View Full Version : 3Dmark05 High and better IQ...
Forgive me if i quoted you at b3d, But I felt it was worth looking into. I opted for higher quality .png files.
That's fine. Did you post the info to the same thread Sazar mentioned earlier?
You have been vindicated a bit OWA, Theres a huge discrepency, I am gonna post your and my findings over at b3d, to see if it can be explained in a moment.
My 6800.
http://members.cox.net/omega1979/6800canyon.PNG
Radeon 9800 Pro
http://members.cox.net/omega1979/aticanyon.PNG
is it me or in the ati screen shot if you look at the water close to the cayon it seems smoother like the water shaders are not being rendered or some kind of OPT in the filtering? correct me if i'm wrong
You are correct dude in the ati shot the water effects end before you get to the boat and in the 6800 shot the water effects go on to infinitety. Ati is rendering less it looks like to me.
That was with old driver set, its been fixed in future drivers.
With 66.51 the "rough shadows" are still there in Farcry.
Note that when topicstarter forces his 6800U doing the Ati path the
shadows become smooth...
http://www.beyond3d.com/forum/viewtopic.php?t=16675
Thanks for checking into it. I was starting to get a little nervous that it might be just me.
I've checked it too. Your findings are correct.
ChrisRay
10-03-04, 08:40 PM
You are correct dude in the ati shot the water effects end before you get to the boat and in the 6800 shot the water effects go on to infinitety. Ati is rendering less it looks like to me.
Actually look again, I made an error there with inconsistent Level of Detail, And one shot had AF and the other did not, The water looks similar after I corrected this issue.
With 66.51 the "rough shadows" are still there in Farcry.
Note that when topicstarter forces his 6800U doing the Ati path the
shadows become smooth...
http://www.beyond3d.com/forum/viewtopic.php?t=16675
I'll have to look for the thread someone said it was fixed, It could very well not be fixed, I havent had a chance to look at it.
ChrisRay
10-03-04, 08:53 PM
For the record, Here are the updated shots, WHich have a consistent LoD/AF settings. the water is rendered correctly on both cards. The Fog however isnt.
http://www.nvnews.net/screenshots/3dmark05/6800canyon.png
http://www.nvnews.net/screenshots/3dmark05/aticanyon.png
Indeed nice for Nv card owners, but not by default. Futuremark should have made a checkerbox where you can put it on, not "off". :lame:
And btw, a little bit off-topic but it seems that Nv uses DST too as a default in FarCry. Check the shadows on the gun:
http://www.xbitlabs.com/articles/video/display/farcry30_4.html
"ATI RADEON X800-series produces smoother shadows in FarCry when compared to NVIDIA GeForce 6800-series."
i agree with the switch.. there should be one to disable Dst on and off at will.. for more acurate comparisons.. but dont agree when it comes to say one technique is better ,that the other. or that Nvidia is doing less workload.. it is only faster because is a feature supported in hardware in NV cards.
ChrisRay
10-03-04, 10:04 PM
i agree with the switch.. there should be one to disable Dst on and off at will.. for more acurate comparisons.. but dont agree when it comes to say one technique is better ,that the other. or that Nvidia is doing less workload.. it is only faster because is a feature supported in hardware in NV cards.
There is a switch to disable DST.
There is a switch to disable DST.
so there is no problem then... it can be turned off... it just that Futuremark want reviewers to buy the professional version for it.
Shamrock
10-03-04, 11:29 PM
I dont use 3dMock, but I am curious as to WHY people are making a fuss about DST, and why didnt Futuremark use Direct X 9c specs? is it because ATI cant use it? or ATI paid more money to NOT use it? Why do they allow FP24, when FP32 is the CURRENT DirectX 9 spec.
I also heard you cant use SM 3.0 unless you get the pro version, is that correct?
And here is a thoughtless conspiracy theory, why is ATI's shots darker, in ALL but one I see ATI shots darker than NV's....remember, what is the best color to render? BLACK
so there is no problem then... it can be turned off...
Lol, i am familiar with the switch, but my point was that the switch should turn DST on, not "off" which is the case now. First, DST is not specified in DirectX 9.0 and is Nv only (in the past, Futuremark stated that they only apply special features if 2 hardware vendors would use it!), and second, it simply decreases I.Q. on shadows. Well, this is imho, i don't like if things are aliased.
On the other side i can understand why they implement DST by default for Nv, to hold balance in the "world of Graphics": the Geforce FX cards would have been completely blown away by the competition (it already is) as the use of DST increases performance up to 50% on FX cards. It wouldn't also done good to the 6800U which already is behind in scores to the X800XT, without DST the gap would be 10/15% bigger which ain't good for the industry.
