• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

PhysX 64-bit driver in September

SlamDunk

Limp Gawd
Joined
Jul 30, 2004
Messages
277
Just spotted this little info bit from the PhysX driver page:


AGEIA is committed to providing 64-bit driver support for PhysX. Currently AGEIA's software development team is planning a September 2006 release for this driver package
 
Not the point I think. A 64 OS like win64 or Vista64 requiers even to emulate win32 games on it a 64bit driver. So if for Gamers with a 64 rig with a 64 OS. playin win32 games. in WOW mode. To just be able to make the win32 PhysX game to be able to use the Ageia card. If it's posible.
And it comes in handy with al those AMD64/EMT64 CPU avaible as a target platform. And Vista release somewhere in the near future.
It could be that future vista games comes more in 64bit variant to.
 
“Why bother with this? Are games even going to run 64bit in the near future?“
Some of us run a 64-bit OS as our main OS. For months I only had windows 64-bit installed. But once I got a PhysX card I had to dual boot to 32-bit if I wanted to use it, once the new drivers come out I can scarp my 32-bit OS and free up lots of hard disk space.
 
AtomicMoose said:
Why bother with this? Are games even going to run 64bit in the near future?

You bitch about the do make 64-bit driver.
Others bitch in the forum because they have no 64-bit driver...
I guess AGEIA is damned it they don't and damned if they do :rolleyes:

Terra - An old fashioned loose-loose situation...
 
SlamDunk said:
Just spotted this little info bit from the PhysX driver page:

I guess we know when PhysX driver version 2.5 is comming then ;)

Terra - In september ;) *L*
 
robberbaron said:
Guess they better tighten their product then if they want to be taken seriously.

There is nothing wrong with the product, don't put words into my mouth.
They got a hell of lot more to sho, than GPU-physics...

Terra - Point the finger at the PR (viral anti-Ageia marketing)...not the product...
 
Terra said:
There is nothing wrong with the product, don't put words into my mouth.
They got a hell of lot more to sho, than GPU-physics...

Terra - Point the finger at the PR (viral anti-Ageia marketing)...not the product...

I guess the joke went over your head. You lose..

Terra - I forgot that I can just put "Terra" in the nice vBulletin signature box I'm provided with rather than signing every single post with an unwitty, summarizing comment.
 
robberbaron said:
I guess the joke went over your head. You lose..

Terra - I forgot that I can just put "Terra" in the nice vBulletin signature box I'm provided with rather than signing every single post with an unwitty, summarizing comment.

Nice...next time you wanna joke...use smileys...that's why they are there...

Terra - Play me again and I will report the post :p ;) .oO(Get it now?)
 
It makes pretty good sense.. making the transition to 64 bit drivers so everything can move over smoother to 64 bit once the 64 bit OS's become a standard, and games/apps are made in 64 bit. I don't see anything wrong with releasing a 64 bit driver. :confused:
 
I don't know a whole lot about the nitty gritty of IC processing and words etc.. Am I correct in thinking that a 64 bit driver would, on some level, allow for faster communication between the CPU and the PPU? That is to say, you could send 'chunks' 64-bits long instead of 32, ultimately saving cycles, right?
 
vidor said:
I don't know a whole lot about the nitty gritty of IC processing and words etc.. Am I correct in thinking that a 64 bit driver would, on some level, allow for faster communication between the CPU and the PPU? That is to say, you could send 'chunks' 64-bits long instead of 32, ultimately saving cycles, right?

It might, if there are any numbers that are too large to fit into a normal 'long'. My suspicion is that the whole point of the PPU is to unload the CPU from doing any physics calculations, hence 64-bitness is a moot point. The PPU is most likely a 32-bit processor, and additionally, since 'long's are expensive to calculate, it would shy away from using them. Additionally, games AFAIK wouldn't require that kind of accuracy, anyways. (Too much processing power required for very little improvement.)

In terms of the communication between the PPU and CPU, speed would be at the mercy of the PCI bus at this point. The PCI bus is very old and can't be accelerated much further, given its 133MB/sec limitation using 32-bits @ 33MHz.
 
Josh_B said:
It might, if there are any numbers that are too large to fit into a normal 'long'. My suspicion is that the whole point of the PPU is to unload the CPU from doing any physics calculations, hence 64-bitness is a moot point. The PPU is most likely a 32-bit processor, and additionally, since 'long's are expensive to calculate, it would shy away from using them. Additionally, games AFAIK wouldn't require that kind of accuracy, anyways. (Too much processing power required for very little improvement.)

In terms of the communication between the PPU and CPU, speed would be at the mercy of the PCI bus at this point. The PCI bus is very old and can't be accelerated much further, given its 133MB/sec limitation using 32-bits @ 33MHz.

