I've got a pretty old computer, a core2 6300 (1.8 Ghz), and both cores are running around 75%. In my case it's clearly gpu limited (geforce 7300) because I got a large dependence on the resolution. But I don't get how the game can be cpu limited on stronger machines, since my old cpu is not used at 100% ?
Anybody that played the Bad Company 2 beta will remember the outcry from the community at the poor fps, rendering it unplayable for most. People crying like little girls left right and center, moaning about poor fps and that it was a shambles etc.... And yet on release, it was a huge improvement (about 200% for me!)
NS2 is still a beta... They have made it very clear that optimizations are currently not very high on the priority list.
I for one would much rather have the eye candy graphics, but perhaps we can get UWE to make NS2 from 10 pixels for all the griefers who want instant 99999fps from a game still in development.
There is no point moaning that your sports car doesn't go very fast if you test drive it half way through he production line.... needs wheels, and engine etc etc....
<!--quoteo(post=1817015:date=Dec 19 2010, 02:21 AM:name=Asraniel)--><div class='quotetop'>QUOTE (Asraniel @ Dec 19 2010, 02:21 AM) <a href="index.php?act=findpost&pid=1817015"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->A 460 works perfectly fine, i have one and get 60-70 fps. I even get like 20 fps with a geforce 140M which is really bad<!--QuoteEnd--></div><!--QuoteEEnd-->
<!--quoteo(post=1817021:date=Dec 19 2010, 02:37 AM:name=nNyxx)--><div class='quotetop'>QUOTE (nNyxx @ Dec 19 2010, 02:37 AM) <a href="index.php?act=findpost&pid=1817021"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->What. You are joking right? GTX 460 is a fantastic card. I get 40-60FPS in NS2 right now with mine.<!--QuoteEnd--></div><!--QuoteEEnd-->
<!--quoteo(post=1817071:date=Dec 19 2010, 06:48 AM:name=Heroman117)--><div class='quotetop'>QUOTE (Heroman117 @ Dec 19 2010, 06:48 AM) <a href="index.php?act=findpost&pid=1817071"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->On my rig i get pretty decent performance. On my GTX 460 and i7 processor i get an average 100 fps in readyroom and 40-60 ingame and 40-50 fps as commander.<!--QuoteEnd--></div><!--QuoteEEnd-->
I would very much appreciate it if you could all back these statements up with a reproducible benchmark. '40-60' fps implies that your average fps is somewhere in between. I find this claim to be wildly implausible and grossly inconsistent with numbers I have achieved with a plethora of different GPUs. The last reproducible benchmark I saw for a GTX460 on Tram fell no-where near a '40-60' average.
The reason I am asking this question so pointedly might be obvious, if you are all truly achieving these numbers, then I need to modify my recording rigs to match your setups. Achieving those framerates would improve the quality of the NS2HD videos immeasurably.
Replying with simply 'oh yeah I get about 40fps on my VooDoo Extreme' does not help me - I need where you were, what you did, for how long, and what exact fps were achieved.
<b>I can't overstate how surprised I am that this many people could be achieving '40-60' fps averages on midrange Fermi GPUs, and that I'm not able to take advantage of this for your NS2HD viewing pleasure.</b>
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
i would be happy to run a benchmark, do you know any ns2 benchmark?
the 60fps are usually on rockdown. i will check again for tram.
If there is no benchmark, we have to define something to test that should include: which map local or remote server marine/alien/commander standing still or moving around
<!--quoteo(post=1817246:date=Dec 20 2010, 01:33 AM:name=Asraniel)--><div class='quotetop'>QUOTE (Asraniel @ Dec 20 2010, 01:33 AM) <a href="index.php?act=findpost&pid=1817246"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->i would be happy to run a benchmark, do you know any ns2 benchmark?
the 60fps are usually on rockdown. i will check again for tram.<!--QuoteEnd--></div><!--QuoteEEnd-->
Thanks for your reply Asraniel. When I say benchmark, I mean any reproducible set of conditions that would allow me to test for these numbers on different cards and cpus. For example, run from marine start to alien spawn and empty your magazine into the hive, in tram or rockdown, running fraps benchmarking for min/max/avg.