To be honest? I think the DST switch in first instance was there to switch on DST, so DST was not the default. And when they saw the last "wonderdriver" from Ati they thought "hey, the gap will be to big". (mag)
Lol, i am familiar with the switch, but my point was that the switch should turn DST on, not "off" which is the case now. First, DST is not specified in DirectX 9.0 and is Nv only (in the past, Futuremark stated that they only apply special features if 2 hardware vendors would use it!), and second, it simply decreases I.Q. on shadows. Well, this is imho, i don't like if things are aliased.
On the other side i can understand why they implement DST by default for Nv, to hold balance in the "world of Graphics": the Geforce FX cards would have been completely blown away by the competition (it already is) as the use of DST increases performance up to 50% on FX cards. It wouldn't also done good to the 6800U which already is behind in scores to the X800XT, without DST the gap would be 10/15% bigger which ain't good for the industry.
To be honest? I think the DST switch in first instance was there to switch on DST, so DST was not the default. And when they saw the last "wonderdriver" from Ati they thought "hey, the gap will be to big". (mag)
interesting theory... but i dont think so.. :)
ATI was behind Nvidia with driver 4.9.. by a nice gap.. just look at hanners/XBitlabs first review.. which Nvidia was in the lead.ATI just rushed up some magical drivers at the last minute , that here people are investigating for ->*missing effects* and lower IQ ..that can increase the performance. and doesnt match the Dx9 reference image. ,the DST is non-issue ,because again you can turn it [off] ,if you have the professional version ,a version that every reviewer should use. another thing is the AA quality ...in my earlier post that seems you have missed? . using your propertly taken 6xaa shots they show more aliasing ..than at NV 4xAA..but just slighty. the missing effects (the fog) seems to be happening too in one of your shot.so the way it stand at the moment ATI is the questioned for its IQ .(with no answers.) not Nvidia ...Nvidia DST diferences already have been explained that can be disable. your screenshots only proof that some of the IQ issues can be fixed through the CP and tweaking. but not that there arent any issues ,that might make Reviewers to incorrectly measure performance with significant diferences in IQ againts ATI like the Bjorn3d/DH shots..so i will like to wait and see how this turn to be. im really sure in a week or less we will have a better picture of this.
I thought ati's newest drivers == 1000 point gain, nvidia's latest drivers == 600 point gain...
both last minute drivers init?
Ahaha, people like nV40 are going to feel so PWNED!
http://www.bit-tech.net/review/359/5
They allready did an image quality comparison at bit-tech between the latest beta drivers from both nV and ATi.. and what shows, no big discrepancies!
ATi seems to have a more blurry effect to textures and the nV beta's seem to be lacking some effects on certain surfaces.
And then, people like nV40 forget that the PCIe version of the X800's were allready scoring those numbers, he just can't accept that it really was a bug in the AGP vertex memory management..
The only real difference bit-tech can find is probably because of the way the different shaders 2.0 and 3.0 are programmed.
so, good luck finding those big cheats in the 4.10 drivers
its nice to see you dont read..:rolleyes: never said ATI cheating or anything.. of the like.. just that people have found missing effects and IQ issues that can have an impact in performance, apart from the DST diferences..that are under investigation in many forums . look at the discoveries of people here.. this thread is not about what NV40 says versus ATI ..dont shoot the messenger. ;)
Not rendering effects is cheating, most notably if it's fog or AF of objects.
Most, if not all of those cases are found by users that.. seem to have user issues.
As it currently stands, the 66.51 forceware drivers have more bugs/cheats in 3dmark05 than the 8.07 catalysts.
But then again, the 4.10 drivers will be WHQL qualified when they arrive today or tomorrotw (or maybe a bit later down this week.) we'll still have to wait like what, two or three months before nV releases a complete WHQL driver package.
And yes, you didn't say ati was cheating... but that they have IQ issues, the same goes for NV offcourse.
Not rendering effects is cheating, most notably if it's fog or AF of objects.
for the record that was told by you not me.. ;)
and yes, you didn't say ati was cheating... but that they have IQ issues, the same goes for NV offcourse.
yes.. like many others ,also noticed the IQ diferences. (when explored more the screeshots) and you will notice them too. but my observations were only about AA quality.. wich were Obvious quality diferences. wich seems to be an ATi issue about the way you enable it with ATI drivers in the Cp. others found banding and missing effects in ATi shots .shadows diferences with Nvidia are DST related i think.. which can be turned off with the Pro version.
Yes, Hexus has a GREAT writeup on image quality, here's their bit on DST
http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD04NzgmdXJsX3BhZ2U9Nw= =
http://img.hexus.net/v2/features/3dmark05_intro/images/dst_diff.jpg
Not rendering effects is cheating, most notably if it's fog or AF of objects.
