Nvidia sucks?

[Sable]

http://www.theinquirer.net/default.aspx?article=32809

GAME developers on an upcoming boat racing benchmark, entitled Rydermark alleged that one of the two graphic vendors is fudging the truth with its per pixel precision. DirectX 9 requires you to use at least 24 bit or full 32-bit precision.
Nvidia doesn?t let developers use more than 16 bit and of course it is much faster than 32 bit precision. The only problem is that 16 bit precision is below the requirements of DirectX 9, so if you use less than 24 you are not DirectX 9 compliant.

If you want to do normal mapping, parallax mapping and water reflection/refraction, your Shader requires 32 bit precision.

Nvidia doesn?t leave you any choice, it's claimed. You simply cannot turn 24 or 32 bit precision on, you are always locked at 16 bit. From a developer and artistic perspective this is really unacceptable but will buy you a few frames here and there.

Developers have also informed us that they have no way to stop Nvidia doing this. The only way is to make the community aware, and that can change some minds. There is more to come and we will try to get you screenshots to really see the difference.

---------------------------------------------------------------------------------------------------

As always, pinch of salt with the story.

If it's only a few fps then why do it? Obviously if the screen shots (when they arrive) don't show any difference then I don't see a problem but if it's degrading image quality then it's a blatent cheat. And one which would bound to be found out. O_o

[akshayt]

inquirer = crap

[Sable]

Originally posted by: akshayt
inquirer = crap

Good excuse for a flame fest though.

Obviously I'm not gonna believe anything until I see some screenshot business.

[josh6079]

Originally posted by: akshayt
inquirer = crap

I thought the X1900XTX was crap?

I do agree with you though in that the Inquirer is just as reliable as a magic 8-ball.

[BFG10K]

That article flip-flops from texture to pixel precision so I sincerely doubt it's correct.

[Gstanfor]

Sounds to me like the developer is misusing the Global Partial Precision flag. FP16 is supported by DX9 as partial precision. You can either _PP hint an instruction or us the Global Partial Precision flag to run all shaders at partial precision.

3Danalyse should quickly show the true situation in any case.

NVIDIA GPU Programming Guide
Page 21

[Stumps]

Originally posted by: akshayt
inquirer = crap

hmmm, the young Padawan is starting to learn

[hardwareking]

how come there isn't a mention of ATI in that article?
Is it cause they have all those features which nvidia don't?

[Sable]

Originally posted by: hardwareking
how come there isn't a mention of ATI in that article?
Is it cause they have all those features which nvidia don't?

That's what they're implying yes.

That article flip-flops from texture to pixel precision so I sincerely doubt it's correct.

Does it? Where does it mention the texture precision?

The thing with this article is that I can see an inq reporter getting confused/messing up the story but surely the benchmarking people would know what they're talking about and they've flagged up the problem originally.

Mind you, now I think of it, why has this never been caught by Futuremark when they've put together their benchmark?

Forget the pinch of salt, time for an entire salt shaker. O_o

[BFG10K]

The texture thing is in the second title, right above Fuad Abazovic's name.

The thing with this article is that I can see an inq reporter getting confused/messing up the story but surely the benchmarking people would know what they're talking about and they've flagged up the problem originally.

If it's true then I can't see how nVidia could get WHQL certification for their drivers, but they appear to do so.

[Sable]

Originally posted by: BFG10K
The texture thing is in the second title, right above Fuad Abazovic's name.

If it's true then I can't see how nVidia could get WHQL certification for their drivers, but they appear to do so.

Christ on a bike!!! Absolutely right, I don't mind them writing utter tosh stories but at least write consistent tosh.


Do you know if this Ryder Benchmark has a website as the only information I can find about it is from theinq or sites regurgitating theinq's story.

[Nirach]

Originally posted by: josh6079

I thought the X1900XTX was crap?

I do agree with you though in that the Inquirer is just as reliable as a magic 8-ball.

Nothing wrong with Magic 8-balls.

You can trust them to be wrong and vague 90% of the time. The other 10% they're broken.

[F1shF4t]

nvidia cards supported full 32bit precition since fx days.
Ati's 9*** and x*** series cards were 24bit precition. Shader model 3 requires 32bit precition i think to be fully compliant.

Nvidia cards can run at 16 bit precition, and it was quicker back in the fx days, i dont know if its the same now.