<!--quoteo(post=1817242:date=Dec 19 2010, 04:24 PM:name=NS2HD)--><div class='quotetop'>QUOTE (NS2HD @ Dec 19 2010, 04:24 PM) <a href="index.php?act=findpost&pid=1817242"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->I would very much appreciate it if you could all back these statements up with a reproducible benchmark. '40-60' fps implies that your average fps is somewhere in between. I find this claim to be wildly implausible and grossly inconsistent with numbers I have achieved with a plethora of different GPUs. The last reproducible benchmark I saw for a GTX460 on Tram fell no-where near a '40-60' average.
The reason I am asking this question so pointedly might be obvious, if you are all truly achieving these numbers, then I need to modify my recording rigs to match your setups. Achieving those framerates would improve the quality of the NS2HD videos immeasurably.
Replying with simply 'oh yeah I get about 40fps on my VooDoo Extreme' does not help me - I need where you were, what you did, for how long, and what exact fps were achieved.
<b>I can't overstate how surprised I am that this many people could be achieving '40-60' fps averages on midrange Fermi GPUs, and that I'm not able to take advantage of this for your NS2HD viewing pleasure.</b><!--QuoteEnd--></div><!--QuoteEEnd--> You forgot to mention that NS2 is written using LuaJIT and UWE is desperately fighting to get any performance from the CPU side.
@ns2hd i think it's peoples exaggeration and/or wishful thinking maybe they take the RR fps right after the game is loaded
Heroman117's case i think is an exception, depending on what i7 he has and if it's OC:d, i would say completely possible. my gtx480 gets utilized upto 1/3, so a 460 which is about ~3/5 (?) as powerful should get utilized ~60%, so there's plenty if room to improve fps by increasing CPU capacity
<!--quoteo(post=1817216:date=Dec 19 2010, 12:25 PM:name=Harimau)--><div class='quotetop'>QUOTE (Harimau @ Dec 19 2010, 12:25 PM) <a href="index.php?act=findpost&pid=1817216"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->New question: Is everyone who is running the game decently running a 64 bit OS?<!--QuoteEnd--></div><!--QuoteEEnd-->
I don't upgrade unless I have a compelling reason to; the last couple of years the games which are the most demanding have been complete bum fodder gamplay-wise(e.g. Crysis).
In short, I'm using a >4 year old graphics card(X1900 with 512 MB memory) and <i>even I'm CPU bound at</i> "ridiculously awful" and "medium" on a Q6600 CPU. On "best" my performance is bound by the 512 MB of memory on my graphics card.
<!--quoteo(post=1817319:date=Dec 20 2010, 06:32 AM:name=Asraniel)--><div class='quotetop'>QUOTE (Asraniel @ Dec 20 2010, 06:32 AM) <a href="index.php?act=findpost&pid=1817319"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->NS2HD: Did as you told, ran from marine start to hive and emptied one clip into it.<!--QuoteEnd--></div><!--QuoteEEnd-->
Thanks heaps for that Asraniel. I'll have to have a go at reproducing those numbers. I to am using the 1090T, 8Gb ram, so it will be interesting to see what I can come up with!
ScardyBobScardyBobJoin Date: 2009-11-25Member: 69528Forum Admins, Forum Moderators, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Shadow
One problem I have is that the 'ridiculously awful' graphics setting is not actually very awful looking. I personally believe that such a setting should have more of a minecraft/pixelized look, where all graphics above the very minimum are sacrificed to squeeze every last fps out of my hardware. The current worst graphics setting is still too good imo, which makes me think that it isn't the very lowest that is possible. I personally prefer smooth gameplay over good graphics and I'd like more options to do so in NS2.
