During Nvidia's fourth-quarter financial results conference call, Nvidia shed a little more light on its acquisition of Ageia and what it plans to do with the firm's PhysX technology. CEO Jen-Hsun Huang revealed that Nvidia's strategy is to take the PhysX engine and port it onto CUDA (Compute Unified Device Architecture), a C-like application programming interface Nvidia developed to let programmers write general-purpose applications that can run on GPUs. All of Nvidia's existing GeForce 8 graphics processors already support CUDA, and Huang confirmed that the cards will be able to run PhysX.

"We're working toward the physics-engine-to-CUDA port as we speak. And we intend to throw a lot of resources at it. You know, I wouldn't be surprised if it helps our GPU sales even in advance of [the port's completion]. The reason is, [it's] just gonna be a software download. Every single GPU that is CUDA-enabled will be able to run the physics engine when it comes. . . . Every one of our GeForce 8-series GPUs runs CUDA."



There are 45 additional comments
Advertisement
Quote this comment Reply to this comment #1 Posted by Xerxes on 16 Feb 2008 - 02:16
That sounds cool, I'm guessing this will only be useful on a 8800 series GPU though. My 8600GTS struggles enough as it is just doing the graphics, doing physics as well would bring it to a crawl!
(10 replies) Quote this comment Reply to this comment #2 Posted by HoochieMamma on 16 Feb 2008 - 02:25
I think my 8800GTX just smiled.
Quote this comment #2.1 Posted by CrashGordon on 16 Feb 2008 - 03:45
So did my G92 GTS
Quote this comment #2.2 Posted by +.Vamp on 16 Feb 2008 - 12:55
So did my 8800GT OC
Quote this comment #2.3 Posted by Tikitiki on 16 Feb 2008 - 15:29
Yep 8800 GTS here
Quote this comment #2.4 Posted by ZombieFly on 16 Feb 2008 - 16:50
mine too! then it started to masturbate so i had to leave the room
Quote this comment #2.5 Posted by solardog on 16 Feb 2008 - 17:31
8800GTS 512oc, the most amazing card Ive ever owned, just became amazinger.
Quote this comment #2.6 Posted by Citrusleak on 16 Feb 2008 - 18:11
My 8800 GT smiled a lot. The only problem is with games like Crysis, where these cards can't even run it at max smoothly.
Quote this comment #2.7 Posted by The_Decryptor on 17 Feb 2008 - 06:05
My 7600GT started crying and wanted to be alone.
Quote this comment #2.8 Posted by Jasur on 17 Feb 2008 - 08:59
(Citrusleak said @ #2.6)
My 8800 GT smiled a lot. The only problem is with games like Crysis, where these cards can't even run it at max smoothly.


8800 GT 512 Golden Sample (G92) runs it at maz just fine.

love this card.
Quote this comment #2.9 Posted by RAID 0 on 17 Feb 2008 - 23:10
Oh yeah. My 6600 is feelin loved! AWW YEAH!
Quote this comment #2.10 Posted by ApuBo on 18 Feb 2008 - 11:25
And my G80GTX cry's happily
(3 replies) Quote this comment Reply to this comment #3 Posted by ivanlai118 on 16 Feb 2008 - 02:25
how about 9 series?
Quote this comment #3.1 Posted by Swordnyx on 16 Feb 2008 - 03:01
NSS. (No sh*t Sherlock)
Quote this comment #3.2 Posted by +Dakkaroth on 16 Feb 2008 - 05:34
(Swordnyx said @ #3.1)
NSS. (No sh*t Sherlock)


Then keep digging, Watson!
Quote this comment #3.3 Posted by IceDogg on 19 Feb 2008 - 07:22
(Dakkaroth said @ #3.2)
(Swordnyx said @ #3.1)
NSS. (No sh*t Sherlock)


Then keep digging, Watson!

ROFLOL I never heard that one before. Awesome
(1 reply) Quote this comment Reply to this comment #4 Posted by RobertH on 16 Feb 2008 - 02:49
Sooo does this mean that PhysX enabled games will now let you turn up the physics when this comes? And all games that support it will be able to give u added eye physics candy?

