AsranielJoin Date: 2002-06-03Member: 724Members, Playtest Lead, Forum Moderators, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Shadow, Subnautica Playtester, Retired Community Developer
Yes the luajit approach does seem to be promissing. As mentioned, dont expect this exact result on your computer. There might be 0 fps increase for some people. If your gpu is slow and you are gpu limited, luajit wont affect you (other than the potentially more stable fps as you can see from the graph posted, which in itself is an improvement). There might also be some unforseen technical problem that could make it impossible to deploy it etc. So dont get too excited yet, its very much work in progress, but at least the progress looks good
Can you please post the specs of your mystical PC. Im sure everyone here want to know what to buy to get constant >60 fps.
Btw, im running NS2 on an i5-2500k@4,4 ghz, gtx670, SSD, enough RAM.
I have tried EVERYTHING to get stable FPS:
- Every possible ingame graphic setting in all possible variations
- Fresh windows install
- Im using a processtamer to boost the foward NS2-Process to high priority and lower the priority of backroundprocesses
- Install the latest drivers of all components
- I create an own NS2- Nvidia profile with all settings to low
- I disabled all no needed services
- Networktraffic shaping
What should i say:
In Endgame it goes down to 35FP sometimes. An these 35 fps feels like horrible 10-15 fps.
Everyone i ask describes the same endgame issues.
This thread should not transform into an technical one wich would be in the wrong forum.
But sorry ezay, i dont believe that you get >60fps in an heavy endgame fight.
I have the version XV of that one (this one being the XVI), which is about the same, except mine has two GTX 660, instead of one GTX 680 on their new version.
I have no idea how and why that new computer runs so well, I get 140 fps an Diablo 3 with everything on Ultra, I can run games people swore are unplayable. I'm just happy because I had such a crappy computer before, and that I saved some money (that you can see on the page ) to get better performances. It's just that my expectations were surpassed.
EDIT: link is in french, but you should be able to read the specs at the bottom of the page just fine.
There is nothing sprecial with this PC. Its a good gaming rig for sure, but do you really think that you can run games with better fps @ stock speed while people complaining about performance with heavy OCd ones?
And i repeat myself ezay, your fps going down to 35 fps between mid and engame in crowded areas like on any other PC.
Except if the FPS displayed in the options menu and my 15 years long of PC gaming eyes are decieving me, no, just no. Wishing it to prove your point just doesn't make it true.
dePARAJoin Date: 2011-04-29Member: 96321Members, Squad Five Blue
Im sure they using some sort of secret Hardwarestuff and implement some secret windowstuning.
UWE should take an eye on this and introduce this rig as higly recommended.
Constant >60fps the whole game, in every situation on an normal 18 slot pub server with default maps. My wishes come true.
I think i throw my crappy rig in the corner.
Except if the FPS displayed in the options menu and my 15 years long of PC gaming eyes are decieving me, no, just no. Wishing it to prove your point just doesn't make it true.
ahh, i see the problem. Don't rely on the menu fps reading. Type fps in the console when you've joined a server and see in the middle of a fight how the fps drops and comes back up.
I'm getting 120 at the start of a game, and this drops to 60 ish late game with dips to 30 or lower in big fights.
I'm really excited to see what uwe can do with this luaJIT work, the initial indications look really promising! Good job!
Yes the luajit approach does seem to be promissing. As mentioned, dont expect this exact result on your computer. There might be 0 fps increase for some people. If your gpu is slow and you are gpu limited, luajit wont affect you (other than the potentially more stable fps as you can see from the graph posted, which in itself is an improvement). There might also be some unforseen technical problem that could make it impossible to deploy it etc. So dont get too excited yet, its very much work in progress, but at least the progress looks good
Just having more stable frame times would be a big improvement. Any FPS increase is a bonus.
14 fps sounds like nothing. IF you have 200 fps, however if you have 60 fps, 14+ fps is massive in that situation
Looking forward to the future
And if you have 10, it's like 240% improvement! jk
Is there a way to join this beta test? I'm seriously botllenecked by CPU this summer. Laptop gets too hot and drops to 1.5 ghz. (tried cooling pad, bios mod to force fan at full speed all the time, gpu undervolt, laptop cleanup, even removed dvd rom and battery for additional space for air to circulate, and this altogether won't get me under 85C, which is when throttling starts).
Except if the FPS displayed in the options menu and my 15 years long of PC gaming eyes are decieving me, no, just no. Wishing it to prove your point just doesn't make it true.
ahh, i see the problem. Don't rely on the menu fps reading. Type fps in the console when you've joined a server and see in the middle of a fight how the fps drops and comes back up.
