What frame rate do you get in UT?

[oldfart]

Here is mine: Same setup. 1024 x 768x x 32. Everything max detail, fog, etc.
Note: To get an accurate benchmark in UT, you MUST set "min desired framerate" in preferences to 0!! If not, UT will turn off details when that number is reached making the score invalid!! This is especially true for UTbench.

UTbench:
D3D 37.3
OGL 34.5

Not impressive numbers, but actual gameplay is very smooth either way.

Thunder:
D3D 70.8
OGL 65.4

PIII 550 @ 733
256 ram
Radeon 64 183/183
Win98 SE

 

[TAsunder]

Yeah UTBench. Actually when I run it 3-4 times (i.e. give UT time to load stuff into memory) my numbers are more like:

D3D: 41
OpenGL: 39.6

not bad

 

[BenSkywalker]

Oldfart-

First off, thanks again for all of your help with Radeons and this patch, it is much appreciated

"TAsunder, Ben, the ini that I'm using that works on the Radeon looks a little different than the one you posted."

Yes, those are not critical settings, I set it that way for a quality/performance reason. For DoPrecache I found that I can pretty much eliminate all disk swapping if I set cache size to 16MB and then enable it. This is my system with 192MB of RAM running a GFDDR, so results may vary.

For Trilinear, I don't play anything I can avoid without it. I also enable anisotropic filtering in display properties. To me it is worth the minimal performance hit to eliminate the mip map "lines" on the screen. This artifact bothers me significantly, I would probably be willing to take 10%-25% hit in performance to get rid of it, but it ends up being much less then that.

I checked out the screens you sent me, they look real nice, thanks again for all your help

TAsunder-

Glad you got it up and running, how do you think it looks? To me it seems like another game on many of the levels, Stalwart and Convex being two of the most dramatic IMHO. Could you post where to get UTBench? I have Thunder and EH1 that I mainly use as they are more fillrate intensive, but I would like to have some numbers for comparison. Unfortunately I'm stuck with a lowly Athlon 550 so it is likely that the numbers above will b!tch slap my rig

How is gameplay for you running the Loki patch? It seems smoother then D3D for me(as long as I precache), although that may be at least partly due to nVidia's killer OpenGL support which always seems smoother then their D3D(the original Unreal benched almost 50% faster for me with D3D, but t was only playable in OpenGL).

Robo-

"eh? wheredat???? wherefore I fidn S3TC software decompressor?"

Here you go. This will work as a plug in for Photoshop(and compatibles). This won't load .utx, but it will decompress S3TC textures so you can recompress them yoursefl(or save them as .bmp or whatever you want to do).

"eh? Whatfore you say that?"

NDA.

 

[oldfart]

Ben, you can get UTbench HERE. Be sure to set min desired framerate to 0 when you run the benchmark.

I guess its true about UT being a CPU benchmark. I just overkclocked my Radeon from the default 183/183 to 200/200. I saw a 0% improvement. I ran a Q3 test and picked up 5-6 fps.

 

[BenSkywalker]

26FPS

Have to hold out for SMP DDR Durons......

I'm now having problems with D3D, I know I hosed my .ini file again(twentieth time or so) trying to troubleshoot another issue, so I can't post comparitive numbers(it was running at around 8.5FPS when I gave up and exited out). I'll post back after I clean rewrite the stinking thing again.

I guess 26FPS isn't too bad for a lowly 550, but I'm used to the ~50FPS benches like EH1 and Thunder. Thanks for the link

 

[oldfart]

Yeah, nothing like UTbench to make you feel like you need an upgrade.

 

[TAsunder]

As far as gameplay goes, it is very smooth. In the past couple of days I did the following and overall UT runs faster now than ever before:

1- went from 224 mb of pc-100 cl3 ram to crucial pc-133 cl2 ram
2- got a radeon
3- went to the UT tweak page listed above
4- installed s3tc and the loki opengl dll

It plays very well.. however I can still notice some slowdown from time to time. There are some levels that play very slowly (but still playable) such as Zeto (I think that's the one) under the challenges. Occasionally in OpenGL mode, UT will freeze for 1-3 seconds for no apparent reason. However, the game looks freaking gorgeous with s3tc. It definitely plays better on the average than when I had a v3 3k, but the minimum frame rate is lower.. I assume this is due to excessive texture memory swapping stuff.

Here's another question: a friend of mine has a duron 700@700 and an asus gf2 gts 64mb. In Q3 his gameplay is very jerky, and in UT with s3tc he gets 20fps average. What's the deal with that?