Granted, it seems like the current problem is the CPU bottleneck, but (presumably) UWE will optimize the CPU side of things so that the GPU becomes limiting again for some people.
endarJoin Date: 2010-07-27Member: 73256Members, Squad Five Blue
<!--quoteo(post=1817319:date=Dec 20 2010, 05:32 AM:name=Asraniel)--><div class='quotetop'>QUOTE (Asraniel @ Dec 20 2010, 05:32 AM) <a href="index.php?act=findpost&pid=1817319"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->NS2HD: Did as you told, ran from marine start to hive and emptied one clip into it. all on max:<!--QuoteEnd--></div><!--QuoteEEnd--> What resolution?
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
all on max also meant resolution on max. have to check what the max resolution is for ns2, but i guess the native resolution of my motitor? 1920x something
Anyways I did the test as NS2HD suggested and these were my results
(GTX 460, 4gig ram, AMD Phenom x4 955 (3.5ghz)
1920x1080 all max graphic settings. vsync is <b>forced on</b>
ROCKDOWN (path right from CC) Frames-1481 Time (ms)- 39521 Min-29 Max-60 Avg-37.474
Junction (top path) Frames - 1041 Time (ms) - 28268 Min -28 Max - 60 Avg - 36.826
Tram (left path) Frames - 1091 Time (ms) - 36901 Min - 22 Max-42 Avg - 29.566
Interesting thing I noticed, is my frames seem to lock at 30fps during large rooms and 60fps during corridors, but like almost exactly, every time. As soon as I hit a big room, bam exactly 30fps solid. I wonder if vsync is doing that to prevent uneven framerates from causing tearing.
NS gets nasty screen tearing without vsync forced on, either way it runs more than fine on my 460, and this is still beta so I am sure there is more optimization to be had, alot more.
<!--quoteo(post=1817440:date=Dec 20 2010, 01:10 AM:name=ScardyBob)--><div class='quotetop'>QUOTE (ScardyBob @ Dec 20 2010, 01:10 AM) <a href="index.php?act=findpost&pid=1817440"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->One problem I have is that the 'ridiculously awful' graphics setting is not actually very awful looking. I personally believe that such a setting should have more of a minecraft/pixelized look, where all graphics above the very minimum are sacrificed to squeeze every last fps out of my hardware.<!--QuoteEnd--></div><!--QuoteEEnd-->
What's in it for UW? They don't want there to be screenshots of their game floating about that look pug ugly.
You already have at your disposal ways to affect an enormous improvement in performance if GPU was the bottleneck. Ontop of the going from best to ridiculously awful you can drop resolution to half of native, which will reduce the workload done by your GPU by an additional factor of almost 4(I say almost, because at higher res cache-locality is slightly improved, which helps performance per pixel a little bit).
The game is very CPU bound and there are only a few graphical features that I can think of that might help if you could turn them off. You can turn off all shadow casting on lights, but that could be cheating as it will brighten areas, particularly sneaky hiding spots; this helps because it avoids a bunch of draw calls associated with generating the shadow map. You can outright remove player models when they die so there is no need for ragdoll physics(looks really fugly).
<!--quoteo(post=1817440:date=Dec 20 2010, 01:10 AM:name=ScardyBob)--><div class='quotetop'>QUOTE (ScardyBob @ Dec 20 2010, 01:10 AM) <a href="index.php?act=findpost&pid=1817440"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->The current worst graphics setting is still too good imo, which makes me think that it isn't the very lowest that is possible. I personally prefer smooth gameplay over good graphics and I'd like more options to do so in NS2.<!--QuoteEnd--></div><!--QuoteEEnd-->
The straight forward way to do it is to just lower texture resolution, lower poly count, lower model poly counts. But why would you want to? It barely makes any difference if you play on medium or ridiculous as it is.
The performance issues in the game have little to do with the graphics engine itself. Lowering screen or texture resolution or even tuning down the lighting won't help much to increase performance (might increase max fps though). All the hickups and micro freezes are mostly based on game and network engine problems that are being fixed.
