Kyle's new thread @[H]

Wow, for once I actually enjoyed one of your posts. Thanks for the link. It was a great read.

Tommy McClain

 

I never understood why a benchmark such as 3Dmark couldn't randomly generate the filename of the benchmark prog and the name its process gets given in windows so no shader replacement is possible.

Dave

 

From what unwinder was saying the drivers do hashing and/or direct full/sub comparisons of the shaders themselves, so the exe name is more of a means of global context detection, not for specific shaders.

 

Slide note, it seems that whatever people where saying over here, the strategic beta account at FM, is in the 2-300K$....

http://www.tomshardware.com/column/20030624/nv_cheating-02.html

 

What difference would that make? Firstly, there would still need to be a main application that you click on to begin the generation - so you're back to square one with the 3DMark/3DBork check. Secondly, even if the process names are different, the processes themselves will still be the same - the drivers could simply check for those patterns, rather than the specific names.

 

Jeez, ED, talk about seeing only what one wants to see........

 

We don't have the exact figures, but, from what we've heard, a company interested in becoming a "Strategic Beta Member" would have to be prepared to pay a yearly membership fee in the vicinity of $300,000!

Heard from whom? nVidia? *chuckle*

Let me tell you, any company that would fork out that kind of dough for 3D Mark membership would have to be staffed by imbeciles...

FutureMark itself, however, has stated a company may become a partner for as little as $5k annually. Quite a jump from $5k to $300K annually, wouldn't you say?

First THG says "we don't really know what it costs (we don't have the exact figures)" but "somebody told us it cost $300K a year, but we can't say who that was, of course (but, from what we've heard...)."

My guess would be that the "what we heard" part came from somebody at nVidia who had to get them to swear not to reveal his identity.

That's like me saying, "I don't know to what exact monetary degree THG is compensated by nVidia each year, but, from what I've heard, it's at least $500K a year." Carries exactly as much weight and credibility.

But let's examine the logical heart of the premise as related at THG, that nVidia was indeed paying hundreds of thousands of dollars a year to FM in order to, and I quote, get "...the only level that grants members any real say in the development process of the test."

Pardon me, but it would appear that nVidia got very little say-so in the development of 3DMk03, which is precisely why the company quit the program. Therefore, something is very, very wrong with the "facts" as THG has related them here.

Rule of thumb: ignore all information prefaced with "we have heard" or "this is what the rumor mill says," etc., as patently false and you'll be much better off.

 

Sure, i'm blind and you people are omniscient, sorry to disturb you dear gods.

 

The figure is incorrect according the to information I have (which came from none of the beta members).

 

Walt, Nvidia pulled out because they didn't get their own way. I guess they need to learn what a CONSENSUS is, instead of the way they think it should be done.

 

The figures might be completely fictitious, but I don't think they have to be. Really, a universally recognized benchmark program is worth that much easily...how much of even just an advertising budget is that?

How about R&D benefits from having another body with input from multiple vendors giving you direct feedback and a specific implementation you can then go test thoroughly and your R&D department can use as another source of input? It is pretty much a technically oriented OEM and individual consumer sales tool combined with R&D feedback at the same time...that price doesn't seem that ridiculous to me. How many extra people could they hire for that much money, and could they be as useful in the various ways that 3dmark can be? Well, atleast as useful as 3dmark can be if you haven't stumbled in comparison to the competition and/or can suspend payment and succeed in distracting from unfavorable results when you have stumbled . I think the answer is no, so I don't disbelieve the figures based on the idea that it is a "ridiculous amount". I do disbelieve the idea that the value is inherently "too much" for the use vendors have gotten out of it, and can potentially continue to do.

Whether 3dmark has, at specific past intervals, or will continue to be, in the future, actually representative of 3d performance...well, that is a separate issue: one that more concerns individual consumers than OEMs (OEMs don't have to care if the numbers are accurate, only if individual consumers, from their perspective, believe those numbers). It continues to disturb me that so much of the "journalism" out there focuses on the first discussion as if it is equivalent to evaluating this separate matter , and do so simply by innuendo and blatant rumor mongering.

