Fanboys fight back
OUR MISSIVE YESTERDAY on how Nvidia was merely playing Ageia for the advantage it paid for has stirred up a hornet's nest of comments and emails.
Close analysis of the feedback shows that it falls broadly into two camps.
The first is that, whilst Nvidia said that it would not be attempting to get the 177.39 driver certified for use in 3DMark Vantage, that is nevertheless the driver it encourages reviewers to use in its 9800 GTX+ reviewers guide - without any mention of the fact the driver isn't approved.
Indeed, the reviewers guide simply says that "Physics is the next big thing in gaming... with the 177.39 driver, the GeForce 9800 GTX, GTX 280 and GTX 260 become physics processors... 3DMark Vantage's built-in physics test (CPU test 2) uses the Physx API for physics simulations. Without GPU physics, the test can run 17 physics operations per second. With Nvidia PHysx enabled, the physics simulation runs 7.5x faster, at 127 operations per second." There's no mention anywhere in there of the fact that if you do that, your test results won't be comparable to any others you've run.
So that's point one - the Nvidia might be decrying its unapproved status after the fact, but it's not exactly been discouraging reviewers from using it. Point two is the fact that the usage model Nvidia is advancing here is totally unrealistic.
In the 3DMark Vantage test, the physics code is passed off to the GPU which uses almost 100 per cent of its available processing power to render the high-quality phsyics and get that 7.5x increase. But in a real game, the GPU obviously has a heck of a lot more to do - like, you know, rendering actual graphics. So whilst that big increase might appear in Vantage, in the real world that is a bogus measurement as the amount of GPU power spare, after taking care of the graphics in a game like Unreal Tournament 3, is going to be minimal.
This looks as much like a failing of 3DMark as it does Nvidia. Once again, we come to a situation where 3DMark simply doesn't reflect what's going on in the real world. And for all the controversy, one is forced to look at the number of games on the shelf right now that are going to see any advantage with Nvidia's fancy-pants physics - and the outlook is not great.
At the end of the day, when it comes to graphics companies and synthetic benchmarketing, I wouldn't trust those results as far I could throw them - whether that throwing motion was rendered with Physx or not. ยต
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
You know nVidia is gonna try to get the best possible scores in anything at this point. DAAMIT opened ITS can of whoop-ass on nVidia, and they know they're in trouble until they can throw something together that's better than the 9800GTX+. What a bunch of wonkers--trying to protect their asses like that. It's quite simple: 3DMark was always used for rasterization scoring. PhysX isn't mainstream and is nVidia-exclusive. Futuremark's "Vantage" licensed PhysX before nVidia bought it. It's clearly unfair; nVidia pussies me off more and more each day.
@Satsuki:

While i applaud your method of trying to avoid using a complete swear word when posting a comment, i think you may have missed the target on this one:

"nVidia pussies me off more and more each day."

[Such words are likely to edited in our family magazine. You can blame me for that one.. MOD]
I resent that remark.

-- 
Andrew Garrard
Publicity Officer to the English Tiddlywinks Association
What was nVidia doing for the 2 years it dominated with the 8800 series?

Keeping prices high, and milking it.. we know.. But, you'd think that in the meantime, they'd use some of the profit to do a bit more R&D, or possibly search for a competent fabrication company.

They had to have known that DAMMIT wasn't going to stay dormant for much longer, especially with the new and improved development team.

Theoretically, with the time they bought, they could have kept on top of the game if they played their cards right. Now they have to rely on spin and deception, rather than genuine numbers.
First off the PhysX test was not ment to be CPU only. It was conceived for the original Aegia PhysX card. Second OFF Nvidia bought out Aegia and have given access to millions to the same capabilities as what Aegia was trying to do except it's using hardware everyone already has. Yes it might be unrealistic as far as the test goes compared to realtime scenarios like gaming & so forth but it does show you the potential. On top of that they have been able to implement it to the ATI cards using the SDK Nvidia is allow access for free. All this whinning is just a bunch of haters who dont have these capabilities and are crying fowl for the fact that they dont have the top score in 3DMark anymore. Why hate on something that will benefit all in the long run when it comes to games. Better Physics because we have the hardware to handle it & we dont need to spend extra money on a separate piece of hardware. You guys should be praising Nvidia for doing exactly what they intended to do with Aegia. Bringing PPU to the GPU.
After reading this follow up article you have put it in proper perspective.

I thought you were a little harsh on the first article, but you have now explained the critical part.

That is, that 3dmark and vantage seperate out their tests. Dedicated to the gpu, ppu and cpu weighing them in on the final score individually.

So, allowing any of these "classes" of equipment to fulfill "dual" roles gives a fake score.

