nVidias rules of optimizing a driver

[gkar1]

You guys are going way overboard and past fair criticism. I see nothing controversial about those slides.

Although the content of the slides seems very fair and reasoned, what is controversial is the source from which they came.

It's like seeing slides of how to run a fair and ethical business model from Enron, or how to properly and respectfully treat women from Mike Tyson.

My sentiments exactly

 

[DemoCoder]

Well, they only have two choices: they can either say nothing to the public about what they are doing internally about this issue, or they can try to communicate it.

What would you prefer they do, just close down the shop and and quit making products? If you've already decided that the company is completely worthless and you hate them, then there is little to be done either way.

But atleast by trying to reach out to people that are sitting on the fence or having an open mind, they may preserve some brand awareness or loyalty until they ship their next driver set and/or NV40.

This BBS is like Democrats and Republicans. Both parties have been caught lying in the past. Each side vehemently denounces the other side and won't listen to anything they say. Each side paints the other in worst times, or argues that the other's crimes are greater evils, etc. But without both parties people would lose choice in elections.


Let us hope NVidia does much better with the NV40 and cleans up their act. If you find yourself hoping that they fail, I think you should examine your principles.

 

[rashly]

Well, they only have two choices: they can either say nothing to the public about what they are doing internally about this issue, or they can try to communicate it.

i dont think this is the right way to communicate it though. they should publically come out and say "yes, we have found problems in our drivers that reduce iq for performance. these problems are currently being fixed, and things like this will be prevented in the future." obviously they arent going to say they were cheating, but a statement like that would be enough to start to get them out of the "dog house". of course they would need to follow up on this promise for it to work, but nvidia has to do better than some slides to get people's trust back.

 

[g__day]

NVidia catch the driver cheat developer - it was the butler - and re-assign him to a new, less well provided for team:

What do we want from NVidia? Their honest best, no more or less.

 

[DemoCoder]

I agree, so that's the kind of criticism I'd like to see. Constructive ideas on what people want from the company, rather than just zingy 1 liners. At the very least, if NVidia is reading this board, they can get a good idea of what people are asking for.

 

[Hellbinder]

Oh come on Democoder...

This is not Rocket Science now is it?? Nvidia needs to Read the board to *Figure out what people want*. Seriously.. You dont think they know what people want?? How bout screw what *people want* and do whats RIGHT.

There is absolutly nothing wrong with taking a hard line on this entire issue. It needs exactly that kind of response. No more Screwing around, No more blaiming everyone else, No more trying to detroy companies with misguided, misinformation filled PR statements. No more *conveniently* engineered to mirror Nvidias wishes Editorials from certain sites... and especially.. ESPECIALLY... No Presentations of what Fair and Honest driver optimizations are as if they are claiming that they set the standard and adhere to it already. Those Slides are no peace offering or claim that they are going to Right the Ship. It is yet another attempt to delute the real truth and Paint them in a completely innocent (and flatley Dishonest) light.

They are not claiming these are their *new* standards. These are their supposed exsisting Guidlines that they are trying to Convince people they follow.

What is needed here is Brevety and HONESTY. Not some High Horse *look at what we do and our lofty standards* Slides that are in truth Complete and Utter Rubbish in the light of their CURRENT and CONTINUEING conduct.

 

[DemoCoder]

HB, I have a hard time believing NVidia could do something that would satisfy you, given your past history.

There is some truth to what you say (don't be a Bill Clinton, admit you had sex with that women and continue on), but the emotional nature of your responses (and the way you cheerlead for ATI) doesn't lead me to expect you'd forgive them, just as Republicans would continue to depise Clinton even if he came clean in the beginning.

Of course, you could accuse me of the same thing, and I used to be an ardent defender of the NV3x (before it came out, based on Nvidia's claims, and people attacking it before there were any benches) and was even there at Nvidia's launch in Vegas, and I am very disappointed with the resulting implementation (but not the idea behind it)

I just don't believe NVidia is the pure evil entity that you seem to make it out sometimes. NVidia is not the first company to get caught cheating on benchmarks and they won't be the last. To me, the more important issue is not the 3dMark issue, but that the NV30 was such a poor implementation.

