NVIDIA's offical response to Futuremark

[Hanners]

Originally posted by ImaginaryFiend
The reason I think they are the only ones capable of cinematic shading is that ATI can only do 64 instruction-long programs. Nvidia can do at least 1024 instructions.

R350 supports an infinite number of instructions (theoretically of course).

I would go through all the other points you got wrong (particularly in your post before last) but it would take way, way too long.

[Zenikase]

Originally posted by ImaginaryFiend
Here's Nvidia's first pixel shader patent. The reason I think they are the only ones capable of cinematic shading is that ATI can only do 64 instruction-long programs. Nvidia can do at least 1024 instructions. Nvidia also has fp32, which cinematic renderers use. ATI also has severe restrictions on the texture lookups, which is unacceptable for renderman-like shaders.

That's not exactly correct. The R300 can have up to 32 texture instructions and up to 128 arithmetic instructions (64 for scalars and 64 for vectors), altogether 160 instructions. NV30 can do 1024 altogether (texture and ALU instructions are shared).

In addition, the professional sector makes up a very small portion of nVidia's sales. Most people just want a video card that will give them pleasant visuals in DX8.1-optimized games (ie, DoomIII, Deus Ex 2, Half-Life 2). Extremely long instructions and FP32-quality color won't occur for at least a good year or so. nVidia's just tooting their own horn.

[TheFrnchTickler]

Originally posted by Hellbinder
ATi did not leak Doom 3. Even Id publcally stated this.

Another case where Nvidia fansites continue to spread misinformation and at the same time try to claim that Nvidia is not evil.. :rolleyes:

I also have *factual* issues with nearly every scentance posted in the above thread..

Example

None of the above is True at all. And thats just to start.

And you guys wonder why i resort to flaming on occasion...

Hmm... not to doubt you, or the person that made this claim, but can either of you back up what you state?

Hellbinder: can you post a link to something before nvidia's time demonstrating these things? (96 or earlier is probably good, since they only had the nv1 before 96)

Imaginaryfiend: same goes for you, can you show any proof (beyond3d article? nvidia whitepaper?)? I notice that this is your first post on the board. It's a little suspicious, not that I'm inclined to believe Hillbinder first, as he has demonstrated his rabid-fanboy status repeatedly.

[deejaya]

Wow, still open :D

I would like to see a link where id publically stated ATi did not leak Doom 3 alpha. Last I can remember was some ATi employee being fired or something similar.

[Hellbinder]

as he has demonstrated his rabid-fanboy status repeatedly.

Great.... :rolleyes:

And no i am not going to prove anything. The information is easily obtainable with a simple Google Search.

If you and others would actually do some *beep* research once in a while you might discover the truth that i am not a *rabbid Fanboy*.

[Hellbinder]

That's not exactly correct. The R300 can have up to 32 texture instructions and up to 128 arithmetic instructions (64 for scalars and 64 for vectors), altogether 160 instructions. NV30 can do 1024 altogether (texture and ALU instructions are shared).

In addition, the professional sector makes up a very small portion of nVidia's sales. Most people just want a video card that will give them pleasant visuals in DX8.1-optimized games (ie, DoomIII, Deus Ex 2, Half-Life 2). Extremely long instructions and FP32-quality color won't occur for at least a good year or so. nVidia's just tooting their own horn.

Which is not really covering the entire story either.

Nvidia cant run at any speed in FP32 color mode which is the Dx9 standard for *cinematic shading*. Further the R350 via its Hardware F-Bufer Technology can process near limitless instructions. In fact they recently Demod several hundred instruction shader running at ove well playable FPS when the FX could only process the same code at less than 3 FPS.

Of Course Nvidia could recompile really long shaders with Cg for the Nv35 as well. But, it is still limited in many ways. In reality the Nv35 will never be able to handle the Raw instruction count of the R350. Its that simple. Nor would they ever be able to render several hundred instruction long shaders at Full percision.

