Futuremark: NVIDIA GPU PhysX not allowed

[Sampsa]

Hi,

I've changed a couple of e-mails between Futuremark CEO Tero Sarkkinen who quickly commented current confusing situation in ORB and Hall of Fame.

NVIDIA's GPU PhysX was launched in the middle of summer and most of FM's staff are on holidays (Most people are here in Finland). Tero said timing couldn't have been worse but they have updated filtering system coming up which will handle the runs which are not following Futuremark's guidelines.

When I asked more specific about NVIDIA's GPU PhysX, he confirmed that it won't be allowed. (Since Vantage's CPU test 2 is designed to measure CPU physics calculations and it is clearly stated in the rules that GPU or driver can't affect the result significantly).

So from now on if you want to run offical 3DMark result or Hall of Fame result with NVIDIA gfx you have to use Futuremark approved WHQL driver and do not install NVIDIA PhysX System Software.

I'm not sure if disabling GeForce PhysX will help (download NV PhysX Tweaker 1.0 (http://www.guru3d.com/news/nv-physx-tweaker-10/)) since it is reported that GPU PhysX installation will replace some 3DMark Vantage's files (source VR-Zone: Nvidia PhysX Driver Overwrites Vantage Files? (http://sg.vr-zone.com/articles/Nvidia_PhysX_Driver_Overwrites_Vantage_Files%3F/5928.html)).

[[XC] gomeler]

Very interesting. Perhaps this will be a move in the right direction towards seeing Vantage adopted by HWBot. Thanks for the heads up, glad to see Futuremark taking a stand :up:

[fubarswe]

we all know how physX affect Vantage.., benching it without it feels very strange to put it mildly...
Thx for info Sampsa :-)

[Fr3ak]

I somewhat like their decision, but on the other hand, having the lower results for extreme overclockers after benching with the PhysX driver might be disappointing.

For the competition between Nvidia and AMD it surely is a step to a fairer battlefield. :)

[George_o/c]

For the competition between Nvidia and AMD it surely is a step to a fairer battlefield. :)

That's right, competition should be fair :up: :)
The only disadvantage is that, scores are going to be a little lower now ... :p:

[youngpro]

i guess it comes down to personal preference,

ive never really cared too much about orb and i will just bench whatever gives me the BEST scores, in this case the physx driver is very innovative for that gpu test and I will be using it ;)

[DonBanana]

For the competition between Nvidia and AMD it surely is a step to a fairer battlefield. :)

Yeah, that is right, but shouldn´t we do it either in an equal way with all the other Benchmarks so that AMD Processors can compete with Intel CPUs :D

[George_o/c]

i guess it comes down to personal preference,

ive never really cared too much about orb and i will just bench whatever gives me the BEST scores, in this case the physx driver is very innovative for that gpu test and I will be using it ;)

What is hwbot's policy on nVidia PhysX drivers ?

[Praz]

What is hwbot's policy on nVidia PhysX drivers ?
I would think the same as Futuremark's since the top 20 need to have a valid ORB link.

[DaMulta]

Boooo if games can use it to off set the CPU in games, then it should be shown inside the benchmark that is SUPPOSED to show how well a video card will do in tomorrows games.....They even said AMD could use it freely which could solve the problem of being unfair.

[NBF]

I can fully understand Futuremark, the main issue is that PhysX replaces a few Vantage files.

If they allow 3rd party file changing in their benchmarks they will open the doors for all kind of cheats.

Only FM may change bench files if it's needed.

[metro.cl]

So what about the PPU cards? those shouldnt be allowed either from what he said.

Could you ask him Sampsa, what would happend to all the guys that bought a separate PPU and now it wont improove the score? arent allowed?

[cadaveca]

PPU should be allowed, as any system can install one, and it doesn't break FM rules, whereas nV phys-X only allows nV cards to use it.

[metro.cl]

Hi,

I've changed a couple of e-mails between Futuremark CEO Tero Sarkkinen who quickly commented current confusing situation in ORB and Hall of Fame.

NVIDIA's GPU PhysX was launched in the middle of summer and most of FM's staff are on holidays (Most people are here in Finland). Tero said timing couldn't have been worse but they have updated filtering system coming up which will handle the runs which are not following Futuremark's guidelines.

When I asked more specific about NVIDIA's GPU PhysX, he confirmed that it won't be allowed. (Since Vantage's CPU test 2 is designed to measure CPU physics calculations and it is clearly stated in the rules that GPU or driver can't affect the result significantly).

So from now on if you want to run offical 3DMark result or Hall of Fame result with NVIDIA gfx you have to use Futuremark approved WHQL driver and do not install NVIDIA PhysX System Software.

I'm not sure if disabling GeForce PhysX will help (download NV PhysX Tweaker 1.0 (http://www.guru3d.com/news/nv-physx-tweaker-10/)) since it is reported that GPU PhysX installation will replace some 3DMark Vantage's files (source VR-Zone: Nvidia PhysX Driver Overwrites Vantage Files? (http://sg.vr-zone.com/articles/Nvidia_PhysX_Driver_Overwrites_Vantage_Files%3F/5928.html)).


