Dont look at average fps, you have to look at how much it drops in combat situations.
Bet its similar to how it plays on my old q6600 oc@3ghz(had the pleasure to play on it last weekend), which manages 60fps at times but usually drops around 25fps in combat - which makes movement and aiming rather difficult. (good enough to have some fun in pubs tho)
Check your VSYNC setting. If you have set it ON, and FPS dip to 50 (or anything below 60), and your monitor is 60Hz, you will actually see it dropping to 30Hz. Use adaptive VSYNC. If you have a strong video card, use tripple-buffered VSYNC. It introduces a few ms of delay, but smooths the game experience a lot, at least for me.
some video cards, AMD mostly, have a pretty high frame to frame times compared to most nvidia cards which causes micro stuttering. for example
Radeon 7990 from powercolor has a very high frame to frame latency which causes microstutter roughly 13ms or higher,
GTX Titan has a frame to frame time of only 2.5ms
what does this mean?
a GTX titan running a game with 50FPS will look like its running smother then a radeon 7990 at 60fps because the frame to frame times is so high
"The first batch of frame time tests show a reflection of what we’ve seen in past reviews but with a bit of a twist. AMD’s HD 7790 still displays a horrible amount of stutter in Far Cry 3, making the game nearly unplayable at some points, but on average it does much better than its higher end siblings. Most games played with a fluidity we haven’t seen from AMD cards in quite some time."
if i had to choose a GTX TItan or a GTX 690 i would go for GTX titan, sure the GTX 690 can put out more frames then the TITAN, but the TITAN will look smoother at 60fps then a GTX 690 would at 80FPS, because of frame times.
"While the framerates of AMD’s high end Crossfire setups is nothing short of spectacular, that doesn’t necessarily translate into an acceptable gameplay experience. On the contrary, in nearly every title, the two HD 7970s or HD 7950s displayed atrocious frame time delivery which resulted in a distinct lack of fluidity and negatively impacted everything from shot accuracy to player immersion."
it's because the FPS drops and it doesn't stay at a consistent 60. I even notice going from 700+ FPS going down to 300 if it drops severely and quickly enough. you can feel it in your mouse movement that it's not consistent.
NS2 has FPS drops introduced by dozens of entities so it's bound to feel different even room to room
it's because the FPS drops and it doesn't stay at a consistent 60. I even notice going from 700+ FPS going down to 300 if it drops severely and quickly enough. you can feel it in your mouse movement that it's not consistent.
NS2 has FPS drops introduced by dozens of entities so it's bound to feel different even room to room
Are you saying you get 700 fps on this game? Or are you referring to a different game. Haha, I guess my fps could stay above 120 on this if I went down to x768 or x600 (maybe)
But yes, it's true, you can for sure notice a dip from 500+ fps to 250ish fps~ there's some line there where all the smoothness becomes so smooth that you stop noticing any flaws, however if it dips quick enough you'll feel it as mouse input lag, among maybe a visual smoothness feel!
Metal ManJoin Date: 2011-11-13Member: 132717Members
edited April 2013
I can't explain it too well but I agree with Fuse. Let me try:
Not in combat, and at the start of the game, I am usually around 50-55 fps, which isn't a terrible value. Fuse is right, 50ish should feel decently smooth, yet it doesn't for me. I wouldn't say it feels quite like 30fps but its pretty close. The real issue for me is that it feels stuttery and choppy when moving. It definitely is playable, but it takes much away from my experience.
In a couple previous threads I came to the acceptance that my CPU wasn't quite up to par and the drop in performance was explainable. But this issue seems different. Is there an aspect of performance or smoothness that isn't synonymous with FPS?
Now I haven't had issues with running any games in a long time, so I may be forgetting my former FPS days. But shouldn't a 50 fps CS or HL mod feel smooth as butter? Near perfect, with no hitches?
And I just wanted to emphasize that there does not need to be any dip in FPS for me to experience this. It occurs initially and throughout. Even in the RR or at starting spawns.
I'm pretty positive there's something w/ smoothness that isn't FPS related
My prime example because CSS 2010 version vs current version, if you compare 150 fps in the original engine that the game was made on, to the latest engine it has been updated to, the smoothness is OFF
You need about 280-300~+ fps in the new version to make it feel like 150 fps in the old version. The difference is the new version does multi-threading, while I believe the old one does not (I'm going to assume this is the primary reason)?
