My Test of GRAW running with and without physX

Fulcrum · 8 May 2006 at 01:26

megatron said:
The last physic based effect I heard of was ragdoll effects (in the ut2004 engine) and thats part of 3dmark03, a graphics power test.

That loads the cpu I think, they had to scale back the physics in CSS due to netcode lag. The demo for css had full physics (like the HL2 game at least) but it also rained dead bodies from the sky when no one had died and other weird stuff

megatron · 8 May 2006 at 02:40

Fulcrum said:
That loads the cpu I think, they had to scale back the physics in CSS due to netcode lag. The demo for css had full physics (like the HL2 game at least) but it also rained dead bodies from the sky when no one had died and other weird stuff

I would have said physics is a client side effect, that is given certain variables everyones PC will calculate the same effect - hence it wouldnt need to be in the netcode.

Fulcrum · 8 May 2006 at 02:44

Hrm yea but when one player causes physics that may or may not affect another player then I guess that becomes the same as bullets fired, etc

I think it still works like that, just simplified to make everything bouncy which is irratating. Its just something I read on steam forums really.

http://forums.steampowered.com/foru...id=431319&highlight=counterstrike+and+physics

SteveOBHave · 8 May 2006 at 07:41

Not that I am trying to discredit the tests that you have been running, but we have to continue to remember a couple of things:

- GRAW is seemingly a very poor implementation of PhysX and as such should be taken as a 'grain of salt' test bed. Benchmarking on such a test bed is unfair on Ageia.

- Ageia cannot either fairly, and I suspect, legally slate Ubisoft for their PhysX implementation and as a matter of course have to provide moral support to them or be seen as being hypocritical after slapping the PhysX logo on the game box.

IMHO I believe that Ageia jumped the gun, and put themselves directly in the line of fire from both intelligent gaming consumers and their competitors without substantial evidence for their 'revolutionary' product. I would write the current PhysX iteration as being a marketing and management failure rather than a failure of the PhysX hardware and development team.

From where I am standing, they have allowed themselves to get in between a rock and a hard place, and the question that remains is - can they survive the marketing failure that is currently PhysX until the gaming monster UT2007 comes out and 'saves' them

VeNT · 8 May 2006 at 11:11

this is sillyness
Wasn't this slated to be like the first 3dfx card?
taking the work AWAY from the CPU rather than making MORE work for it?
I'd rather have seen it take all the physics calcs away from the CPU/GPU(don't think any go on there) and have nothing EXTRA on screen rather than have so much crap on there that it slows my PC down to a crawl!
whats the point?

SteveOBHave · 8 May 2006 at 11:28

VeNT said:
this is sillyness
Wasn't this slated to be like the first 3dfx card?
taking the work AWAY from the CPU rather than making MORE work for it?
I'd rather have seen it take all the physics calcs away from the CPU/GPU(don't think any go on there) and have nothing EXTRA on screen rather than have so much crap on there that it slows my PC down to a crawl!
whats the point?

As above - it's the implimentation of the card by the games developers, not the card itself. A game needs to be told that the card is there, it's not going to find it by itself.

And I think you mean 'Slated to be the first PPU card'.

Dutch Guy · 8 May 2006 at 11:40

I would not expect the framerate to drop, I would have expected the opposite actually, I thought the PhysX card to take over stuff the CPU usually does so that the CPU can do other stuff.

If the framerate drops, even when using a PhysX card, why bother with it, what is it supposed to do anyway, surely not just draw a few square blocks on the screen that a CPU can do as well?

HEADRAT · 8 May 2006 at 11:46

I think the problem is that the PPU might be doing all the physics stuff (e.g. interactions between many differrent objects) but the GFX card still has to render all the stuff flying about on screen hence the FPS drop. The actual "physics" calculation are being done by the PPU but obviously that doesn't help the GPU render them on the screen

HEADRAT

Darren555 · 8 May 2006 at 11:57

[H]ardOCP have a new Ageia PhysX article - no surprises with their conclusion.

SteveOBHave · 8 May 2006 at 12:09

Darren555 said:
[H]ardOCP have a new Ageia PhysX article - no surprises with their conclusion.

Thanks for that - that was a really well balanced review and has a brilliant message at the end that should be taken notice of...

http://enthusiast.hardocp.com/article.html?art=MTA1MiwxLCxoZW50aHVzaWFzdA==

Dutch Guy · 8 May 2006 at 12:31

HEADRAT said:
I think the problem is that the PPU might be doing all the physics stuff (e.g. interactions between many differrent objects) but the GFX card still has to render all the stuff flying about on screen hence the FPS drop. The actual "physics" calculation are being done by the PPU but obviously that doesn't help the GPU render them on the screen

HEADRAT

Yes, I understand that, but the objects are hardly graphically intensive, the thing that might slow it down is physics but that is hadled by the PhysX card.

ihatelag · 8 May 2006 at 12:38

Any positive results from any reviews so far?