If so, then thats great. Something good came out of them buying PhysX. =)
Quote this comment #4.1 Posted by WICKO on 17 Feb 2008 - 04:11
Only if you have 2 videocards unfortunately. And if you did, you would more likely want to use SLI than physx.
Quote this comment Reply to this comment #5 Posted by +GreyWolfSC on 16 Feb 2008 - 02:50
Sounds cool. I have an 8600GT. Too bad my favorite PhysX game (Uru) is shutting down...
Quote this comment Reply to this comment #6 Posted by El-Diablo on 16 Feb 2008 - 02:59
hrmm, this sounds interesting i'd like to know more about this.
Quote this comment Reply to this comment #7 Posted by zeta_immersion on 16 Feb 2008 - 03:20
yeah, same here sounds interesting
(1 reply) Quote this comment Reply to this comment #8 Posted by vetneufuse on 16 Feb 2008 - 03:22
Didn't CUDA exist before the 8xxx seriese? I remember reading about it when I had a 6xxx seriese card...
Quote this comment #8.1 Posted by rajputwarrior on 16 Feb 2008 - 03:36
yeah but they never implement it at the gpu level until the 8 series, i am wondering what kind of performance hit this will have, especially on a 8800gt which i have
Quote this comment Reply to this comment #9 Posted by iCeFuSiOn on 16 Feb 2008 - 03:57
Ooh, nice!
(1 reply) Quote this comment Reply to this comment #10 Posted by Skyfrog on 16 Feb 2008 - 04:24
Wasn't it better to have a seperate chip doing all the physics work rather than putting that load onto the GPU? Seems a bit like a step backwards to me, but I could be wrong I guess.
Quote this comment #10.1 Posted by toadeater on 16 Feb 2008 - 06:15
(Skyfrog said @ #10)
Wasn't it better to have a seperate chip doing all the physics work rather than putting that load onto the GPU? Seems a bit like a step backwards to me, but I could be wrong I guess.


SLI.
Quote this comment Reply to this comment #11 Posted by Ronnie Sunde on 16 Feb 2008 - 04:46
it will probably be a CHIP in the future, for now they will enable physics in game with a software update...

you dont have to run physics in the games if you dont want too, or you could use a lower resolution and use physics...

be happy you will get something for free that you can try out... if not you have not lost anything.
Quote this comment Reply to this comment #12 Posted by +Harlem39s Finest on 16 Feb 2008 - 05:32
kewl
(2 replies) Quote this comment Reply to this comment #13 Posted by IntelliMoo on 16 Feb 2008 - 06:58
Bad move! hehe....

(From http://www.next-gen.biz/index.php?option=c...&Itemid=61)

Edge: You’ve more or less already placed your cards on the table about this, but what do you make of discrete physics cards, like Ageia’s PhysX?

Gabe Newell: I think that’s a horrible idea. At the same time that the distinction between the GPU and CPU is going away the PPU guys want to come in and define a new set of abstractions, where we have memory and data that’s really far away from the CPU and GPU... How do I tell when something breaks, or gets pushed by a monster? All these decisions I have on my CPU have to sit around until they are resolved on the PPU and GPU, and you end up with a physics decelerator. This is the reason you want a homogenous architecture.
Quote this comment #13.1 Posted by Kushan on 16 Feb 2008 - 15:00
(IntelliMoo said @ #13)
Bad move! hehe....

(From http://www.next-gen.biz/index.php?option=c...&Itemid=61)

Edge: You’ve more or less already placed your cards on the table about this, but what do you make of discrete physics cards, like Ageia’s PhysX?

Gabe Newell: I think that’s a horrible idea. At the same time that the distinction between the GPU and CPU is going away the PPU guys want to come in and define a new set of abstractions, where we have memory and data that’s really far away from the CPU and GPU... How do I tell when something breaks, or gets pushed by a monster? All these decisions I have on my CPU have to sit around until they are resolved on the PPU and GPU, and you end up with a physics decelerator. This is the reason you want a homogenous architecture.


