UT2K3 DM-Icetomb performance feedback needed ASAP!

[Behemoth]

and ati's AA *always* looks better than nvidia's ;)

[The Baron]

Originally posted by Behemoth
and ati's AA *always* looks better than nvidia's ;)
Well, not if you're if you're looking at an alpha texture. Then anything that uses supersampling will look better.

Problem is alpha textures are all 1998ish.

[Behemoth]

Originally posted by The Baron
Well, not if you're if you're looking at an alpha texture. Then anything that uses supersampling will look better.

Problem is alpha textures are all 1998ish.
damn, you seem to know everything :)

[The Baron]

Originally posted by Behemoth
damn, you seem to know everything :)
No, I just have a good memory. Dave Baumann, Anthony Tan, Uttar, those guys knows everything.... but one day, I'll catch up... ;)

[fivefeet8]

Why would Tri Linear filtering be disabled in UT2k3 anyways? It doesn't decrease performance that much does it?

Tri Linear filtering is still enabled in other PC games? Or no?

[saulin]

And NVIDIA'S *always* looks worse in UT2003 because of the lack of trilinear filtering.

lol sorry dude if I upset you.

Anyways so what you are saying is that the image I posted looked so much worse than the ATI one?

Or are you just so full of sh!t?

IMO they look pretty damn close. Unless you really found the textures to be crap lol. And remember that was 65% image quality since it was reduce to less than 100KB.

You can't BS me by saying Nvidia's IQ is crap cause the shots I posted just prove that they are not crap.

Crap would be a GF 2 image quality. Even GF4 image quality. But believe me FX IQ is better than GF4 IQ. I know so because I did own a GF4 and a GF2 and GF1.

And also I did have a Radeon 9800 pro on my system for testing. And you know what?

My FX killed it in pretty much everything since it did overclock better. Of course The 9800 pro does kill it in DX9 games. But for now all I play is DX8 :)

Here have more fun reading this review

http://www.anandtech.com/video/showdoc.html?i=1821&p=23

I know you guys are gonna debate this results since you are always bitching when Nvidia wins in some benchmarks.

The truth is that Nvida will always win in some benchmarks. Your Radeon won't out perform it in everything. Is a fact.

The truth is that when I tested the Radeon 9800 pro my FX totally killed it in UT 2003 even at default speed. These new maps are the only ones I see that kill the FX. On the other hand looks like the Dets 5 give the performance back.

And I think I made it clear that even at the cost of IQ lol. The FX still looks kick ass and not like crap. Or prove me wrong. Tell me the shots I posted look like crap because why? Oh yeah because is not using Trilinear filtering right?

[StealthHawk]

Originally posted by saulin
You can't BS me by saying Nvidia's IQ is crap cause the shots I posted just prove that they are not crap.

Crap would be a GF 2 image quality. Even GF4 image quality. But believe me FX IQ is better than GF4 IQ. I know so because I did own a GF4 and a GF2 and GF1.

No, gf4 IQ is at worst the same as gfFX quality. They have the same AF and the same FSAA quality. Since driver 44.03 NVIDIA has disabled trilinear filtering and trilinear AF in UT2003. As of 51.75 they have moved this "optimization" to all D3D games.

http://www.3dcenter.de/artikel/ati_nvidia_treiberoptimierungen/index3_e.php

You can see above that in some situations NVIDIA's handling of AF kills IQ.


MikeC produced some videos with AF off and you could tell that trilinear wasn't being done. The videos aren't being hosted anymore, but the comments will probably be of some value.

Discussion (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=15805)


B3D has posted screenshots(I'm too lazy to search) where you can see the mipmap levels because trilinear is not on.


Conclusion: It doesn't even matter what NVIDIA's IQ looks like compared to ATI's IQ. The point is that with old drivers, NVIDIA's new IQ looks worse than NVIDIA's old IQ.

[The Baron]

Oh God. Why did you link Doom 3 benches? We all know they're about as representative of the performance of the final game as version 0.02 was. We all know that NVIDIA was using a path with lower IQ than ATI. We know it's not running in high quality and is instead using medium quality. But yet some of us helpfully ignore all of that and continue to proclaim the NV35's Real Ultimate Power based on the PR Doom 3 benches.

If you look at the floor in the distance in your pics, it is blurry. Why? Because NVIDIA does not apply trilinear filtering at all to UT2003. ATI, helpfully, does not have this problem. They apply trilinear filtering to the first texture stage in UT2003 when anisotropic filtering is enabled, so that trilinear is pretty much all you need. They do real trilinear if you enable anisotropy from within UT2003. NVIDIA NEVER does trilinear in UT2003 when anisotropic filtering is enabled anywhere. It JUST AIN'T THERE.

