Originally posted by: TheSnowman
sure enough, and a big white mess of snow isn't exactly the optimal texture to show the difference either. regardless, cod is an opengl game and as i stated above both ati and nvidia do do full triliniear in opengl. so no Acanthus, you havn't proven anything but the fact that you don't even understand what you are talking about.
What game do you suggest that has explicit bilinear and trilinear settings. Out of all the games i have Call of Duty was the only one with the option.
BFG-
I'd suggest an optomotrist
And they have this lovely bridge for sale I heard.
This ties in with your last quote. Futuremark in the case of PowerVR is doing everything in their power to make certain that they can not be compared in any realistic sense with anyone else. If PowerVR does the exact same thing for 3DM that they do for every game, why the hell shouldn't it be considered valid? They do not care about being fair, they do not care about creating a level playing field for the sake of comparison. What their agenda is I'm not sure, but it quite clearly has nothing to do with making a synthetic bench that can be used to reasonably compare the IHVs.
DooM3 10% slower and HL2 10% slower or both at full speed, which would you want? In order to get both to full speed they need to use app detection in this scenario. You really going to try and tell me that you wouldn't want full speed?
By using a custom driver profile for DooM3. You would be willing to give up ~30% performance(for this hypothetical) for no reason other then it requires a driver profile for the game? With D3 in particular odds are quite high that you will own a decent sized pile of games based on the engine over the next few years(I'd bet on it actually), and the D3 profile will almost certainly offer benefits to any of the games based on the engine also.
If I was having hard locks with nV's drivers you can bet your @ss I'd be blasting TWIMTBP.
Do you recall which forum it was on? I'd like to see one of these hard lock nV bugs. The fog issue I have heard from several people, they also have a problem with current drivers and shadows/flickering in RiseOfNations, but no hard locks.
Do you recall which 4x.xx it was? I think I have all of the official ones on my HD(a good chunk of the betas also), I'll install them and give SS:SE another play through(fun game and short).
TheSnowman-
For 'has never been that good', BFG is happy with bilinear<shrugs>. I've always been anal about texture filtering, that's why I'm not running a FX right now and one of the reasons I dumped the R300 board I had(the slew of driver issues was the main factor however).
There is no AF applied at certain angles at all, it is straight isotropic filtering. I think it was XBit that had a review up of S3's new part, check out the half pipe filtering test they have and look at ATi's images.
They have AF selected in the application there. SeriousSam(both of them) have a menu option for AF which does allow you to get AF+full trilinear.
That's not just UT2K3, that's every D3D app with nV right now. Running AF under D3D you don't get trilinear as of now, only brilinear. I need a bit more hands on time with it before I pass full judgement, but a lot of people couldn't pick out PVR's box filter and it jumped out of the screen and slapped me across the face......hard
I'd suggest an optomotrist
And they have this lovely bridge for sale I heard.
This ties in with your last quote. Futuremark in the case of PowerVR is doing everything in their power to make certain that they can not be compared in any realistic sense with anyone else. If PowerVR does the exact same thing for 3DM that they do for every game, why the hell shouldn't it be considered valid? They do not care about being fair, they do not care about creating a level playing field for the sake of comparison. What their agenda is I'm not sure, but it quite clearly has nothing to do with making a synthetic bench that can be used to reasonably compare the IHVs.
DooM3 10% slower and HL2 10% slower or both at full speed, which would you want? In order to get both to full speed they need to use app detection in this scenario. You really going to try and tell me that you wouldn't want full speed?
By using a custom driver profile for DooM3. You would be willing to give up ~30% performance(for this hypothetical) for no reason other then it requires a driver profile for the game? With D3 in particular odds are quite high that you will own a decent sized pile of games based on the engine over the next few years(I'd bet on it actually), and the D3 profile will almost certainly offer benefits to any of the games based on the engine also.
If I was having hard locks with nV's drivers you can bet your @ss I'd be blasting TWIMTBP.