In NS2 the difference between 70 to 140 fps obviously is huge when you enter a small room and get these kind of frames, you'll feel it immediately. BUT it doesn't feel like 140 frames would in say a smoother game, also your input never feels "as" responsive. However that not feeling responsive could be related to my maxed out FOV, since I noticed in another title, my mouse felt more responsive with a lower fov compared to a higher one (since my sensitivity is low/very low) even when the framerate remained the same!
Interesting, but I can't really explain it further than that
ScardyBobScardyBobJoin Date: 2009-11-25Member: 69528Forum Admins, Forum Moderators, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Shadow
After reading through this frame rating discussion, I've begun to wonder if certain hardware setups are having some of the problems described in the article. Specifically, runt or dropped frames. It may be the reason why some people think 60fps in NS2 feels more like 30fps in other games.
Amd crossfire setups suffer from frame time issues. Amd is working on it and has it just about fixed for single gpu setups. Afaik, they are working on it game by game so I don't know if ns2 has been fixed.
After reading through this frame rating discussion, I've begun to wonder if certain hardware setups are having some of the problems described in the article. Specifically, runt or dropped frames. It may be the reason why some people think 60fps in NS2 feels more like 30fps in other games.
I've been looking into getting a cheap capture card so I can test if this is the case in NS2. Their method seems straight forward, but the hardware they used is expensive (4x256GB SSDs in RAID 0!!!!).
If the problem was hardware related, everyone wouldnt be having the problems, generally it would specifically be crossfire users at this point (AMD has massively improved frame rate timings over the past year on single cards). It would probably be better to just make a small lua mod to record each frame's time client side, and then you can make a graph from that (maybe have it log ents and a couple other relavent stats). If your seeing super consistent frame times in the engine (you wont in the live build) you could then test with fraps to get a general displayed frame time comparision, and if there was a disparity then maybe it would be worth looking at specific hardware/driver issues. Then you would want to use something like GPUView or FCAT.
If the problem was hardware related, everyone wouldnt be having the problems, generally it would specifically be crossfire users at this point (AMD has massively improved frame rate timings over the past year on single cards). It would probably be better to just make a small lua mod to record each frame's time client side, and then you can make a graph from that (maybe have it log ents and a couple other relavent stats). If your seeing super consistent frame times in the engine (you wont in the live build) you could then test with fraps to get a general displayed frame time comparision, and if there was a disparity then maybe it would be worth looking at specific hardware/driver issues. Then you would want to use something like GPUView or FCAT.
That's a fair point. The plog function does record the frametimes, but there is no way to analyze it other than using perfanalyzer.py (e.g. you can't export the data to something like a csv or xml yet). It may be useful just to graph the FRAPs benchmark results of NS2 by the frametime percentile or even the frame variance.
I've been looking into getting a cheap capture card so I can test if this is the case in NS2. Their method seems straight forward, but the hardware they used is expensive (4x256GB SSDs in RAID 0!!!!).
Perhaps you could capture just a single row of pixels. All the other data appears to be unnecessary.
Fraps frametime recordings are not super accurate, simply because they do not take into account some of the other parts before the frame is displayed to the user. I believe the UpdateClient() event is called on each frame client side - you could probably just use something like this.
local cFPSDataLog = { }
local function OnUpdateClientFPSCounter(deltaTime)
local FPSData = {dT = deltaTime, ents = Shared.GetEntitiesWithClassname("Entity"):GetSize(), T = Shared.GetTime()}
table.insert(cFPSDataLog, FPSData)
end
Event.Hook("UpdateClient", OnUpdateClientFPSCounter)
//Event.Hook("UpdateServer", OnUpdateClientFPSCounter)
local function DumpClientFPSData(name)
local filename = string.format("config://%s.log", name or "fpsdata")
local cFile = io.open(filename, "r")
if not cFile then
cFile = io.open(filename, "w+")
if cFile == nil then
return
end
for i = 1, #cFPSDataLog do
cFile:write(string.format("%.5f,%s,%.5f\n",cFPSDataLog[i].dT, cFPSDataLog[i].ents, cFPSDataLog[i].T))
end
end
io.close(cFile)
end
Event.Hook("Console_dumpfps", DumpClientFPSData)
Vsync at triple buffer made my game appear smoother but it actually changed the rate at which I could turn / aim, and it made the game feel slower somehow, even though the FPS was good and the game appeared smooth I could not aim worth a crap and kept dying. I still don't now whats wrong with it but it seems like a buggy feature at least with my video card, a Radeon HD 6850 1gb.