<!--quoteo(post=1817524:date=Dec 20 2010, 09:44 AM:name=noproblem)--><div class='quotetop'>QUOTE (noproblem @ Dec 20 2010, 09:44 AM) <a href="index.php?act=findpost&pid=1817524"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->Does anyone think I might be able to get more stability by disabling SpeedStep?<!--QuoteEnd--></div><!--QuoteEEnd-->
Unlikely. The problems many older games had with multicore and CnC/SpeedStep was related to unsafe usage of timers. Microsoft's recommended practice is to use the QueryPerformanceCounter and QueryPerformanceFrequency timers, call them from one specific thread and set the affinity of that thread to a specific CPU to minimize risk of problems(for some timers, e.g. RDTSC, there may be a small ofset between the different cores, even though Microsoft has tried to adjust for this).
QueryPerformanceCounter chooses what it thinks is the best timer(sometimes that's RDTSC, sometimes it's a timer on the motherboard that runs at a specific frequency, independent of CPU), does whatever adjustments it needs to in order to deal with CnC/SpeedStep. Problems with the timer are rare and have usually been traced back to a bug in BIOS of that particular make and model of motherboard.
<!--quoteo(post=1817525:date=Dec 20 2010, 02:59 PM:name=NurEinMensch)--><div class='quotetop'>QUOTE (NurEinMensch @ Dec 20 2010, 02:59 PM) <a href="index.php?act=findpost&pid=1817525"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->The performance issues in the game have little to do with the graphics engine itself. Lowering screen or texture resolution or even tuning down the lighting won't help much to increase performance (might increase max fps though). All the hickups and micro freezes are mostly based on game and network engine problems that are being fixed.<!--QuoteEnd--></div><!--QuoteEEnd-->
I also wanted to say this previously. When I create a SP game, the performance seems completely fine. I haven't done any benchmarking yet, but it doesn't feel stuttery at all. When playing online though, the performance is miserable. Ping is great, performance not so much.
I too think, that before trying to tweak settings or even change hardware, the network code should get more optimized (not completely optimized, this can wait - but right now, this "hurts" the most imho).
Edit: If anyone is interested:
On <b>Rockdown</b>, <b>1680x1050</b> (and <b>16x</b> forced HQ<b>AF</b> via CCC),<b> lowest details</b>. Running from Marine start via West to Hive and emptying a full mag into hive:
<!--coloro:#FF0000--><span style="color:#FF0000"><!--/coloro-->Edit<!--colorc--></span><!--/colorc-->: Altough there is still some periodic stuttering which isn't visible in the averaged FPS, as you can clearly see in the following graph: <a href="http://www.abload.de/img/rockdownyupo.png" target="_blank">http://www.abload.de/img/rockdownyupo.png</a>
<!--quoteo--><div class='quotetop'>QUOTE </div><div class='quotemain'><!--quotec-->Interesting thing I noticed, is my frames seem to lock at 30fps during large rooms and 60fps during corridors, but like almost exactly, every time. As soon as I hit a big room, bam exactly 30fps solid. I wonder if vsync is doing that to prevent uneven framerates from causing tearing.<!--QuoteEnd--></div><!--QuoteEEnd--> That's VSync. VSync is syncing your FPS to the refreshrate of your monitor. This means, in simple terms, that the GPU is delivering either 1 Frame per "refresh" of your screen, every 2nd refresh, every 3rd and so on. I suppose you are using a TFT with 60Hz. So you either have 60FPS, 30FPS or 15FPS and so forth. If your GPU is capable of delivering 59FPS, you only get 30. It is not advisable to use VSync if your computer isn't able to deliver at least 60FPS - otherwise you will experience bad slowdowns.
