3DMark Vantage benchmark

[tzetsin]

ORB - Compare This is the system i just put together, completely stock, no overclock at all, on anything.

[cadaveca]

You get an inflated score due to nV Phys-x driver.

[tzetsin]

its not exactly 'inflated', it makes a huge difference in the cpu physics tests. i've never tested this cpu without that driver, but my q9450 @ 3.0ghz only gets a little over 33 ops in the physics tests where the e8400 gets over 100 at 3.0ghz. so you cant exactly say its inflated because the driver actually makes a huge difference, saying its inflated sounds like its cheating the test somehow.

[cadaveca]

Considering Furturemark does not allow scores using such drivers into the Hall of Fame, I'd almost call it cheating too. I hear what you are saying, beleive me, but the fact of the matter is that if Futuremark doesn't consider it valid, neither do I.

The problem comes in when trying to compare scores...even though I have a real Phys-X PPU, 2x3870x2's, and QX9650, you'd think I'd have a hope of getting close to, if not better than, your score. In fact, if those drivers worked on my cards, I'd definately be faster.

Here's my score, posted on the last page...notice my AI score is almost 4x yours...

ORB - Compare

But because your Phys-X tests are run on GPU, and not cpu, we cannot directly compare scores, and hence my calling your scores "inflated".

Here's what Futuremark says about your score:

Please note that this result does not comply with the rules defined in the Driver Approval Policy for 3DMark Vantage and will not be visible on default result search on the ORB or on the Hall of Fame. Please ensure that you are using approved display drivers and system software versions provided with 3DMark Vantage or use the "Disable PPU" option.

And here's what they say about mine:

Please note that this result has been produced using display drivers which have not been validated or approved by Futuremark and will not be visible on default result search on the ORB or on the Hall of Fame.

Note that they do not say my drivers break the rules, but that yours do. It's very clear.

[tzetsin]

hrm, i can see what you mean, but i was under the impression that the physx driver doesnt support the gx2 am i wrong? i installed it when i was grabbing the latest beta for the gx2, just saw it there grabbed it, installed it, but it said i wasnt supported during install, then i saw the score and now i wonder. i agree that it must be because of that driver, but im not sure how?

[cadaveca]

nVidia decided that they would not comply to the programming manual for Vantage(since it's release this guide has said cpu tests must be run on cpu or PPU, not on GPU, in order for scores to be comparable across different systems), and all drivers now feature Phys-X support for every 9-series and newer card; I do beleive that even the 8800 series can run the same test the same way now too.

I don't mean to single you out, but it's a big issue in the overclocking world where this test has been held back from global ranking @ HWBOT due to this and other issues. It's not anything you've done, but whayt nV does with thier driver.

If you install Phys-X driver, you can disable geforce accelleration for Phys-X, or you can do as Futuremark suggests, and "disable PPU" in vantage.

[tzetsin]

i can see how this could be a big issue, i mean, whats the difference really if i have a ppu or a gpu or the cpu do the job? just to get this straight, your allowed a ppu in vantage, and the ppu is allowed to effect your score? but the gpu isnt?

[cadaveca]

It works this way because PPU cannot draw graphics, merely does calculations that a cpu does(this seperates what a PPU and gpu do), and in that specific test, they have increaased teh workload(from 8 threads to 16), so as to make the score comparable. However, the gpu is capable of many more threads, and the test is not programmed for it...basically during that test with those drivers, it is doing what nV thinks it should be doing, and not Futuremark. So furtureamrk has banned all such results using those drivers from the Hall of Fame and default score searches(you must look for the scores; noone will ever see your score by jsut browsing scores).

Now, my result is also invalid, but because I am using a driver that Futuremark has yet to approve. Every driver does this the first few days after release. But your driver changes the way the test is run...mine does not.

PPU increases cpu score, as it only can do cpu calculations...900-1200 points due to the increased cpu resources(PPU). Your driver increases cpu score...but you have not added cpu resources...and it increases scores by THOUSANDS...

[tzetsin]

hrm, your right then, if the debate is over weather we compare apples and apples then futuremark is correct in its position, however for me personally, (and this is just me personally, dont bash me for it pls, cause its not in the same spirit of fairness that other people may use vantage for) i use futuremark benches to see if my computer is faster than yours, and it says it is (im just using you as an example, i actually just want to see where it rates in the grand scheme) i really dont care how it does it, my pc is more capable than yours is, and since you have the same opertunity to buy the same equiptment that will to the same tasks, i dont see why its a problem. So, if you want to compare apples to apples, then i have no argument, your perfectly right, but for those who compare to see if theirs is better than whoever's, then the argument will naturally be quit opposite.

so i guess my position would be that my pc is capable of 18000 marks, and it is, because it is doing exactly what it takes to get that score, but because of "how" its getting it, its not valid to futuremark for comparison purposis, but it doesnt change the fact that a computer with nv physx IS doing a better job of the tests. so i guess its this standpoint that has caused the debate eh?

[cadaveca]

Yeah, and in the end my pc is much faster in all aspects than your is...even though you have a higher Vantage score. This is why I called your result inflated...your E8400@ 3ghz, or my QX9650@ 4ghz...your two gpus VS my 4...

And when it comes to games with Phys-X, the only thing that yours might be faster in...you'd still be slower because you'd be using gpu resources for Phys-X, and on my system all gpu resources are available for graphics...Phys-X would run on my PPU...unlike yours.

So, in the end, the problem is that even though you got a higher score, your system is in no way faster than mine in any application, and as such, Vantage has become a "broken" benchmark, incapable of providing the measurement you are looking for.