(im talking Nv30/35 not Nv40)

[ImaginaryFiend]

Originally posted by Hellbinder
Great.... :rolleyes:

And no i am not going to prove anything. The information is easily obtainable with a simple Google Search.

If you and others would actually do some *beep* research once in a while you might discover the truth that i am not a *rabbid Fanboy*.

Google Search is exactly what I did to find the Nvidia patents that proved my points.

As for the timing of each of Nvidia's inventions and when ATI came out with their version, it's common knowledge, but you can check Tom's Hardware for the history.

Maybe you're not rabid, but given your lack of any evidence (and what I believe to be the wrongness of your position), you appear uninformed.

[TheFrnchTickler]

Originally posted by Hellbinder
Great.... :rolleyes:

And no i am not going to prove anything. The information is easily obtainable with a simple Google Search.

If you and others would actually do some *beep* research once in a while you might discover the truth that i am not a *rabbid Fanboy*.

No, the burden of proof lies on he who makes the claim. Imaginaryfiend made a claim, and you made a counter claim. I simply came along and asked if either of you can prove you're right. It's not my job to prove you're wrong, it's your job to verify your claim or have no credibility (and be relegated to the fanboy label).

Have a nice day.

[ImaginaryFiend]

Originally posted by Hellbinder
Which is not really covering the entire story either.

Nvidia cant run at any speed in FP32 color mode which is the Dx9 standard for *cinematic shading*. Further the R350 via its Hardware F-Bufer Technology can process near limitless instructions. In fact they recently Demod several hundred instruction shader running at ove well playable FPS when the FX could only process the same code at less than 3 FPS.

Of Course Nvidia could recompile really long shaders with Cg for the Nv35 as well. But, it is still limited in many ways. In reality the Nv35 will never be able to handle the Raw instruction count of the R350. Its that simple. Nor would they ever be able to render several hundred instruction long shaders at Full percision.

(im talking Nv30/35 not Nv40)

That's true about the F buffer. My original point was that Nvidia came out way ahead of ATI on pixel shader capability. They're still ahead in precision and texture accesses but behind on total instruction count. NV40 vs. R400 (or whatever) remains to be seen.

As for ATI's recent demo, of course Nvidia's card runs an ATI demo slower. Demos are designed that way. I promise. I've done it. Nvidia often shows demos that run more slowly on ATI cards. Dawn, for example.

This actually gets back to the main issue that Nvidia's and ATI's hardware is so different (more different than Intel vs. AMD, especially in functionality) that critical code sections
will always be hardware-specific, except when game companies don't have the resources to do it. Thus, the only valid way to compare is by comparing speed of equivalent rendering results, which is what games provide and 3DMark doesn't.

[Hellbinder]

Google Search is exactly what I did to find the Nvidia patents that proved my points.

As for the timing of each of Nvidia's inventions and when ATI came out with their version, it's common knowledge, but you can check Tom's Hardware for the history.

Maybe you're not rabid, but given your lack of any evidence (and what I believe to be the wrongness of your position), you appear uninformed.

You are completely 100% full of ****.

when a product was released is not an indicator of who Developed the Technology. It is really really irritating for me to have to supposedly Defend the *Truth* from someone like you who is just making **** up as you go along. :rolleyes:

Even though i am completely aware that you looked a couple pattents up. Let me tell you that many patents owned by Nvidia, ATi, Intel, IBM etc etc were aquired becuase they Purchaced the IP from or Purchaced the Company holding the IP.

[Hellbinder]

That's true about the F buffer. My original point was that Nvidia came out way ahead of ATI on pixel shader capability.

Again, when a product is released has nothing to do with who invented the technology.