<!--quoteo(post=1817532:date=Dec 20 2010, 02:15 PM:name=Soylent_green)--><div class='quotetop'>QUOTE (Soylent_green @ Dec 20 2010, 02:15 PM) <a href="index.php?act=findpost&pid=1817532"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->Unlikely. The problems many older games had with multicore and CnC/SpeedStep was related to unsafe usage of timers. Microsoft's recommended practice is to use the QueryPerformanceCounter and QueryPerformanceFrequency timers, call them from one specific thread and set the affinity of that thread to a specific CPU to minimize risk of problems(for some timers, e.g. RDTSC, there may be a small ofset between the different cores, even though Microsoft has tried to adjust for this).
QueryPerformanceCounter chooses what it thinks is the best timer(sometimes that's RDTSC, sometimes it's a timer on the motherboard that runs at a specific frequency, independent of CPU), does whatever adjustments it needs to in order to deal with CnC/SpeedStep. Problems with the timer are rare and have usually been traced back to a bug in BIOS of that particular make and model of motherboard.<!--QuoteEnd--></div><!--QuoteEEnd-->
Wow, that completely went over my head, but thanks, I'll take your word for it! :)
Just for the heck of it i ran 2 fraps benchmarks for you guys since im in the last year of my 3 year upgrade cycle so i have a "old" system right now.
Both are 1600x1200, 16xAF 4xAA high quality on build 159 recorded with the fraps benchmarking option.
Im running:
Core i7 920 2.66ghz 6GB of ram Geforce GTX 275
First test: Local Server Rockdown, runthrough from marine start to alien hive and emptying 1 clip: Frames Time (ms) Min Max Avg 1554 39211 24 54 39.632
Second test: Armory chamber.net #2, joined in half way with 8 to 12 players, Marine flamethrower victory (the benchmark can't be reproduced, but does give an indication for current real life situations): Frames Time (ms) Min Max Avg 31619 1203528 11 54 26.272
That was a full blown battle - all the way up to scorching the enemy hive with multiple marines setting it on fire. In the list of FPS those moments usually hover around 19 FPS on my rig. The most noticable drop in the list of FPS i got was around the point where the 11th and 12th player joined. That alone took about 5 FPS off my average - while they where in the ready room. From that point onward i think my average was closer to 22 fps, but before, it stayed at 20+ at all times.
I hope that gives some more to compare with. IMO, My rig for build 159 happens to be the absolute minimum to run it on high settings. If servers didn't become unstable with more then 10 players i'd probably go with no AA or a lower res. I wouldn't try 1920x1080 either, 1600x1200 is stretching it already.
But all in all i've played through games on worse settings (pwned quake 4 with 20 fps @ 640x480 on a pentium 4 ^^) So it's really not all that bad considering the current state. I'll bet FPS will double before release, and if it's true that the engine isn't multithreaded yet i'd almost expect it to triple.
Or at the very least, not drop too hard when player counts are upped.
Comments
That beeing said, both my computers are 64bit
People crying like little girls left right and center, moaning about poor fps and that it was a shambles etc....
And yet on release, it was a huge improvement (about 200% for me!)
NS2 is still a beta... They have made it very clear that optimizations are currently not very high on the priority list.
I for one would much rather have the eye candy graphics, but perhaps we can get UWE to make NS2 from 10 pixels for all the griefers who want instant 99999fps from a game still in development.
There is no point moaning that your sports car doesn't go very fast if you test drive it half way through he production line.... needs wheels, and engine etc etc....
Bottom line:
Be patient :D
I even get like 20 fps with a geforce 140M which is really bad<!--QuoteEnd--></div><!--QuoteEEnd-->
<!--quoteo(post=1817021:date=Dec 19 2010, 02:37 AM:name=nNyxx)--><div class='quotetop'>QUOTE (nNyxx @ Dec 19 2010, 02:37 AM) <a href="index.php?act=findpost&pid=1817021"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->What. You are joking right? GTX 460 is a fantastic card. I get 40-60FPS in NS2 right now with mine.<!--QuoteEnd--></div><!--QuoteEEnd-->
<!--quoteo(post=1817071:date=Dec 19 2010, 06:48 AM:name=Heroman117)--><div class='quotetop'>QUOTE (Heroman117 @ Dec 19 2010, 06:48 AM) <a href="index.php?act=findpost&pid=1817071"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->On my rig i get pretty decent performance. On my GTX 460 and i7 processor i get an average 100 fps in readyroom and 40-60 ingame and 40-50 fps as commander.<!--QuoteEnd--></div><!--QuoteEEnd-->
I would very much appreciate it if you could all back these statements up with a reproducible benchmark. '40-60' fps implies that your average fps is somewhere in between. I find this claim to be wildly implausible and grossly inconsistent with numbers I have achieved with a plethora of different GPUs. The last reproducible benchmark I saw for a GTX460 on Tram fell no-where near a '40-60' average.
