What in the world is up with Nvidia?!?!

[lordtyranus]

I can think of a small handful and even then almost half of them are as fast, or quite close, on nV3x hardware as they are on R3x0. The shader hype so far has been just that.

Which titles are those? Certainly not Farcry and HL2.

 

[Marsumane]

Originally posted by: TheSnowman

Originally posted by: BenSkywalker

Even quite a bit worse then the R3x0 parts which is saying a lot(the R3x0 parts are unuseable with any real shader load already- I can almost hit 10FPS with my R9800Pro 😛 ). It's completely and utterly useless with heavy shader loads(the R3x0) to completely utterly and totally useless with heavy shader loads(the NV3X).

Omg, an old part with lower clockspeeds half the pipelines of newer chips can't keep up in a benchmark made to stress the latest graphics cards? What the hell was Ati trying to pull!?! Then there is my xt-pe that can't even break 30fps in any of the tests, I suppose it is trash too, eh? :roll:

Originally posted by: Marsumane


Actually the reason is for different drivers. ATI released drivers specifically to take advantage of 3dmark and raised their top cards by around 15-20%. I suspect it may have something to do w/ catalyst ai (i think it was in those newer drivers) and its general optimizations (NOT app specific cuz they said they wouldnt do that for 3dmark) and maybe a few others. ATI is hurting for a good win and needed people to think that the nv4x archetecture isnt as good in some situations (like nv did w/ doom 3). So basically i call it a tie and say its a driver war and who cares cuz its just 3dmark and buy the cheaper card.

Is that how it is? Could you please explain why Ati's pci-express cards don't need the new drivers to get their high scores then?

I really dont see how catalyst AI couldnt be in drivers for pci express as well as agp. And actually i was just speculating. Someone in another forum told me that they didnt include catalyst ai in their drivers for 3dmark so i may be wrong. I actually found out that may be due to an error that didnt let them address over 128mb of memory in their 256mb cards. I dont know if this is true either, but xbit has some info on that one also. They also had suspected that something may be amiss here due to such the large increase (15-20%) but didnt find anything thus far and are calling the scores legit. Time will tell i guess.

 

[kylebisme]

The thing is AI doesn't have anything to do with benchmarking aplications like 3dmark, Ati has made it clear that they reserve AI strictly for games. Besides, AI is always optional so the boost in performace is always visable simply by testing with it both on and off. The reason the new drivers give such a boost is because they fixed a problem with memory managment on their 256mb agp cards, hence the reason older drivers don't show the same imapact on performace on pci-express cards.

Oh and lordtyranus, I don't think Far Cry and HL2 are in the half Ben wants to talk about. 😉

 

[Pete]

Originally posted by: Gamingphreek
Well will drivers do anything. I mean people are paying like 130 for 9600XT's and i payed 198 for my 5900XT, and its getting the crap beat out of it. Is there anything that can happen BESIDES me upgrading. I was going to hold out until like Q3 of next year, however i dont know if ill be able to.

-Kevin

It's not like your 5900XT wasn't faster in every game up til now, especially with AA+AF. I agree, sell your card while it's still got some decent resale value, and save up for a 6600GT or so, or maybe trade for a 9700P. Otherwise, it was inevitable that the 9600XT's superior (for DX9) vertex and pixel (remember the FX's temp register problem?) shaders would eventually trump the FX cards in heavy DX9 benchmarks, but neither card are really fast enough to give a hoot. The 9600XT technically comes out on top in 3DM05, but neither card will win with games that oriented toward DX9 shaders.

 

[FuFighterStan]

Don't forget certain leaked Nvidia drivers provide quite the hefty boost over the "official" ones on 3dmark05 scores as well

 

[BenSkywalker]

Which titles are those? Certainly not Farcry and HL2.

I'll even spot you TombRaider: AoD to go along with FarCry- which other released game are you talking about? Been listening to the shader hype from people for two years now- quite telling when the staggering list of released titles is so huge, a massive shader revolution to be sure.

Oh yeah, Halo and DooM3.

 

[ZimZum]

Originally posted by: BenSkywalker

Which titles are those? Certainly not Farcry and HL2.

I'll even spot you TombRaider: AoD to go along with FarCry- which other released game are you talking about? Been listening to the shader hype from people for two years now- quite telling when the staggering list of released titles is so huge, a massive shader revolution to be sure.

Oh yeah, Halo and DooM3.

Colin Mcrae 2004 makes pretty extensive use of shaders.

 

[Pete]

Ben, Hardware.fr's X700XT review shows the 5900XT in a less-than-flattering light with their games selection. The 5900XT averages 70% of the performance of a 9800P at 10x7 4x8, 16x12 0x0, and 16x12 4x8.