As you clearly demonstrate, by saying that in real world the gpu would not be 100% available if at all for physics calculations.

Maybe now futuremark need to change the physucs test so that it runs in a graphics intensive capacity, so that the gpu is appropriatly loaded before being given an artificially high number
Who cares what drivers reviewers use? You shouldn't be looking at the overall score in vantage while reading a video card review anyways. If you take a peek at some of the reviews lately you will notice that the PhysX does absolutely nothing for the GPU score, that's the score that matter in a video card review right?? Unless of course the reviewer happens to be using the exact same memory, cpu and mobo combo you have...
You guys are getting carried away over some computer parts? 
That's sad, really.
How about all of you wake up and join in a debate and learn about something that will impact our lives greatly!
I'm talking about "Peak Oil"
Go and research it, then debate it.
At this point and time because of our wasteful ways and because folks aren't aware of peak oil, means that there won't be any more computers in a few years time! Start by researching and finding out what you can do to help us prevent disaster.

This comment isn't SPAM, but it's actually here so that people can become aware of the energy crises we face now. Instead of worrying about computers, you should worry about where you'll be getting your gas in a few years, heating oil, plastics, medicine, food, shelter. All of these things made possible because of OIL!!
Esp. your computers and the games you play, only exist because of OIL!!
I find ironic that those benefited from the effort may be on the other camp. After reading that Aegeia API can be run on a 4850 wouldn't be possible to run it on an integrated chipset like the 780g? Event if it can't (be it because missing functionality or politics) it becomes obvious that such work can be offloaded to the integrated graphics leaving a high end card doing it's job and the low end doing physics. Now is AMD's job to make the same implementations on the Havoc engine for its crossfireX platform and Nvidia's on it's integrated chipsets.
Why are people complaining about this so much? Both video card camps are GOING to bring hardware/gpgpu physics accelleration very soon, and anyone whining about this is delusional. This will only help gamers. Free physx? Thank you! And to the author, my 8800gt doesnt even break a sweat on UT3 with everything maxed out; there is PLENTY of GPU power left for physics, as proven by the new drivers. This is just the beginning of something that is going to eventually become the norm for both amd/nvidia, and you are ALL going to love it. It's change fore the better, not 'cheating'. Accept it.
What I really want to know if Nvidia has any load balancing in their latest PhysX DLLs? :-) And has anyone you ran any PhysX games with it? I would love to see what happens! :-)

I didn't really agree with the reasoning in the first article (or couldn't follow it) but the results indeed does not reflect real work performance. And yes, the weighting will need to be adjusted or a new formula is needed for a overall score. That is why I agreed Nvidia needs to get drivers certified.



Ace, I think it is a good thing in itself but when Nvidia (or AMD or anyone) plays with benchmark results AND get caught then all hell should break loose otherwise more and more benchmarks will become pointless (or even more irrelevant than they are already). As end users, we should draw a line somewhere. As for Aegia, well before you had a PhysX card that was dedicated so it wasn't an issue then. Just my 2 cents. :-)
The creators of Futuremark are one of the few companies out there who are demonstrably on a par with Nvidia when it comes to FUD and generally talking cr*p.

Based on their dubious history of bias in benchmarking and the lack of parity with real world findings and Futuremark scores, Futuremark is not a benchmarking scheme I would consider useful for measuring anything, nor being of any consequence to either the IT industry or the real world...

They really are completely irrelevant and anyone who makes purchasing decisions based on Futuremark results really is heading for disappointment, unless their intended application is of course running Futuremark.

In which case they'll be delighted.
What matters to me is the FPS I get in my favorite games, or the games that are considered hip at the moment (like Crysis).
FuturMark is just an artificial benchmarketer that no longer has much relevance to Real World gaming.
And I have had enough of the fudging merry-go-round in the graphics arena. I've been hearing the same arguments for the past 15 years. It's really old by now.
I do look at benchmark scores - but only the game scores. That's what I base my purchasing decisions on.
Wonder how the nvidiots would react if Futuremark made an Dx 10.1 patch for Vantage and made it count in the total score!
If anyone should be blamed here it's Futuremark for writing a pure physics benchmark in the first place, If the test is designed to measure the pure physics performance of the machine without a high 3d graphics load (as the test does) then its their' own stupid fault.

Admitedly nvidia is being a little sly fox about it, but as far as the legality issue is concerned you could quite satisfactorily put the whole thing down to "load ballancing", If the graphics hardware qualifies for physics processing and is not being used, there is no reason NOT to use it, It is Futuremark's responsability to load the graphics card sufficiently to grant the simulation merit