Most, if not all of those cases are found by users that.. seem to have user issues.
Not rendering the mist has been confirmed by others as well. I don't appear to have any "user issues". ChrisRay confirmed what I found and I got someone else on another forum (that has the XT) to confirm it with his card. If you have the pro version of 3DMark05, it should be something that you can easily confirm.
Edit: I'm not sure if you're talking about the mist stuff I brought up or not since you also have another discussion going. So, if you were talking about something else, just ignore my response. :)
[quote]
the DST is non-issue ,because again you can turn it [off]
Well heck, let's put 3Dc in there too by default, and make a switch so the Pro users can turn it off. Ofcourse this is ridicilous, hardware vendor specific features which affects the reference image negative, and especially if they are patented by one vendor (so the others can't use it) and are non DirectX9 (which is the case with DST) doesn't belong in there by default.
(Fine if you can turn it on afterwards, or fine using by default when it doesn't affect IQ, but it does so it shouldn't be the default mode.)
jbirney
10-04-04, 03:55 PM
[quote]
Well heck, let's put 3Dc in there too by default, and make a switch so the Pro users can turn it off. Ofcourse this is ridicilous, hardware vendor specific features which affects the reference image negative, and especially if they are patented by one vendor (so the others can't use it) and are non DirectX9 (which is the case with DST) doesn't belong in there by default.
(Fine if you can turn it on afterwards, or fine using by default when it doesn't affect IQ, but it does so it shouldn't be the default mode.)
Good summary. 3Dmark is suppose to be netural and not use any non-dx9 vender specific paths. Including DST and having it on has no place in a DX9 benchmark. I am sure if 3Dc was there instead of DST and on by default, people would have posted differently in this thread :)
[quote]
Well heck, let's put 3Dc in there too by default, and make a switch so the Pro users can turn it off. Ofcourse this is ridicilous, hardware vendor specific features which affects the reference image negative, and especially if they are patented by one vendor (so the others can't use it) and are non DirectX9 (which is the case with DST) doesn't belong in there by default.
(Fine if you can turn it on afterwards, or fine using by default when it doesn't affect IQ, but it does so it shouldn't be the default mode.)
Remember.. the PRo version is the one FUturemark wants people to buy for professional benchmarking ,for more freedom and more settings for comparison.so this is a moot point already. it wil be better ofcourse if FM gave us more options ,and skip their pro one. :) your question about 3Dc can be answered here ... (http://www.beyond3d.com/forum/viewtopic.php?t=16710&postdays=0&postorder=asc&start=0&sid=f681acbfede7846f607fb083375006a5) . there is a great misconception about that , thanks to the apples vs oranges comparison of ATI marketing papers. 3Dc and DXt5 have the same advantages .is just that 3dc have slightly more precision. its not an all around solution and if it is not wisely used the qualiy can be lower.think about that before asking that feature to be used in your games..is not something that every game Developer will want to use for many reasons explained there.
3Dc is just a slightly improved compression technique of DXT5 , but not a all around solution for every case.not something that every game Developers will want to use for many reasons explained there.
Imho vendor specific features are brought in by the game developers when the vendor comes with $$$ (Nv's twimtbp/Ati's gitg), it has nothing to do with a game developer liking it or not. :) (Except when all vendors provide the feature, in that case dollars can't influence a developer for liking the feature or not. He just does or doesn't.)
With this in mind, it's a big questionmark for me why Futuremark implement DST in the Nv path by default, the shadows are simply not matching the reference image, it's only helping NV's scores.
Hexus.net stated: "Futuremark have gone on record in the past saying they'd not support vendor-specific extension to DirectX in their benchmarks."
But maybe 3DMark05 has become a TWIMTBP game? :)
Diamond_G
10-04-04, 05:38 PM
Been lurking in this thread for a while. Was wondering why everyone is comparing NVidia images to ATI images. Shouldn't you guys be looking at how close each IHV is to the reference image? If ATI doesn't look like it the they are wrong if NVidia doens't look like it they are wrong. Simple. So now that we know both are wrong, who is the least wrong (you know partial credit). Someone should render every single frame using the reference shots and put them together as an uncompressed movie so we can see what it is supposed to look like, after viewing on our cards.
Been lurking in this thread for a while. Was wondering why everyone is comparing NVidia images to ATI images. Shouldn't you guys be looking at how close each IHV is to the reference image?
I think Futuremark does this when a vendor provides a driver for approval. In the end, when we take 1000 Ati screenshots and 1000 Nv screenshots we will see that Ati matches the ref images (for instance) 70%, and Nv does too. I'm gessing there must be some kind of method. Else they wouldn't get approval.
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.