 

[Rage187]

Originally posted by: FuFighterStan
Don't forget certain leaked Nvidia drivers provide quite the hefty boost over the "official" ones on 3dmark05 scores as well

Those arent leaked, those are approved by Futuremark to use w/ 3dmark05.

 

[MercenaryForHire]

9500Pro with Beta Drivers

- M4H

 

[FuFighterStan]

Originally posted by: Rage187

Originally posted by: FuFighterStan
Don't forget certain leaked Nvidia drivers provide quite the hefty boost over the "official" ones on 3dmark05 scores as well

Those arent leaked, those are approved by Futuremark to use w/ 3dmark05.

If they're not on Nvidia's website, they're leaked. I believe the Futuremark approved drivers were 66.51 anyway, while the latest drivers out are 66.70

 

[Rage187]

....

 

[Blastman]

I suspect just about every game being released these days has a ?fair? amount of shaders --probably still mostly DX8 class integer, but nevertheless, lots of shaders is the future of games. The much stronger DX9 shader engine on the R300 cards likely helps it run better in heavier DX8 shader situations too.

10x7 ?HL2 -source

5950U - DX8.1 ?98.2
9800XT - DX9 ?106.8
9800XT -DX8.1 ?126.6

5700U - DX8.1 ?57.5
9600XT - DX9 ?61.0
9600XT - DX8.1 ? 75.5

Both the 9800 and 9600 have sizable leads running just DX8 class shaders.

3Dmark2005 is probably a little overkill as it uses a pure floating-point DX9 shader path -- and I suspect games will use a mix of DX8 (integer) and DX9 (FP) shaders for years to come. Although the DX9 shaders are going to be the ones limiting performance so it?s probably good to see how a card performs there.

Any type of FP shader load on the FX cards really hurts performance. They are basically being treated as DX8 class cards -- which means missing out on some effects. Which may or may not be a huge deal. The DX8 path looks pretty good on HL2. On my 9600XT, depending on how the game runs, I might run it in DX8 and use 8AF, rather than DX9 and no AF. Same for Farcry, what?s the point of having nice shader-rendered water when it gets blurry after 10 ft because of no AF. :disgust:

 

[ronnn]

Originally posted by: FuFighterStan
Don't forget certain leaked Nvidia drivers provide quite the hefty boost over the "official" ones on 3dmark05 scores as well

I love nividia leaked drivers. 😉 Every new release has a leaked driver that sometimes becomes official and othertimes seem to just dissipate. They always have a hefty boost. Now ATI has a beta driver and that sounds much better. Strange that a beta would be fm approved.

 

[gururu]

at least for ATI, I think beta, for the most part, simply implies non-WHQL certified. but they certainly made sure to sneak in there 4.10s to FM, as much as they do to avoid leaks.

 

[FuFighterStan]

Originally posted by: Rage187
....

Problem?

 

[Ackmed]

I dont think I would call my PC a "Monster" if it had a 5900 in it.

PS2.0b hits the 5900 very hard, and 05 doesnt use much 1.1 from what I can tell.

 

[ronnn]

Well to be honest it doesn't do much for my 9700pro either. Watching that balloon galleon jerking accross the sky at 1fps did not instill confidence in my card. What does the x800xt pe get, 10fps? If this is the new games they better be quick with the newer hardware. These synthetic tests are really a fun time waster.

 

[BenSkywalker]

Colin Mcrae 2004 makes pretty extensive use of shaders.

If you define extensive by has a few then I guess. In eventual terms shaders should be replacing texture maps- that will be extensive.

Pete-

Ben, Hardware.fr's X700XT review shows the 5900XT in a less-than-flattering light with their games selection. The 5900XT averages 70% of the performance of a 9800P at 10x7 4x8, 16x12 0x0, and 16x12 4x8.

Remember your nV naming. The 5900XT was only 66% of the price of the R9800Pro. Besides that, I'm talking about how useless DX9 shader hardware has been despite the hype. How much have you missed having it Pete?

Blastman-

I suspect just about every game being released these days has a ?fair? amount of shaders

If by fair amount you mean miniscule then perhaps. You can count the amount of shaders utilized in 99% of games being released right now with your fingers- this is two years after all of the PR BS stating how huge they were going to be. It hasn't happened yet and it won't for some time.

The much stronger DX9 shader engine on the R300 cards likely helps it run better in heavier DX8 shader situations too.

I can throw benches back showing the boards running Halo neck and neck running DX9 class shaders- and all of them sucking wind doing it. The R3x0 boards are complete crap for running anything resembling a medium shader load. I've been running my R9800Pro now for about eight months, and the extremely small amount of shader limited titles I've purchased(read- every one I know about) all run very poorly on my R9800Pro.

The DX8 path looks pretty good on HL2. On my 9600XT, depending on how the game runs, I might run it in DX8 and use 8AF, rather than DX9 and no AF. Same for Farcry, what?s the point of having nice shader-rendered water when it gets blurry after 10 ft because of no AF.