The reason I am asking this question so pointedly might be obvious, if you are all truly achieving these numbers, then I need to modify my recording rigs to match your setups. Achieving those framerates would improve the quality of the NS2HD videos immeasurably.
Replying with simply 'oh yeah I get about 40fps on my VooDoo Extreme' does not help me - I need where you were, what you did, for how long, and what exact fps were achieved.
<b>I can't overstate how surprised I am that this many people could be achieving '40-60' fps averages on midrange Fermi GPUs, and that I'm not able to take advantage of this for your NS2HD viewing pleasure.</b>
the 60fps are usually on rockdown. i will check again for tram.
If there is no benchmark, we have to define something to test that should include:
which map
local or remote server
marine/alien/commander
standing still or moving around
the 60fps are usually on rockdown. i will check again for tram.<!--QuoteEnd--></div><!--QuoteEEnd-->
Thanks for your reply Asraniel. When I say benchmark, I mean any reproducible set of conditions that would allow me to test for these numbers on different cards and cpus. For example, run from marine start to alien spawn and empty your magazine into the hive, in tram or rockdown, running fraps benchmarking for min/max/avg.
The reason I am asking this question so pointedly might be obvious, if you are all truly achieving these numbers, then I need to modify my recording rigs to match your setups. Achieving those framerates would improve the quality of the NS2HD videos immeasurably.
Replying with simply 'oh yeah I get about 40fps on my VooDoo Extreme' does not help me - I need where you were, what you did, for how long, and what exact fps were achieved.
<b>I can't overstate how surprised I am that this many people could be achieving '40-60' fps averages on midrange Fermi GPUs, and that I'm not able to take advantage of this for your NS2HD viewing pleasure.</b><!--QuoteEnd--></div><!--QuoteEEnd-->
You forgot to mention that NS2 is written using LuaJIT and UWE is desperately fighting to get any performance from the CPU side.
i think it's peoples exaggeration and/or wishful thinking
maybe they take the RR fps right after the game is loaded
Heroman117's case i think is an exception, depending on what i7 he has and if it's OC:d, i would say completely possible.
my gtx480 gets utilized upto 1/3, so a 460 which is about ~3/5 (?) as powerful should get utilized ~60%,
so there's plenty if room to improve fps by increasing CPU capacity
besides, i can get over 100fps in RR as it is
Did as you told, ran from marine start to hive and emptied one clip into it.
all on max:
Min, Max, Avg
12, 47, 36.811
Same on rockdown
Min, Max, Avg
41, 63, 50.058
my hardware:
phenom x6 1090
Geforce 460
8GB ram
Windows 7 64bit
Is everyone who is running the game decently running a 64 bit OS?<!--QuoteEnd--></div><!--QuoteEEnd-->
I have Win 7 Pro 64
In short, I'm using a >4 year old graphics card(X1900 with 512 MB memory) and <i>even I'm CPU bound at</i> "ridiculously awful" and "medium" on a Q6600 CPU. On "best" my performance is bound by the 512 MB of memory on my graphics card.
Did as you told, ran from marine start to hive and emptied one clip into it.<!--QuoteEnd--></div><!--QuoteEEnd-->
Thanks heaps for that Asraniel. I'll have to have a go at reproducing those numbers. I to am using the 1090T, 8Gb ram, so it will be interesting to see what I can come up with!
