UT2K3 DM-Icetomb performance feedback needed ASAP!

[Hellbinder]

Actually I messed up on the hothardware one. That only shows a 5900 not a 5900U. (Wich happen to be priced similarly)

Here is the correct one..

http://www.hothardware.com/hh_files/S&V/asus_fx5900u(4).shtml

(Yes i skipped the Gunmetal results in both cases... becuase dammit its a freaking Biased out the a$$ Benchmark period. If you have to ask why ill explain later. The results are pretty close but it pisses me off)

For those of you keeping count, we ran 34 different tests with these card. In 19 of the tests, the GeForce FX 5900 Ultra based V9950 Ultra outperformed the competition. In 15 of the tests, the Radeon 9800 Pro was the victor. Winning in 56% of the tests doesn't seem so bad, until you realize the 9800 Pro we tested with, is available for about $200 less than the V9950. Ouch...

Uh huh... Of course I threw out the cheesey Gunmetal benchmark resulsts. Still Ati lead most of the D3D based tests they showed.

OpenGL.. Serious Sam "Little Troubble" demo is so freaking "App Detected" by Nvidia its not even funny. It has been for a long long time. Lets see some actual game play with Fraps results. I can Gurantee you who will come out on top.

[StealthHawk]

Originally posted by Hellbinder
:rolleyes: He ran the tests in HQ mode. It says Right on it.

Secondly the 500mhz e-Vga card never worked right from day one. They released a bios for it to lower clock speeds and its not been heard from since.

I thought that eVGA released a new BIOS and cards worked properly at 500MHz? That is what I remember reading here a few months back.

Also note that I am Quite sure than Nvidia is Cheating in Splinter Cell by App detecting this specific level (always used for benchmarking). B3D and other places have Nvidia getting pummeled in Splinter cell. I also Question their Nascar 2004 (EA fix is in). I have a my doubts about IL2 as well.. But...

B3D actually ran Splinter Cell with anti-detect and scores were the same. So it appears that NVIDIA actually wasn't cheating in Splinter Cell(in the past at least). I'm surprised too, but it's actually one instance where NVIDIA is clean. Amazing, but true.


P.S. For those of you who don't quite get it, many of us are claiming that the flyby benchmarks in UT2003 may very well be rigged by NVIDIA. It seems odd that NVIDIA cards perform so well in comparison to ATI cards in the most publically benchmarked flyby modes but are basically on equal footing in the aft looked at flybys. Is something fishy going on? Looking at the dramatic increases* from new drivers it's very likely.

*Look at reviews with early drivers then watch scores shoot up with newer drivers. For example, look at old [H] reviews.

[fivefeet8]

Originally posted by Behemoth
yes this is sucky IQ, in my defination:
http://www.nvnews.net/vbulletin/attachment.php?s=&postid=198620


Strange how you showed the one shot with the fog missing. BTW, I think it actually looks better without fog. But not show the other shots I did with no gaping difference at all.

Yes, I do think that Nvidia shouldn't disable Trilinear filtering completely and let the user's choose it. But honestly, from a quality standpoint, trilinear and non trilinear IQ aren't that different save the floor textures. But that's just my opinion. In the end, if we can play our games the way we want, then that's really all that matters. For some ATi is how they like it, for other(a few others :D) Nvidia is fine for them.

And let me re itterate what was acknowledged by my 51.75's/45.23's dets comparison shots of UT2k3 DM-ICetomb/DM-rustatorium shots. Very little IQ loss if any, 2x-3x performance gain. I also posted the shots in other message boards and the consensus was the same.

It's gonna be a while before I can do my experiment. Family problems and all.

[Spotch]

Screen Capture Detection! Screen Cqapture Detection! These screenshots are useless! This was alleged as being in the Det 50 drivers by the evil Gabe. Somebody please prove or disprove it so we can stop wasting time on the IQ question.

Somebody take a picture of the screen and post it for IQ comparisons! :D

[Flavius]

BTW, I think it actually looks better without fog So? The point here is that the fog is SUPPOSED to be THERE coming out of that ceiling light, as a level design decision by the map maker! So any card that doesn't show the level exactly as it was meant to be played has bad IQ! Missing details = bad IQ! It's not just based on what you CAN see!

And yes, I think Nvidia is detecting UT2K3 Flybys and optimizing for them. My ATI flybys were WAY closer to real world performance in the game than my FX flybys were. Nvidia beats ATI in the flybys, but ATI beats Nvidia in game? Something real fishy there!