Have they fixed AF with shaders yet? Since the enormous avalanche of shader heavy titles(heh) is here I haven't picked up a shader limited game since D3 and AF was screwed up with every title when used in conjunction with shaders(although- due to an old bug in ATi's drivers it used to work in Halo, they fixed the bug and it screwed up the AF again).

 

[Blastman]

BenSkywalker

Not this Halo stuff again.

NV cards have a separate shader path for Halo, and still aren?t rendering the PS2.0 effects in that game. The FX cards probably run default to a DX8 path for games these days. And, the built in Halo timedemo benchmark does NOT represent the performance in the game -- oczone ran a custom Halo (Fraps) benchmark and found the 5700U performance tanks compared to the 9600XT. You can?t compare the shader load between ATI and NV cards in Halo.

If by fair amount you mean miniscule then perhaps. You can count the amount of shaders utilized in 99% of games being released right now with your fingers- this is two years after all of the PR BS stating how huge they were going to be. It hasn't happened yet and it won't for some time.

You?re just skirting the issue. The shader programming model is being used in games and is the future of games. Like I say, pretty well all the new games like MOH Pacific Assault, Joint Operations are all using shaders. You don?t think a GPU with a much stronger shader engine is going o be better equipped for the new games coming out in the next year?? --give me a break.

 

[BenSkywalker]

NV cards have a separate shader path for Halo, and still aren?t rendering the PS2.0 effects in that game.

Really, why don't you explain out which PS 2.0 level shaders nV is missing that ATi is running(just double check real quick on your end to make sure you can't run that particular shader on PS 1.4 level hardware).

You can?t compare the shader load between ATI and NV cards in Halo.

Of course not, because..... it's been out for a while and we actually know about it. Much better to talk about next years games when discussing shaders.

You?re just skirting the issue.

I am?

You don?t think a GPU with a much stronger shader engine is going o be better equipped for the new games coming out in the next year?? --give me a break.

I could pull this same quote from a year ago or two years ago for that matter. You and the other PR followers have been saying the same sh!t for two years now and I'll ask you the same question all of them have refused to answer- Where are all these games? Two fvcking years. It got old a long time ago, now it's pathetic.

 

[Keysplayr]

I really wish a thread would once in a while, come to a point. There's no point to this argument. None. Who cares who's video D**K is bigger? If you do, you've got much bigger problems than a slower video card.

 

[Pete]

Remember your nV naming. The 5900XT was only 66% of the price of the R9800Pro. Besides that, I'm talking about how useless DX9 shader hardware has been despite the hype. How much have you missed having it Pete?
For a few months, yes, a 5900XT was much less than a 9800P ($185 vs. $250, IIRC--and the XT included CoD at the beginning). The price difference narrowed a few months later, though. DX9 shader hardware may be useless, but ATi's DX9 GPUs brought faster AF and nicer AA, as well. They weren't exactly one-trick ponies, but they also weren't the end all, be all at the $150-200 price point once the 5900XT hit the market (and given the games released since then). I'm not arguing that the 5900XT wasn't a good card at its price point. I was just pointing out that there are a few current games (more than Halo, FC, and TR) that show the 5900XT noticably behind the 9800P. I wasn't making a comment about its desirability or the wisdom in buying one.

I'll let you know how personally disappointed I am in the 9700/9800's DX9 speed once I get a hold of one. 🙂

 

[Blastman]

Take a look at the second set of Halo shots at oczone -- the ice gun is a PS2.0 effect and is not rendered by the 5700U.

Bablefish translation from Portuguese ?

?In this another part of the game has something that jumps to the sight. The effect of the personage when it is invisible seems to function in way correcto in the 9600XT, since for who the Halo in the Xbox played, the effect is equal. Only that in the version for PC, the effect it requires PS2.0, and same in the FX5700U this effect is not to function, and same that if forces the use of PS2.0 in the game the effect does not function.

 

[lordtyranus]

I could pull this same quote from a year ago or two years ago for that matter. You and the other PR followers have been saying the same sh!t for two years now and I'll ask you the same question all of them have refused to answer- Where are all these games? Two fvcking years. It got old a long time ago, now it's pathetic.

Don't kid yourself. There were more reasons than DX9 to buy a 9xxx card over a 5xxx. Your "all cards suck at shaders" comments are rather foolish, considering the XTPE is the best we have at running them. When will cards be good enough for you to run shaders? 2008?

Remember your nV naming. The 5900XT was only 66% of the price of the R9800Pro. Besides that, I'm talking about how useless DX9 shader hardware has been despite the hype.

If 5xxx owners did not already regret their decision beforehand, they would have certainly regretted it this January when Farcry demo was released. This was 6 months before the NV40/R420.