Granted, it seems like the current problem is the CPU bottleneck, but (presumably) UWE will optimize the CPU side of things so that the GPU becomes limiting again for some people.
Did as you told, ran from marine start to hive and emptied one clip into it.
all on max:<!--QuoteEnd--></div><!--QuoteEEnd-->
What resolution?
The 460 runs NS2 absolutely fine.
Anyways I did the test as NS2HD suggested and these were my results
(GTX 460, 4gig ram, AMD Phenom x4 955 (3.5ghz)
1920x1080 all max graphic settings.
vsync is <b>forced on</b>
ROCKDOWN (path right from CC)
Frames-1481 Time (ms)- 39521 Min-29 Max-60 Avg-37.474
Junction (top path)
Frames - 1041 Time (ms) - 28268 Min -28 Max - 60 Avg - 36.826
Tram (left path)
Frames - 1091 Time (ms) - 36901 Min - 22 Max-42 Avg - 29.566
Interesting thing I noticed, is my frames seem to lock at 30fps during large rooms and 60fps during corridors, but like almost exactly, every time. As soon as I hit a big room, bam exactly 30fps solid. I wonder if vsync is doing that to prevent uneven framerates from causing tearing.
NS gets nasty screen tearing without vsync forced on, either way it runs more than fine on my 460, and this is still beta so I am sure there is more optimization to be had, alot more.
What's in it for UW? They don't want there to be screenshots of their game floating about that look pug ugly.
You already have at your disposal ways to affect an enormous improvement in performance if GPU was the bottleneck. Ontop of the going from best to ridiculously awful you can drop resolution to half of native, which will reduce the workload done by your GPU by an additional factor of almost 4(I say almost, because at higher res cache-locality is slightly improved, which helps performance per pixel a little bit).
The game is very CPU bound and there are only a few graphical features that I can think of that might help if you could turn them off. You can turn off all shadow casting on lights, but that could be cheating as it will brighten areas, particularly sneaky hiding spots; this helps because it avoids a bunch of draw calls associated with generating the shadow map. You can outright remove player models when they die so there is no need for ragdoll physics(looks really fugly).
<!--quoteo(post=1817440:date=Dec 20 2010, 01:10 AM:name=ScardyBob)--><div class='quotetop'>QUOTE (ScardyBob @ Dec 20 2010, 01:10 AM) <a href="index.php?act=findpost&pid=1817440"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->The current worst graphics setting is still too good imo, which makes me think that it isn't the very lowest that is possible. I personally prefer smooth gameplay over good graphics and I'd like more options to do so in NS2.<!--QuoteEnd--></div><!--QuoteEEnd-->
The straight forward way to do it is to just lower texture resolution, lower poly count, lower model poly counts. But why would you want to? It barely makes any difference if you play on medium or ridiculous as it is.
Unlikely. The problems many older games had with multicore and CnC/SpeedStep was related to unsafe usage of timers. Microsoft's recommended practice is to use the QueryPerformanceCounter and QueryPerformanceFrequency timers, call them from one specific thread and set the affinity of that thread to a specific CPU to minimize risk of problems(for some timers, e.g. RDTSC, there may be a small ofset between the different cores, even though Microsoft has tried to adjust for this).
QueryPerformanceCounter chooses what it thinks is the best timer(sometimes that's RDTSC, sometimes it's a timer on the motherboard that runs at a specific frequency, independent of CPU), does whatever adjustments it needs to in order to deal with CnC/SpeedStep. Problems with the timer are rare and have usually been traced back to a bug in BIOS of that particular make and model of motherboard.
I also wanted to say this previously.
When I create a SP game, the performance seems completely fine. I haven't done any benchmarking yet, but it doesn't feel stuttery at all.
When playing online though, the performance is miserable. Ping is great, performance not so much.
I too think, that before trying to tweak settings or even change hardware, the network code should get more optimized (not completely optimized, this can wait - but right now, this "hurts" the most imho).