I could tolerate Nvidia's PR or cheating if in fact, the NV30 actually was a good performer. Since it's not, it makes it worse, because the cheats are not covering up a good product by embellishing it alittle, they are covering up an underperforming part. If the NV40 turns out to be an amazing product, most of this will be forgotten.

If the NV40 turns out to suck, then NVidia will turn to PR and media exercises to try and prop up sales, and we will all be even more angered by the circus trying to distract us from a bad product again. At that point, the brand may suffer irrepairable harm.

I'm just saying that it's too early to tell whats going on, so let's be a little open minded about what they are saying. It's a critical time for them, maybe there's some hint in their words that they are actually changing, even if it is not direct?

(e.g. they are not admitting directly that they are cheating, but maybe these slides are actually a "New Process" they have instituted, even though they try to make it look like it was there all along)

 

[croc_mak]

Well said democoder...

The main issue NV is trying to desperately cover up is the weakness of NV3X architecture and implementation....As long as the debates and the forum wars are centered around 3dmark03, Kyle bennett and Nvidia's PR department, NV is doing good damage control...

If the general impression of the public is that 5900 is more or less the same level as 9800, NV succeeds. As long as NV can delay DX9 pixel shader benchmarks and games and divert attention to trilinear/aniso and AA cheats NV wins too

 

[Hanners]

I think DemoCoder is right to say that the slides alone don't tell us a great deal, without knowing what else was said and explained a long side them.

They could have said 'This is the process we should use... But we're not going to, because we want to cheat you all! Muahahahaha...'

They could have said 'This is the process we have just put in place to prevent the current problems from occuring again'

They could have said 'This is the process we have always had in place, which proves that all the allegations are wrong and we have never, or will never, cheat'

I think we'll just have to wait and see which one of these scenarios plays out... I'm hoping that it's going to be the second one.

 

[Sharkfood]

Well, they only have two choices: they can either say nothing to the public about what they are doing internally about this issue, or they can try to communicate it.

Funny, those slides don't fall into either category. They are fairly standard (and reasonable) defining traits of a mindset... one that I think most people may agree with. But they are non-committal and have absolutely nothing to do with NVIDIA corporation or their drivers.

If you've already decided that the company is completely worthless and you hate them, then there is little to be done either way.

Or on the apparent other side of the fence would be- If you've already decided that a company can do no wrong, you'll like them and defend them no matter how readily apparent or visible their flaws may be.

This BBS is like Democrats and Republicans. Both parties have been caught lying in the past. Each side vehemently denounces the other side and won't listen to anything they say. Each side paints the other in worst times, or argues that the other's crimes are greater evils, etc. But without both parties people would lose choice in elections.

Actually, I've found this BBS is more like scientists and revisionists. The scientists keep pointing to measurable data, the revisionists trying to continually fillibuster that A = Z. Just saying "both parties have been caught lying in the past" makes this pretty clear. An optimization that works at all camera angles, completely clears buffers, and requires a diff/delta as the difference is impossible to notice running side by side is somehow categorized as the "the same lie" next to something that leaves trails on camera motion, has absolutely disgusting black artifacts and fails to perform full buffer clears.

I'm just saying that it's too early to tell whats going on, so let's be a little open minded about what they are saying. It's a critical time for them, maybe there's some hint in their words that they are actually changing, even if it is not direct?

Your first sentence is correct and I agree. The second is totally unfounded. Anything sourced from NVIDIA at this time should be looked at with neutrality and skepticism until there is an underlying factual, measurable outcome. Until then, no credit can be given, nor any additional blame be denounced. We know what we have now. We don't know what we will have down the road. But a set of slides that have nothing measurable or tangible to go with them certainly do not award any level of praise or commendation.

 

[MuFu]

You guys are going way overboard and past fair criticism. I see nothing controversial about those slides.

Although the content of the slides seems very fair and reasoned, what is controversial is the source from which they came.