V-sync renders the frames before they are sent, looks smooth, but gives so much input lag that it should NEVER be used in a game like this, unless of course you like playing handicap and losing
Perhaps you could capture just a single row of pixels. All the other data appears to be unnecessary.
Besides not knowing how to do that with their setup, I believe its possible to do without needing SSD overkill. However, the write speed on the SSD/HDD will effectively limit the resolution you can benchmark.
My generic 1TB 7200RPM HDD has a roughly 100MB/s or so write speed, which would seem to limit my benchmark resolution to 720p or less, according this this formula (8 bit @ 1280 x 720 @ 59.94fps = 105 MB per/sec).
Doing some back of the envelope calculations, it looks like to do a 20min, 1080p benchmark mark, I'd need SSD or HDD with a minimum data rate of 125MB/s and at least 150GB of space. Though, I'd like to have some breathing room, so something that has a write speed of at least 200MB/s and 240GB of space would be better. Looks like my cheapest option would be a Samsung 840 250GB at $169 at least according to Tom's Hardware. Add in the roughly $110 for the cheapest capture card that would work means dropping at least $280 to do this correct (already got a second computer I can record to). So cheaper than RAIDed SSDs, but still means dropping some decent cash.
I know what the TE does mean cause i suffer from the same issue. actually on my card a 68 FPS would run smooth - going below 61 even if its just 1-6 frames just start to feel like playing on a low 30 fps.
in example BF3 starts to stutter @ lower value as 48 fps cs and CS:S are games wich are still smooth with 51 FPS.
i dont know what causes the problem in NS2 that it just looks stuttery and laggy below 61 fps - anyways there are alot of maps wich are wasting tons of performance - mineshaft runs best on my comp. while docking runs like shit from the very early on.
V-sync renders the frames before they are sent, looks smooth, but gives so much input lag that it should NEVER be used in a game like this, unless of course you like playing handicap and losing
I'd rather take v-sync over having screen tearing.... screen tearing makes my aim far worse than having v-sync on.... even the tiniest amount of screen tearing bothers me to no end.
Comments
Unless you are getting micro freezes caused by not enough vram, or the server is lagging.
Bet its similar to how it plays on my old q6600 oc@3ghz(had the pleasure to play on it last weekend), which manages 60fps at times but usually drops around 25fps in combat - which makes movement and aiming rather difficult. (good enough to have some fun in pubs tho)
thats a buddism symbol
the nazi symbol is mirrored. trust me im german xD
i really dont want to post any pictures here but you can easily find it on google.
Radeon 7990 from powercolor has a very high frame to frame latency which causes microstutter roughly 13ms or higher,
GTX Titan has a frame to frame time of only 2.5ms
what does this mean?
a GTX titan running a game with 50FPS will look like its running smother then a radeon 7990 at 60fps because the frame to frame times is so high
do you have an AMD video card?
hardwarecanucks.com/forum/hardware-canucks-reviews/60303-amd-radeon-hd-7790-review-7.html
"The first batch of frame time tests show a reflection of what we’ve seen in past reviews but with a bit of a twist. AMD’s HD 7790 still displays a horrible amount of stutter in Far Cry 3, making the game nearly unplayable at some points, but on average it does much better than its higher end siblings. Most games played with a fluidity we haven’t seen from AMD cards in quite some time."
if i had to choose a GTX TItan or a GTX 690 i would go for GTX titan, sure the GTX 690 can put out more frames then the TITAN, but the TITAN will look smoother at 60fps then a GTX 690 would at 80FPS, because of frame times.
http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/60166-nvidia-gtx-titan-vs-sli-crossfire-15.html
"While the framerates of AMD’s high end Crossfire setups is nothing short of spectacular, that doesn’t necessarily translate into an acceptable gameplay experience. On the contrary, in nearly every title, the two HD 7970s or HD 7950s displayed atrocious frame time delivery which resulted in a distinct lack of fluidity and negatively impacted everything from shot accuracy to player immersion."
NS2 has FPS drops introduced by dozens of entities so it's bound to feel different even room to room
Are you saying you get 700 fps on this game? Or are you referring to a different game. Haha, I guess my fps could stay above 120 on this if I went down to x768 or x600 (maybe)
But yes, it's true, you can for sure notice a dip from 500+ fps to 250ish fps~ there's some line there where all the smoothness becomes so smooth that you stop noticing any flaws, however if it dips quick enough you'll feel it as mouse input lag, among maybe a visual smoothness feel!