Do you recall which forum it was on? I'd like to see one of these hard lock nV bugs. The fog issue I have heard from several people, they also have a problem with current drivers and shadows/flickering in RiseOfNations, but no hard locks.
Do you recall which 4x.xx it was? I think I have all of the official ones on my HD(a good chunk of the betas also), I'll install them and give SS:SE another play through(fun game and short).
TheSnowman-
For 'has never been that good', BFG is happy with bilinear<shrugs>. I've always been anal about texture filtering, that's why I'm not running a FX right now and one of the reasons I dumped the R300 board I had(the slew of driver issues was the main factor however).
There is no AF applied at certain angles at all, it is straight isotropic filtering. I think it was XBit that had a review up of S3's new part, check out the half pipe filtering test they have and look at ATi's images.
They have AF selected in the application there. SeriousSam(both of them) have a menu option for AF which does allow you to get AF+full trilinear.
That's not just UT2K3, that's every D3D app with nV right now. Running AF under D3D you don't get trilinear as of now, only brilinear. I need a bit more hands on time with it before I pass full judgement, but a lot of people couldn't pick out PVR's box filter and it jumped out of the screen and slapped me across the face......hard
kylebisme
Diamond Member
- Mar 25, 2000
- 9,396
- 0
- 0
scroll down the page on that b3d link and you will see images from xmas test app and there is anistropy at all angles.
but i wasn't refering to serious sam, again scroll down the page a bit.
Originally posted by: BenSkywalker
That's not just UT2K3, that's every D3D app with nV right now. Running AF under D3D you don't get trilinear as of now, only brilinear. I need a bit more hands on time with it before I pass full judgement, but a lot of people couldn't pick out PVR's box filter and it jumped out of the screen and slapped me across the face......hard
Acanthus asked for a game to see what is going on in, i gave him a game. are you just speed reading looking for arguments or something?
oh and your futuremark considers pvr to be cheating argument is bunk. they have no reason to have anything against tile based rendering, futuremark just isn't going to waste their time certifying drivers for a dx7 level card to be used in a dx9 level benchmark.
BFG10K
Lifer
- Aug 14, 2000
- 22,709
- 3,003
- 126
So FutureMark is supposed to accept cheating drivers because they cheat in other games too?
FutureMark's rule is "no cheating in 3DMark allowed". The rule isn't "no cheating in 3DMark allowed unless you also cheat everywhere else too".
I want full speed but I don't want the problems caused by application detection. Say a new version of Doom 3 is released and the optimizations fall over. What then? I can wait until the next drivers come and hopefully put performance back to how it was or I can use the old version of Doom3 and be forced to endure the problems that have been fixed in the newer version.
Now if ATi managed a 5% gain without app detection, that would be better in the long run since it wouldn't matter about future Doom3 patches and it woudn't require the driver devs redoing the application detection each time the game was patched.
You're making the flawed assumption that it's not possible to match or better such a speed gain by legitimate means.
If it benefits more than Doom3 then it's more than just simple application detection.
Yes, it was at Ars.
I think it was 44.43 but I'm really not sure as it was almost two years ago now.
SS:SE isn't that short. In fact it's quite a decent length, especially if you explore all the secret areas and play it on the hardest difficulty level.
Uhhh, download that app and check it out for yourself and then compare it to whar XBit was looking at.
And they enable AF through the test app there too.
BFG-
If you would stop kicking dogs this conversation would be easier
You have nothing remotely resembling any proof that PVR is cheating in anything, and yet you are still jumping to that conclusion defying all logic to the contrary.Site me an example. Show me a single time that that has happened and I'll start applying weight to it. PowerVR has been doing app specific optimizations since they started in the PC sector, you should have no problem whatsoever listing a slew of examples of this it were a real issue. It certainly can't be any worse then 'ATi's let's break a game' per driver release(this month- Tron 2.0)
I'm stating that the optimal driver path for D3 is not going to be the same as the optimal driver path for SeriousSam3 and you can quote me on that
My first play through the game took me ~4.5 hours on one notch above Normal. Currently I'm playing Morrowind GOTY(PC version, already have the original XBox version although that doesn't have Tribunal and Bloodmoon), at the rate I'm going I should finish it off in just a shade under 200 hours(not a typo, two hundred). SS:SE is damn short. Anything under 20 hours is short, under 10 hours is way too short for a $50 game(although for $20 I can live with it). Get away from shooters some time and you'll realize how sadly lacking they are in length
You posted the link
The page that has the AF info on it @B3D has a download link for the file, it's real small too(less then 100K).I have the app.