[StealthHawk]

Originally posted by fivefeet8
Yes, I do think that Nvidia shouldn't disable Trilinear filtering completely and let the user's choose it. But honestly, from a quality standpoint, trilinear and non trilinear IQ aren't that different save the floor textures.

That may or may not be true. For example, I can't tell the difference between ATI's 8x Performance AF(bilinear) and 8x Quality AF(trilinear) from static screenshots, or from the ground in FPS games. There may be a difference, but I'm not sure. I haven't exactly recorded movies to find out. But I can tell the difference when looking at walls.

[OWA]

Quite often.. Nvidia is winning some of these tests because Fog is removed. This goes all the way back to the Nv30 at UT2003. It continues today. Especially the 50 series drivers which completely remove Fog for speed.
When comments like the above are made, is the fog that is being discussed a smoke effect / "real" fog (heh, like the fog in the movie The Fog) or are you talking about the technique games use to avoid having to render a lot of detail in the distance (view distance setting) or is it something other than those two?

[The Baron]

Smoke-type effect. Something that impairs visibility but requires a lot of rendering power and such to compute.

[fivefeet8]

Considering that the whole DM-Icetomb map is slow with the 45.23's anywhere you ran to, why would the missing roof fog make such a difference and need to be removed? The rest of the map didn't have the fog and it still ran poorly with the 45.23's. With the 51.75's, the map is running 67 fps on average with your Icetomb benchmark.

BTW. I've gotten the screenshots posted:

51.75's or 45.23's. (http://www.geocities.com/pao_vang5f8/Screenshots/PC.htm)

2 shots are from DM-ICetomb with different dets. 2 shots from DM-Rustatorium with different Dets. Which shot is with which driver?

[Flavius]

Fivefeet8, this thread was never about average FPS, it was about MIN FPS. I don't care if I average a playable 45FPS if there are spots on the map that drop me to 6 FPS! I consider than unplayable. And this is about playability at high quality settings. Some people would actually like to play UT2K3 instead of just benchmarking it. There are other places in Icetomb where that fog exists, NOT just in that chamber, and the fog does cause big dips. Why? Ask Nvidia, not me.

[fivefeet8]

Originally posted by Flavius
Fivefeet8, this thread was never about average FPS, it was about MIN FPS. I don't care if I average a playable 45FPS if there are spots on the map that drop me to 6 FPS! I consider than unplayable. And this is about playability at high quality settings. Some people would actually like to play UT2K3 instead of just benchmarking it. There are other places in Icetomb where that fog exists, NOT just in that chamber, and the fog does cause big dips. Why? Ask Nvidia, not me.

Guess you didn't look at my DM-Icetomb fraps bench posted on this thread earlier. The minimum fps I got was 36. And that's with 8 bots running around other places in the map. Even using Baron's Icetomb bench demo, the lowest fps I saw was 30. And these benches were done at 1024x768, 4xAA, 8xAF. Show me a shot where there is fog because even using the 45.23's, I only saw a lot of fog in the chamber. Not much else.

And you did ask what minimum FPS we were getting with DM-Icetomb right? How are we suppose to find that out without running some benches? BTw, I did actually play it while I was benchmarking with fraps. You can do both.

[Flavius]

Guess you didn't look at my DM-Icetomb fraps bench posted on this thread earlier Guess YOU haven't been reading this thread from the beginning. The 5900U has a 4X slower FPS in some spots of Icetomb than a 9800P at the same exact details settings and similar system specs. Confirmed by multiple people. We're talking 40FPS compared to 10FPS here. This was before 50.xx were leaked. I don't care about Nvidia's new ultra cheating, IQ killing, unreleased drivers. I give up. :rolleyes:

[fivefeet8]

Originally posted by Spotch
Screen Capture Detection! Screen Cqapture Detection! These screenshots are useless! This was alleged as being in the Det 50 drivers by the evil Gabe. Somebody please prove or disprove it so we can stop wasting time on the IQ question.

Somebody take a picture of the screen and post it for IQ comparisons! :D

From comparing my shots while actually playing the game and going to the same scene, there is NO screen capture detection.

[fivefeet8]

Originally posted by Flavius
Guess YOU haven't been reading this thread from the beginning. The 5900U has a 4X slower FPS in some spots of Icetomb than a 9800P at the same exact details settings and similar system specs. Confirmed by multiple people. We're talking 40FPS compared to 10FPS here. This was before 50.xx were leaked. I don't care about Nvidia's new ultra cheating, IQ killing, unreleased drivers. I give up. :rolleyes:

Actually I did, and I posted scores from the 51.75 dets which have increased performance in those two maps 2 to 3 times. And if you really think they killed IQ with these drivers, look at my link above and tell me which do you think are the shots from the IQ killing 51.75's. :rolleyes:

One more thing, I'm not even arguing the point that the 45.23's are very slow with these 2 maps. It's been proven by others and my own testing.