Gabe isn't the be-all and end-all of games. Plus, this is about using the GPU to perform physics, removing the need for a separate card, so it doesn't really apply anyway.
Quote this comment #13.2 Posted by Smigit on 16 Feb 2008 - 18:19
(IntelliMoo said @ #13)
Gabe Newell: I think that’s a horrible idea. At the same time that the distinction between the GPU and CPU is going away the PPU guys want to come in and define a new set of abstractions, where we have memory and data that’s really far away from the CPU and GPU... How do I tell when something breaks, or gets pushed by a monster? All these decisions I have on my CPU have to sit around until they are resolved on the PPU and GPU, and you end up with a physics decelerator. This is the reason you want a homogenous architecture.
Well with NVidia putting this all onto the Video card I guess it's going in the direction of being more homogenious as Gabe put it.
Quote this comment Reply to this comment #14 Posted by Sazz181 on 16 Feb 2008 - 07:43
That's great news!
Quote this comment Reply to this comment #15 Posted by ajua on 16 Feb 2008 - 08:54
that is good news.
the main thing here will be the performance/eye candy hit current cards would take when doing physics...

my 8000gts 320mb has already returned the investment but this could do a little more for free.
Quote this comment Reply to this comment #16 Posted by sPudz on 16 Feb 2008 - 11:16
this sounds good to me, saves you £150 on one of those PhysX cards i guess... we'll just have to wait and see performance wise but i expect it shouldnt be too bad if at all
Quote this comment Reply to this comment #17 Posted by -Vendetta- on 16 Feb 2008 - 12:05
AWESOME!!!! No regrets with getting my 8-series card.
(2 replies) Quote this comment Reply to this comment #18 Posted by Smigit on 16 Feb 2008 - 14:03
I personally don't see this being a hugeeee bonus for current card owners. By the time NVidia has this out and games are making use of the new feature I'm sure the 8800's will be at a point where you'll probably be looking to disable it to maintain a framerate anyway.
Quote this comment #18.1 Posted by notuptome2004 on 16 Feb 2008 - 16:20
(Smigit said @ #1
I personally don't see this being a hugeeee bonus for current card owners. By the time NVidia has this out and games are making use of the new feature I'm sure the 8800's will be at a point where you'll probably be looking to disable it to maintain a framerate anyway.




this feature has been out for what i think 2 years in the ageia hardware add-on but now it is coming to the GF8 cars via software from nvidia and i doubt anyone would disable it as it would free up physx being done on the CPU for the most part on games made for the Physx add-on hardware
Quote this comment #18.2 Posted by Smigit on 16 Feb 2008 - 18:07
Sure but how many current games really take alot of advantage of Ageia currently? I can think of maybe a handful but not to many. Any future games are likely to strain the 8800 cards as it is given that they havent really had any upgrade in the past 18 months performance wise (GTX is still more or less the top tier product besides the slightly faster Ultra). With dual cores and now quad core CPU's becoming more mainstream, I think that for gaming currently the video card is the main bottle neck (especially as you scale the resolution up to higher resolutions) in which case I'd rather have the CPU handling the physics if I had to pick between the CPU and GPU.

I realise that it's getting a solution that could only be done in hardware up to now onto a software platform more people can make use of it (to which end it's good), but I feel theres not many current games that take advantage of it and for future titles the 8800's will likely struggle enough without the extra workload. For a card lineup that's 18 months old I'm unsure how useful it's ultimately going to be and the real benefits will be seen on future cards (to which end I'd love to see them incorporate it at a hardware level).
Quote this comment Reply to this comment #19 Posted by Citrusleak on 16 Feb 2008 - 18:15
It's gonna be a nice addition to current cards. I really am beginning to think ATi is going to have to have a miracle to ever be able to catch up to the nvidia price/performance ratio and just catch up w/ speed overall.
(3 replies) Quote this comment Reply to this comment #20 Posted by BrainDedd on 16 Feb 2008 - 18:28
As far as I am aware Cuda is a separate part of the 8 series chip, so essentially it's free physics.
I am still waiting for the software to use the hardware mp4 encoding supposedly built into my 8 series though.
Quote this comment #20.1 Posted by WICKO on 16 Feb 2008 - 18:47
Unfortunately this is not the case. CUDA allows programmers to write programs which actually run on the GPU itself. I think the point here is that an 8-series card can now be used as a physics card, rather than a video card. So most likely you will need 2 8-series cards, or 3 if you want SLI+physics.