kylebisme
Diamond Member
- Mar 25, 2000
- 9,396
- 0
- 0
i mean the link to "what xbit was looking at", xbit looks at a lot of things. as for Xmas's aa tester, sence you have it you should be able to quickly see for yourself that it goes by controlpanel settings and doesn't give you any af when the controlpanel is set to application preference.
BFG10K
Lifer
- Aug 14, 2000
- 22,709
- 3,003
- 126
They use application detection don't they?
Build 340 of 3DMark03.
That's true and it''s probably be the case that ATi will put the bias onto Doom3 by the time it rolls out.I'm stating that the optimal driver path for D3 is not going to be the same as the optimal driver path for SeriousSam3 and you can quote me on that
4.5 hours? It took me 12-15 on the hardest diffculty.
Did you just race through the entire game or did you actually take the time to explore everything? How many sceret levels and areas did you find? Did you kill all of the hidden monsters?
I tend to finish shooters at least five times and the most I've finished a game is Quake (eleven times). How many times do you finish your games?
That?s true if they are legitimate optimizations. And I could possibly see ATI eventually being forced into application specific optimizations simply to equal out whatever NV is doing. But the real question here is how NV is achieving these gains. You assume NV?s application detection?s are inserting legitimate optimizations. But if NV has legitimate application specific optimizations they can use -- why did they resort to clip planes in 3Dmark2003????? At this point I?m going to view NV?s application detection as hacks and with great skepticism unless they can convince me otherwise. .
When I set app preference in the control panel I get whatever level of AF I select in the app. Big slider, top says 'Max Aniso', does it for me(although I know my drivers are working properly).
BFG-
So?
A slight performance hit is the dire consequnce you are so afraid of....?
Any you would be opposed to having a switch that loads the optimal profile for SS3 built in? Instead you would rather dance around to different drivers for optimal performance?
I found probably about half of the secrets(first play through), still haven't found them all in that one. Remember at the that came out I was reviewing games, keeping GB going and working my 9-5, I wasn't going to quadruple my time with the game when I had so many others waiting(I didn't write the review for that one BTW).
Depends on the game greatly. Max Payne I played through exactly once, and I wouldn't do it again if you paid me the $50 I paid for the game. Disgustingly poor piece of trash. Take the original Zelda and I'd say its easily over a hundred(I got so I could go through the game and play the second and then repeating quests each one taking me about 3-4 hours without a single death). Back in my younger days when the NES was around I didn't have the disposable income for games I do now, and I had more time. Right now I have eight or nine games I haven't touched yet, that will likely grow to over a dozen by the time I finish Morrowind(my wife and kids all game too) and DooM3 and HL2 are looming. More then likely, I won't play through any of them more then once for some time(most of them I'll never play through more then once actually) until we hit a lull in terms of decent games hitting and then I'll play back through the ones I really like. Occasionaly this gets screwed up, last vacation I took I had a stack of games and started with KoTOR(XBox, when it first shipped) intending to play through all the games I had backed up. Beat KoTOR three times, hardly touched the others.
For shooters unless they are really good I play through them once or twice. There are exceptions, HL I have played through probably a dozen times, Halo half a dozen, SS and SS:SE probably four times each(not including getting bored and firing up the final level- on either one, for some quick reflex fun). Exploring them to find everything isn't something I tend to do, not on a shooter anyway. I don't have enough time and rarely is it worth the effort(a few of the SS/SS:SE secrets I really liked- the movie set and Santa I really got a kick out of).