Radeon had Pixel shader 1.0 at the Time of the GF2. Or were you even aware of that. which by your logic makes ATi the inventors of Pixel shader Technology. Which makes the Radeon always 1 step AHEAD of Nvidia in Pixel shader generation. :rolleyes:

But that is not the case. Becuase Pixel shaders, Vertex shaders, Cube maps and nearly all 3D technologies were developed by Groups of people in think tanks, or at Colledges, or at Pure Technology companies. T&L was Fist developed for a retail product By Rendition who now makes Ram for Micron. S3 had T&L designs going long before Nvidia but never introduced them. PowerVR offered Full Hardware Geometry Engine over a year before Nvidia in their ARcade Devision.

The list goes on.

dont you people get that I am not trying to say that it Was ATi over Nvidia here???

[Miester_V]

FutureMark: One step foward (patch), one GIANT step backwards.

Nvidia: Questionable market strategy at first...Now de-thrones Microsoft as ultimate evil corporation.

I'll be updating my sig soon. ;)

[Miester_V]

Originally posted by ChevyBlazerBoy
Yes! Nvidia Rocks!

:super:
*cricket* *cricket*

Feels sorry for you. All alone.

[StealthHawk]

Ok, time to take a deep breath and take a step back.

Someone answer this question, please. nvidia and ATI's driver optimizations for 3dmark03 are still invalid, are they not? Futuremark said yesterday that nvidia's optimizations weren't cheats, but also weren't allowed for 3dmark03.

Does this mean they will recall patch 330? I don't think so. Then why does everyone say that nvidia is free to cheat in 3dmark03 now? Because as I see it, that isn't the case at all.

[Miester_V]

Just for fun, I'd like to see Ati make a FutureMark 'optimized' driver set and use it to run the benchmarks. And when Nvidia sees that ATi used their own strategy against them and BEAT them, they can't cry 'CHEAT!' anymore. Hahah lol. What a predicament.

[Hellbinder]

Stealthhawk...

Here is what you dont understand.

If the extremes that Nvidia went to are only hardware *optimizations* then it does not really matter whether Futuremark says they disprove or not. It is literally open season to do whatever you can get away with until another patch comes out, then find another way to *optomize*, then another patch comes out.. etc etc etc...

If there is no cheating. Then its up to IHVs and their fanbase and partners to decide what is legit or not. Which now at this time means its completely open season anything goes. Becuase Frankly 3dmark03 just became nothing more than another run of the mill Application like any other game.

Catalyst Maker already posted a public statement that they are officially suspending all work on Features etc in drivers and will concentrate *Exclusively* on 3dmark *optimizations*. He speculates that this will result in a 25% increase in performance. I saw another post in a private forum that the ATi D3D driver Guys where jokingly putting together a driver that replaces every single texture in 3dmark with the ATi logo.

Now these guys are jesting a little you understand but it makes the point. If its not a *Cheat* and all these things are specific hardware *optimizations* then anything goes and Who care wether 3dmark approves of it or not.

Dont you see the ramifications?

[muzz]

Originally posted by Mulciber
that was blatant sarcasm :rolleyes:

No sa.........

[ImaginaryFiend]

Even though i am completely aware that you looked a couple pattents up. Let me tell you that many patents owned by Nvidia, ATi, Intel, IBM etc etc were aquired becuase they Purchaced the IP from or Purchaced the Company holding the IP.


You're right that companies license patents from each other and they also buy rights to own and prosecute patents. Nvidia bought 3dfx's patents, for example. Microsoft bought SGI's graphics patents, too. But if you would bother to look the patents up, as I did, you would discover that every one I cited had entirely Nvidia employees as inventors. This ought to fit your criteria for specifying who invented it. If it's good enough evidence for the US government and the whole legal and business communities, it ought to be good enough for us.

Originally posted by Hellbinder
Again, when a product is released has nothing to do with who invented the technology.