I'm getting 120 at the start of a game, and this drops to 60 ish late game with dips to 30 or lower in big fights.
I'm really excited to see what uwe can do with this luaJIT work, the initial indications look really promising! Good job!
Using the "fps" in console technique, the lowest FPS drop I had was 51 on late game with exos and so on, on 16 players server, using every settings on High.
Other than that the average seemed to be about 70~ on late game.
Hi all,
We're all aware of the performance issues and are doing our best. In a sense, every week is a performance jam for us. We're always trying things and open to new ideas, and some things work out and some things don't. Believe me, we're not happy about the game's performance either.
--Steve
Have you tried making a serious attack on "performance" issues that do not affect average frame rates?
Like input latency. Can you sample input more than once per frame to try and keep movement physics snappier and more precise even at lower framerates?
Like frame time variance. 60 FPS average with every other frame taking twice as long does not feel like 60 fps.
When you spin around quickly frame times increase; as it happens, spinning around quickly is something that often occurs in combat.
The framerate rate drop during combat is large, and enormous during heavy combat, and that's precisely the time you least want to drop frames. Can you strip back the most CPU demanding effects during combat, especially heavy combat?
Just being around lots of players seem to be a chief culprit, structures less so. Would some kind of LoD system allow you to get big savings on the animination system or whatever it might be that causes poor performance when lots of players are around?
One of the things we're working on now is getting better information on what is being slow when something is slow. The game is really complex, and it's difficult to predict what's actually causing slow down, especially on various machines. It's easy to say when the game slows down, but what exactly is causing that in the code? Less clear.
We did address combat effects with Low Quality particles - but during heavy combat, it's still very CPU intensive.
I don't know how you would LOD animation without affecting hit-reg.
Using the "fps" in console technique, the lowest FPS drop I had was 51 on late game with exos and so on, on 16 players server, using every settings on High.
Fyi,
As beautiful as it is, you'll want to turn off ambient occlusion, (or at the very least put to medium) as it can easily eat up 15-20 fps on the best of rigs - if you are in fact running everything on high.
Yo, I remember reading that ns2 had like double the amount of something time. Like Half-life 1 had 50 ms, but ns2 takes into account 100ms. Have there ever been tests trying 50ms to see if it can work just as well or better?
Using the "fps" in console technique, the lowest FPS drop I had was 51 on late game with exos and so on, on 16 players server, using every settings on High.
Fyi,
As beautiful as it is, you'll want to turn off ambient occlusion, (or at the very least put to medium) as it can easily eat up 15-20 fps on the best of rigs - if you are in fact running everything on high.
I checked after reading your post, it seems I came to the same conclusion some times ago, since I already have it on medium. Texture streaming is on "off" too, though it's mainly because I have no idea what this is and how it would affect my FPS.
IronHorseDeveloper, QA Manager, Technical Support & contributorJoin Date: 2010-05-08Member: 71669Members, Super Administrators, Forum Admins, Forum Moderators, NS2 Developer, NS2 Playtester, Squad Five Blue, Subnautica Playtester, Subnautica PT Lead, Pistachionauts
edited May 2013
@ezay it keeps your graphic card's memory (VRAM) at a much lower amount (~300 MB constant usage over potentially 900MB-1GB) by streaming textures in real time instead of loading them in memory before entering the game.
There are benefits to both systems.. for one, texture streaming cuts down on loading times to a notable degree.. but yet accessing and loading textures constantly can lead to the occasional texture popping, compatibility issues, or lack of performance if you do not need it.
Basically, if you have a mid to low end GPU you should be using texture streaming, but if not and you have a decent amount of VRAM (close to 1GB) you won't need it, and in fact shouldn't use it as you will get a smoother experience - yet if you were a low end gpu you would get a smoother experience with it on.
Pretty situational
Also, should be noted that portions of texture streaming are now merged / quasi forced on for everyone.
Thanks for explaining, that was kinda what I understood from the name of the option. Considering I'm always the first to finish loading the map, I guess I don't need to use it.
And while saying 40 fps feels like 15 fps in other games might be an exaggeration, there are some issues that may cause the quality of the fps to drop when the client gets loaded - mostly because frame time may start to vary a lot when there is a lot of stuff happening, so those 40 fps does not mean every frame is 25ms - one or two may clock in at 75ms while most clock in at 20ms...
Added b_fps to Better NS2. b_fps is a companion to fps which is customizable and has an experimental, alternative FPS indicator which may more accurately, or in some way, represent perceived performance (just released, so I've not actually used it in a real game yet :P).