Blastman-
Why wouldn't I?
If nV can figure out how to put clip planes in every game without being visible all the power in the world to them. ATi and nVidia both publicize the fact that they try and reach that goal(HSR- HZ etc) and PowerVR's entire architecture is designed around doing exactly that. Not drawing what isn't seen in real time is ideal, not something you try and avoid. The 3DM2K3 cheat that they did use was an overzealous optimization done by someone who shouldn't have been doing it. Since that time if you look at where nV's parts are performing after FM's official supported testing methods(approved drivers/patch etc) NV's performance is up significantly from prior to the static clip planes cheat. There are an awful lot of ways of optimizing without cheating even in on a per app basis.
BFG10K
Lifer
- Aug 14, 2000
- 22,709
- 3,003
- 126
What does it say in FutureMark's rules about detecting 3DMark? A great big don't do it.
20% isn't slight. Besides, that's not really what this is about. This is about how fragile application detection and optimizations are.
And where are the switches for nVidia's profile that allows users to control it? Detecting "UT2003.exe" is not a profile or a switch, it's a cheat.
I wouldn't dance at all; I'd use the latest version.
Because all evidence points to the contrary.
Most of fiasco since the NV30 has been dodgy at best and blatant cheating at worst and at this stage nVidia has almost zero credibility. I mean you really have to wonder when their PR puts up optimization guidelines and the very next driver release breaks all of them.
So Futuremark is now a deity unto which you worship? Think for yourself and rationalize the logic, or complete lack of in this case, behind them excluding PVR.
FM went so far as to recode their shaders to work around nV's compiler to slow down the FX with the last patch(feel free to check that out for yourself, they reordered code as nV's optimizations worked with all apps that used the same code). Why don't you find a single game that this has happened to. If these optimizations are one hundredth as fragile as you make them out to be you should be able to list dozens of games easily. I've already pointed you to the parts you should focus on, I've done everything for your end of this discussion except found any evidence of app specific optimizations being fragile with the sole exception of when a company tries to break them. Normal application updates will not have this sort of impact, and I have several years a slew of drivers and hundreds of games to back me up on this. You have FM.
The UT2K3 specific optimization is long gone. And yes, detecting UT2K3 would be a profile(by definition actually). You still have not given a single good reason why this should be considered a cheat except that FM has told you how you will think and you follow them blindly, or talk of fragile optimizations that break so easily without giving a single example of this happening unless the developers attempt to do exactly that explicitly.
Except it doesn't. Even using FutureMark solely, who seem to tell you how to think, nVidia's performance is up ~30%-40% without clip planes.
Has almost zero credibility to you. Do you realize that nVidia has increased their marketshare a decent amount since the launch of the NV30? They have grown in strength in the industry in reality, very far removed from what you think is happening. They have lost credibility with you and people who think like you but that is really a very short term issue. As soon as the situation changes and the IHV you favor is doing the same thing it will be fine.
One other point I want to bring up to you. Do you realize that in the professional market having a driver profile that is entirely based around what application you are using is a major feature? Not talking about $200-$500 consumer parts, but the $2K-$5K professional parts. Exactly what you are lamenting is not only tollerated but people pay quite a bit extra to make sure they get it. For the record, ATi does it right along with anyone else. You want to start blasting E&S, SGI, HP, Sun et al for doing what you are showing such opposition to? Please start blasting ATi for the same thing in all of your posts from now on, or explain in detail why you don't think they should be. And for the record, ATi focuses their app specific profiles for the same applications that are most commonly used in pro 3D benches.
The KEY word there was EXPECT.
Question, do you have any supporting evidence for what you just said? You know, you sound a lot like a ATIboy to me. Blind fanboyism? I am just as blind as you are, we are just going to have to wait and see.
Well, I remember AMD releasing a .13u named the thourobred which was the same core as the .15u palomino. New manufacturing process, that's all you got. Not enough info for your thought. Keep daydreaming.