Radeon had Pixel shader 1.0 at the Time of the GF2. Or were you even aware of that. which by your logic makes ATi the inventors of Pixel shader Technology. Which makes the Radeon always 1 step AHEAD of Nvidia in Pixel shader generation. :rolleyes:

"Pixel shader 2.0" and beyond refers to programmable pixel processing, the new DX9 stuff that Nvidia and ATI both separately developed and ATI and Nvidia worked with Microsoft to define the DX9 interface to it. The "Pixel Shader 1.0" term is a DirectX 8 term. Before DX8 it was just called register combiners. Either way, PS 1.0 through 1.4 are essentially register combiners, an Nvidia patented technology. David Kirk, Nvidia's chief scientist, is the principal inventor. All the other inventors are Nvidia employees.

Register combiners were first introduced in the Geforce 256 in August 1999. The Geforce 2 GTS came out in early May 2000 and the Radeon in May or June 2000. The GF2 made nearly no changes to the register combiners. The Radeon was nine months after the Geforce 256, the first machine to use register combiners, or "Pixel Shaders".

But that is not the case. Becuase Pixel shaders, Vertex shaders, Cube maps and nearly all 3D technologies were developed by Groups of people in think tanks, or at Colledges, or at Pure Technology companies. T&L was Fist developed for a retail product By Rendition who now makes Ram for Micron. S3 had T&L designs going long before Nvidia but never introduced them. PowerVR offered Full Hardware Geometry Engine over a year before Nvidia in their ARcade Devision.

The list goes on.

dont you people get that I am not trying to say that it Was ATi over Nvidia here???

First, we need to distinguish between T&L and programmable vertex processing. Both have existed for decades. Jim Clark, founder of SGI, had a 1984 Siggraph paper on a T&L engine. SGI and other graphics hardware has had it forever. Some of them were even programmable, but only in microcode by the manufacturer.

In the PC graphics era, several companies have had T&L on the same card as the rasterizer. 3D Labs might have been the first. Diamond was early also, with the FireGL 5000. Rendition was also in there, as you say. However, these need to be distinguished from what is currently done - having T&L on the same chip as the rasterizer, not just the same card. This was first done for PC graphics by Nvidia with the Geforce 256, hence its name (Ge for geometry). Nvidia patented having it on the same chip and all the inventors listed are Nvidia employees. Again, this was nine months before the Radeon. The S3 Savage 2000 was introduced between the Geforce 256 and the Radeon.

The next step is programmable vertex processing, which was developed entirely by Nvidia - with Microsoft, the OpenGL ARB, and ATI coming in later in the design to work on interfaces to define vertex programs. Please see the Siggraph 2001 paper on the programmable transform engine by Erik Lindholm, an Nvidia architect, and other Nvidia employees. Programmable vertex processing was also one of the patents I listed, with Erik Lindholm being the main inventor.

You're right that cube maps were not invented by Nvidia. I didn't say they were. They were first used in the early 80s in software renderers at universities and animation companies, as you say. I said they were first put in hardware by Nvidia. Again, this was in the Geforce 256 vs. the Radeon era. If Nvidia hadn't put cube maps in their hardware Microsoft wouldn't have exposed them in DX7 and ATI wouldn't have implemented them as soon. The original point still stands.

You have Nvidia to thank for several graphics technologies and your gaming life is better because of Nvidia. That doesn't mean they can do no wrong but I think it means that they will be worthy of respect in the future as they have been in the past.

[digitalwanderer]

Originally posted by Hellbinder
Stealthhawk...

Here is what you dont understand.

If the extremes that Nvidia went to are only hardware *optimizations* then it does not really matter whether Futuremark says they disprove or not. It is literally open season to do whatever you can get away with until another patch comes out, then find another way to *optomize*, then another patch comes out.. etc etc etc...

If there is no cheating. Then its up to IHVs and their fanbase and partners to decide what is legit or not. Which now at this time means its completely open season anything goes. Becuase Frankly 3dmark03 just became nothing more than another run of the mill Application like any other game.

Catalyst Maker already posted a public statement that they are officially suspending all work on Features etc in drivers and will concentrate *Exclusively* on 3dmark *optimizations*. He speculates that this will result in a 25% increase in performance. I saw another post in a private forum that the ATi D3D driver Guys where jokingly putting together a driver that replaces every single texture in 3dmark with the ATi logo.

