Multi-threaded Physx. Will this end current Physx?

[EastCoasthandle]

Pure PhysxInfo.com Fluidmark 1.2 will be released soon showing how well Phsyx works when multithreading optimizations are enabled. Results are below:



From the graph a Q9400 is able to out pace a GTX275 using the multithreaded optimizations for Physx. However when 2 GPUs are used they out pace both. However, when mult-core physx is off the results are far less. Currently it's is believed that the current version of physx doesn't use this. This only adds to the speculation that the current version of Physx gimps CPUs.

[Solaris17]

how does it not already use multithreaded physx if it uses a graphics cards shaders which are like hundreds of mini cores? I think they mean that software physx isnt muli threaded and only uses one CPU core. Which i would believe. But I do not buy those performance numbers. maybe it does scale and increase dramatically if its enabled for CPU's. but i seriously doubt they could outpace a grfx card. Unless that is some kind of poisoned graph were they attempt to run software and not hardware physx on a GPU?

[digibucc]

well the score is giving way to much of an increase solely based on PHYSx capability. no games use physx to the extent that that much more power would really do anything. yet.hopefully, never.

physx is a horrible money ploy. i wish nvidia would just go standard with physics, so no more games would require it to make leaves float or something simple.

[EastCoasthandle]

The multi-threaded optimizations didn't come from nv. It's coming from Fluidmark, the maker of Fluidmark 1.2 (soon to be released).

The people at Fluidmark are claiming that their version of Physx will utilize all 4 cores (multithreaded optimizations). In the example below 100%.

[Fitseries3]

cant wait to see how this does on my machine once i get it runnin.

[DrPepper]

Originally Posted by Fitseries3

cant wait to see how this does on my machine once i get it runnin.

Holy f'n shit. I want to see that too.

[AhokZYashA]

i want to see that...
must be fckin awesome numbers.

[DirectorC]

Will this end current Physx?

No?

[dr emulator (madmax)]

well i'm seriously considering buying something like a 3850 pci with my x58 ud7 and then seeing what problems people run into with this card,
why ?
well i keep seeing niggly problems with the 5870 that need sorting and the hd3850 isn't that bad of a card, it runs well enough on this old gal
(ye ye i know i only have agp for now)and the i7 will give me the core processing power i need so i wait with baited breath

[dr emulator (madmax)]

no

[JATownes]

Originally Posted by Fitseries3

is that a ___________ reference?

http://en.wiktionary.org/wiki/niggly

To clear up any confusion...I don't think Dr. would say something like that.

[Fitseries3]

just makin sure.



not that i'd get upset about it if it was.

[bobzilla2009]

current physX never even took off you can't end what didn't begin.

[EastCoasthandle]

Originally Posted by DirectorC

No?

Take a look at the CPU results. That's why I asked the question. If you examine it, the CPU results are higher the the non multithreaded GPU results.

[GotNoRice]

You know, I was lucky enough to have the opportunity to work with Ageia back in 2006 before the original PPU was even released and what was interesting is that back then the PhysX engine was very multi-threaded. I recall running a demo and having it easily max out all 4 virtual cores of the Dual-Xeon rig I was running at the time (dual processors each with HT). And when I got my Q6600 in 2007 it could max out all of those cores as well. The PhysX engine was designed from the very beginning to run well on the CPU.

It was only after Nvidia bought Ageia that the PhysX engine magically lost it's ability to take advantage of more than one core

[KainXS]

well of course, gotta make money some how you know.

if a C2Q can do that, I can't even guess what a I7 would do.

[EastCoasthandle]

Very interesting post from GotNoRice indeed.

[JATownes]

Interesting read regarding NVidia removing multi-threading from Phsyx:

“Our PhysX SDK API is designed such that thread control is done explicitly by the application developer, not by the SDK functions themselves. One of the best examples is 3DMarkVantage, which can use 12 threads while running in software-only PhysX. This can easily be tested by anyone with a multi-core CPU system and a PhysX-capable GeForce GPU. This level of multi-core support and programming methodology has not changed since day one. And to anticipate another ridiculous claim, it would be nonsense to say we ‘tuned’ PhysX multi-core support for this case.”


http://news.softpedia.com/news/NVIDI...X-132681.shtml

Thought you all might be interested.

[EastCoasthandle]

Which game using physx allow for over 90% consistent CPU usage on all 4 cores? Notice I didn't ask for 100%.

[RejZoR]

So, from the chart, C2Q 9400 is faster than GTX 275 when it comes to hardware PhysX capabilities.
No wonder they are blocking multithreading in games, imagine what would a Core i7 920 and above do to the poor NVIDIA... but in the end, it's a lame and sad attempt to cripple anything non NVIDIA.
Sigh...

[Zubasa]

Originally Posted by KainXS

well of course, gotta make money some how you know.

if a C2Q can do that, I can't even guess what a I7 would do.

Even worst, anything that can run on a C2Q might as well also run on a $100 Athlon II X4.
The Athlon II might need higher clocks, but you get the idea.

[Mussels]

i should add this to my other PhysX thread...

[Wile E]

Originally Posted by EastCoasthandle

Which game using physx allow for over 90% consistent CPU usage on all 4 cores? Notice I didn't ask for 100%.

Tell the game devs to stop being lazy.

And these graphs prove not a damn thing. How do we know multi-threaded Fluidmark was optimized properly to take advantage of the cpu plus a single card? Could be a simple programmer's mistake.

Short version = Need more data.

[shevanel]

can the fluidmark guys legally do this?

how would this affect real world use if the driver isnt written to allow mulithreaded physx?

I wouldnt doubt if nvidia botches the cpu from being able to use more than one core.

[Mussels]

Originally Posted by shevanel

can the fluidmark guys legally do this?

how would this affect real world use if the driver isnt written to allow mulithreaded physx?

I wouldnt doubt if nvidia botches the cpu from being able to use more than one core.

itsp ossible they picked a task physX can accelrate, and then just wrote their own multithreaded code to do that task - not using physX itself, just doing the same task on a CPU.