 

Not a question of omniscience at all, but of logic.

 

Hey, D, that's such a glowing picture of the cost ratio to benefits of belonging to the FM program that it really stretches the limits of credibility to imagine an IHV ever quitting the program regardless of cost...

Of course cost isn't the issue at all as whatever the cost was nVidia stayed in the program for years and paid it. All of this nonsense nVidia churned up about the "hundreds of thousands of dollars" it has cost the company (nVidia's comments never specified over what period of time the money was allegedly paid, or even whether this sum was representative of money paid directly to FM as contrasted with sums it may have spent internally figuring out the best methods of cheating it) seems merely an attempt to deflect attention from the real reasons nVidia quit the program--because it could *not* control its development and the latest iteration of the program did not portray its products as nVidia desired.

IMO, nVidia's current position seems directly related to its inability to steer the benchmark--the "cost" angle being merely a smokescreen. Apparently whatever level of membership nVidia paid for it was not sufficient to directly control the formation of the benchmark--contrary to what THG alleges happens in the program if only a company spends enough money.

 

Agreed. Consensus, no--control, yes.

 

Even if the $300,000 is correct, it's a drop in the ocean for a company with hundreds of millions in the bank. Being able to point to being at the top of the 3DMark scores (as Nvidia did for many years) and get *one* OEM contract could make that money back.

How much money did ATI make by grabbing that recent Medion OEM win in Germany? How many other sales did ATI get becuase they could point to being the only people with a capable 3Dmark2003 card?

It's obviously *very* important to Nvidia because of all the effort they have put into getting high scores on 3DMark (cheats included).

 

In my uninformed (outside 3D industry looking in) opinion, $300K isn't all that unreasonable a price for a controlling interest in one of, if not the, major 3D benchmarks.

 

It's gotta be cheaper than paying their driver guys to write "optimizations" and all the reviewers they've had to "convince" to write glowingly about 'em!

 

If the $300,000 number is even close to true then why would anyone actually develop just average joe blow software??? If I get into programing I'm going to make the baddest A$$ benchmark program and let the big [N] cheat... (Ooops I mean optimize!!) however they want for that kind of loot... Hell I may even let Trident get in there and throw a couple of "optimaaaaazations"...

 

However, whatever money nVidia spent per year, be it $25K annually or the exhorbitant sum of $300K, it did not buy them controlling interest in 3DMk03--which is why nVidia quit the program (obviously.)

The THG comments allege that $300K buys companies the rights to manipulate the benchmark--obviously not true as per nVidia's example (whatever it actually paid.) Had nVidia been able to buy its influence then it would never have quit the program since it would have been able to force vendor-specific paths in the benchmark which would have had the effect of representing the performance of nVidia hardware in less than 2% of all published software--and nVidia would not have quit. Right?

I mean, the fact that they quit the program and have attempted to besmirch it is proof that whatever they spent it did not allow them to dictate to FM as to the formation of the benchmark. Therefore, THG's suggestion that FM's benchmark software is "up for sale" seems patently false on its face.

 

People keep forgetting that ATI and NVIDIA were not the only ones that were paying the max to have their suggestions included in the benchmark. I suspect that any of NVIDIA's suggestions were probably over-ruled by majority vote. Meaning that it wasn't just FM and ATI that didn't want their suggestions included. Had the process been like the OpenGL forum, they probably wouldn't gotten their way either. So spending the large amount money itself doesn't guarantee that you get your suggestions included. Something tells me NVIDIA had a rude awakening when they found this out.

So what specifically caused the rift? I suspect the interpretation over Microsoft's DX9 precision specs. NVIDIA probably wanted there to be shaders that used all 3 precisions(FX12, FP16 and FP32). When Microsoft made the minimum FP24, FM and other beta members pretty much concurred. This is what I believe led to NVIDIA's insistence that games will optimize for their hardware and it wasn't fair for them to be handicapped in the benchmark. I'm not actually saying I agree with NVIDIA, but I think it does help explain why NVIDIA had such a problem with the benchmark.

Tommy McClain