PPU should be allowed, as any system can install one, and it doesn't break FM rules, whereas nV phys-X only allows nV cards to use it.

1st the Owner words are pretty clear.

2nd PhysX doenst block ATI cards is AMD who has to invest in making the work just like NVIDIA did.

[YMAA]

Good for Futuremark. You level that playing field!

[cadaveca]

1st the Owner words are pretty clear.

2nd PhysX doenst block ATI cards is AMD who has to invest in making the work just like NVIDIA did.


Well, current nV phys-X driver breaks PPU function(nessecary files needed only found in VGA driver), so it's a two-sided street ATM. Unless you have nV VGA, no Phys-X functionality can be used, period. nV needs to fix that before ATI can do ANYTHING. But yeah, I totally agree...I said thing would be like this when this issue first came up, for the smae reason we hear now...no real suprise here.

[Eldonko]

At least the rules are clear now.

[CraptacularOne]

About time. Artificially inflating their scores is all Nvidia was doing.

[Marios]

How about running 3DMark Vantage with an Ageia physics card?
This does affect the outcome and the physics test is intentionally put there by Futuremark.
In the meanwhile, an Nvidia driver does the same job by using the GPU.
There is no difference between an ageia card doing physics or the GPU doing physics.
Why is there a physics driver installed on our PCs with 3DMark Vantage anyway?
Something is fishy I think and you know what I mean.

On the other hand I expect to see larrabee GPU cores integrated on the nehalem CPU. Well .... define the term CPU now.
This measure discourages AMD to include Ageia and or Havok physics hardware acceleration on their GPUs.
Though we are going to see real word improved gaming from those technologies in the near future from all three rivals (Intel - AMD - Nvidia)

[Sampsa]

Why are you guys asking about AGEIA PhysX physics card over and over again?!?! 3DMark Vantage has been supporting it since the beginning and will continue supporting it in the future. We are talking here about PhysX on GPU which is being banned!

[Marios]

There is no difference. In both cases we get hardware physics acceleration.

[cadaveca]

The difference is that Physics on the gpu does not seperate gpu/cpu performance into seperate categories. A PPU cannot render graphics on it's own, so it is considered part of the cpu sub-system.

The whole point of that test is to measure CPU sub-system performance in rendering physics, and to isolate it from gpu performance. It says so right in the design documents, as I posted in the first thread we had here, when they started to pull the scores.

[DaMulta]

I'm going to buy a AGEIA PhysX physics and be done with it.

That way I can do physX and not be pissed about them taking it out.

[Marios]

The difference is that Physics on the gpu does not seperate gpu/cpu performance into seperate categories. A PPU cannot render graphics on it's own, so it is considered part of the cpu sub-system.

The whole point of that test is to measure CPU sub-system performance in rendering physics, and to isolate it from gpu performance. It says so right in the design documents, as I posted in the first thread we had here, when they started to pull the scores.
Why is it OK for a CPU to do the job of a PPU and not a GPU?
Unless we get zero points from a CPU doing the job of a PPU we should allow a GPU doing the job of a PPU. Are we serious now?
Are we going to define the PPU GPU and CPU terms now?
Who cares about terms as soon as something out there does the same job?

[YukonTrooper]

Boooo if games can use it to off set the CPU in games, then it should be shown inside the benchmark that is SUPPOSED to show how well a video card will do in tomorrows games.....They even said AMD could use it freely which could solve the problem of being unfair.
You'd definitely have an argument if PhysX was used in all games. Considering it's not, it shouldn't be used in benchmarks.

[cadaveca]

Why is it OK for a CPU to do the job of a PPU and not a GPU?
Unless we get zero points from a CPU doing the job of a PPU we should allow a GPU doing the job of a PPU. Are we serious now?
Are we going to define the PPU GPU and CPU terms now?
Who cares about terms as soon as something out there does the same job?
You got it backwards though. The test is for CPU Physics which the PPU accelerates. It jsut so happens that Futuremark has added to the number of threads of that test when using a ppu, but they have not adjusted for the gpu, and hence the huge score increase.

In other words, that test does what nV thinks it should do on the gpu, not FUturemark, and as such, Futuremark cannot guarnatee the validity of the scores. As Someone said, it's about Futuremark adding the functionality to the app, not nVidia.

There are already GPU Physics test incorporated into the other tests. That specific test merely inflates CPU scores, not gpu scores. If nVidia had used this to accelerate the gpu cloth physics in the Jane Nash test, then there'd be no issue, as that test measures gpu performance, and is coded for such, but the cpu test is not coded to be run properly on a gpu.

[Marios]

I thought FM included a physics test to accelerate physics innovation and wider physics support from games.
Test calibration is not an excuse to allow CPU doing physics and exclude the GPU.
Physics acceleration is not a cheat. It is real and we expect better gaming in the future from this.
I cannot see your point about calibration, since all three rivals are able to include physics hardware acceleration.
3DMark Vantage X (1920x1200) is not greatly affected from the CPU or the PPU score.
This is a graphics card benchmark and should reflect real world gaming.
It is not a CPU benchmark anyway.