About the PowerVR, have heard that they were releasing a new chip, but have never actually dealt with one besides the Dreamcast GPU, which was awsome. Great graphics, great power, rather look at that than the Graphics Synthesizer.
How is 9700 Pro top shelf, when they just finished releasing their new 9800 Pro. 9700 Pro is old stuff.
L O L
Valid drivers? Futuremark validated 52.16, but says that it can only be compared to other Nvidida cards. What kind of BS is that? Was there a purpose to not valid other drivers then, or even to look for Nvidia cheats? Futuremark is BS.
Yeah, well, that time that ATI misplaced a shadow in Splinter Cell, I believe that it was a cheat to get better performance. -Couldn't it be a driver issue? Nvidia has many as does ATI.
Obviously, XGI thought that their IQ was so high, that they needed to lower it so that it may be comparable to the competition. How is that a cheat? You know what, its a bad cheat. GO to your room. Even prior knowledge and such is not cheating. What about when Developers optimize code for a piece of hardware is that cheating, unfare play, I think that shouldn't be done either. The P4 sucked until certain benchmarks and apps started coding for it, oops Intel cheats now. Its not handling things in real time, because the application is detecting the chip.
Could it at all be possible that they were trying out different ways to create the same image with better performance. Now, we won't call NTSC cheaters for implementing interlacing won't we? They could just have been seeing the reaction from the public, and hey, for such a POS card, it's better to have something than nothing at all.
You know why the arguement that it has a negative impact on is stupid. The competition can do it too. Why is it called cheating anyway, both can do it and then we will ultimately see who is faster. You can't cheat on this. There is no unfair advantage if anyone can do it.
BFG10K
Lifer
- Aug 14, 2000
- 22,709
- 3,003
- 126
If you can't follow an argument then don't ask questions about it. I am sick and tired of your constant dancing around issues and ignoring facts.
PowerVR has violated one FutureMark's rules. Therefore PowerVR is not certified. Put any spin on it you like, dance all you like but that isn't going to change that fact.
They did no such thing. They made trivial changes which should not have made any changes to genuine optimizations. In fact we had this discussion before and you made the ludicrous claim that it's quite normal for any compiler to make such ridiculous assumptions.
Excuse me? Were you asleep when the links to Unwinder's anti-detection tests were posted? Were you also asleep when custom benchmark results were posted showing nVidia taking a nose-dive in benchmarks they usually dominated?
No, you weren't asleep. The reason why you didn't see it is because it's not possible to post any evidence to someone who thinks nVidia can do no wrong.
Ah, so it's an optimization now is it? Not even a hack? It's been elevated by yourself yet again to make nVidia look good?
Really? I thought you said it was a hack? Now it's a profile? Interesting.
The next time we have this discussion I suspect you'll elevate it to a generic optimization, claim that there was no application detection to begin with and that in fact some ATi PR monkey was responsible for spreading the myth (possibly even Gabe!).
It most certainly does with you being the sole exception.
(1) So have ATi.
(2) And?
Says who?
Yes and you can usually control the profile in the control panel in case it doesn't work. Now show me how to control nVidia's "profiles" (and no, renaming the executable isn't "controlling" them).
And that's exactly why it happens. Now tell me, who in the consumer space is paying for nVidia to cherry pick their applications and code them so they make their hardware look better than it really is? Having a dozen or so professional applications is a lot different to dealing with 365+ games a year.
I'm done with this, I really am. I can see there is simply no way to ever change your pro-nVidia bias. I don't know what level of interest you have the company but there is simply no convincing you otherwise, not even in the face of overwhelming public evidence.
So post a reply if you like but like I said, I'm done with this because there is simply nothing more to be achieved here.
BFG10K
Lifer
- Aug 14, 2000
- 22,709
- 3,003
- 126
And that changes nVidia's cheating how exactly?
I admit though, it is a little strange that FutureMark is listing those drivers as valid. I'm guessing if they didn't then there would be no certified drivers at all from nVidia so they've chosen the ones that cheat the least.
Driver issues that magically cause screenshot programs to produce AA when there is none?
