Performance benchmarks for build 154

Avalon

The system:

Windows 7 Ultimate 64bit
Gigabyte P35 DS3L
Intel E8400 Core2 Duo @ 3.78ghz
4GB DDR2 RAM @ 900Mhz
Visiontek Radeon HD4850 512MB stock
Audigy 2
Ati Catalyst 9.11 (November '09)
1650x1050 resolution, all high game settings, application settings in driver
NS_Tram

I use fraps for my benchmarking purposes. My run involves a 60 second walk from the center of marine spawn to the alien hive room. Stopping by the tall pillar near the entrance with the hive in sight, I unload 2 rifle clips into the hive. Then, I pull out my pistol and unload two clips again. Finally, I pull my rifle back out and unload two more clips. Run ends right after this.

Build 154
Min: 14fps
Avg: 24.8fps
Max: 39fps

Build 153
Min: 10fps
Avg: 25.3fps
Max: 34fps

Build 152
Min: 13fps
Avg: 25.2fps
Max: 31fps

The game definitely feels faster, but I am getting a lot of stuttering which I think is causing a lot of near minimum frame rates, dragging my average down. As you can see, my max fps is noticeably higher than say, build 152, so the performance is there...

I also can confirm occlusion culling seems to be working properly. When I walk into a wall, my fps goes up generally by 10-15fps, whereas it would stay the same previously, so the game is definitely only drawing what it can see now.

If the stuttering can be fixed, I think the game would run quite well for me.

_Thresh_

Win 7 64
i7 950
X58A-UD7
6GB 2000MHz RAM
Crossfire Radeon 5870

Crossfire disabled
1900 x 1200- 61 fps no stuttering

Crossfire enabled
1900 x 1200- 45 fps increased stuttering on shadows

Wheeee

question for you, did you test with shadows disabled? Shadows disabled gave me a good ~10-15 fps boost.

_Thresh_

Disabling shadows eliminated stuttering, had no effect on frame rate

Zek

For what it's worth disabling shadows has no effect on my performance, either FPS or the stuttering. I also watched my FPS number as the stuttering happened and the stuttering doesn't seem to impact FPS at all for me - it could be running a steady 30fps and I would still see a lot of blatant skipped frames.

Windows 7
Gigabyte P35 DS3L
4gb DDR2 RAM
8800 GT
Onboard Sound

Obraxis

Win7 64bit
i920
6GB DDR3
GeForce 470

Average up from 30 fps in 153 to 40 fps.

Lows of 20fps, highs of 68 fps.

Noticeable improvement! Seems smoother. New GUI not based on Flash works MUCH better, and LOOKS much better! Much cleaner feel to it.

Bugs:

*Went to slash a Crag, game froze, engine did not. Console said something about sound initialisation problem.
*Occlusion is sometimes drawing intermittently the 'glow' for the Hive/CC techpoint where it goes slightly strobe-like.
*Slight stuttering at random points.
*Delay to switching weapons on internet server, mostly from gun to pistol. Not tested on LAN yet.
*GUI fades away as alien when not using energy, but not as marine. Consistency?

That's all for now. once I play around a bit more I'll post up some more. Good work!

Jiriki

You should post your map you're playing. Rockdown has considerably lower rates than Tram.

I get around 30-60 (25 in com mode) FPS in ns2_rockdown, around 40-60 in a lot of places.

Cpu: Intel Core2 Duo E8500, OC'ed to 4.2Ghz
Gpu: Ati Radeon HD4870
Ram: 4096MB DDR2

Narcil

q9550
8gb ram
4870x2
win7 64bit
______________

I get 15-25 fps with crossfire enabled

and 15-25 with crossfire disabled

so something not working when it comes to using both cores.

Also I get flickering shadows, so I have to turn them off with r_shadows

Avalon

Updated with resolution and settings, as well as the map (ns_tram). My shadows were enabled.

If disabling shadows to stop the stuttering does not improve minimum fps as you say, then something is wrong with my game as I am not showing any fps improvement (on paper) aside from max fps. My shadows don't flicker, but the stutter is bad.

Underwhelmed

In HL there was a console command called timedemo that let you do a benchmark based on a recorded demo file. I don't know if there's an equivalent of that in NS2 yet, but if there is, I'd suggest using that instead for the sake of consistency.

Avalon

<!--quoteo(post=1806182:date=Nov 18 2010, 06:28 AM:name=Underwhelmed)--><div class='quotetop'>QUOTE (Underwhelmed @ Nov 18 2010, 06:28 AM) <a href="index.php?act=findpost&pid=1806182"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->In HL there was a console command called timedemo that let you do a benchmark based on a recorded demo file. I don't know if there's an equivalent of that in NS2 yet, but if there is, I'd suggest using that instead for the sake of consistency.<!--QuoteEnd--></div><!--QuoteEEnd-->

Unfortunately we don't have anything like that, so I'm stuck doing repeated runs in order to average out my min, avg, and max to try and make it as close to normalized as possible.