[k|ngp|n]

Tell your futuremark friend that they need to do a better job of "banning" the scores sampsa. FM made the decision to remove the scores awhile back but still can't seem to get it right. :down:
The whole point was to make the scores more comparable across platforms and now the scores aren't even comparable across the same platform lol.

[DaMulta]

You'd definitely have an argument if PhysX was used in all games. Considering it's not, it shouldn't be used in benchmarks.

Well from going by what your saying, if it's not in all games. They should just remove the PhysX test all together from the test....

[YukonTrooper]

Well from going by what your saying, if it's not in all games. They should just remove the PhysX test all together from the test....
No, no. If you want to run a separate Physx test that's fine. However, in a "competitive" environment where benchmarks are compared and documented, Physx should be disallowed. I agree with their decision 100%.

[DaMulta]

Well AGEIA has been bought, and soon you will not be able to buy those cards anymore. The test will still use AGEIA PhysX cards. So for the people that still have them down the road(a year or 2 from now) when it becomes almost imposable to find. Will have an unfair advantage as well because the whole test will not just be done on the CPU for those people.

The only way to really keep the AGEIA PhysX test in the program fairly(because AGEIA PhysX was bought out) would to allow it to be done on the GPU. AMD could implement this into their cards as well if they want to.

[Sampsa]

Tell your futuremark friend that they need to do a better job of "banning" the scores sampsa. FM made the decision to remove the scores awhile back but still can't seem to get it right. :down:
The whole point was to make the scores more comparable across platforms and now the scores aren't even comparable across the same platform lol.

New updated filter should be up and running in couple of days which should recognize PPU PhysX and GPU PhysX. Now there seems to be some results which shouldn't be there. Lets hope they will get all issues fixed soon so we can concentrate to benchmarking :)

[Sampsa]

The only way to really keep the AGEIA PhysX test in the program fairly(because AGEIA PhysX was bought out) would to allow it to be done on the GPU. AMD could implement this into their cards as well if they want to.

That's not how things work, designing a 3DMark benchmark tool is huge and long process and the "problem" they had with Vantage was that NVIDIA bought AGEIA when benchmark was already on finishing stages.

[DaMulta]

Very true, and I'm not saying that Vantage took them a long time to produce(I know it did). What I'm saying is that your not going to be able to buy AGEIA PhysX cards for very much longer. Then those people that have them will be the only ones to see the advantage if they don't allow GPUs to take the place of those cards.

[Anemone]

They shouldn't be called Futuremark, they should call themselves Ancienthistorymark, because clearly they have absolutely nothing to do with advancing future abilities in hardware. I have a good many reasons to have no love of Nvidia right now, but this decision is ridiculous. Replacing Vantage files? Foul and that should be something no one is allowed to do. But beyond that, if you have hardware to accelerate a process, any process it should be quite useful to do it. How else do you advance the industry? You don't want to know what 5 year old cards could do! You want to know what a card will do someday when the programming becomes common. This should have been pressure on ATI to get some code that could keep up or shut up and deal.

You want comparables? Filter out the test out of the final score, plain and easy. But no, not Futuremark. They don't care in the slightest about the end product, they only care that getting to the end product is done in some ridiculous outdated way, as some demonstration that the future of gaming is being demonstrated by such a test.

Nvidia are idiots, and can dig their own pit (proof is happening right now even). But if there is hardware acceleration available to complete a test faster that's not a foul. Replacing files, that's bad. But the acceleration itself is fine. If you have a wish to keep it fair, or judge "world records" on a fairness basis, just exclude the test and take a score without it.

Heck they are both such smart companies, I'm thinking Nvidia and Futuremark are siblings who just fight all the time and are both dumb as posts.

[STEvil]

Would be better if they just released a patch that ran the physics test twice. Once on CPU once on GPU.

[fireice2]

what about those using dual GPU's where one is used as a physx card?

[Macadamia]

what about those using dual GPU's where one is used as a physx card?

Currently it's not supported, but wouldn't it apply too?

SLI boost in game test + PhysX scores in CPU test = same double standard going nowhere.

[TheGoat Eater]

This whole debacle leaves a bad taste in my mouth:( and I am happy to continue benching the older versions and AQ3 :up:- I am a '01 addicted... I think i am starting to enjoy it more than AQ3. :D

[Blacky]

Would be better if they just released a patch that ran the physics test twice. Once on CPU once on GPU.

I soo agree with this, it could put a end to this drama finally -_- by other hand I think is lame because nvidia invested alot of money on this technology is not their fault AMD/ATi don't have it, so its unfair to call it "unfair" or "cheating" because nvidia gets a boost with their own technology in my opinion...

[fireice2]

Currently it's not supported, but wouldn't it apply too?

SLI boost in game test + PhysX scores in CPU test = same double standard going nowhere.

Yes not supported for now, but what about if Nvidia already fully implements such solution? It would be unfair to those using Nvidia based Physx solutions.

