• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

I need some help with UT....

B

BenSkywalker

Diamond Member
This should be simple for those that know how to do it/where to to get it, I need some UT benching tips and good benchmarks along with how to run them.

The Flyby is OK, but I was hoping for something a bit more fillrate intensive and less CPU dependant for one bench and a "Crusher" for another. Where can I download any of these types of benches and then how do I run them?😛

Just bought the game yesterday, I was taking a stand and not purchasing it until Epic fixed Unreal which I didn't realize they did a month ago😱 (much better with the latest patch btw)

Thanks for all help tips anyone can offer🙂
 
You can take a look at S3's super high resolution levels. The textures in this put the nVidia TnL demos to shame. They will bring your CPU to a crawl. Only problem is finding them. I used to have em' but I lost em' in a reformat and I haven't found them again yet.
 
Would they be on S3's web page? I'll dig around and see if I can find them. Does anyone know if the compressed textures will work on any other card besides an S3/MetaL board? According to UT it will work with any board that has DXT1 or S3TC but I've seen it mentioned several times that it only works with MetaL.

Thanks for the reply Trinitron, I've found a couple of benches over at Rev's though both are more CPU intensive then the Flyby, all I need now is to find a fillrate bench or figure out how to create one myself. Definately want to check those S3 levels also.
 
just play a round of LAVA with 15 other bots, that should give real world benches...use timedemo 1 command to initiate fps benching
 
dl-

That would work good for figuring out my own systems performance, I'm more interested in trying to find/create a good bench for UT that is more fillrate intensive then all the others which rely far too much on CPU power to make for a good bench. I've been looking at how the engine works and seeing what circumstances make it so heavily dependant on the CPU and think I know how to get a good bench that scales with fillrate demands. I'm not truly interested in the in game performance as I know I have no problems handling it at all, I want a good D3D bench and there aren't any. Right now I'm using about a dozen different D3D benches because none of them are close to being as good as Q3 in terms of scaling with both FP and fillrate power.

Leon-

"Ben, you can record your own demo🙂 Just use demorec command"

That's what I'm talking about😀 OK, so how do you stop recording when you are done and is there a default .dem file that it saves the demo to or does it offer for you to name it yourself?? Perhaps the UT/system file??

I am quite illiterate in terms of the Unreal console commands, never had a good reason to learn them. Thank you very much for the demorec tip😀, as soon as I figure out/someone tells me the rest(stoppiing/file) I'll be all set.
 
DLing now, thanks🙂

I figured out how to stop recording demos and you are asked to name files before you start recording so I have that covered (stop demo for those wondering).

I'll post back as soon as I have some UTBench numbers, I hope its' fillrate intensive(well, as much as possible). I've been working on getting a decent bench and have figured out it is much easier to create a "Crusher" then it is a "Timedemo1" 😛

Thanks again for the link🙂
 
Yeah, UTBench is what most use. AnandTech does not for whatever reason, so their numbers are completely different from everyone else's - much higher.

I find that a 40 fps bench in UTBench is very playable (like a 60 fps in Q3), but in the most complex (well-designed) scenes it can still slow down enough to be annoying. Under 35 it is irritatingly slow, even for this casual gamer. I wonder why the huge descrepancy in numbers? Is the UTBench that much more complex than demo001? Or is just because of the engine?
 
UTBench is definately a good CPU bench, but Rev's Wicked400 puts it to shame in that respect(I get 50% higher FPS with UTBench). Hmmm, I guess I'll have to create one of my own.

Has it been posted on any web sites yet how to enable use of the high quality(ultra high quality compressed) textures without having an S3 board?

Curious as I don't recall seeing it.
 
"UTBench is definately a good CPU bench, but Rev's Wicked400 puts it to shame in that respect(I get 50% higher FPS with UTBench)."

Can you explain? I'm not sure what you mean. Are you saying that it's CPU limited?

