Is PhysX worth the extra?

[mercyground]

http://www.fudzilla.com/content/view/17525/1/

"Thanks to the dedicated Geforce 9800GT, Radeon HD 5870 managed to score
over 100fps even with PhysX effects set at high, and impressively enough, not even the GTX 285-9800GT combination managed to beat this."

So "support reasons" or just the fact that once you strip physx off NV... ATI still kicks ass?

/me shrugs.

Open Physics. Do it on CPU or GPU but open standard or there will be no proper usage of physics in games for a good while. It will just be added "fluff".

Fermi had better be as good as NV are PRing. Otherwise i'd HATE to be an investor.

[scaryjim]

... once you strip physx off NV... ATI still kicks ass?

Potentially that's a slightly unfair test though, since it's pitting different generation cards against each other. The 5870 *should* be much faster than a GTX285 if the rest of the subsystem is equivalent.

<edit>ignore that rubbish about MSAA, they're tesing Batman so no AA on ATI cards *doh*</edit>

[Pob255]

HardOCP proved that you get better performance in software than on cards. Even on Batman ATI cards did better than NV's when driver hacked.

Ummm Hacked drivers, means they are not running it on software (ie emulating it on the cpu) they are using hacked drivers to run physX on the graphics card rather than the cpu.

Because of the nature of the gpu it can run physics calculation processes better than the cpu.

[Georgy291]

would i be breaking the rules if i asked how these hacked drivers actually worked?

[badass]

Potentially that's a slightly unfair test though, since it's pitting different generation cards against each other. The 5870 *should* be much faster than a GTX285 if the rest of the subsystem is equivalent.

This is the problem with getting too caught up in the details.

What generation the cards are is completely irrelevant. Who cares if it's unfair on Nvidia?
The GTS 285 costs between the cost af an HD 5850 and an HD 5870 yet is slower or at best similar in performance to the HD 5850 with worse energy consumption. Get yourself a Geforce 9600GT and the 5850 costs the same as the GTX285, yet it can drive 5 monitors at once and is still faster.

Nvidia's top card needs to lose £70 off it's price to be competitive.

[KowShak]

would i be breaking the rules if i asked how these hacked drivers actually worked?

physX can use a "spare" nVidia GPU in the system to do the physics work and you need a second graphics card to do the graphics. The physX drivers had a limitation added to them so that if the second card was not an nVidia card they would refuse to work, i.e. buy two nVidia cards or you get no hardware physics. The hacked drivers were modified so they'd allow the physX to be accelarated on an nVidia card when the second card (the one doing graphics) was a non-nVidia card. There was no technical reason as far as I can see for restricting the physX to systems with two nVidia cards that was purely a marketing decision. Do you remember how you "needed" an nVidia motherboard chipset for SLI to work, where as later on Intel chipsets became SLI certified too?

[Pob255]

Not quite, you can run PhysX off a single card, it uses some of the steams to do it.
How much of an impact this will have depends on the gpu, how many streams it has and how many can be spared.
Which was another part of the thinking behind this as a couple of years back when this all got started gpu's were under used, much like quad core cpu's, the games where just not using all the cores and where more dependent on shear speed and memory bandwidth.

I know there where hacked drivers floating about the let you run physX on an ati card no need for any nvidia in a pc.
Don't know if that still works though.