The situation is different on an intel chipset where SLI not normally possible.

[Kunaak]

just another example of how little 3dmark actually relates to gaming.

if a videocard can be used as a PPU, and games can take advantage of it, and the benchmark has a specific area for PPU testing....

wheres the problem?

[Anemone]

if a videocard can be used as a PPU, and games can take advantage of it, and the benchmark has a specific area for PPU testing....
wheres the problem?

QFT

In my mind this would be like disabling CF or SLI from the scores and say that really, since 99% of the market only has a single GPU, then you shouldn't be able to run two or for to boost scores. At what point does a legitimate hardware ability become something you don't want to allow? ATI could do this. They don't want to spend the resources on it. And frankly, with the bugs in Nvidia drivers sometimes, I think I can see that point of view, hehe. But if hardware can do a think better, then that is ok as long as it's not circumventing the "proper" code paths from operating.

Replacing files in the Futuremark software was a no-no. And that still stands as the thing "done wrong" on this issue. Having a suddenly higher score is like SLI before there was Crossfire. You don't have something comparable tough cookies.

[ACE76]

About time. Artificially inflating their scores is all Nvidia was doing.

How is it artificial? They are doing physics work with the driver and hardware implemention in their new GPUs right? Maybe I'm not understanding this whole physX situation properly.

[Shocker003]

Futuremark´s 3dmark benching is not a true test of system preformance as far as physx is not allowed. While gaming UT3 makes use of physx for example(i thought we run these benches to know how our rigs will fare while gaming or running certain programs)

I presume Intel is flexing it´s powers again... or maybe am wrong. I will bet that Larabee will be accepted by Futuremark for physics rendering(havoc), but Nvidia´s physx ain´t good enough for them. Something fishy is going on. If AMD/ATI is willing to support physx they should only ban the use of gpu physx physics for a while until ATI gpu are fully supported(fair play). I hate the fact that we have upgrade our cpu + motherboard for physics, instead of just installing a free Nvidia´s physx program. :shocked: I don´t think Nvidia is cheating as they planned using gpu to run physics after taking over AGEIA or should they simiply throw their invested money in a drawer until Intel is thur with their Larabee(HAVOC), so that we can say fair play.

[vladimir]

just another example of how little 3dmark actually relates to gaming.
if a videocard can be used as a PPU, and games can take advantage of it, and the benchmark has a specific area for PPU testing....
wheres the problem?

:up: I whole heartedly agree! According to FM's thinking, the next thing FM should ban will be the first 8-core CPUs since I'm sure one company will have this before the other.

[ahmad]

just another example of how little 3dmark actually relates to gaming.

if a videocard can be used as a PPU, and games can take advantage of it, and the benchmark has a specific area for PPU testing....

wheres the problem?

But isn't there a difference when you run ONLY physx on your GPU, and not intense 3D + physx?

[MoF]

I have understood this like this:

NVidia GPU PhysX is not allowed since it runs CPU tests with a GPU. Not because of unfair advantage etc.

If I am correct, there is just no other way to approach this than to not allow that.


Edit: And I had understood this wrong. It uses Ageia PhysX so it should use GPU PhysX too.

[Praz]

But isn't there a difference when you run ONLY physx on your GPU, and not intense 3D + physx?
Exactly. During the CPU test if the card had to render game-like graphics the score would be nowhere close to what we are seeing. As such the score is completely artificial as the card would not be capable of the same performance in a gaming environment.

[Talonman]

Where was all the crying when ATI cards produced a higher score in 3DMrk06, but slower in actual game play. Why didn't the 3D makers fix that...

That violated everything that was just in the Universe...

Now Nvidia spend millions to get our GPU's to do PhysX calculations, to aid in Game speed, and the 3DMark boys cry unfair?

To add insult to injury if your PC can do Physx calculations on another card, it counts, and is considered fair....

LOL

We seem to be always waiting for ATI to catch up in speed or features...

If your CPU is allowed to farm Physx calculations off to another card, don't call it cheeting if Nvidia is doing ours as a feature of the GPU.
We don't care if the card starts with an A or N!!

looks Vantage is for benching tomorrow's games, as long as ATI can do it too.

[MoF]

Where was all the crying when ATI cards produced a higher score in 3DMrk06, but slower in actual game play. Why didn't the 3D makers fix that...

That violated everything that was just in the Universe...

Now Nvidia spend millions to get our GPU's to do PhysX calculations, to aid in Game speed, and the 3DMark boys cry unfair?

To add insult to injury if your PC can do Physx calculations on another card, it counts, and is considered fair....

LOL

We seem to be always waiting for ATI to catch up in speed or features...

If your CPU is allowed to farm Physx calculations off to another card, don't call it cheeting if Nvidia is doing ours as a feature of the GPU.
We don't care if the card starts with an A or N!!

looks Vantage is for benching tomorrow's games, as long as ATI can do it too.

As I already said, the way I understand this is that NVidia PhysX runs a CPU test on GPU, which is just wrong. Simple. It has nothing to do with it being better or worse than Ati.