As for the 50%... at 800x600 maxed I get about a 16% increase with UTBench over Wicked400. Numbers are 40.4 and 34.7 respectively. (V3 179 with Celeron 897)

By the way, if you're building your own demo, check out DM-SuburbanDelux at http://www.planetunreal.com/nalicity/reviews/dm-suburbandelux.html. It's gorgeous but it's unplayable because it's far too slow on my machine even with just a couple of bots. I don't why it's unplayable, but it is. (I don't understand the terminology, but it's supposed to have over 500 polygons in the most complex scenes, when the recommended number by the mappers is far less than 200 for best playability.) I don't know if this would suit your needs.

 
I'm good at UT, just a moron when it comes to benchmarking...how do you use UTbench?(feel free to skip the "Unzip the file" step, I'm not that big of a moron)
 
~
timedemo 1
demoplay utbench
~

If you're on the menu screen and see no demo but hear the music, then press ESC to get to the right screen
 
Eug-

I'm on an Athlon 550 and GF DDR, the benches seem to run a bit differently. I'll have to rerun them both using 640x480 16bit and check my scores again.

So has it been posted anywhere how to use the high quality textures on CD2 on any sites yet on a non S3 board?

Also- Does anyone know what level that shot Sharky used a while back was in??-

http://www.sharkyextreme.com/hardware/reviews/video/diamond_viper2/9.shtml

Trying to find that "box", any ideas?
 
BenSkywalker - Did I get you thinking again about UT benchmarks <grin>? The problem I have with Anandtech's UT benchmarks is that the recorded demo is not available for download. That means that I can't try the same demo on my system nor can I view it to see what elements are in it. I really wish they would make it available.

I would suggest that Tim Sweeny would be a good person to contact. He appears to be fairly accessable and should be able to give the best information on benchmarking UT.

Michael

ps - here's some links that might help:

http://dynamic.gamespy.com/~tooreal/unrealtournament.cgi
http://www.3dspotlight.net/tweaks/ut/ut-tweak-13.shtml

The second link has great info if you read from the very beginning

Here's the beginning:

http://www.3dspotlight.net/tweaks/ut/index.shtml
 
Ben, im kinda an idiot when it comes to fill rate and things to do with video cards, but i have a suggestion if your trying to make a fill rate intesive benchmark. Their is a map called ancient, I think that this may be what your looking for, the map is very nice, lots of eye candy, with very high ceilings, and lots of light sources (i think) the map has kinda fallen out of favor with many people because it took alot of power to make the thing run at exceptable framerates. Here is the link, ill see if i can make a demo from it myself. (http://www.planetunreal.com/dl/dlredirect.asp?file=\planetunreal\nalicity\utctf\ctf-ancient.zip&amp;url=ftp://sunsite.doc.ic.ac.uk/packages/idgames2/planetquake/planetunreal/nalicity/utctf/ctf-ancient.zip
 
I fooled around a little with UT benchmarking myself today, just to see what my new V5 5500 could do. Since FSAA is the one big advantage, I tried the cityintro benchmark with max FSAA enabled.

C566->850, ABIT BX6R2, 128 meg PC100 memory (I think CAS3 but might be CAS2, I don't feel like rebooting right now to check), Mx300 sound card with A3D enabled (cause that's how I would play the game), a couple of NICS, a voice/fax modem, USB posts being used, etc.

All tests 4X FSAA and 800x600

MaxFPS set to 0 and all other settings on &quot;high&quot;, vsync disabled, use 3dfx set in the D3D settings, latest drivers from 3dfx.

Demo run three times and 2nd score taken.

LOD slider at center, card not overclocked.

GLIDE Average 43.85 Min 23.12 Max 90.82

D3D 16 Average 51.4 Min 30.63 Max 90.16

D3D 32 Average 26.4 Min 11.78 Max 40.99

The interesting thing here is that D3D is faster than GLIDE for the V5. I had thought that it was strange that D3D was used in the Anandtech V5 review, but that actually may have helped the score (there was a bug in D3D where resolutions over 1280x960 went into single chip mode, I don't know if 3dfx fixed that yet in the new drivers).