Not in combat, and at the start of the game, I am usually around 50-55 fps, which isn't a terrible value. Fuse is right, 50ish should feel decently smooth, yet it doesn't for me. I wouldn't say it feels quite like 30fps but its pretty close. The real issue for me is that it feels stuttery and choppy when moving. It definitely is playable, but it takes much away from my experience.
In a couple previous threads I came to the acceptance that my CPU wasn't quite up to par and the drop in performance was explainable. But this issue seems different. Is there an aspect of performance or smoothness that isn't synonymous with FPS?
Now I haven't had issues with running any games in a long time, so I may be forgetting my former FPS days. But shouldn't a 50 fps CS or HL mod feel smooth as butter? Near perfect, with no hitches?
And I just wanted to emphasize that there does not need to be any dip in FPS for me to experience this. It occurs initially and throughout. Even in the RR or at starting spawns.
Any feedback is greatly appreciated
My prime example because CSS 2010 version vs current version, if you compare 150 fps in the original engine that the game was made on, to the latest engine it has been updated to, the smoothness is OFF
You need about 280-300~+ fps in the new version to make it feel like 150 fps in the old version. The difference is the new version does multi-threading, while I believe the old one does not (I'm going to assume this is the primary reason)?
In NS2 the difference between 70 to 140 fps obviously is huge when you enter a small room and get these kind of frames, you'll feel it immediately. BUT it doesn't feel like 140 frames would in say a smoother game, also your input never feels "as" responsive. However that not feeling responsive could be related to my maxed out FOV, since I noticed in another title, my mouse felt more responsive with a lower fov compared to a higher one (since my sensitivity is low/very low) even when the framerate remained the same!
Interesting, but I can't really explain it further than that
Here is a pretty recent video about the whole subject.
https://www.youtube.com/watch?feature=player_embedded&v=CsHuPxX8ZzQ
Very interesting. Thanks for posting that.
Perhaps you could capture just a single row of pixels. All the other data appears to be unnecessary.
local cFPSDataLog = { } local function OnUpdateClientFPSCounter(deltaTime) local FPSData = {dT = deltaTime, ents = Shared.GetEntitiesWithClassname("Entity"):GetSize(), T = Shared.GetTime()} table.insert(cFPSDataLog, FPSData) end Event.Hook("UpdateClient", OnUpdateClientFPSCounter) //Event.Hook("UpdateServer", OnUpdateClientFPSCounter) local function DumpClientFPSData(name) local filename = string.format("config://%s.log", name or "fpsdata") local cFile = io.open(filename, "r") if not cFile then cFile = io.open(filename, "w+") if cFile == nil then return end for i = 1, #cFPSDataLog do cFile:write(string.format("%.5f,%s,%.5f\n",cFPSDataLog[i].dT, cFPSDataLog[i].ents, cFPSDataLog[i].T)) end end io.close(cFile) end Event.Hook("Console_dumpfps", DumpClientFPSData)My generic 1TB 7200RPM HDD has a roughly 100MB/s or so write speed, which would seem to limit my benchmark resolution to 720p or less, according this this formula (8 bit @ 1280 x 720 @ 59.94fps = 105 MB per/sec).
Doing some back of the envelope calculations, it looks like to do a 20min, 1080p benchmark mark, I'd need SSD or HDD with a minimum data rate of 125MB/s and at least 150GB of space. Though, I'd like to have some breathing room, so something that has a write speed of at least 200MB/s and 240GB of space would be better. Looks like my cheapest option would be a Samsung 840 250GB at $169 at least according to Tom's Hardware. Add in the roughly $110 for the cheapest capture card that would work means dropping at least $280 to do this correct (already got a second computer I can record to). So cheaper than RAIDed SSDs, but still means dropping some decent cash.
in example BF3 starts to stutter @ lower value as 48 fps cs and CS:S are games wich are still smooth with 51 FPS.
i dont know what causes the problem in NS2 that it just looks stuttery and laggy below 61 fps - anyways there are alot of maps wich are wasting tons of performance - mineshaft runs best on my comp. while docking runs like shit from the very early on.
I'd rather take v-sync over having screen tearing.... screen tearing makes my aim far worse than having v-sync on.... even the tiniest amount of screen tearing bothers me to no end.