Edit: I'll note it here too, before someone comments about this. I had understood this wrong. CPU Test 2 does use Ageia PhysX too so it should also use PhysX on a GPU.

[RejZoR]

But when Ageia PCI card was running that test, that was fine? Well Ageia PhysX wasn't CPU either... I don't see anything wrong with Physx running on GPU.
NVIDIA just replaced Ageia own PPU with their hardware. What's wrong with that? After all they have all right to do taht since they own Ageia technology now...

[MoF]

But when Ageia PCI card was running that test, that was fine? Well Ageia PhysX wasn't CPU either... I don't see anything wrong with Physx running on GPU.
NVIDIA just replaced Ageia own PPU with their hardware. What's wrong with that? After all they have all right to do taht since they own Ageia technology now...

Again, as I have understood this (and I may be wrong), Ageia PhysX does not run that test as it is CPU test. Which would still make Nvidias solution wrong.

But as I said, I may also be wrong.

Edit: And I was wrong. It uses Ageia PhysX.

[Talonman]

It's just sad....

The only thing wrong about doing PhysX calculations on a GPU, is the 3D boy's attitude.

3D makers Street Cred is dropping...
They are in bed with ATI...
Nvidia's multi million doller PhysX project is starting to pay off, and upping Vantages score... (Which tests for your systems PhysX speed)
Let's change the rules now, and get them scores back down to ATI level... Hurry!!!
We made sure ATI won the 3DMark06 game, even thow the cards were slower. We can't let Nvidia win the Vantage war...
That is our new baby and everything... Don't worry ATI, we will get you back on top. We know what to do....

I wonder if they asked Nvidia for input, or just ATI on this idea....
My money says they only talked to ATI about this...

Once again we will have Nvidia cards running faster, but ATI scores looking better than they actually are on Vantage...
They are hurting the value of their score... Soon nobody from either camp will want to run Vantage anymore.

We are seeing the beginning of the end for 3DMark right now!
Talor made, to make your systems PhysX speed not count, and ATI cards to appear as fast...

I would rather they delete all PhxsX testing off their benchmark, than to allow a PhysX test to run, but not be allowed to calc on a Nvidia card.

Let us decide what part of our system actually does the PhysX calculations. They should only be in the measureing business. Not picking out our hardware, and deciding what is valid to calc on, and what isn't...
That should be my call only!

Vantage tests both CPU, and GPU scoring both. In the name of 'This is a CPU Test Only', let's bone all Nvidia customers.

Mabey the 3DMark boy's have enouigh clout in the market to have all PhysX code removed from all games too....

That would fix Nvidia's little red wagon!!!

CPU speed depends on so many other components of the system... RAM, GPU, cooling...
Let's not pretend Vantage isn't already effected by your total system specs. Picking on Nvidias PhysX enabled GPU's, and call that cheating just nuts!!

[MoF]

- nevermind -

[MoF]

- nevermind about this either -

[MoF]

Ok, I take my words back, quote from futuremarks test descriptions:

CPU Test 2: Physics The Physics Test features a heavy workload of future generation game physics computations. The scene is set at an air race, but with an unfortunately dangerous configuration of gates. Planes trailing smoke collide with various cloth and soft-body obstacles, each other, and the ground. The smoke spreads, and reacts to the planes passing through it. The physics test takes advantage of the AGEIA PhysX physics accelerator, if found on the system.

So it does use Ageia PhysX and not CPU alone.

This would mean that if NVidia could do this PhysX calculation without modifying 3dmarks files (which at this point it seems not to be able to do), I don't see any problem in it either. Except of course the fact that it might be unrealistic for the GPU to be under such a low load that it could up the scores so much.

[MoF]

Anyway, I don't still see this as favoring Ati, at least if this is accurate:

http://www.techpowerup.com/64042/PhysX_Runs_On_RV670_Scores_22_000_CPU_Marks_in_3DM ark_Vantage.html

http://www.ngohq.com/news/14219-physx-gpu-acceleration-radeon-hd-3850-a.html

http://www.ngohq.com/news/14254-physx-gpu-acceleration-radeon-update.html

[Talonman]

So when I run Vantage after the ATI patch is applied, when Crash'N'Burn runs, I am going back to crawl speed in my PhysX enabled Nvidia system. This is due to the 3DMark Boy's calling that cheating, and won't be officially recorded, even though it adds speed to actual games...

But if I have a 3rd party PhysX card like Ageia, and my calculations are not being done on an Nvidia GPU, Crash'N'Burn will run like the wind, and the 3DMark Boy's give it their blessing, and encourage you to submit your score to their database for recording? :shrug:

Sure, That's fair... and not ATI slanted at all... :shakes:
My goodness, they boned the Nvidia boy's so bad on 3DMark06, they aught to let if anything, one round go Nvidia's way. Nvidia did after all give us a PhysX processor as a feature of the GPU. That should count, and does in my book!