It's like seeing slides of how to run a fair and ethical business model from Enron, or how to properly and respectfully treat women from Mike Tyson.

Exactly. Any monkey could have written those slides. There is nothing particularly wrong with them apart from the manner in which they were conceived.

I agree that alot of the current grief is simply due to IHV favouritism and frustration at nV apparently getting away with what they are doing at the moment. The "burn in hell!" attitude gets a little tiresome but "taking the hard line", as HB put it shouldn't be discouraged. Such things have swung both ways in the past and an overly neutral and considerate opinion would be counter-productive. In fact, I believe by far the most damaging reactions we've seen so far have been from nVidia die-hards. I think discussing the content of the slides is even more pointless than the continued fuss regarding driver hacks. IMHO, they amount to absolutely nothing whatsoever. Call me the eternal pessimist, but I cannot see any kind of satisfactory resolution to this issue until NV40 is launched, by which time such an event may be largely irrelevant.

I find it hard to believe nVidia didn't intentionally give [H] the apparent option of posting these slides, knowing they would. Internal documents like this usually look so much better than public announcements because they cut through the BS, but these days it's so hard to be sure. Take this nV internal mail, for example:

Our shader optimization was a legitimate form of tuning as described by recent postings from John Carmack and Tim Sweeney. While Futuremark still does not see it this way, the statement below shows that they should not have called our optimizations cheats and that we are both open to on-going discussions about how to best benchmark modern processors like GeForce FX...

...Please feel free to use these statements but refrain from adding your own interpretation or color. We are trying to get on a more healthy track with Futuremark.

Whether such things are due to a breakdown of communication within the company or a deliberate misinformation programme, the result is that we as consumers/developers/reviewers are still "talking to the hand"; the hand that never really seems to know what the rest of the body is doing.

MuFu.

 

[martrox]

I've own many nVidia cards, and have been happy with them....mostly! (FSAA has always been pretty poor). However.... this is not the first time they have lied, cheated and used unethical behavior towards their major rival. I believe it was Dave (or maybe Rev?) that told the story of the 3DFX former employee that, upon starting to work at nVidia after the demise of 3DFX said "we knew they were cheating, but never realized by how much!" And, anyone else remember how nVidia attacked the Kyro when it first hit the market? So, Demo, I'm sorry, but nVidia has a long track record of these things. No I don't want nVidia to go out of business, and neither do most of the people that condemn them. It's just that, after years of this kind of thing, and after all their "supporters" screeming in your face that nVidia "can do no wrong" it's tough to be appear objective.... especially to those nVidia "supporters"! This time, though, nVidia has finally been bit in the butt by their cheats and lies, and even some of their own fans have "seen the light."

No, I don't want nVidia to go out of business. But, they do need to come clean about all of this. They need to clean house. It's time for a major philosophical change to the way they do business. They need to come clean about what they have done and establish a track record of honesty ....well, at least establish the track record on honesty! If this doesn't happen, and they have another product line that's not competitive - and, lets face it, if the drivers for the FX series were not full of "optimizations" the FX series would be seen for exactly that! - then they do begin to look like they have the "3DFX Curse".

 

[Bouncing Zabaglione Bros.]

Whether such things are due to a breakdown of communication within the company or a deliberate misinformation programme, the result is that we as consumers/developers/reviewers are still "talking to the hand"; the hand that never really seems to know what the rest of the body is doing.
.

The main problem for me is that Nvidia is now a completely marketing led company. I've seen and worked in companies that are run that way, and I've seen the Chairman of the Board inform the whole company that their *primary* task is to support and follow the needs and requests of the Sales and Marketing departments. Everything else is secondary.

To me, this means that these slides (or their release) could very likely be nothing more than a marketing spin, both for internal and external consumption. And once again, as we have seen so many times, these are not officially released by Nvidia, but sneakily leaked through the backdoor of a friendly website. This means that Nvidia doesn't have to comment or support it's words, and they (at least in times past) gained a veneer of honesty by having an "independent" website parrot the Nvidia marketing spin.