NVIDIA's IQ is poor in regards to AA. That's the only place where their IQ is naturally poor. They are making their IQ worse by implementing application-specific cheats that decrease IQ for speed.

And re this:
Tri Linear filtering is still enabled in other PC games? Or no?
Not as of 51.75. Oh, I love the wondrous Det 50s already.

[StealthHawk]

Originally posted by fivefeet8
Why would Tri Linear filtering be disabled in UT2k3 anyways? It doesn't decrease performance that much does it?

Tri Linear filtering is still enabled in other PC games? Or no?

NVIDIA is performing an extra aggressive version of ATI's AF. ATI does trilinear AF on texture stage 0 and bilinear AF on all other texture stages. NVIDIA does trilinear AF on texture stage 0 and 2x bilinear AF on all other stages. Looking at benchmarks the only place where performance is gained by this optimization is in games where IQ is lost.

The difference between the two IHVs is that you can get full trilinear AF on ATI cards, whereas there is absolutely no way to get full trilinear AF on NVIDIA cards. Also, when AF is off NVIDIA cards are doing the tri/bi mix unlike ATI cards. NVIDIA has been doing this since 44.03 for UT2003. With 51.75 they are doing this in all D3D games.

[saulin]

You guys amaze me...

Of course the floor is a bit blurry when looking far and so is in the Radeon image. And if you compare both images you'll see the difference is like almost nothing.

That difference in AF is something you would not notice while running like crazy in the game.

Also who said you have to use the latest beta driver always?

Heck I'm waiting to see if the Dets 50s final will be worth it. If the image quality is gonna suck then of course I'll stick with the older drivers.


Also if you can't notice the IQ difference between the GF4 and the FX I'm not sure how you can tell the diff between the Radeon and the FX then.

The Gf2 GTS had better image that the GF 256 and the GF4 had much better IQ and the FX does indeed have better IQ. When I first saw my desktop on the FX I did notice the difference.


Want one more review?

http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD01OTE=

I guess soo many people can't be all wrong or can they?


So to summarize this:

- FX IQ is more than descent in fact I could not get you guys make my FX IQ look like trash.

- Speedwise the 9800 Pro kills the FX in DX9 and in some of the current games. However I can see the FX has the lead in many current games.

- Nvidia is indeed *****ing up with the new Dets 50s. I'll just wait and see how the final product ends. I doubt they would sacrify soo much IQ for performance on the final product.

- With my overclocking I'm really not disappointed with the performance I get out of my card. I'm happy with the IQ is pretty damn good. I bought my card specially for Emulation (http://www.ngemu.com) and there it beats the crap out of the Radeon anytime.


So there you have it. I just wanted to let you guys know that the FX IQ is not as bad as it is said to be by Nvidia haters. I wanted to let you know that it performs better than the Radeon in many current games.

And I never said Nvidia has better IQ and that it will rule the DX9 world so don't even try to bring that up. I did say it does have better AF and that seems to be true in some games. As you saw in the link I posted. Also the driver I used for UT 2K3 are not the ones that are said to have the best IQ. In fact I read a couple of days ago that since 44.03 well 44.03 have the best IQ.

The End.

[The Baron]

Also if you can't notice the IQ difference between the GF4 and the FX I'm not sure how you can tell the diff between the Radeon and the FX then.
*hits forehead*

They're the same. In everything. There is absolutely no image quality difference between the two. They have the same AF, they have the same AA, they have the same *everything.*

[saulin]

OMG you are so wrong dude. You notice it specially in 2D.

And how can it be the same when the FX can do 12X FSAA in D3D and 16X FSAA in OpenGL, well not officially but can be done. When the GF4 can not.

NVIDIA to improve 2D image quality on NV30
by Julio @ 2:20 AM - [Comments]

http://www.3dspotlight.com/



Quote:

Nvidia is working on this problem since the time of Geforce 3 TI 500 when we first time urged them to do something about this and we saw nice progress on Geforce 4 TI cards and learned that the real improvement will come in NV30 where engineers will pay special attention to this problem.

The company realises that the problem comes from some of their 15 or more partners who still have the freedom to use capacitors, resistors, RF filer parts or any other small component when making that Geforce cards. Some use cheap low quality components, you'll be horrified to learn.