What would you be willing to bet if they get the ATI PhysX hack to work, the 3DMark boys would change their test back to the origional again, all in the name of 'Fairness'. :rofl:

When the 3DMark makers officially get into the business of patches designed to hurt the performance of systems only running one brand of GPU, it isn't a good thing for either camp. That's one slippery slope to start down, while still trying to maintain credibility. May the 3DMark rules go to the highest bidder!!

I know we have ALLOT of ATI fans here, but trust me when I say that this won't be received well by the huge Nvidia customer base. As soon as Crash'N'Burn returns to a snails pace on their system, they will be like what happend?
We were just starting to forgive them for putting that test in there in the first place. Suddenly it made sense...
Wait until they find out the 3DMark Boys just didn't want to count Nvidia's calculations... 'Your calculations are no good here!!!' ;)

Never mind that the old score woud better show actual game speed, we just don't want Nvidia GPU's to count... That's gunna play well.
Sweet!! Brace for impact... :D

There will be thousands waiting to throw some heat into this fire...

[Marios]

Ok, I take my words back, quote from futuremarks test descriptions:

CPU Test 2: Physics The Physics Test features a heavy workload of future generation game physics computations. The scene is set at an air race, but with an unfortunately dangerous configuration of gates. Planes trailing smoke collide with various cloth and soft-body obstacles, each other, and the ground. The smoke spreads, and reacts to the planes passing through it. The physics test takes advantage of the AGEIA PhysX physics accelerator, if found on the system.

So it does use Ageia PhysX and not CPU alone.

This would mean that if NVidia could do this PhysX calculation without modifying 3dmarks files (which at this point it seems not to be able to do), I don't see any problem in it either. Except of course the fact that it might be unrealistic for the GPU to be under such a low load that it could up the scores so much.
Modifying 3dmarks files? There is no such modification going on by installing latest Ageia drivers, just like installing Beta GPU drivers.
Ageias drivers are now Nvidias property. Technically speaking any benchmarks run with non FM approved drivers should be treated just like the benches run with beta Ageia drivers.

[Liquid3D]

Yeah, that is right, but shouldn´t we do it either in an equal way with all the other Benchmarks so that AMD Processors can compete with Intel CPUs :D

Very important point made here.... :shocked:

[Talonman]

Makes you wonder if they would have sold as many copies of Vantage, if they told the Nvidia boy's up front that they were going to disable the PhysX abilities of their cards.

[Praz]

About the only thing left to post in this thread. lol


http://fliiby.com/images/_original/clja6gwrke.gif (http://fliiby.com/file/16145/clja6gwrke.html)http://fliiby.com/images/_original/clja6gwrke.gif (http://fliiby.com/file/16145/clja6gwrke.html)http://fliiby.com/images/_original/clja6gwrke.gif (http://fliiby.com/file/16145/clja6gwrke.html)http://fliiby.com/images/_original/clja6gwrke.gif (http://fliiby.com/file/16145/clja6gwrke.html)

[ahmad]

No use beating a dead horse, but people are still not understanding how this works. So I will make my final statement, not trying to defend ATI or nVidia, just using plain old logic. PhysX test:

1. Uses CPU by default.
2. If you payed extra for a physx card, then you can take advantage of a dedicated physics processing unit, therefore your video card resources are not used when physics is working.
3. Hacking the test to make it run on nVidia hardware in the case where the GPU is idle.

Option 1 by default is valid. Option 2 by default and in collaboration with Futuremark, is valid. You paid extra to get a dedicated physics processor that will handle physics calculations independant of the video card. Option 3, according to Futuremark should not be a valid option because it does not indicate real world physics performance of your CPU, or your physX unit, or your video card!

Unless nVidia made it so if you choose to do it, you could have one card in SLI configuration do the physX, and the other do the GPU.

But in a single card setup, GPU cycles will not be wasted on physics calculations.

Coming out crying saying this favours this or that is illogical, and the bottom line is nvidia is doing whatever they can to get a higher score in vantage, even where it doesn't make sense.


Ok, I take my words back, quote from futuremarks test descriptions:

CPU Test 2: Physics The Physics Test features a heavy workload of future generation game physics computations. The scene is set at an air race, but with an unfortunately dangerous configuration of gates. Planes trailing smoke collide with various cloth and soft-body obstacles, each other, and the ground. The smoke spreads, and reacts to the planes passing through it. The physics test takes advantage of the AGEIA PhysX physics accelerator, if found on the system.

So it does use Ageia PhysX and not CPU alone.

This would mean that if NVidia could do this PhysX calculation without modifying 3dmarks files (which at this point it seems not to be able to do), I don't see any problem in it either. Except of course the fact that it might be unrealistic for the GPU to be under such a low load that it could up the scores so much.

[Talonman]

Funny how if Nvidia is being baned from being able to do PhysX calculations, because "it does not indicate real world physics performance of your CPU, or your physX unit, or your video card", but yet still will give you a better real world idea of how your PhysX games will play on your system.... Odd!! :p:

What if your PhysX processor is your GPU...

Again ... so sad....

[MoF]

Hopefully this "issue" gets fixed in a reasonable way when summer holidays are over...