SteveOBHave · 8 May 2006 at 12:41

ihatelag said:
Any positive results from any reviews so far?

Seemingly not, however there is the small problem of not having anything reliable to benchmark the hardware

LOL

ihatelag · 8 May 2006 at 12:54

SteveOBHave said:
Seemingly not, however there is the small problem of not having anything reliable to benchmark the hardware LOL

Wasn't there anything that came with the product to do benchmarking with? Like a program to showcase the results...sort of like the ATI's monkey demo or the NVIDIA's Mad Mod Mike demo.

SteveOBHave · 8 May 2006 at 13:07

ihatelag said:
Wasn't there anything that came with the product to do benchmarking with? Like a program to showcase the results...sort of like the ATI's monkey demo or the NVIDIA's Mad Mod Mike demo.

Not that I am aware of. I'm not sure I'd trust a benchmarking application developed by the company that makes the hardware itself

What I mean is that there has been no reliable game created that utilises the hardware the way Ageia intended, and therefore we're unable to form an educated opinion on the harware itself.

ihatelag · 8 May 2006 at 13:15

SteveOBHave said:
Not that I am aware of. I'm not sure I'd trust a benchmarking application developed by the company that makes the hardware itself

What I mean is that there has been no reliable game created that utilises the hardware the way Ageia intended, and therefore we're unable to form an educated opinion on the harware itself.

Yeah I have to agree...shame about the Cellfactor demo + PhysX ...another game with the FPS hit. http://forums.overclockers.co.uk/showthread.php?t=17572420 In the first picture, looking on the right hand list, the FPS is 10...which is not that great...I just hope UT2007 + PhysX actually delivers something good! Otherwise the developers of HavocFX will not have much of a challenge or motivation to deliver something really good :s

[ui]ICEMAN · 8 May 2006 at 16:27

Ok been doing a lot of work on Cell Factor with different driver revisions etc. The beta AGEIA drivers give me huge issues with stability and actually degrade my performance in Cell Factor.

Using the 84.37 TweaksRus drivers and the latest non beta drivers for the PhysX card, I can finally attain good fps in Cell Factor on high graphics mode. What does cripple it however is turning on per pixel motion blur, with this option on, 10-15fps is all I can achieve on Quad and normal SLI. By "good" fps, i'm not talking 50+ but for the most part its 35-45 dipping occasionally to 25 during very large effects.

SteveOBHave · 8 May 2006 at 16:38

[ui]ICEMAN said:
Ok been doing a lot of work on Cell Factor with different driver revisions etc. The beta AGEIA drivers give me huge issues with stability and actually degrade my performance in Cell Factor.

Using the 84.37 TweaksRus drivers and the latest non beta drivers for the PhysX card, I can finally attain good fps in Cell Factor on high graphics mode. What does cripple it however is turning on per pixel motion blur, with this option on, 10-15fps is all I can achieve on Quad and normal SLI. By "good" fps, i'm not talking 50+ but for the most part its 35-45 dipping occasionally to 25 during very large effects.

That's quite frightening, and in Quad SLI too... Out of my price range, and no game has been good enough for me to justify such a large upgrade! A games producer would be mad to release a game based soley on such a young piece if hardware... IMHO

ihatelag · 8 May 2006 at 16:39

Thanks for the info Iceman, I guess FPS problems are here to stay for a while. The only good thing from the test is that you can get 35 - 45 fps..which is okay....but not optimum for gaming.

megatron · 8 May 2006 at 16:45

[ui]ICEMAN said:
Ok been doing a lot of work on Cell Factor with different driver revisions etc. The beta AGEIA drivers give me huge issues with stability and actually degrade my performance in Cell Factor.

Using the 84.37 TweaksRus drivers and the latest non beta drivers for the PhysX card, I can finally attain good fps in Cell Factor on high graphics mode. What does cripple it however is turning on per pixel motion blur, with this option on, 10-15fps is all I can achieve on Quad and normal SLI. By "good" fps, i'm not talking 50+ but for the most part its 35-45 dipping occasionally to 25 during very large effects.

You have an Aston Martin DB7 (hopefully with working ejector seat) and u still find time to play pc games? :eek: