This is from Tom's Hardware. Is this possible? Looks like 9800 XT and Pro are miles ahead of 9700 Pro. Can anyone confirm/deny the accuracy of this? It is very hard to believe, almost double the 9700 Pro. Hard to believe because in other game benchmarks there appears to be maybe a 25% improvement.BTW, how do you post images here without getting the lil link thingey? I used to be able to post images directly.Noel

It's the smoothness of FS2004 which is important, not the frame rates.I don't care very much for frame rates anyway but what's this fraps thingy of 14.1 for an FS5600 card? I don't go below about 25 fps with this card even at a complex airfield with the FSD Porter loaded.DavidIntel D865GBF moboIntel Pentium 4 3.2 Ghz 800 processor200 Gb Seagate hard disk 7,200 rpm1 Gb PC3200 DDR 400 RAMNVidia FX5600 256 Mb video card with 52.16 driversDirectX 9.0aCreative Labs Sound Blaster Audigy 2 sound cardLG Electronics 17" LCD monitorWIndows XP Professional

I don't think it's either/or David. When you have frames less than about 13 or so, you can kiss smoothness good bye. Esp when it's down to 8-9. On the other hand, as Tevye would say, you can have frames of 40 and it can be stuttering bigtime. This is why I say, it isn't either/or. The other HUGE factor frames give, in the context of lack of stutters, is a much better flight model IMO. I remember how all planes handled so much better in FS2002 when I got this machine.And the other thing is that it is extremely hard to evaluate performance in this sim. I don't care what anyone says, the EXACT situation must be duplicated, and for the life of me there are times when clouds look virtually the same, in the same area, yet clearly one situation is hugely more impacting than another.Noel

"And the other thing is that it is extremely hard to evaluate performance in this sim. I don't care what anyone says, the EXACT situation must be duplicated, and for the life of me there are times when clouds look virtually the same, in the same area, yet clearly one situation is hugely more impacting than another."True that. Especially if you're still using the xml-defined autogen and vector objects since they could be slowing you down because they haven't been purged when benchmarking one card, even though the scene is identical.Keeping in mind that the sim was set to 1600x1200 with 4x FSAA/8x Aniso, the memory bandwidth requirements were probably quite high. It's possible that simply switching to 4x AF or 2x FSAA would completely level the field, as CPU performance becomes more important. I would expect the 9700 Pro to score closer to 30 FPS in an IDENTICAL situation to the 9800, so something isn't quite right with those.To tell the truth, I've seen as many different results from FS as I've seen benchmarks using the sim. Anywhere from 60 FPS with the GeforceFX being twice as fast as the Radeons, to results in the 20 FPS range with the Radeons way ahead.

>Keeping in mind that the sim was set to 1600x1200 with 4x>FSAA/8x Aniso, the memory bandwidth requirements were probably>quite high. It's possible that simply switching to 4x AF or 2x>FSAA would completely level the field, The only way to "level the field" when testing an ATI vs. Nvidia trial is to disable ALL Autogen and run with perfectly clear weather as the Nvidia card will be using AA on all autogen trees/radio towers as well as clouds while the ATI card will not be using AA on those objects, also cut the AI traffic to zip as its another variable. Then Simply save a flight video of a quick flight, reboot and restart the sim for each test and run it three times and then do an average.Cheers!

_______________________________________________________________This is from Tom's Hardware. Is this possible? Looks like 9800 XT and Pro are miles ahead of 9700 Pro. Can anyone confirm/deny the accuracy of this? It is very hard to believe, almost double the 9700 Pro. Hard to believe because in other game benchmarks there appears to be maybe a 25% improvement.________________________________________________________________There's very little difference between a 9700Pro/9800NP/9800Pro/9800XT. The 9700PR0 = the 9800NP; Maybe 5 percent at stock speeds as you go from the 9800NP to the 9800Pro and again from the 9800Pro to the 9800XT. All three cards perform very well. Those numbers in the Tom's H review are not right. Not even close. With my 9700Pro, I see very little difference in framerate going to a higher resolution and therefore run everything at 1600x1200x32 res. Better performance running 4xAA at 1600x1200x32 than running 6xAA at 1280x1024x32 on my rig. Go on over to rage3d.com and read up on the various cards and there performance differences. Many questions asked similar to yours and the answer is always "you're foolish to upgrade if you have a 9700Pro". If you have a 9700Pro and want a 9800Pro, just overclock the 9700Pro and you got it. I don't recommend overclocking anything though and always run at stock speeds. From the Call of Duty benchmark:" The ATI DX9 cards are CPU-limited at low resolutions, so we decided to omit the benchmark results at 800x600. In fact you can see that we