[Talonman]

I wonder if a Class Action Lawsuit could be started for selling us a program that was advertised that it woud measure our system's PhysX processing speed.

Now that feature is being removed....

Can all Nvidia customers get their $20 bucks back?

Let's teach them that applying patches to only hurt Nvidia's cards score is costly, and all just to make ATI look better again...

It's like rooting for the underdog is being taken to an entire new level. I would't even want to win that way.

Shame on them!!! I shouldn't have to buy an Ageia PhysX card, when my Nvidia 280 does that for me already, and does it well!

[cadaveca]

Good luck on that one. nV modified the software to run on thier vgas, there's no basis for a lawsuit here, unless you want it to go the OTHER WAY around...

[Praz]

This thread has gone from sorry to complete stupidity.

[Talonman]

I was thinking strictly from a features standpoint, and false advertising.

They could probably avoid this if they renamed the product to 3DMarkATI. At least we would have known what we were getting... And that Nvidia cards were officially on the out's.

Mabey we can get a new benchmark made that will take into consideration our entire system speed, PhysX included. Even if it runs on a Nvidia card!!!

We need a true Gamers Benchmark, not one in bed with ATI...

[STEvil]

Good luck on that one. nV modified the software to run on thier vgas, there's no basis for a lawsuit here, unless you want it to go the OTHER WAY around...

They can get nailed for it.

[Talonman]

I hope so... Big Time!!

[cadaveca]

I'd love to see it happen. Maybe that's what they all need to shake 'em up a bit.

I could care less about the whole thing really anyway. I'm still gonna bench how I always have, using whatever drivers I think are best, and not what Futuremark thinks is best for me.

I mean really....so what if the scores are not included in the top 10...most people here ocmplaining won't ever be there anyway, so I don't understand the complaints, as it really only affects a few select people WORLDWIDE...


Lawsuit. :rofl: The bench still runs with whatever drivers you install in your system, and you can run whatever version you like. Since they are doing things filtered, you can still compare against other people running the same specs, but any scores using Phys-X on teh gpu for cpu tests are excluding from global ranking. And they havea valid arguement as to why, even had it covered when the software was published. nV is running cpu tests on thier vgas...I mean, it's a great way to compare an nV vga for phys-X vs the cpu, and really has no bearing on ATI users at all.

However when you do that comparison, you have to keep in mind that a vga can run many more threads than a cpu. In order to test how fast a thread is completed, you must run a sufficient number of threads to load the gpu properly. Vantage does this currently when you add a Phys-X card, because it's simple...one single card only means just one additional parameter to worry about.

But change that Phys-X to a nV vga...how can you compare 260 to 280, or 8800GT to 8800U? Do they run the same number of threads? Are they being fully loaded, or are they only partially working becuase they complete the data so fast, and must await cpu for further info?

Anyway, there's no lawsuit against Futuremark here...all they have done really is restored the original functionality of the ORB, and have done NOTHING to change Vantage itself. All the features are there, and they have advertized nothing falsely. Vantage works just as it did when released, and the ORB does too, now.


But Futuremark could possibly sue nVidia for modifying the software, and creating an extra bit of work, and thereby financial setback. I don't see nV losing anything financially here...so they are asking for what sort of comphensation? They lost what?


You lost what?

:stick:

[Talonman]

An accurate representation of total system speed, including PhysX.

[cadaveca]

if you think that's accurate...lol..please go ahead with your lawsuit, and make sure you got some math to back it up.


:horse:


Software liscence agreements don't work like that anyway. They, as publisher/writer, are free to update the program as they see fit.:fact: You don't own it, you merely have a right to use it.


And guess what, they didn't update anything.:poke:so what did they do exactly, that affects you?

And they didn't remove those scores, they can still be viewed, so how'd you lose what you are asking for? Where's the burden for you from not being able to publish in the top 10? I mean guys at the top, ok, but you're just whining about stuff to fit in with the top guys.

[STEvil]

Lawsuit would be for modifying their code without permission.

The catch though is that the file nVidia replaces isnt made my FM anyways so its really a moot point there. They could make the case that the DLL is acting like the "MMO Glider Bots" that Blizzard dislikes and modifying the output of the program.. which is partly true, but really if FM wanted to take legal action they'd have done it a while ago.

[Talonman]

They may not loose in a cort of law for ruling that Nvidias PhysX calculations shouldn't count, and Ageia's PhysX Card should... But in every just persons heart in the world that holds truth and honor sacred, the shame of it all will ring true.

Thanks 3DMark for selling us Nvidia guy's a benchmark, that would measure our systems PhysX calc's accurately, but then change it to not let it be reported, putting the air of cheating into the mix. Mabey they will also fix thoes bogus ATI slanted 3DMark06 scores too while their at it? Naaaa, just kidding... We both know that won't happin! Their heart just wouldn't be in it. ;)

Good for Agea!! Their cards still count and will be reported as valid numbers.

Mabey after ATI get's their cards in the game, PhysX calculations will be fair game to measure again, even on a Nvidia made card too?
We can only hope...

