Dienstag • 22.08.2017 • 15:02 Uhr
 Neuigkeiten
 Artikel
 Downloadbereich
 Bildergalerie
 Forum
 PowerVR FAQ
 Suchmachine
 Mitarbeiter
 Impressum
 SGX544MP angekĂĽndig...
 Der aktuelle Stand
 zweiter SGX Artikel ...
 Die Videofähigkeite...
 neue Treiber fĂĽr Po...
 neuer Treiber mit Ha...
 SGX - Aufbau, Techn...
 neuer Serie5XT Grafi...
 Apple und Intel inve...
 Mehrere weitere Lize...
 SGX - PowerVRs Shade...
 Interview mit ImgTec...
 Quack - das andere D...
 Hercules4500 AGP4x-M...
 Optionen des KYRO Tr...
 FSAA auf KYRO
 Nummer 5 lebt!
 Startpartition von W...
 MBX - Mobile Grafik
 Die Grafikfähigkeit...
 Vivid! - PowerVR Gra...
 Videologic Neon 250
 Geforce 6800 - the n...
 Beyond3d
 BluesNews
 Chrome-Center
 Hartware
 Imagination T.
 ParaKnowYa
 Planet 3dNow!
 PowerVR
 PowerVR-Ext.
 PVR Generations
 Tom's Hardware
 Videologic
 VoodooExtreme

3D Center


PowerVR

Imagination Technologies

Geforce 6800 - the next generation

There were many that claimed that UT2k4 is what UT2k3 should have been all along. I can’t entirely disagree, because the addition of Onslaught maps, new tactical game play, vehicles etc. are exactly where the first excels compared to the latter.

UT2k4 contains many maps of it’s predecessor, some of the official UT2k3 bonus pack and of course adds many new ones.

I’ve read quite a few arguments on how to benchmark either/or game. There’s one side that claims that in UT2k3 flyby scores do not represent final game performance, while botmatch scores do; as there’s the opposite that claims that flybys are more accurate since botmatches are way too CPU bound. The truth – as in many other occasions – lies somewhere in between.

You don’t play the game without any opponents (or bots if you’re in single player mode), but it is true that botmatches are even more CPU bound than the game actually is. First I used UMark to test the game with.

FLYBY:

4xAA / 16xAF

CTF-Citadel @ 1280x1024 : 107.28

















































100%
CTF-Face3 @ 1280x1024 : 38.90

















































36%
DM-Inferno @ 1280x1024 : 84.55

















































78%

CTF-Citadel @ 1600x1200 : 72.78

















































100%
CTF-Face3 @ 1600x1200 : 27.56

















































37%
DM-Inferno @ 1600x1200 : 55.13

















































75%

Just looking at the flyby results at 1280*1024, one could easily think that the game is actually playable with full details enabled in that resolution. Now let’s see how things look like in botmaches with 12 bots.

BOTMACH:

4xAA / 16xAF

CTF-Citadel @ 1024x768 : 58.55

















































97%
CTF-Face3 @ 1024x768 : 41.28

















































68%
DM-Inferno @ 1024x768 : 59.97

















































100%

CTF-Citadel @ 1280x1024 : 53.69

















































100%
CTF-Face3 @ 1280x1024 : 30.29

















































56%
DM-Inferno @ 1280x1024 : 37.03

















































68%

Playability is right on the edge on 1280 in these maps, yet I’m still only half way through in my attempt to illustrate reality.

Time to see how an ONS map fairs with 16 bots this time. For that I borrowed 3DCenter’s Primeval demo. Again full in game settings and 4xAA/16xAF enabled:

1280*960*32
16.574 fps

That was a tough case; granted the demo is attempting to illustrate one of the absolute worst-case scenarios, yet Onslaught maps are that demanding on systems/GPUs in reality.

I then tried to reduce details both in the driver control panel as in the game. 2xAA/4xAF this time and in the game physics detail= low, dynamic mesh LOD= high, character shadows = blob.

1280*960*32
24.652 fps

Quite a bit better, yet I’m still staring at purple and red framecounters. Last resort: same reduced settings, lower resolution.

1024*768*32
43.443 fps

I played after that through a couple of ONS maps and actually used 1152*864*32 and it was perfectly playable with the above settings. Sadly enough none of the available so far time demos, or benchmarking applications include that resolution to test it with.

One more thing I did was to revisit UMark and give Primeval with 16 bots enabled and max details another run, to see if I get similar results here too. Note that the application doesn’t support 1280*960, but just 1280*1024, meaning that results aren’t 100% comparable.

Primeval
16 bots
Maximum detail
4xAA / 16x AF

ONS-Primeval @ 1280x1024: 15.75

















































100%

For the end I tried one of the very popular CTF maps Moon Dragon.

UMark
1280*1024*32
Maximum detail
4xAA / 16xAF

CTF-MoonDra @ 1280x1024: 47.52

















































100%

To verify that score, I fired up the game this time with FRAPS enabled and played a fair time with the exact same settings in Moon Dragon:

CTF-Moon Dragon
Frames: 8844 - Time: 205140ms - Avg: 43.112 - Min: 19 - Max: 96

UT2k4 is a very demanding FPS game. It absolutely needs both a powerful CPU and GPU and definitely more than 512MB host ram, which isn’t that much different with recent demanding FPS games either. My goal though was to illustrate an entirely different picture than usual flyby scores or in-door benchmark demos that get usually used do.

Seite 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
 Geforce 6800 - the next genera...
 17 Seiten
 verfasst von Ailuros
 Donnerstag - 26.08.2004 - 18:20 Uhr
[1]  Introduction & Systemsetup
[2]  Antialiasing & Anisotropic Filt...
[3]  2. Anisotropic filtering
[4]  Synthetic Applications
[5]  F1 2002
[6]  Rallisport Challenge
[7]  Serious Sam: The Second Encount...
[8]  Need For Speed: Underground
[9]  Unreal Tournament 2004
[10]  Far Cry
[11]  Hybrid Super-/Multisampling Mod...
[12]  Hybrid Super-/Multisampling Mod...
[13]  2D, DVD / VIDEO Playback
[14]  Summary
[15]  Softmode
[16]  doom3
[17]  DVD