A Few Notes on Future NV Hardware

[Dio]

W-buffer, as far as I can remember, doesn't have the problem that regular Z-buffer has meaning it distributes linearly over the distance while Z-buffer has a non-linear distribution over the same range...

Just because it has a different distribution doesn't make it better. W-buffer has several problems, most notably 1. that it's never been consistent across different hardware platforms and 2. that the close-up Z distribution is often lacking leading to W-artifacts. These are clearly seen on the original Unreal Tournament, which on some cards had obvious stair-step effects on the GES BioRifle.

FP numbers do not have a linear distribution of precision over the range either...

There is a Jim Blinn paper that talks a lot about this called W Pleasure W Fun, I found this link to it: http://www.computer.org/cga/cg1998/g3078abs.htm

BTW, if they totally eliminate W-buffering they could go with 3x3 Matrix transforms and simplify the architecture of their shaders

You still need to generate and interpolate W. Coordinates (colour, texture etc.) don't interpolate perspectively correct in screen space, but they do if they are multiplied by 1/w first.

 

[TheMightyPuck]

The GeForceFU isn't available to the public so benching it with unavailable drivers doesn't pose a problem at all IMO.

 

[Evildeus]

Life is like that, Joe. Putting things through its "proper paces" usually is a tough decision on both parties.

Rev, do you approve of [H]'s and other web-sites testing GeForceFX with detonator drivers that are not available to the public?

The 42.68 is available on the net!

 

[gokickrocks]

hardocp used 42.67
but like you said 42.68 is available, and the only thing it increases is the points in 3dmark03

hwextreme says

- It's confirm that the new Detonator 42.68 drivers will increase 3DMark03 scores over 100-200 points but there will be visual artifacts in a few of the benchmarks.

- Users running older versions of 3DMark won't notice any improved performance.

 

[Doomtrooper]

http://www.nvidia.com/view.asp?IO=winxp-2k_41.09

Wrong, if you can't download them off Nvidia's webpage and are not WHQL then they shouldn't be used.

 

[Joe DeFuria]

The 42.68 is available on the net!

Is it available from nVidia!?

 

[Evildeus]

No and? when someone is using beta on Ati boards are ou both whinning? :?

 

[Doomtrooper]

No..see my post above.

Here is the archive:

http://www.nvidia.com/view.asp?IO=winxp-2k_archive

 

[Doomtrooper]

No and? when someone is using beta on Ati boards are ou both whinning? :?

Please check reviews that were done on all 9700's, they are using drivers downloaded off ATI's webpage, I also only run offical ATI drivers.

Including this sites review.

 

[Joe DeFuria]

No and? when someone is using beta on Ati boards are ou both whinning?

My stance on drivers. Read it for yourself.

http://www.beyond3d.com/forum/viewtopic.php?p=7388#7388

 

[Doomtrooper]

Yes a very good stance on graphic card reviews here at Beyond3D, using leaked unsupported drivers in reviews should be left to the PR cronies like Kyle and [H].

 

[Evildeus]

No and? when someone is using beta on Ati boards are ou both whinning? :?

Please check reviews that were done on all 9700's, they are using drivers downloaded off ATI's webpage, I also only run offical ATI drivers.

Including this sites review.

Where did i say 9700? I think the first reviews in july where with non public available drivers

 

[Evildeus]

The 42.68 is available on the net!

Is it available from nVidia!?

No but it is available to the public as i stated. That was your question if i remind correctly and the correspondant answer.

Rev, do you approve of [H]'s and other web-sites testing GeForceFX with detonator drivers that are not available to the public?

Then, yes it's not available from Nvidia, so what? The GFFX is not public yet or am i missing something?

 

[Joe DeFuria]

The point is, Doom is not being hypocritical. You asked if we would be "whining" if this happened in ATI reviews....and the fact is it didn't. So raising that accusation without any basis isn't fair, now is it?

And no, the first hands-on reivews (where reviewers got a card and were able to test it in house) it were in August, just as the Radeon 9700s were shipping, with the shipping drivers.

In any case, I'm disappointed that we're talking about non-released drivers, when the real question is about the hardware itself...will we actually see any 5800 Ultras for sale any more?

 

[Evildeus]

The point is, Doom is not being hypocritical. You asked if we would be "whining" if this happened in ATI reviews....and the fact is it didn't. So raising that accusation without any basis isn't fair, now is it?

And no, the first hands-on reivews (where reviewers got a card and were able to test it in house) it were in August, just as the Radeon 9700s were shipping, with the shipping drivers.

In any case, I'm disappointed that we're talking about non-released drivers, when the real question is about the hardware itself...will we actually see any 5800 Ultras for sale any more?

I don't know i'm not working for Nvidia. Ask them

I'll rephrase then. The first bench on the 9700 were on non official drivers

 

[Joe DeFuria]

No but it is available to the public as i stated. That was your answer if i remind correctly.

Sigh...

So what you're saying is it's a "leak" of a private driver, and not a driver made publically available by nVidia. That's what I thought. And read my link about my thoughts on that matter...

Then, yes it's not available from Nvidia, so what? The GFFX is not public yet or am i missing something?

What you are missing is that these are UNIFIED drivers. Boy, does nVidia love to brag about those....There are officially available, WHQL candidate drivers available at nVidia's site.

With WHQL, we can have SOME level of confidence that there's no GROSS cheating going on, like say, trying to force FP16 when an app requests FP32.

For whatever reason, nVidia has not released this set to the public. If they are good enough to send to web-sites to publish benchmarks on, they should be good enough for the public to download and use.

 

[Joe DeFuria]

I'll rephrase then. The first bench on the 9700 were on non official drivers

Sigh...actually you don't even know THAT. For all we know, they were using official drivers at the time. ATI has unified drivers too, you know.

In any case you point is irrelevant, because it has no bearing on this discussion about web reviewers using drivers...

 

[Evildeus]

Joe,

First, often IHVs put some beta on their site non available on Nvidia site, as official drivers, so what?

Then, yes they are not certified drivers, and till then we won't be able to know what trick they used if any. But i stay correct on your first question. You are raising another question, and for that no, i can't respond till i don't know.

 

[Evildeus]

I'll rephrase then. The first bench on the 9700 were on non official drivers

In any case you point is irrelevant, because it has no bearing on this discussion about web reviewers using drivers...

I said ati boards not 9700s board, so prove me that i'm wrong. Even if it's on the website of Ati or Nvidia, till it's beta, that won't respond to your implicite question.

 

[Joe DeFuria]

First, often IHVs put some beta on their site non available on Nvidia site, as official drivers, so what?

If that's the case, then fine. The point is, SOME PUBLIC SOURCE (be it nVidia, or the board manufacturer) must make the driver available. It should NOT be some "mystery driver" handed over from nVidia to a web site, nor should it be some internal build driver handed to DEVELOPERS that was leaked. So what is the source of these drivers?

I said ati boards not 9700s board, so prove me that i'm wrong.

Whatever.... you want me to prove that ATI "never" gave unreleased drivers to web-sites to use? Don't you think that it might be a little more logical for you to find ONE case where they DID?

And if they DID, then that was WRONG. So what exactly is your point?