[STEvil]

I cant disagree that the situation seems to have been handled poorly, however when you've been around a while you learn to sit back and let things play out some.

If I ran futuremark i'd release a patch that ran the physics benchmark portion once for each accelerator in the system (1x CPU, 1x Physics card, 1x GPU) and gave seperate scores for each. Maybe they're doing that and just not talking (making them look bad) or maybe they're just clamming up (making them look bad). There's a win situation, but unfortunately Futuremark (and most game dev's) are poorly connected to the enthusiast community so all we can do sit back and race our PC chairs while whining to each other about how they dont do this or that or the other thing...

To test physics + gaming performance you would have to also give more numbers.


Realistically Vantage is a synthetic performance measurement tool because you cant measure the effect on performance of enabling options (other than AA/AF that is).

[Lestat]

i get over 16k with my 8800 GTX and my extra special Q9650
there's no nvidia physX support for the 8800 GTX yet, so ..
if they saw that score would they immediately throw it out ?

i had a physX card installed with my Q6600 and was only getting 15k points....

[CraptacularOne]

Don't forget your Agea Physx card helping the score. :rolleyes:

[Talonman]

Is this going to be adding more performance to Nvidia GPU's that 3DMark Vantage is going to miss out on measureing too?

http://www.nordichardware.com/news,7979.html

"NVIDIA has something big coming, according to its own estimates something about the same size as SLI when it launched. It's not the introduction of the rumored Radeon HD 4870X2 killer, the 55nm G200b core. That would be a bit too weak. There is a new PhysX driver that will enable PhysX on all CUDA-capable GPUs that will launch soon, but it should arrive sooner than that. Instead, the bang comes from updated CUDA support and improved drivers.

SLI will be updated, connectivity will be updated, CUDA looks like it will go commercial, and better quality and performance overall. The last part may also include improved scaling in various SLI configurations. Whether it will have the same impact as SLI had remains to be seen. More news should follow as we approach the next Big Bang".

Can ATI cards also not run CUDA?

Hoping for SLI on intel mobo's too...
I would jump on a second 280 for sure. :)

[CraptacularOne]

The new X58 chipsets that support Nehalem will be able to use SLI similarly to how the Skull Trail board did. You will likely never see SLI support on any current gen board, so if you want an Intel board and SLI, you will have to swap motherboards anyway.

[Talonman]

Mabey, but there won't be enouigh X58 boards out to call this a Bang...

They came so close with the GX2 by putting an SLI chip on the GPU.

I wonder if a magic driver might do the rest? One 280 with SLI chip on board, and my current 280 with Big Bang driver installed.

Or better yet, a new SLI bridge with SLI chip on board. Keep the 280's as is.

A guy can dream... ;)

I still love my Maximus...

It's probably just driver related:
http://www.tweaktown.com/news/9858/nvidia_s_next_big_bang_coming_soonindex.html

I bet it's cuda about to flip the graphics processing world upside down.

I wonder how much performance a new CUDA driver could add to a single 280.. :p:
There might be an entire new high gear on my GPU yet to find? Shift baby!!

It would be fun to have it measured, if it will indeed make games run faster.

[Talonman]

What are the odds that the 3DMark boys got the word about CUDA being released later this year, and new it would be a Big Bang of a performance increase, and wanted to make sure to not measure it...

I bet not measureing PhysX calculations on Nvidia cards was just the tip of the iceburg...
CUDA on Nvidia cards is probably out too?

Will Agea's Physx cards be able to run CUDA? Im sure if they can, Vantage will count their calculations as valid.

Is CUDA a Physx feature, or a totally seperate animal?

UPDATE: I think ATI cards can run CUDA:
http://www.tomshardware.com/reviews/nvidia-cuda-gpu,1954-6.html
"But, Brook’s critical success was enough to attract the attention of ATI and Nvidiahttp://en.wikipedia.org/wiki/Nvidia , since the two giants saw the incipient interest in this type of initiative as an opportunity to broaden their market even more by reaching a new sector that had so far been indifferent to their graphics achievements".

I think ATI might be in the game with CUDA?

Looks like CUDA does not support SLI if this is still accurate.
"If you want to use several GPUs for a CUDA application, you’ll have to disable SLI mode first, or only a single GPU will be visible to CUDA".

[Marios]

http://www.fudzilla.com/index.php?option=com_content&task=view&id=8600&Itemid=35
How about Havendales? Do they count as CPU, GPU or what?

[Talonman]

A truly excellent question!!

Using FM logic, if CPU calculations being done on a GPU is cheating, doing GPU calculations on a CPU must shurly be a clean foul too. :yepp::cool::p:

[NaMcO]

I don't really care about 3DMark, but if the calculations can be made (and are in real life) by the GPU, then by all means why not? This is a step towards the past if you ask me.

[Bellisimo]

i guess it comes down to personal preference,

ive never really cared too much about orb and i will just bench whatever gives me the BEST scores, in this case the physx driver is very innovative for that gpu test and I will be using it ;)

too bad it's a CPU test