Now these guys are jesting a little you understand but it makes the point. If its not a *Cheat* and all these things are specific hardware *optimizations* then anything goes and Who care wether 3dmark approves of it or not.

Dont you see the ramifications?
Ok, I GOTTA get a copy of the 3dm2k3 with the ATi logo replacement thingy...too much screenshot fun!

Originally posted by ImaginaryFiend
You have Nvidia to thank for several graphics technologies and your gaming life is better because of Nvidia. That doesn't mean they can do no wrong but I think it means that they will be worthy of respect in the future as they have been in the past.
They are going to have to earn that respect, the same way they've earned the communities scorn and disgust.

[Zenikase]

Correct me if I'm wrong, but aren't they called "pixel shaders" for their programmability? The major differences between PS1.4- and PS2.0/3.0-type shaders is that the latter is capable of floating-point precision and branching/conditional code support. Essentially, that's it. The pixel shader found in NV2x isn't fixed function (like register combiners), but it doesn't offer the flexibility of its successor.

[SmuvMoney]

Originally posted by frenchy2k1
Actually, this is probably not such a bad thing. Now, instead of just spending our days running useless benchmarks to compete for bragging rights, we could return to playing games...


OT, but I have never actually run or installed 3DMark03. I stopped running 3DMark01 sometime in early 2002. That being said, let me get back on topic...

I think there is a place for both syn(thetic) and game benchmarking. Syn benching can be used to show theoretical potential - PS 2.0 shaders and such. Meanwhile, game benching show a closer correlation to actual potential/ability based on game engine. Both are needed to assess a card's ability.

However, how can I be sure that nVidia (or to be fair, ATi, Sis, Matrox, etc) won't take some of these "less than noble" application optimizations and use them in game benchmarks? I welcome the valid application optimizations - it's the invalid ones (aka known as cheating but nVidia didn't see me type that :angel: ) that concern me. Worse, If you can strongarm your way into getting away with it on a syn benchmark, what stops it from occurring on a game benchmark? What happens that potential based on benchmark is shown to be a fallacy once you start the game? Heck how do you know if the game benchmark you run is even truly valid anymore? At least FM had dev tools to make sure things were legit, which is how nVidia got into this into the first place. Will games be able to have this same type of functionality? Do we want to burden the game devs as such? They have enough deadlines and resource management issues to deal with trying to create the game - a game engine benchmark suite could be considered superfluous if push comes to shove.

Maybe I should have changed the title to "The Day The Objective Benchmark Died..."


Seriously, as stated in the PR, all the other bench are using processor specific optimisations. The graphic chips are getting closer and closer to general purpose processors, with shaders used instead of fixed treatment. Intel wins in most benchmarks because they are using their extensions SSE2, do you see AMD complain? They won in FP before thanks to their architecture. Graphic is just becoming the same way.


My concern here is the potential lack of objective benchmarking for GPUs. I do agree with you that different CPUs and GPUs do use their own methods and optimizations to achieve the same thing. However, that doesn't eliminate the need to bench based on a standard - be it an API, a set workload, a set/minimum IQ, etc. If you optimize in such a way that you violate a standard that you're testing against (reduced IQ/precision, hardwired clipping planes), then the optimization isn't valid, i.e., the "ch-" word. Optimization while keeping within the rules and standards is valid.


And dont talk about 3DFX. They were the first to optimize this way: nobody remember the miniGL drivers? What do you think those were?


Well nVidia is trying to push their own initiative/standard that benefits them the most instead of a third-party industry-wide standard - what is the difference between that and what 3dfx did? Not to talk about 3dfx, but you brought them up. :D In fact, wasn't it nVidia that helped prevent that from happening the first time? It seems that nVidia is following the road that 3dfx attempted to pave not so long ago - just using different terms...
I'd rather all IHV's optimize their cards based on an external standard, not one IHV's standard.