"The only way to "level the field" when testing an ATI vs. Nvidia trial is to disable ALL Autogen and run with perfectly clear weather as the Nvidia card will be using AA on all autogen trees/radio towers as well as clouds while the ATI card will not be using AA on those objects"The only time NVidia will AA the Autogen trees is when using any of the mixed MSAA/SSAA modes such as 4xS, 6xS etc. When using normal modes such as 4x, they both use plain MSAA = no AA on alpha textures.

>"The only way to "level the field" when testing an ATI vs.>Nvidia trial is to disable ALL Autogen and run with perfectly>clear weather as the Nvidia card will be using AA on all>autogen trees/radio towers as well as clouds while the ATI>card will not be using AA on those objects">>The only time NVidia will AA the Autogen trees is when using>any of the mixed MSAA/SSAA modes such as 4xS, 6xS etc. When>using normal modes such as 4x, they both use plain MSAA = no>AA on alpha textures.JimmiG,That is simply not True Period, under any AA setting from 2x on up Nvidia cards AA Alphas. There was a driver that however did just what you describe and that was one of the 28.xx drivers.You must suffer from short-term memory loss! :-lolI posted to you several pics a while ago showing you this.These are brand new, from the crappy 52.16 drivers:800x600x32 No AA (look at the pine):http://www.frontiernet.net/~pleatzaw/images/800x600noAA.jpg800x600x32 2xAA same area:http://www.frontiernet.net/~pleatzaw/images/8x6_2xAA.jpgAnyone with a GF3 on up can duplicate this easily.Some folks just pass on what they have read that someone else has written without testing it themselves - I'm usually not one of them.What I stated above about leveling the playing field, hope you understand now.

Multi-Sampling, by its nature, can not do AA on alpha textures.There is one way. As far as I know, only supported by OpenGL 1.2+ at this time, you can use the alpha coverage mask call, to convert pixel shader output to an anti-aliased coverage mask. This can cause compatibility issues however, and I don't think there's a similar function for DirectX at this time. I'm also not sure this works on PS 1.x chips like NV25 but I could be wrong (I'm not an OpenGL programmer). Basically, this approach (hack) is not efficent enough to be used in most cases and I doubt that even if there is a similar function for DirectX, that Nvidia would force it through the drivers.If you see AA on alpha textures, you're NOT using pure MSAA. Only a combination (*XS) or plain SSAA (in which case, performance will be very low) is able to AA alpha textures.ATI chose not to include any legacy SSAA or MSAA/SSA combined modes, in favour of the faster MSAA modes. That was a good decision IMO. Alpha textures are a thing of the past anyway.edit: I did some research, and NVidia has renamed the FSAA modes in the drivers to deceive their customers:"*Important* The 50 series driver(5x.xx) has had some massive changes to FSAA modes. On a GeForceFX, 4x Super Sampling, 4x+9tap, 6x, and 8x have all been removed from the driver. You cannot get any of these modes. The control panel is also misleading. Quincunx has been renamed "2xQ." 12x (also known as 16x in D3D in certain drivers that exposed it in the control panel) is known as "6x" in the control panel. The old 6x is gone. The "real" 8xS is known as "8x" in the control panel. The old 8x is gone, and good riddance. I applaud NVIDIA for replacing old modes which hardly differentiated themselves with higher quality modes. I do not however, condone the misleading naming schemes."So, when you select 8X, you are actually selecting 8XS (MSAA combined with some level of SSAA). This is probably the reason why you were a bit confused over the AA modes.

