not whatever they've been fiddle fucking with for the last 5 years.
By offloading certain things onto the video card directly, if you watch this years CitCon they demonstrated that and the massive performance gains.
@rooster1012 I did. It's still not in the game. So promises, promises, but no actual progress.
Can the main tread run on multiple cores ? Or they have to optimize it to run as efficiently as possible on one core ?
if this exists, i beleive we can expect optimisations soon?
My current rig is an 7800X3D 64gb ddr5 6000 and a 4070 (playing at 1440p) and still get woeful fps in 3.24.3
Things should improve at least a bit once it's off the ptu!
@@tenpoundfortytwo lol
Define woeful? I have similar specs I have no issues with 60 fps
Thanks for your testing and work done by CIG. I would like to share its content in text form on a Chinese forum, with full credit to you. Is that okay?
But weren't you doing those testswith DX11 in this video?
Hello i need ur help i have 4k screen but as we know even the best hardware is not sufficient ibhave 5700x3d with 6700xt 32gb@3200mt now when i set the reselotion to 1440p tje game window turn smaller its not full screen mode and im pretty sure it still 4k cus fps stays same as native 4k on while showing full screen how do i make it 1440p(2560x1440) on full screen
I'd recommend asking on the discord- someone should be able to work out what's going on!
The Exclusive fullscren mode was removed in 3.24.2 only leaving borderless which doesn't scale to fit if you're using a resolution lower than native. This has been fixed with the new "Fullscreen" borderless mode in the 3.24.3 PTU which will scale to fit to your screen. If you want to fix it for LIVE currently you will have to change your desktop resolution to 2560x1440p.
@ tried but still it showed in small screen (u can see the windows desktop)
@ when u say fix u mean replace my 4k to 2k physically?
Now we just need the servers to perform.
immensely hopeful video, thank you very much sir 🫡
There seems to be an issue with DLSS in the last few patches, at least on my machine. No matter what upscaling quality preset is used the r_displayinfo resolution will show the game is at 66% (quality mode). This has been pretty negatively effecting my performance as I used to use DLSS balanced at 1440p which is now not being applied even when selected in game settings! Wonder if anyone else has this issue or if theres a workaround? (Tried both Vulkan and DX11)
Any of you smart fellas know if you can make a macro of a sort to switch gfx settings on the fly without going into menus?
hm didnt work for me .... is it ptu only?
Yep!
The CPU is not the issue it is the code the game is built on. It is an unoptimized mess at the moment. If it were not we would need 10 more years of CPU advancements to see good framerates.
Really they just need to implement PresentMons CPU Wait, CPU Busy, GPU Wait and GPU Busy metrics. Every third-party frame time or FPS tool already uses their API calls / framework.
yes! more tools to be mad at the bad performance.
🤣
I still play this game at high settings 1440p on my gtx1080 😂
Courage
my user.cfg is not working anymore, anyone know how to fix it?
Just tried this. Doesn't pull anything up for me.
same, i think its ptu or something only.
Sorry didn't make it clear enough in the video-its in the ptu, will be coming when 3.24.3 goes live!
what's MT mean?
The Green line? What, below the blue line? :P
I keep photo mode on the clouds all the time and I don't really notice a performance difference. But I'm on a 13900KS/4090 RTX w/ 64 Gb of RAM
Tried with 4.0, that don't works, yet.
Hmm interesting that it's in 3.24.3 ptu but not 4.0!
i play on photomode, 3440x1440, works good enough
p.s. only available in the ptu... not current live.
I already know my performance is dire 😂
Doesn't really quite matter since we are all going to get 10fps at Orison anyways
With AMD GPU, can see all that info in Adrenalin. In my case the GPU is weaker (12GB RX6750XT with 5900X CPU) so the CPU isn't working as hard.
Dont need a tool to see my frame rate that got cut in half with their absurd graphics update that removed the fullscreen. Rtx 4080 with i9 got down to 29-32 in orison cloudview from 52-65.
very nice
does forcing a limit on the frames in the USER.CFG to 60 give more time to the CPU for the main thread?
Optimization is the issue. Not your CPU, or GPU
How about offloading all of the graphics to the stupidly expensive GPUs we own?
Cool!
Can we really trust this tool considering performance do change a lot depending on servers? On one server i can get 70fps on Lorville but at another i drop down to 25. Maybe it has more stuff to render in due to player activities, not sure but.
7800x3d and 4090 here at 4k, 64Gb ram @6000 and the frames still suck. This engine is a mess.
Wow their engine is really resource inefficient, to bottleneck a 4080 in QT.
Wow cig released something that worked the first time. Impressive. Just a shame the game is optimized like sh and designed for pc's that will be manufactured in 2032.
Cool. Still should finish and fix their buggy ass game. This is a trivial thing that RTSS has done for a decade lmao.
It is so disappointing to hear. I was really hoping multithreading to save the game and It’s horrible performance. So TLDR, the game was always designed poorly and only now possibly they will start to optimize. But I have 0 hope of that. These devs don’t seem to care about performance in this tech demo.
Well considering my CPU is the 7800X3D I know for sure it would be my GPU that would be the bottleneck. I have the 6800 XT but am waiting for RX 8000 cards to drop to upgrade. Be it an RX 7000 reduced in price or if the 8000s are good enough for their price I'll snag one of them. Switched from Nvidia during the first gen of RTX and with their prices I just wont go back.
Sitting in the San'tok.yai's cockpit in the middle of space is a gpu limit with dlss on. Something wild is wrong with that ship.
At last!! a USEFUL qol tool that will help people make informed choices for future hardware upgrading.