Cheat, optimization, or pixel shader program shuffling? WHO CARES. As far as I have seen, nothing else is different from the 45.23's and 51.75's for Ut2k3 save the missing roof fog in IceTomb and the huge performance increase.

[Flavius]

Haha, ok, maybe they did increase UT2K3 to proper levels, that would be good, but not enough for me to go back after all this. And they DID lower IQ. I can tell EASILY that UT20032003i.jpg is from the 51.xx drivers. It is missing texture detail in certain areas.

[fivefeet8]

Originally posted by Flavius
I can tell EASILY that UT20032003i.jpg is from the 51.xx drivers. It is missing texture detail in certain areas.

So your saying that the first Icetomb shot is from the 51.75 drivers and the second Icetomb shot is from the 45.23? ok. Could you by any chance point out the missing texture detail?

Lets wait and see what other people say as well. How about the Rustatorium shots?

[Flavius]

The Rustatorium shots are terrible. Too dark and chaotic to tell anything. But I stand by my Icetomb claim. This exercise is a bit pointless since there is already a really good web article showing how the 51.xx has *obviously* lowered the IQ.

As for pointing it out... there is missing texture detail in Icemtomb shot on left and right walls.

[fivefeet8]

Originally posted by Flavius
This exercise is a bit pointless since there is already a really good web article showing how the 51.xx has *obviously* lowered the IQ.

As for pointing it out... there is missing texture detail in Icemtomb shot on left and right walls.


It's a test of percieved IQ and to what extent has IQ been lost that it's actually noticeable. Many people have claimed that the IQ loss with the 51.75's in UT2k3 is quite noticeable. I'm simply testing that claim. As for your texture detail in the walls, in that scene there are water falls running down the walls. As the shots are not taken at the same exact time, there are differences in how the water looks as it's falling. Make sure you have distinguished between the two.

[aapo]

Originally posted by Flavius
And they DID lower IQ. I can tell EASILY that UT20032003i.jpg is from the 51.xx drivers.

Odd. To me it seems UT20032003i1.jpg has slightly worse quality. But it's hard to tell, because for some reason the textures are different. Shouldn't the textures be the same? Why do they depend on drivers? :confused:

EDIT: Uh-huh - I think I was looking at the waterfalls. Can't see much else, only couple of slight differences in the lightning, but they might be originated from the different viewing positions. So, zero differences in the still shots for me. :)

[Sphinx]

Because - The Way it meant to be...

[StealthHawk]

Originally posted by aapo
Odd. To me it seems UT20032003i1.jpg has slightly worse quality. But it's hard to tell, because for some reason the textures are different. Shouldn't the textures be the same? Why do they depend on drivers? :confused:

The pictures aren't from the same frame. And shouldn't we be using pngs instead of jpgs when trying to point out IQ differences?

[fivefeet8]

Originally posted by StealthHawk
The pictures aren't from the same frame. And shouldn't we be using pngs instead of jpgs when trying to point out IQ differences?

Yeah. Because it's hard to get the same frame after switching to different drivers and then try to position them exactly the way the other shots were taken. They are pretty close to the same frame like the other shots that was posted earlier. And differences in IQ were also claimed from them. Hm. Maybe I'll make a Ut2k3 demo and use it to get the same frames. I'll post it later, but tell me what you think which drivers the shots came from.

The JPEG compression was done at maximum quality with photoshop on all the shots. There shouldn't be much difference from PNGs. I'll try posting some PNG's later if you want. But the files sizes might be larger.

[saulin]

Sorry but you guys are full of it. You talk trash about major IQ differences and now you can't even tell the difference.

Get a life guys.

I just give up. Nvidia IQ is more than just nice!!! is awesome.

BTW Nvidia isn't the only one cheating.

ATI cheats in 3DMark and cheats in Aquamark and in UT 2K3 just so you know.

[hovz]

i think the i1 jpg looks better, the tetxures on the bottom left part of the ledge are sharper, and the chains seem less blurry, and it just seems sharper in general

[The Baron]

Originally posted by saulin
ATI cheats in 3DMark and cheats in Aquamark and in UT 2K3 just so you know.
They certainly don't cheat in 3DMark. Or Aquamark, as far as anyone knows. And UT2003? Nope, not there either.