>Multi-Sampling, by its nature, can not do AA on alpha>textures.>There is one way. As far as I know, only supported by OpenGL>1.2+ at this time, you can use the alpha coverage mask call,>to convert pixel shader output to an anti-aliased coverage>mask. This can cause compatibility issues however, and I don't>think there's a similar function for DirectX at this time. I'm>also not sure this works on PS 1.x chips like NV25 but I could>be wrong (I'm not an OpenGL programmer). Basically, this>approach (hack) is not efficent enough to be used in most>cases and I doubt that even if there is a similar function for>DirectX, that Nvidia would force it through the drivers.>>If you see AA on alpha textures, you're NOT using pure MSAA.>Only a combination (*XS) or plain SSAA (in which case,>performance will be very low) is able to AA alpha textures.I believe you refer to ATI's "Smoothvision" ver2.0x which can work in GL if ATI enabled it but for itt incompatablity with many game titles.This is not and has not been the case with Nvidia hardware for quite some time. In D3D any setting of AA will fully AA the alpha mask inside and outside lines included; that includes All of the following AA settings:2xAA, QunxxAA, 4xAA, 4xsAA, 6xAA and 8xAA -Fully tested.You are either referencing the wrong material or something that is well out of date.Just because you read it don

Please download and run this program:http://www.users.on.net/triforce/d3d_fsaav...saaviewer-3.zipAnd no, I'm not trying to send you a trojan ;)Then post a screenshot of the program when using the 2x mode you used for the FS screenshots. I would like to see what's going on with the samples.

>Please download and run this program:>http://www.users.on.net/triforce/d3d_fsaav...saaviewer-3.zip>>And no, I'm not trying to send you a trojan ;)JimmiG,I suggest you do your own research before you stick your neck out any futher, I know exactly were I stand with regards to Nvidias use of AA filtering of Alph texture within FS.If you have such questions you may want to try your own testing or get in touch with someone at Nvidia, Or if you like I can pass on your question to Nvidias Software Quality Assurance lab and they my answer it for you if they wish.>Then post a screenshot of the program when using the 2x mode>you used for the FS screenshots. I would like to see what's>going on with the samples.Not sure why you not use ver-4.0 as 3.0 had some bugs and could not test high resolutions, none the less I will be happy to post a screen shot for you...tell me what version 3.0 or 4.0 if 4.0 what DX version? what Driver? I am using 53.03 at the moment but there is no change of AA in D3D from 52.16s...what AA setting would you like?What do you hope to see, as it is AAx2?Take care.

I did use a Geforce3 Ti before getting a Radeon 9700 Pro, and I used drivers up to the early 30.x series. I never saw any AA on alpha textures in any app, including FS2002, unless I used a hybrid SS/MS AA mode. At which point did NVidia add this supposed functionality? Never saw a sig nof it in either the 20 or 30 Detonator series.If you can point me to an article that explains/proves that NVidia's MSAA does work on alpha textures, even on the old Geforce3, I'd appreciate it. Considering it's a problem ATI has been trying to solve for well over a year both through drivers and hardware, someone must have mentioned if NVidia found the magical cure to the problem. ATI haven't found a way to add it in software, and no such feature exists even in hardware, even in the latest R360 chip (third refresh cycle of R300). All articles I've seen only mention what I already knew: MSAA HARDWARE can not AA alpha textures. Using pixel shaders to detect edges and apply AA, deteting alpha test and using SSAA for those pixels, other hackery? Yes. Pure MSAA? No.You can use a new version of the FSAA tester if there is one. I guess my bookmarks are a bit outdated.I want to see how many texel sample points you're getting with the 2x MSAA modes just to make sure you're not actually getting SSAA or combined MS/SS AA when you sellect 2x, or other supposed MSAA modes.If you only got one texture sample point, how can you AA alpha test textures anyway? Does it switch to some kind of pseudo-SSAA mode whenever alpha test is detected as being used? Sounds like lots of work; lots of problems to work around, plus you'd need fairly powerful hardware just to detect alpha test and apply SSAA just for those pixels...in the end normal SSAA for the whole scene might even be faster. All that for a feature that is almost never used in newer games and only in a handful of older titles. They should have spent that time trying to eek out more PS 2.0 performance out of their NV3x instead...not through 16-bit hackery, of course ;)This would explain why there's a performance hit when big clouds are used in FS on a Geforce with AA, though...but then you're effectively using SSAA for the whole scene anyway. Or have they done it in some other way?