Until Nvidia publicly state the aims in their slides, and vow to support their lofty ideals, and explain why they have not been doing so in the past, it's just more smoke and mirrors from a company that has nothing but marketing spin to sell.

 

[DemoCoder]

I disagree. These slides look like typical management rules that are usually not adhered to.

I've been in lots of companies that have rules like this given in similar powerpoint slides:

o all methods in your code must be documented
o each method must have test cases written that cover all conditionals
o a functional requirements document should exist for the application
o UML diagrams for major architectural pieces
o adhere to corporate coding style (formating, etc)
o have code reviews, security reviews, etc

They all look good on paper, but in practice, programmers hate being forced to code under a process because everyone thinks differently. I personally hate writing documentation.

I can very well imagine some well meaning nVidia project lead coming up with these slides, and trying to ensure rigorous adheral, but reality is different than hope.

Just look at ISO9000 and SEICMM, who wants to put up with that bureacracy and paperwork?

I think someone at Nvidia thought it was a good idea for internal guidelines to be published, not knowing what the backlash would be.

The nail on the head is that the NV3x was disappointing. Their PR exaggerations wouldn't be so annoying if they had a killer product.

Right now, I think the best thing Nvidia is shipping is nForce.

 

[DemoCoder]

Actually, I've found this BBS is more like scientists and revisionists. The scientists keep pointing to measurable data, the revisionists trying to continually fillibuster that A = Z.

Scientists don't always go looking for data that fits their hypothesis and exclude data that doesn't. I prefer a better analogy: this board is full of prosecutors and defense attourneys. The prosecutors only look for evidence that the suspect is guilty of a crime, the defense seeks to prove their clients innocent. Both make their case primarily through verbal argument. Every once and a while, a science expert is called to the witness stand, either on the side of the prosecutor, or the defense, to prevent evidence. The evidence may or may not be a smoking gun, but further influences the arguments on both sides. However, no verdict is ever rendered, it's an endless trial.

We've had two trials here: Quack and 3dMarkCheat, and in these cases, isn't it interesting that the prosecutors and defense attourneys happened to switch roles? Those who were "scientists" looking for evidence to prosecute Nvidia were instead looking for evidence to exhonerate ATI. If you were consistent, then you'd either consistently be a defense attourney, or consistently be a prosecutor (e.g. look at how Consumer Reports, and consumer/corporate watchdog groups work)

In no way are the people on this board true scientists, objective, and interested only in the truth. They are interested more or less in scoring points in a pathetic little tug of war.

And of course, no one believes anyone can be an independent. You must take sides in this war. If I post that I disagree with Microsoft's HLSL compiler architecture, the "prosecutors" assume therefore I must be hawking Cg, and I must endure ten thousand words of nonsense, instead of my real point, which was OGL2.0's approach is better. If I say that Vanilla is better than Chocolate, there must be some pro-Nvidia hidden message in there, and again, the subject will turn to irrelevency.

Which is why of course, I spent the last 7 months avoiding participating in any thread which discussed Nvidia or ATI.

Anything sourced from NVIDIA at this time should be looked at with neutrality and skepticism until there is an underlying factual, measurable outcome.

See, there's the rub. Neutrality and skepticism is "I highly doubt these are adhered to, or existed previous to the debacle, or..." It is not "Nvidia sUx0rz! Bwahaha, look at Nvidia fall on their face, tee hee hee..."


I don't have a problem with well reasoned commentary, but I find the sound bite denigrations to be distasteful to say the least, and not at all "scientific"

 

[Bouncing Zabaglione Bros.]

They all look good on paper, but in practice, programmers hate being forced to code under a process because everyone thinks differently. I personally hate writing documentation.

I can very well imagine some well meaning nVidia project lead coming up with these slides, and trying to ensure rigorous adheral, but reality is different than hope.

You put up with it if it is your job to do so. If management really believed in these things, they would be enforced. Nvidia is not some startup run by half a dozen programmer-geeks that do what they fancy from day to day. I'm sure it wasn't the programmers that decided to put all these cheats into their drivers and keep them there over the last year or two - it's a management decision in direct opposition to these slides.