Someone who knows more about this should post something.

[LOUISSSSS]

x1900xtx is crap
if this is true then it'd come out elsewhere, besides the inquirer. dont really trust anything they write

[akugami]

Originally posted by: Sable
Mind you, now I think of it, why has this never been caught by Futuremark when they've put together their benchmark?

Forget the pinch of salt, time for an entire salt shaker. O_o

I've got a pallette of 50lb bags of salt in the backroom if you need that.

I find it highly unlikely this story is true. ATI and nVidia dissects each other's hardware and software looking for any chinks in the armor and anything they can copy. There are a lot of knowledgeable reporters out there that also scrutinize the hardware. Then there are game developers who are looking at the hardware and trying to get the most out of it. Out of all those people, you'd think someone would have noticed the Geforce 7 series is not DX9 compliant way before now (assuming the story is true, which I doubt).

I don't mind reading the Inq. They have nuggest of truth in some of the things they report and of course it's a rumor site. This just seems like something they wrote after smoking some Wacky Weed.

[apoppin]

Nvidia doesn?t let developers use more than 16 bit . . .

i find that really [really] hard [no - impossible] to believe.

i need *confirmation* before i give ANY credit to this theinq BS "news".


even if i wanted to believe it
:Q



happy 4th of July!

[everyone, everywhere . . even Canada and Texas]
:laugh:

[hans030390]

I'm pretty sure Nvidia likes developers to use 16 bit when it can be. You have to admit, it's probably not practical to have 32 bit running all the time when the same effect can usually be had by 16 bit. I'm sure with new features in Shader Model 3 and all that, which require 32 bit, they would let developers use 32 bit.

Even if it is true, it doesn't bother me. I've never seen an IQ difference between ATI and Nvidia, so even if Nvidia is doing this 16 bit thing, it looks the same as ATI...so I can't complain. Everything looks good to me.

[LittleNemoNES]

Originally posted by: LOUISSSSS
x1900xtx is crap
if this is true then it'd come out elsewhere, besides the inquirer. dont really trust anything they write

WHAT? Let the flames begin!

[Nelsieus]

Developers have also informed us that they have no way to stop Nvidia doing this. The only way is to make the community aware, and that can change some minds. There is more to come and we will try to get you screenshots to really see the difference.

Why would developers go to theINQ if they expected to be taken seriously, and why would they have waited all these years to finally say something.

Most of the people at B3D call this BS, so I'm inclined to go along with them.
http://www.beyond3d.com/forum/showthread.php?t=31786

[fierydemise]

Waiting for confirmation from a more credible source

[otispunkmeyer]

if this came out round the time of the FX series i would of believed it.

those cards could do either 16bit or 32bit and they were horrendous at 32bit, this along with other pieces of misguided desgin gave the FX series ****** DX9 performance. for good speed you had to go to 16bit...hence having to run in DX8.1 mode instead of DX9 mode on many games.

i thought the NV40 was designed from the ground up with 32bit speed in mind? and of course G7x is a progression of that rather quite stellar design.

[ronnn]

Fuad is very good at getting your attention and he certainly covers all bases - so he can get at least some of his stories right. This one is highly unlikely, but does futher erode the public preception of nvidia IQ. Nvidia's famous pr better get cracking.

[josh6079]

To answer if it was true:

why would they have waited all these years to finally say something.

because G80 is almost here and the 7 series has already had a good run. If they would have let this be known at its launch, I don't think it would have been pretty.

I still think this is BS. (I had two 7800's, it better be BS. Otherwise its "five across the eyes" if I ever see an Nvidia rep)

[Elfear]

Originally posted by: josh6079
Otherwise its "five across the eyes" if I ever see an Nvidia rep

Lol. That struck my funny bone for some reason. Thanks josh6079.

The story sounds doubtful to me too. Like someone else mentioned, both graphics companies scrutinize each other very closely and I can't imagine this hasn't come to light earlier if it were true.

[otispunkmeyer]

Originally posted by: Elfear

Lol. That struck my funny bone for some reason. Thanks josh6079.

The story sounds doubtful to me too. Like someone else mentioned, both graphics companies scrutinize each other very closely and I can't imagine this hasn't come to light earlier if it were true.

http://www.thebestpageintheuniverse.net/c.cgi?u=beat