 

[Eug]

26 FPS??? Ouch.

I find UT very annoying to play on average maps once I drop below 30 fps in UTBench. You're running an Athlon 550 and a DDR, is that correct? I'm just wondering. What resolution? By the way, I find the numbers in UTBench much more accurate for "real life". I ran EH1 and it was a like a walk in the park and did not compare at all to real play online. In fact, when I play, often it's MORE intense than UTBench and the whole setup slows to an absolute crawl even though I've set mine up to get >40 FPS in UTBench. If I were able to get 60 fps somehow in UTBench then I think I'd be relatively happy, but that ain't gonna happen. My numbers are listed above in another post... Glide and Voodoo rock when used with a fast CPU at medium resolutions, but I must admit that the comparison is not fair because of the limitations of both Glide and the video card. However, nonetheless, after getting used to these speeds I don't think I could ever accept the speeds you're getting for anything other than testing and benchmarking purposes.

 

[BenSkywalker]

Eug-

"You're running an Athlon 550 and a DDR, is that correct? I'm just wondering. What resolution?"

1024x768 32bit color everything cranked to max(including S3TC and all ~preferences options set to true), that's the way I play.

I don't have a real problem playing through most boards, even the reactor level with 16bots on. For me, EH1 and Thunder are closer to the FPS I see in game, I have more kills with the Sniper Rifle then any other weapon, I have more HeadShots then I do multi kills, the way I play(same as Quake3, it just works much better with the slower pace of UT) luckily I don't see my FPS drop as badly as they do running UTBench.

TAsunder-

Try upping your cache setting to 16MB-64MB and dropping the detail level of the player models. I'm not sure off the top of my head what the tweak sites say about the "DoPrecache" option, but from my own testing it is better to have it enabled then disabled when running the S3TC textures.

Upping the Precache amount and enabling it all but eliminated the stalls that you mention for me, I was having the same problem but now it is every once in a great while I get a quick "hiccup" that lests for a fraction of a second. With the original patches(4.02/FG's) it was as bad as twenty to thirty seconds at times(ouch!). I'm running 192MB of RAM, you have more then I do so perhaps with a large enough precache you can completely eliminate the stutters(they are nearly gone for me, and better then the ~10FPS for three or four seconds I get once in a great while with D3D).

 

[Eug]

<< EH1 and Thunder are closer to the FPS I see in game >>


Yeah, maybe I should rephrase that. For me EH1 better simulates the average numbers while playing, but UTBench better simulates the worst case scenario with online play. For me the worst case scenario is the most important (for obvious reasons). I haven't tried Thunder.



<< I have more HeadShots then I do multi kills, the way I play(same as Quake3, it just works much better with the slower pace of UT) luckily I don't see my FPS drop as badly as they do running UTBench. >>


Hmmm... It does seem the default bot speed for UT is slow, but that can be adjusted. However, I generally don't play with bots anymore because real people are much faster, more unpredictable, and much more fun to frag. In any case, I find online the pace of both Q3 and UT are similar, but it depends on the map design and size.

 

[RoboTECH]

<giggles>

all that trouble, and all ya gotta do to get the baddest ass UT performance is to pick up a 5500

hehehe....

BTW Ben, thanks a bunch for the link. Gonna check it out this weekend and see what I can do.

 

[TAsunder]

I have my cache size (that's the one u set in advanced properties under &quot;cachesizemegs&quot; right?) at 256 meg... the halts occur once every 45 minutes or so, usually at the beginning of a level (thought not at the VERY beginning).

 

[BenSkywalker]

&quot;I have my cache size (that's the one u set in advanced properties under &quot;cachesizemegs&quot; right?) at 256 meg... the halts occur once every 45 minutes or so, usually at the beginning of a level (thought not at the VERY beginning).&quot;

I have found that upping mine too high makes matters worse. It will &quot;pause&quot; less frequently, but do so for quite a bit longer. With mine set for 16MB I get hiccups every now and then, but they are a fraction of a second. Upping it to say 128MB reduced the time between the hiccups but they became significatnly longer, multiple seconds. Dropping it down to 4MB, the default, crteated choppy FPS far too frequently(like when entering a room for the first time).

 

[RoboTECH]

<< I have more kills with the Sniper Rifle then any other weapon, I have more HeadShots then I do multi kills, the way I play(same as Quake3, it just works much better with the slower pace of UT) >>



ahhhh....so yer a bastard camper, eh? hehee.....