That is total and utter garbage.
How is Intel cheating if developers choose to optimize their applications for their architecture? Do you even understand the relationships between code that shares something vs code that is shared?
Yes, it would be possible if the person lacked even basic understanding of the issues at hand.
Except NTSC has precisely nothing to do with anything being discussed here.
\
Actually that is false. They have done nothing but give up marketshare to ATI in all but one sector. Their overall marketshare numbers are less than 4 percent apart
I've given up trying to figure out FM already.
cheat ( P ) Pronunciation Key (cht)
v. cheat·ed, cheat·ing, cheats
v. tr.
To deceive by trickery; swindle: cheated customers by overcharging them for purchases.
To deprive by trickery; defraud: cheated them of their land.
To mislead; fool: illusions that cheat the eye.
To elude; escape: cheat death.
v. intr.
To act dishonestly; practice fraud.
To violate rules deliberately, as in a game: was accused of cheating at cards.
Informal. To be sexually unfaithful: cheat on a spouse.
n.
An act of cheating; a fraud or swindle.
One who cheats; a swindler.
Law. Fraudulent acquisition of another's property.
Botany. An annual European species of brome grass (Bromus secalinus) widely naturalized in temperate regions.
Now tell me that ATI doesn't fit into one of these definitions.
The developers are cheating AMD, because it is unfair to them. Cheating is something that only one person can do making it unfair to the competition.
cheat ( P ) Pronunciation Key (cht)
v. cheat·ed, cheat·ing, cheats
v. tr.
To deceive by trickery; swindle: cheated customers by overcharging them for purchases.
To deprive by trickery; defraud: cheated them of their land.
To mislead; fool: illusions that cheat the eye.
To elude; escape: cheat death.
v. intr.
To act dishonestly; practice fraud.
To violate rules deliberately, as in a game: was accused of cheating at cards.
Informal. To be sexually unfaithful: cheat on a spouse.
n.
An act of cheating; a fraud or swindle.
One who cheats; a swindler.
Law. Fraudulent acquisition of another's property.
Botany. An annual European species of brome grass (Bromus secalinus) widely naturalized in temperate regions.
Now tell me that ATI doesn't fit into one of these definitions.
What did I misunderstand?
I was relating NTSCs Interlacing as a cheat in order to give the illusion that there are more frames than there are. Nvidia using that AA in one frame and not in the other, if frames are fast enough may not be noticeable and may appear to look smoother than none at all, just like interlacing showing half the image on one frame and the rest on another.
You blast XGi without merit I defend them.
You blast PowerVR without merit I defend them.
You blast NV without merit I defend them.
This equals pro nV bias in your eyes, interesting.
They reordered the shader instructions. Go over to B3D and check it out, they have FM coders on the boards over there.
And when did Unwinder start releasing game patches? Your assertion has been that a game update breaking the optimizations is the reason why they aren't any good. Give an example of it.
You start calling ATi's AF a hack and I'll start using the term again for nV. NV's UT2K3 'cheat' was still doing more filtering then ATi's AF does on the overwhelming majority of games(although not UT2K3 in particular).
Everyone is. nV pulls it off with the most popular titles as does ATi(compare the 9800 to the Ti4200 in NWN).
And you take the 12 most popular game engines and you cover ~90% of actual game unit sales.
Show me where this has happened from a game update. You keep talking about this like it has happened or because if you try to you can break it. Why not go through and deoptimize the graphics engine of the game, makes as much sense.
It's not one company as you seem forget quite easily. Your stance is actually far more damaging to PowerVR then it is to nVidia and I've been defending both of them against your attacks.
I've told you already that this was going to be a pointless conversation. You want to blast everyone but ATi, I'm not going to change the stance I've held for years, including when ATi was on the hot seat, because they are the popular bandwagon now.
Reever-
Check the full market numbers and ATi is still in third place with a ways to go to number two. nVidia just managed to get in to first place a couple of quarters ago.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 22K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 400
Facebook
Twitter