Metal ManJoin Date: 2011-11-13Member: 132717Members
edited May 2013
PLEASE READ UWE! I'll try to help out the best I can from my personal experiences.
NOTHING used by the GPU is lowering performance for most people complaining. My performance literally drops when I change the resolution from highest to lowest. There is almost no difference between all visual settings on max vs min/off. Obviously this issue has been hammered with the answer "CPU Bottleneck." So, moving on.
The most dramatic FPS change occurs when people leave the RR. I average around 60-70fps in a full ready room (18 man server) and it soars to 110-120 when everyone joins the match. Why is a full ready room (with no complex things like infestation, combat, structures, etc.) tanking FPS so much? The ready room is void of all technical aspects. The ONLY thing affecting performance is the PLAYER COUNT and the interactions between units in PROXIMITY. UWE PLEASE begin your purge by investigating the netcode or whatever surrounding this. Does a Full ready room require that much from a CPU and, if so, why?
I have also noticed that my performance is almost TOP NOTCH inside of a gorge tunnel, regardless of how many units are in there. Many games I will chill in a gorge tunnel for a majority of the game just to avoid the headaches created from the performance. The smooth inner linings seem to be immune from the chaos created from the majority of the map. Can anyone explain to me why the gorge tunnels run so much smoother? If there is a simple answer, is there any way to apply that to the rest of the map? Additionally, why do gorge tunnels with structures and units and combat seem to run smoother than a full ready room. Is the gorge tunnel located logically in an isolated section? I think by exploring these questions we can find the path to better performance.
Any response would be greatly greatly appreciated. TY so much!
The most dramatic FPS change occurs when people leave the RR. I average around 60-70fps in a full ready room (18 man server) and it soars to 110-120 when everyone joins the match. Why is a full ready room (with no complex things like infestation, combat, structures, etc.) tanking FPS so much? The ready room is void of all technical aspects. The ONLY thing affecting performance is the PLAYER COUNT and the interactions between units in PROXIMITY. UWE PLEASE begin your purge by investigating the netcode or whatever surrounding this. Does a Full ready room require that much from a CPU and, if so, why?
As SteveRock said, it's easy to point out the situation where these issues happen, but not so easy to find them in the code. Might be anything related to entity interactions/game logic, graphics, animation, physics, networking or even sound, that is, almost the entire engine. The fact that they're improving the profiler before doing any changes to the engine and game scripts demonstrates that they're looking into the issue but they're not sure where exactly it is happening, and a profiler will (hopefully) help locate the bottleneck.
That's not to say you should not describe situations where the game slows down. They help a great deal when testing changes, because they give a real scenario in which to test.
matsoMaster of PatchesJoin Date: 2002-11-05Member: 7000Members, Forum Moderators, NS2 Developer, Constellation, NS2 Playtester, Squad Five Blue, Squad Five Silver, Squad Five Gold, Reinforced - Shadow, NS2 Community Developer
Both the ready rooms and the gorge tunnels are located far away from everything else (run with cheats and type "location" in console), so the number of entities that needs to be updated and rendered are very small ("cents" (client entities) console command). Likewise, r_stats shows that the number of draw calls in a gorge tunnel is about 200 compared to the usual 1000-2000 in a standard location.
The most complex entity to run is actually players (cysts/infestation are only a problem because there are so many of them, individually they are quite cheap), so having 18 players in the readyroom at 70 fps then hitting 110+ when they leave is not surprising - you are going from 14ms per frame to 9ms per frame (and you are also probably renderer limited at that).
If you packed in 24 players in a gorge tunnel, your fps would drop as well.
I find that even in a 18-20 man postmatch Ready Room filled with every lifeform I'll have 50-60 fps but walking around a local server empty map I drop to ~35 in some of the larger rooms (South Tunnels and Hub on Tram, Landing Pad on Docking). My Wait for GPU is always 0ms.
Both the ready rooms and the gorge tunnels are located far away from everything else (run with cheats and type "location" in console), so the number of entities that needs to be updated and rendered are very small ("cents" (client entities) console command). Likewise, r_stats shows that the number of draw calls in a gorge tunnel is about 200 compared to the usual 1000-2000 in a standard location.
The most complex entity to run is actually players (cysts/infestation are only a problem because there are so many of them, individually they are quite cheap), so having 18 players in the readyroom at 70 fps then hitting 110+ when they leave is not surprising - you are going from 14ms per frame to 9ms per frame (and you are also probably renderer limited at that).