If management are serious about these "rules" you discipline your staff, make adhereing to the guidelines mandatory, and take away their bonuses if they don't do it. It's important because without the comments, documentation, clearly written code (so that it's function is obvious) you end up with un-managable code that no one knows how to fix. It's the old "what happens if your programmer gets run over by a bus?"

Just look at ISO9000 and SEICMM, who wants to put up with that bureacracy and paperwork?

I'd agree with that because often ISO9000 is not suitable for programming, it's origins in manufacturing not translating to well to software engineering. It's especially frustrating when it is adopted for no other reason than to give marketing another checkbox, or because "our competitors have it". I found that ISO9000 reflected what the company wanted their processes to be, rather than what their processes actually are.

However there are better standards, and working standards per se are a good thing, making programmers work well together as a team, rather than a group of independent individuals all knocking up their own bits of code in secret and using different techniques and styles.

The nail on the head is that the NV3x was disappointing. Their PR exaggerations wouldn't be so annoying if they had a killer product.

Right now, I think the best thing Nvidia is shipping is nForce.

Agreed. Nforce is a great product and improving all the time, which makes the GFFX graphic cards all the more disappointing. Someone pointed out previously that while ATI were playing second fiddle to Nvidia, at least ATI could point to their good IQ, and AIW products.
Now Nvidia has lost their speed crown, they've lost all they had to boast about, leaving them with this dirty markting war and nothing else.

 

[MikeC]

Well, they only have two choices: they can either say nothing to the public about what they are doing internally about this issue, or they can try to communicate it.

My take is that those slide are serving as a prequel for upcoming chat and Q&A sessions with NVIDIA. It appears to me that NVIDIA wants to publicly discuss recent events that they've been criticized for.

 

[volt]

Well, they only have two choices: they can either say nothing to the public about what they are doing internally about this issue, or they can try to communicate it.

My take is that those slide are serving as a prequel for upcoming chat and Q&A sessions with NVIDIA. It appears to me that NVIDIA wants to publicly discuss recent events that they've been criticized for.

You think they would love to chat with us ?

 

[ClyssaN]

Let's just hope that they don't give us the PR Talk ...

 

[WaltC]

Well, they only have two choices: they can either say nothing to the public about what they are doing internally about this issue, or they can try to communicate it.

I would not categorize the reproduction of what are represented to be internal-consumption-only Power Point slides by way of a third-party web site as meeting the standard for "public communication" on these issues.

nVidia's had a long, long time to:

A) Make a genuine public statement, such as:

"We recognize that our previous driver sets have contained certain improper optimizations and henceforth this will no longer be the case."

Even something as vague and as general as that would be much more than the company has managed to "communicate" over the past months. I mean, nVidia had no trouble writing clear statements about what it didn't like about 3DMk03, did it? The company's public response to this issue simply doesn't yet exist.

B) Issue a new set of official drivers correcting the problems--not leaked--official. Had plenty of time there, too.

Something's still very much wrong with this picture.

What would you prefer they do, just close down the shop and and quit making products? If you've already decided that the company is completely worthless and you hate them, then there is little to be done either way.

Doing what is suggested above--which they haven't done--seems like the proper suggestion.

But atleast by trying to reach out to people that are sitting on the fence or having an open mind, they may preserve some brand awareness or loyalty until they ship their next driver set and/or NV40.

nV40 has no relevance here whatever.

This BBS is like Democrats and Republicans. Both parties have been caught lying in the past. Each side vehemently denounces the other side and won't listen to anything they say. Each side paints the other in worst times, or argues that the other's crimes are greater evils, etc. But without both parties people would lose choice in elections.

Irrelevant and meaningless. A convenient attempt to dodge responsibility.

Let us hope NVidia does much better with the NV40 and cleans up their act. If you find yourself hoping that they fail, I think you should examine your principles.

Let us hope nVidia manages to ascertain the fundamentals involved in "cleaning up its act" long before nv40 ships.