Nvidia's answer to this problem is that with the NV30 it will start a new quality control process where it will make strongly encourage its partners to use the same parts that Nvidia uses for the reference design.

Read more: The Inquirer.

[StealthHawk]

Originally posted by saulin
You guys amaze me...

Of course the floor is a bit blurry when looking far and so is in the Radeon image. And if you compare both images you'll see the difference is like almost nothing.

That difference in AF is something you would not notice while running like crazy in the game.

ROFL. "There is no difference in IQ but just in case there is you wouldn't notice it anyway."

Also who said you have to use the latest beta driver always?

Just to reiterate, every driver since 44.03 does not do trilinear in UT2003. That includes two official releases, 44.03 and 45.23, and all beta drivers in between and since then. 44.03 is the driver that shipped with gfFX5900. It's not just the latest beta driver that has IQ problems.

[StealthHawk]

Originally posted by saulin
OMG you are so wrong dude. You notice it specially in 2D.

And how can it be the same when the FX can do 12X FSAA in D3D and 16X FSAA in OpenGL, well not officially but can be done. When the GF4 can not.

WTF are you talking about. The GeForce3 can do 12x FSAA in D3D :rolleyes:

Discussion (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=15821)

[saulin]

???

I've been using aTuner for a long time and I never got those options untill I got my FX.

Yes using the same version too.

The best I got on my Ti 4600 was 8XS not 12X or 16X.

BTW can you confirm that the GF3 or GF4 can do 16X?

A quote from that same discussion link you posted.

I wished 4xS was available in OpenGL. That is much better then 4x multisampling especially in motion. Except for my GF3 Ti200 at 240/500 core/mem the performance hit almost makes it unusable. For older games especially like flight simulators this is great stuff. ATuner works great by the way and wished Nvidia would have had these settings in their drivers for the GF3 owner crowd. I do get some lag though when using 4xS so maybe that is why Nvidia decided not allow GF3/4 owners the luxury of these settings.

So I guess Open GL does not have the extra modes then?

I never saw 16X on a GF4 that is for sure. I know cause I use OpenGL a lot for emulators and I went like WTF when I saw all the extra modes on my FX.

[saturnotaku]

Originally posted by saulin
- Speedwise the 9800 Pro kills the FX in DX9 and in some of the current games. However I can see the FX has the lead in many current games.


Then what's going to happen when DirectX 9 becomes current? Say in two weeks when Halo hits the shelves?

NVIDIA's AF might look better in some cases, but the performance hit doesn't justify the IQ increase in most cases. And NVIDIA still hasn't fixed the lack of trilinear filtering in UT2003. And this was a major sore spot/contributing factor to my going back to ATI (along with the DX9 performance, inferior AA quality, etc).

[saulin]

You guys continue to amaze me after I posted screen shots for you to compare and all you could say is. Well far ahead there is some blurry spot. For fu@k sakes the Radeon has the blurry spot as well.

saturnotaku I don't think the FX will be able to handle all DX9 games. At least not with FSAA and AF.

At least Doom3 and Quake 4 I would think they would run fine.

Both companies should have new cards soon BTW soo the FX 5900U or the Radeon 9800 pro won't be the only solution for DX9 games and future games then.

You guys talk like these cards will be here for years. In fact I don't think even the Radeon 9800 Pro can last that long in future games if they will require the power HL2 does or more.

So you saw HL2 running at 60 fps in a benchmark?

Well what about in all other maps? What will happen when the action gets so intense in the game. I bet your FPs will drop like a wh0re on her knees.

[saturnotaku]

Originally posted by saulin
At least Doom3 and Quake 4 I would think they would run fine.

The FX will run Doom 3 just fine on its lowered precision codepath.

Well what about in all other maps? What will happen when the action gets so intense in the game. I bet your FPs will drop like a wh0re on her knees.

UT2003 4xAA, 4xAF, 1280x1024 seems to run just fine for me with a whole lot of bots.

[saulin]

UT2003 4xAA, 4xAF, 1280x1024 seems to run just fine for me with a whole lot of bots.

It runs fine for me too with those settings. Of course that is not including the new maps like DM-Icetomb. Performance goes to hell there. Performance is back in the new Dets 50s still to be released. Of course I said I will wait for the final product where I hope I won't see much IQ loss. Or else fuc* the new dets. I would stick to the current drivers I'm using.

However HL2 may be a different story. Remember it is a DX9 game and it does demand a lot of horse power. I really would not espect UT 2K3 speeds on that game even if I had a 9800 pro.