http://techreport.com/discussions.x/14147

Read the second quotation.

Last edited by WICKO on 16 Feb 2008 - 18:54
Quote this comment #20.2 Posted by ZombieFly on 16 Feb 2008 - 19:55
(WICKO said @ #20.1)
Unfortunately this is not the case. CUDA allows programmers to write programs which actually run on the GPU itself. I think the point here is that an 8-series card can now be used as a physics card, rather than a video card. So most likely you will need 2 8-series cards, or 3 if you want SLI+physics.

http://techreport.com/discussions.x/14147

Read the second quotation.


i think you missed the word "Potentially" when you read that!
Quote this comment #20.3 Posted by WICKO on 17 Feb 2008 - 04:09
(ZombieFly said @ #20.2)
(WICKO said @ #20.1)
Unfortunately this is not the case. CUDA allows programmers to write programs which actually run on the GPU itself. I think the point here is that an 8-series card can now be used as a physics card, rather than a video card. So most likely you will need 2 8-series cards, or 3 if you want SLI+physics.

http://techreport.com/discussions.x/14147

Read the second quotation.


i think you missed the word "Potentially" when you read that!


What? Even if it is a possibility, it still doesn't change the fact that CUDA is *not* a second chip which will process physics. I think you missed my point

In fact, even the mention of it being a possibility (and how they said they think it will encourage people to buy a second GPU) implies that a video card can only do one or the other.

Last edited by WICKO on 17 Feb 2008 - 04:14
Quote this comment Reply to this comment #21 Posted by IceBreakerG on 17 Feb 2008 - 02:50
I'm thinking that this will help nvidia push their new Hybrid SLI technology as well. Since the onboard GPU for Hybrid SLI is supposed to be like a 8200 series gpu, it will be able to support CUDA as well. I'm not sure how intensive the physics are though. I'm also not sure how the performance/power of the 8200 compares to a standard PhysX card (it may be quite a bit more powerful for all we know). So this may give a new reason to look into Hybrid SLI. Guess we just have to wait and see what nvidia says about it.
Quote this comment Reply to this comment #22 Posted by carmatic on 17 Feb 2008 - 16:48
in the short term, this allows me to use my x38 mobo with a pair of 8800's so the second x16 pci express slot doesnt go to waste... since it isnt going to do sli anyway, might as well make the other card do physics....

i wonder, when the card is doing physics, will it output anything at all? it might show really funky stuff on the screen...

reminds me of this http://www.nvidia.com/object/tesla_gpu_processor.html ... its basically a 8800 gtx , with no video outputs and twice as much ram... tesla + cuda + physx = ?
didnt nvidia also have something called the 'quantum effects technology' that uses the g80 to do physics?

also, in general the cuda thing should also allow the supported graphics cards to be more flexible, like the distributed computing projects should be able to use the nvidia gpus , up to now people like folding@home could only use ati's gpus

also, unlike the physx cards , the gpus are general purpose and not specialized for physics...

Last edited by carmatic on 17 Feb 2008 - 17:07
[1]

Commenting has either been disabled on this article or you are not logged in. Click here to login or register, its free!

Note: Anonymous commenting is disabled in order to keep the quality of responses to a high standard.


Scroll to the Top
....
My Preferences
....
Communicating with server
Loading
Please Wait...
....
Loading
 X 
....