Please list the data that need to be transfered between the PPU and CPU? :)

Terra - Will that stupid "PCI-bus-myth" not die?!
 
vidor said:
I don't know a whole lot about the nitty gritty of IC processing and words etc.. Am I correct in thinking that a 64 bit driver would, on some level, allow for faster communication between the CPU and the PPU? That is to say, you could send 'chunks' 64-bits long instead of 32, ultimately saving cycles, right?
Now its no easy stuf.

First a 64bit integer are proscest in one stroke. Where on IA32 it must be done in two with the overhead. Altho it looks like one line in C++
_int64 someint= 10000;
someint=someint*someint;

int is 32bit in AMD64 and EMT64 and IA32.
But I bet the machine code generated for IA32 is much longer and what overhead Compared to the 64bit version.

But pointers are twice as big altho only 48bits are actual used for Mem addressing.
So fooling around very much with pointers make it slower under AMD64/EMT64 compared to 32bit pointers of IA32.

So depends how PPU is used. But the performance more depends on CPU performance where the PPU is set up in 64bit fashion CPU computing game logic for loading the PPU and when the data is retrurnd for further 64bit CPU computing on the results. To make the final frame to compute for the GPU.

Yes I do wonder if the PPU goes faster with 64bit. no idea.
 
SuperGee said:
Now its no easy stuf.

First a 64bit integer are proscest in one stroke. Where on IA32 it must be done in two with the overhead. Altho it looks like one line in C++
_int64 someint= 10000;
someint=someint*someint;

int is 32bit in AMD64 and EMT64 and IA32.
But I bet the machine code generated for IA32 is much longer and what overhead Compared to the 64bit version.

But pointers are twice as big altho only 48bits are actual used for Mem addressing.
So fooling around very much with pointers make it slower under AMD64/EMT64 compared to 32bit pointers of IA32.

So depends how PPU is used. But the performance more depends on CPU performance where the PPU is set up in 64bit fashion CPU computing game logic for loading the PPU and when the data is retrurnd for further 64bit CPU computing on the results. To make the final frame to compute for the GPU.

Yes I do wonder if the PPU goes faster with 64bit. no idea.

I think the difference will be negligeable, at best. :)
 
Terra said:
Please list the data that need to be transfered between the PPU and CPU? :)

Despite the PPU being a discreet processor, and most likely allowing bus mastering, it would still have to interact with the CPU to complete a frame, n'est ce pas?

(I'm only suggesting... no flames req'd, please.)

Also, any links to a beta version, even? I want to buy a PPU but I am running XP x64!
 
Terra said:
Please list the data that need to be transfered between the PPU and CPU? :)

Terra - Will that stupid "PCI-bus-myth" not die?!

well, I can't imagine that the PPU just magically generates data out of nowhere...so it's gotta go over the pci bus somewhere along the way ;)
 
Terra said:
Please list the data that need to be transfered between the PPU and CPU? :)
Vertex, and/or perhaps transformation data (although that would be bizarre imo) must be sent to memory and subsequently the cpu.

This will account for most of the bandwidth used, but there will also be some overhead in whatever protocol is being used.
 
There will be a lot of static physx data stored localy in the PPU Mem.
because that 128MB is not for show alone.
So updates to that static data and PhysX event per frame are send over the PCI bus each frame. Because PhysX is very event driven. The load is very spiky. So not average FPS count but the Dips where a heavy PhysX event occurse like a explosion with a lot of interacting objects nearby.
Cloth hair will give more a continues load if in vision..
Where the event driven Physx spikes on top of that. The sum of all PhysX features used.
 
Dillusion said:
So any new news on this driver?
Yeah, really.

I want teh driver now. Without it, I have no reason to get a card. I want to play with it NOW! :cool:

NO COMMENTS ABOUT PPU USEFULNESS PLEASE.

If you don't like PPUs, you can keep your comments to yourself.
 
64-bit WinXP drivers are out: http://www.ageia.com/drivers/drivers.html


Release Notes:
Release Date: 8-September-2006
OS Support: Windows XP (32 & 64-bit), Windows Vista RC-1
Hardware Support: AGEIA PPU (All)

AGEIA PhysX Engines Included With This Release:
2.5.1, 2.5.0, 2.4.4, 2.4.1, 2.4.0, 2.3.3, 2.3.2, 2.3.1

Changes In 2.5.1:

* Microsoft Windows XP 64-bit now supported.
* Microsoft Windows Vista 32-bit runtime now supported.
* SDK Enhancement: Improved mesh cooking
* SDK Enhancement: Optimized narrow phase collision detection
* SDK Enhancement: Optimized broad phase collision detection
* SDK Enhancement: Optimized dynamics solver

Download Full Release Notes (PDF)
 
Back
Top