My Scores at 1280x1024 4XAA and 4XAF (http://members.shaw.ca/navasa/12x10-4x4x.html)

[StealthHawk]

Originally posted by saulin
anyways this article should explain it better.



http://firingsquad.gamers.com/hardware/imagequalityshootout/page3.asp

Oh well I'm off to bed...

laters dudes.

How does this article corroborate your assertion? They don't even test IQ in UT2003. There is also a bug with ATI cards in Nascar2003 where detail is lowered because it doesn't autodetect all the memory or something. Anyway, I think we all know and acknowledge that ATI's AF doesn't look as good at certain angles.

Originally posted by saulin
The Gf2 GTS had better image that the GF 256 and the GF4 had much better IQ and the FX does indeed have better IQ. When I first saw my desktop on the FX I did notice the difference.


Want one more review?

http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD01OTE=

We are not talking about 2d quality(filters), we are talking about 3d scene quality. And BTW from the conclusion of the Hexus review it says "decent 2d quality," emphasis mine.

Originally posted by saulin
???

I've been using aTuner for a long time and I never got those options untill I got my FX.

Yes using the same version too.

The best I got on my Ti 4600 was 8XS not 12X or 16X.

You need to use newer drivers.

BTW can you confirm that the GF3 or GF4 can do 16X?

I don't think I saw that option, but I don't really remember.

So I guess Open GL does not have the extra modes then?

It has 8xS in OGL which is probably better or just as good as 16x anyway. Is 16x FSAA really a commodity? Considering that there is barely any difference at all between 12x and 8xS in D3D...8xS is actually better than 12x. Do you really need 16x FSAA? Is 16x FSAA really better?

[mrsabidji]

I couldn't get my friend's CDs this week-end (had to format and re-install XP, for some reason), so no benchmarking for me right now. Sorry guys.

mrsabidji

[Behemoth]

Originally posted by saulin
You guys continue to amaze me after I posted screen shots for you to compare and all you could say is. Well far ahead there is some blurry spot. For fu@k sakes the Radeon has the blurry spot as well.
why dont you post some screenshots on icetomb with both IQ and speed comparable to 9800pro? because you cant. with fx5900u you can only get either sucky IQ/fast incomplete trilinear fps with 51.75 or ok IQ(still no full trilinear)/slooooww fps with 45.23 or previous drivers.


saturnotaku I don't think the FX will be able to handle all DX9 games. At least not with FSAA and AF.
but nvidia is still saying otherwise. they are still making up dx9 numbers and pissing people off.


At least Doom3 and Quake 4 I would think they would run fine.
one would expect more than that for a top dollor card.


Both companies should have new cards soon BTW soo the FX 5900U or the Radeon 9800 pro won't be the only solution for DX9 games and future games then.
there is probably no solution to nv3x fake dx9 peformance that has been advertised by nvidia.


You guys talk like these cards will be here for years. In fact I don't think even the Radeon 9800 Pro can last that long in future games if they will require the power HL2 does or more.
my geforce 1 and 3 are still with me and functional, when i buy a computer i expect it to run for years.


So you saw HL2 running at 60 fps in a benchmark?
forgot


Well what about in all other maps? What will happen when the action gets so intense in the game. I bet your FPs will drop like a wh0re on her knees.
then i will turn some details options off and the game is still playable. whereas fx card would drop like a real ***** on her knees even with all dx9 features off.

[The Baron]

Saulin, don't circumvent the swear filter.

[fivefeet8]

Originally posted by Behemoth
with fx5900u you can only get either sucky IQ/fast incomplete trilinear fps with 51.75 or ok IQ(still no full trilinear)/slooooww fps with 45.23 or previous drivers.

I don't think the Fx cards have sucky Iq at all. THey may not be better than ATi IQ, but you can't say they suck either. All this time you've been trying to prove your points, I've been playing the game and the IQ is definately not sucky. Lack of Trilinear filtering does lose some detail from distant ground texture, but everything else looks fine. From comparisons on this thread and other sites, the IQ of both cards is comparable with the ATi being a little better. A little better doesn't mean the other card's IQ sucks.

And the IQ difference from the 51.75's are minimal at best in UT2k3 compared to the 45.23's. I think if your expecting IQ loss, you're going to find it no matter where the shot came from. Let me do a little experiment. I'll post up 2 Ut2k3 shots in the same scene. You tell me which shot are from which drivers. 51.75's or 45.23. I'll post them later. And it's not going to be the missing fog in DM-icetomb. :D

[hovz]

ok, just make sure ur not staring at the sky or a wall