If you packed in 24 players in a gorge tunnel, your fps would drop as well.
The cyst problem can be easily solved - it has been solved in the balance mod, I believe. Increase cyst range so less cysts are needed on the field.
I actually wonder if the game was taken away from LUA 3 years ago and hardcoded to C/C# if there would be anywhere near the amount of problems as there is now.
I actually wonder if the game was taken away from LUA 3 years ago and hardcoded to C/C# if there would be anywhere near the amount of problems as there is now.
There would be no problems at all, because there would be no game.
The devs have said many times that the rapid and flexible development workflow made possible by using Lua was a crucial factor in producing the game in as little time as they did with as few people as they had.
I don't play NS2 much anymore because I can play it pretty good but the game feels out of reach for my computer and it's not as fun to play a game that makes you feel like you have constantly worry about your hardware.
Love NS2 still play it but I'm too poor to really play it and enjoy it with my computer.
ScardyBobScardyBobJoin Date: 2009-11-25Member: 69528Forum Admins, Forum Moderators, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Shadow
I decided to do some benching to try to figure out this frame variance issue. Data is from an alien round on a 24p server on Summit with 1080p/all graphics on low/off. Round was a 36min slog fest so it should give a decent amount of potential frame issues. Here are the raw fps and frametime data:
FPS:
Frametime:
The frame variance is calculated by subtracting the current frametime from a running average of the previous 1, 5, 10, 20, 100, and 1000 frames (e.g. Variance-20 = Running average of previous 20 frametimes). Here is the frame variance data:
Frame Variance:
So what I suspect is making NS2 fps seem lower than other games even if the values are the same is the huge spikes in the frame variance on the right side of the graph. They only occur over a small fraction of the frames, but can be pretty huge and likely to show up as stutter or poor-looking fps.
Comments
You can find the specifications here: http://www.materiel.net/ordinateur/materiel-net-duck-nukem-xvi-pc-gamer-88594.html
I have the version XV of that one (this one being the XVI), which is about the same, except mine has two GTX 660, instead of one GTX 680 on their new version.
I have no idea how and why that new computer runs so well, I get 140 fps an Diablo 3 with everything on Ultra, I can run games people swore are unplayable. I'm just happy because I had such a crappy computer before, and that I saved some money (that you can see on the page
EDIT: link is in french, but you should be able to read the specs at the bottom of the page just fine.
There is nothing sprecial with this PC. Its a good gaming rig for sure, but do you really think that you can run games with better fps @ stock speed while people complaining about performance with heavy OCd ones?
And i repeat myself ezay, your fps going down to 35 fps between mid and engame in crowded areas like on any other PC.
UWE should take an eye on this and introduce this rig as higly recommended.
Constant >60fps the whole game, in every situation on an normal 18 slot pub server with default maps. My wishes come true.
I think i throw my crappy rig in the corner.
thx for the tip
ahh, i see the problem. Don't rely on the menu fps reading. Type fps in the console when you've joined a server and see in the middle of a fight how the fps drops and comes back up.
I'm getting 120 at the start of a game, and this drops to 60 ish late game with dips to 30 or lower in big fights.
I'm really excited to see what uwe can do with this luaJIT work, the initial indications look really promising! Good job!
Just having more stable frame times would be a big improvement. Any FPS increase is a bonus.
14 fps sounds like nothing. IF you have 200 fps, however if you have 60 fps, 14+ fps is massive in that situation
Looking forward to the future
And if you have 10, it's like 240% improvement! jk
Is there a way to join this beta test? I'm seriously botllenecked by CPU this summer. Laptop gets too hot and drops to 1.5 ghz. (tried cooling pad, bios mod to force fan at full speed all the time, gpu undervolt, laptop cleanup, even removed dvd rom and battery for additional space for air to circulate, and this altogether won't get me under 85C, which is when throttling starts).
Using the "fps" in console technique, the lowest FPS drop I had was 51 on late game with exos and so on, on 16 players server, using every settings on High.
Other than that the average seemed to be about 70~ on late game.
One of the things we're working on now is getting better information on what is being slow when something is slow. The game is really complex, and it's difficult to predict what's actually causing slow down, especially on various machines. It's easy to say when the game slows down, but what exactly is causing that in the code? Less clear.
We did address combat effects with Low Quality particles - but during heavy combat, it's still very CPU intensive.
I don't know how you would LOD animation without affecting hit-reg.
As beautiful as it is, you'll want to turn off ambient occlusion, (or at the very least put to medium) as it can easily eat up 15-20 fps on the best of rigs - if you are in fact running everything on high.