>I did use a Geforce3 Ti before getting a Radeon 9700 Pro, and>I used drivers up to the early 30.x series. I never saw any AA>on alpha textures in any app, including FS2002, unless I used>a hybrid SS/MS AA mode. At which point did NVidia add this>supposed functionality? Never saw a sig nof it in either the>20 or 30 Detonator series.Not sure what you are trying to say here. You saw it, but only in some "hybrid" AA mode? But it at the same time is also a "supposed" function?I believe it has been supported since it was "voted in" by both Nvidia and ATI -OGL 1.3(2001)DX7.1(?)2000).>If you can point me to an article that explains/proves that>NVidia's MSAA does work on alpha textures, even on the old>Geforce3, I'd appreciate it. You can register at various developer sites and you will have access to many docs if you want more info, try ATI, Nvidia or Microsoft directly. Considering it's a problem ATI>has been trying to solve for well over a year both through>drivers and hardware, someone must have mentioned if NVidia>found the magical cure to the problem. ATI haven't found a way>to add it in software, and no such feature exists even in>hardware, even in the latest R360 chip (third refresh cycle of>R300). All articles I've seen only mention what I already>knew: MSAA HARDWARE can not AA alpha textures. Using pixel>shaders to detect edges and apply AA, deteting alpha test and>using SSAA for those pixels, other hackery? Yes. Pure MSAA?>No.That is false, ATI has had hardware enabled AA of alpha since at least 9500, it isn't available because the "smoothvision2.0" team deemed it as it was programmed a compatibility risk for older game titles as well as a performance issue and that is very common knowledge as ATI has shared it publicly when they were pushed as to why the lack of it upon the cards release. >You can use a new version of the FSAA tester if there is one.>I guess my bookmarks are a bit outdated.>I want to see how many texel sample points you're getting with>the 2x MSAA modes just to make sure you're not actually>getting SSAA or combined MS/SS AA when you sellect 2x, or>other supposed MSAA modes.In the image the red squares represent the geometry sampling positions and the green the texture sampling positions:Lokk at the alpha mask, it also Looks like the Alpha is correctly Antialiased here as well:http://www.frontiernet.net/~pleatzaw/images/FSaatest.jpgNo AA:http://www.frontiernet.net/~pleatzaw/images/noAA.jpg4xAA:http://www.frontiernet.net/~pleatzaw/images/4xAA.jpg4xsAA:http://www.frontiernet.net/~pleatzaw/images/4xsAA.jpg8xAA:http://www.frontiernet.net/~pleatzaw/images/8xAA.jpg>If you only got one texture sample point, how can you AA alpha>test textures anyway? Not sure what you describe, we have alpha mask enabled, C?But if you want the most accurate results then Do what the big boys do:Test it in the actual application that you question as synthetic test are just that -synth.As in go back and look at the screen shots I first showed of the alpha textured trees, what do you call that? It is called Antialiasing, no?Does the ATI card do that? No.Sorry, but I am getting a little tired of this.>Does it switch to some kind of>pseudo-SSAA mode whenever alpha test is detected as being>used? Sounds like lots of work; lots of problems to work>around, plus you'd need fairly powerful hardware just to>detect alpha test and apply SSAA just for those pixels...in>the end normal SSAA for the whole scene might even be faster.>All that for a feature that is almost never used in newer>games and only in a handful of older titles. The discussion though is based on FS8/9 essentially a DX7 title.>They should have>spent that time trying to eek out more PS 2.0 performance out>of their NV3x instead...not through 16-bit hackery, Exactly how many DX9 game can you actually purchase that use PS2.0? :-lolWhen DX9.1 arrives (Jan/feb?)you may look at this differently.>This would explain why there's a performance hit when big>clouds are used in FS on a Geforce with AA, As well as trees, radio towers, some aircraft and panel parts etc. etc.That has been my whole point for since how long? :-rollAs when the two card are viewed in game side by side the ATI-95-9800p(faster) has shimmering objects when set at best quality LOD while using its great AA while the Nvidia TI-FX cards (slower) are busy AAing the whole scene, hence No shimmering objects such as trees and such when using good AA modes at the highest default LOD settings even at 8xAF. So the comment goes:To level the playing filed when testing an ATI vs. Nv card autogen should be turned off, the weather should be cleared as well as the AI and ATC for obvious reasons.Or do you still differ? :)

Well besides my many typos..."Lokk at the alpha mask, it also Looks like the Alpha is correctly Antialiased here as well"Sorry, FSAA-viewer does not seem to use an alpha mask, but the most accurate result is from in game testing anyway and is easily tested.