Edit:
If anyone is interested:
On <b>Rockdown</b>, <b>1680x1050</b> (and <b>16x</b> forced HQ<b>AF</b> via CCC),<b> lowest details</b>. Running from Marine start via West to Hive and emptying a full mag into hive:
Min: 46
Max: 72
Avg: 59.6
High Detail:
Min: 44
Max: 68
Avg: 56.9
Q9550@3.76
4GB RAM
5850 1GB slightly OCed
Win7 x64
<!--coloro:#FF0000--><span style="color:#FF0000"><!--/coloro-->Edit<!--colorc--></span><!--/colorc-->: Altough there is still some periodic stuttering which isn't visible in the averaged FPS, as you can clearly see in the following graph:
<a href="http://www.abload.de/img/rockdownyupo.png" target="_blank">http://www.abload.de/img/rockdownyupo.png</a>
<!--quoteo--><div class='quotetop'>QUOTE </div><div class='quotemain'><!--quotec-->Interesting thing I noticed, is my frames seem to lock at 30fps during large rooms and 60fps during corridors, but like almost exactly, every time. As soon as I hit a big room, bam exactly 30fps solid. I wonder if vsync is doing that to prevent uneven framerates from causing tearing.<!--QuoteEnd--></div><!--QuoteEEnd-->
That's VSync.
VSync is syncing your FPS to the refreshrate of your monitor. This means, in simple terms, that the GPU is delivering either 1 Frame per "refresh" of your screen, every 2nd refresh, every 3rd and so on.
I suppose you are using a TFT with 60Hz. So you either have 60FPS, 30FPS or 15FPS and so forth. If your GPU is capable of delivering 59FPS, you only get 30. It is not advisable to use VSync if your computer isn't able to deliver at least 60FPS - otherwise you will experience bad slowdowns.
QueryPerformanceCounter chooses what it thinks is the best timer(sometimes that's RDTSC, sometimes it's a timer on the motherboard that runs at a specific frequency, independent of CPU), does whatever adjustments it needs to in order to deal with CnC/SpeedStep. Problems with the timer are rare and have usually been traced back to a bug in BIOS of that particular make and model of motherboard.<!--QuoteEnd--></div><!--QuoteEEnd-->
Wow, that completely went over my head, but thanks, I'll take your word for it! :)
Both are 1600x1200, 16xAF 4xAA high quality on build 159 recorded with the fraps benchmarking option.
Im running:
Core i7 920 2.66ghz
6GB of ram
Geforce GTX 275
First test: Local Server Rockdown, runthrough from marine start to alien hive and emptying 1 clip:
Frames Time (ms) Min Max Avg
1554 39211 24 54 39.632
Second test: Armory chamber.net #2, joined in half way with 8 to 12 players, Marine flamethrower victory (the benchmark can't be reproduced, but does give an indication for current real life situations):
Frames Time (ms) Min Max Avg
31619 1203528 11 54 26.272
That was a full blown battle - all the way up to scorching the enemy hive with multiple marines setting it on fire. In the list of FPS those moments usually hover around 19 FPS on my rig. The most noticable drop in the list of FPS i got was around the point where the 11th and 12th player joined. That alone took about 5 FPS off my average - while they where in the ready room. From that point onward i think my average was closer to 22 fps, but before, it stayed at 20+ at all times.
I hope that gives some more to compare with. IMO, My rig for build 159 happens to be the absolute minimum to run it on high settings. If servers didn't become unstable with more then 10 players i'd probably go with no AA or a lower res. I wouldn't try 1920x1080 either, 1600x1200 is stretching it already.
But all in all i've played through games on worse settings (pwned quake 4 with 20 fps @ 640x480 on a pentium 4 ^^) So it's really not all that bad considering the current state. I'll bet FPS will double before release, and if it's true that the engine isn't multithreaded yet i'd almost expect it to triple.
Or at the very least, not drop too hard when player counts are upped.