Aww yissss gogogogogo NS2 PERFORMANCE WOO YEAH AWRIGHT YEAH WOW GOGOGO.
Yo, I remember reading that ns2 had like double the amount of something time. Like Half-life 1 had 50 ms, but ns2 takes into account 100ms. Have there ever been tests trying 50ms to see if it can work just as well or better?
I checked after reading your post, it seems I came to the same conclusion some times ago, since I already have it on medium.
There are benefits to both systems.. for one, texture streaming cuts down on loading times to a notable degree.. but yet accessing and loading textures constantly can lead to the occasional texture popping, compatibility issues, or lack of performance if you do not need it.
Basically, if you have a mid to low end GPU you should be using texture streaming, but if not and you have a decent amount of VRAM (close to 1GB) you won't need it, and in fact shouldn't use it as you will get a smoother experience - yet if you were a low end gpu you would get a smoother experience with it on.
Pretty situational
Also, should be noted that portions of texture streaming are now merged / quasi forced on for everyone.
Added b_fps to Better NS2. b_fps is a companion to fps which is customizable and has an experimental, alternative FPS indicator which may more accurately, or in some way, represent perceived performance (just released, so I've not actually used it in a real game yet :P).
NOTHING used by the GPU is lowering performance for most people complaining. My performance literally drops when I change the resolution from highest to lowest. There is almost no difference between all visual settings on max vs min/off. Obviously this issue has been hammered with the answer "CPU Bottleneck." So, moving on.
The most dramatic FPS change occurs when people leave the RR. I average around 60-70fps in a full ready room (18 man server) and it soars to 110-120 when everyone joins the match. Why is a full ready room (with no complex things like infestation, combat, structures, etc.) tanking FPS so much? The ready room is void of all technical aspects. The ONLY thing affecting performance is the PLAYER COUNT and the interactions between units in PROXIMITY. UWE PLEASE begin your purge by investigating the netcode or whatever surrounding this. Does a Full ready room require that much from a CPU and, if so, why?
I have also noticed that my performance is almost TOP NOTCH inside of a gorge tunnel, regardless of how many units are in there. Many games I will chill in a gorge tunnel for a majority of the game just to avoid the headaches created from the performance. The smooth inner linings seem to be immune from the chaos created from the majority of the map. Can anyone explain to me why the gorge tunnels run so much smoother? If there is a simple answer, is there any way to apply that to the rest of the map? Additionally, why do gorge tunnels with structures and units and combat seem to run smoother than a full ready room. Is the gorge tunnel located logically in an isolated section? I think by exploring these questions we can find the path to better performance.
Any response would be greatly greatly appreciated. TY so much!
As SteveRock said, it's easy to point out the situation where these issues happen, but not so easy to find them in the code. Might be anything related to entity interactions/game logic, graphics, animation, physics, networking or even sound, that is, almost the entire engine. The fact that they're improving the profiler before doing any changes to the engine and game scripts demonstrates that they're looking into the issue but they're not sure where exactly it is happening, and a profiler will (hopefully) help locate the bottleneck.
That's not to say you should not describe situations where the game slows down. They help a great deal when testing changes, because they give a real scenario in which to test.
The most complex entity to run is actually players (cysts/infestation are only a problem because there are so many of them, individually they are quite cheap), so having 18 players in the readyroom at 70 fps then hitting 110+ when they leave is not surprising - you are going from 14ms per frame to 9ms per frame (and you are also probably renderer limited at that).
If you packed in 24 players in a gorge tunnel, your fps would drop as well.
The cyst problem can be easily solved - it has been solved in the balance mod, I believe. Increase cyst range so less cysts are needed on the field.
The devs have said many times that the rapid and flexible development workflow made possible by using Lua was a crucial factor in producing the game in as little time as they did with as few people as they had.
Love NS2 still play it but I'm too poor to really play it and enjoy it with my computer.
FPS:
Frametime:
The frame variance is calculated by subtracting the current frametime from a running average of the previous 1, 5, 10, 20, 100, and 1000 frames (e.g. Variance-20 = Running average of previous 20 frametimes). Here is the frame variance data:
Frame Variance:
So what I suspect is making NS2 fps seem lower than other games even if the values are the same is the huge spikes in the frame variance on the right side of the graph. They only occur over a small fraction of the frames, but can be pretty huge and likely to show up as stutter or poor-looking fps.
No but honestly this sounds good.
You can cap FPS with tools like MSI Afterburner, saves my graphics card drawing 100% power during main menu
\:D/