After all, if nvidia spends time to optimize their drivers on popular games, more power to them. Sure, they look better at UT2k3 Flyby, but the game benefits also. 3Dmark is only here for bragging rights! (and so is Quake3 now, with its 300+ fps...)


See my point above for why this has potential for backfiring...I agree with the principle, but it is the execution that determines whether or not this is beneficial to the user playing the game. If said game benchmark runs at 100 FPS with tons of stuff onscreen (enemies and such) in a given spot, but standing in the spot in an empty map yields one quarter that framerate using the same settings, something is gravely amiss.

All in all, I see the potential of this being a bad thing for the industry as a whole. All standard, objective benchmarks - game and syn - are potentially suspect for invalid optimization where benchmark performance is a higher priority than ingame engine performance.

So much for trying to keep this short...my bad... :D

[StealthHawk]

Originally posted by Hellbinder
Stealthhawk...

Here is what you dont understand.

If the extremes that Nvidia went to are only hardware *optimizations* then it does not really matter whether Futuremark says they disprove or not. It is literally open season to do whatever you can get away with until another patch comes out, then find another way to *optomize*, then another patch comes out.. etc etc etc...

If there is no cheating. Then its up to IHVs and their fanbase and partners to decide what is legit or not. Which now at this time means its completely open season anything goes. Becuase Frankly 3dmark03 just became nothing more than another run of the mill Application like any other game.

Catalyst Maker already posted a public statement that they are officially suspending all work on Features etc in drivers and will concentrate *Exclusively* on 3dmark *optimizations*. He speculates that this will result in a 25% increase in performance. I saw another post in a private forum that the ATi D3D driver Guys where jokingly putting together a driver that replaces every single texture in 3dmark with the ATi logo.

Now these guys are jesting a little you understand but it makes the point. If its not a *Cheat* and all these things are specific hardware *optimizations* then anything goes and Who care wether 3dmark approves of it or not.

Dont you see the ramifications?

This is only dangerous to people who don't read between the lines. Granted, there will be a lot who don't. I don't care, let the sheep believe what they will.

If 3dmark says with one hand that nvidia is not cheating, but says with the other hand that their optimizations are not allowed, then it obviously means they are cheating. I don't think that needs to be spelled out for anyone to understand.

How is it different if Futuremark still alledged nvidia was cheating? nvidia could still find ways to cheat around 3dmark03 patches. nvidia had a suprising amount of support aroudn here through this whole debacle. Do you honestly think that more people would care if nvidia did it repeatedly? Of course, they already have.

[SlyBoots]

http://www.limerick-leader.ie/issues/20020323/box.html

I like the last paragraph>

"At the time a leading Wall Street Banker said: "We must shift America from a 'needs' to a 'desires' culture. People must be trained to desire; to want new things even before the old has been entirely consumed. Desires must overshadow needs".

They have succeeded.;)

[solofly]

In case you guys didn't know, this thread was linked from within rage3d forums found here...

http://www.rage3d.com/board/showthread.php?threadid=33689928&perpage=20&pagenumber=1

PS
Had a good laugh reading that thread btw. Rage3d is the biggest joke of a forum on the Net in my opinion or in other words waste of time. Sites like that give ATI a bad name...

[ChrisRay]

Originally posted by solofly
In case you guys didn't know, this thread was linked from within rage3d forums found here...

http://www.rage3d.com/board/showthread.php?threadid=33689928&perpage=20&pagenumber=1

PS
Had a good laugh reading that thread btw. Rage3d is the biggest joke of a forum on the Net in my opinion or in other words waste of time. Sites like that give ATI a bad name...


Hey now. No need to insult the rage3d community. I think some of the people there are over zealous, But as a whole they aren't all that bad


Just because someone owns an ATI card does. Not make them an ATI zealot. Just because someone owns an Nvidia card, Doesn't make them an Nvidia zealot.

we all people here :)