There was barely any differences between the three runs at each setting, which gives me some confidence that the benchmark ran ok.

Note - these benchmarks are not from a &quot;clean&quot; machine and NAT32 plus RC5 were running in the background.

Michael
 
Thanks for all the replys guys, I only have a quick minute now but I will respond properly tomorrow. I have read through the UT tweaking article and will dl those extra levels to check them out also. Thanks again🙂

Michael- You have a private message.
 
Michael-

&quot;Did I get you thinking again about UT benchmarks <grin>?&quot;

Hehe, yep, it was you who got me into my current fixation(well, the bench fixation ended up leading into the other one).🙂

&quot;I would suggest that Tim Sweeny would be a good person to contact. He appears to be fairly accessable and should be able to give the best information on benchmarking UT.&quot;

I've been looking into how the engine works and I may end up having to build my own level if I want to get a fillrate bench out of UT, I haven't seen one yet that doesn't strain the CPU moreso then the graphics card. Perhaps Sweeney could help me out with comparable ratios of CPU/fillrate limits of different levels, I have noticed that he tends to be very prompt when any web sites email him a question, even the smaller sites.

Thanks for the link to those pages, I read them through and made note of any parts that I didn't have down yet.

For benches, I need to run up a batch, but first I need to uninstall and reinstall the game(texture reasons) to get some more screenshots.

Snoop-

&quot;Ben, im kinda an idiot when it comes to fill rate and things to do with video cards, but i have a suggestion if your trying to make a fill rate intesive benchmark.&quot;

Fillrate benches tend to be very easy to run(high FPS at low res), they just drop FPS with each resolution bump. There are a few exceptions but that tends to be due to overdraw. I'm dling the file you linked now though to check it out🙂

Eug-

OK, the other file is DLed, now I'm going after Suburban Deluxe. I'll tell you one thing though, the poly counts that guy mentions are far too low, Quake2 uses about 3,000 polys on screen at once and Quake3 uses roughly 10,000. Once I have it I'll take a look but the numbers sound way off.
 
By the way, I was checking the UT on-the-fly specs and it states SuburbanDelux has around 700+ polygons in the most complex areas with 1350+ nodes as well, whatever that means. That's in the living room.

In contrast, I had an extremely hard time finding spots in Gothic (which is used for UTBench.dem) of hitting either 180 polys or 400 nodes.

I don't know what these numbers mean, but in any case, the numbers do jive with what the reviewers say, as measured by the UT software itself.
 
Eug-

How do you activate the on the fly stats?

I work with 3D models, and looking at the characters alone I would have them in the 300-400 poly range each, just a single player model. For Q3 it is roughly 10,000 polys on screen at a time, that doesn't include overdraw as mentioned in the thread. Clearly that is only a rough estimate as the particular area you are in, what you are looking at and the number of bots can significatnly alter that number.

If you have 3DMark2K run the high poly test, a reasonably powerful CPU running raw poly code can push in the eight million polys per second range, 80,000 polys per frame at 100 FPS. A GeForce, running a properly optimized engine can push well over ten million polys per second, 100,000 polys per frame at over 100FPS, a GF2 is even faster.

Not trying to argue with you at all, just that thread is extremely misleading and the numbers aren't matching up with what is appearing on screen. When I get some spare time I'll hack up a NV15 stye level for UT with massive poly counts and see what the engine does🙂

Does anyone know how to export the level models to a standard 3D format?? That would be the easiest way to check, I can select an area and post the exact number of polys for an entire level. I know that the editor can import at least .dxf, can it export as well?
 
~
stat fps
~

You may also want to check UnrealEditor to figure out what the numbers mean, because the mappers seem to talk about different polys for different things. eg. bots don't count, etc. Just relaying the info.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top