Sign in with
Sign up | Sign in
Your question

SLi Physics??

Last response: in Graphics & Displays
Share
January 6, 2007 5:49:09 PM

hello,
I am looking for a new mobo with SLi feature..
and while i was searching for a cheap board,
I remembered nVidia's idea of SLi Physics.
From my understanding, SLi required same graphic card..?

SLi physics = same gpu required?
or is it like ATi's triple play,
where the card for physics can be different card then for graphics?

does anyone know if current nforce 4 sli will support sli physics?
or nforce 4 is obsolete?

More about : sli physics

January 6, 2007 6:10:16 PM

sli physics isnt all its cracked out to be...which is why we silently saw the campaign for it dissapear. I wouldnt worry about it at this time.

Some of the info present is promising, but SLI 7xxx series still can process "true" physical calculation.
January 6, 2007 6:22:28 PM

actually with this new design taking by nVidia, i actually think they are going to go after physics processing.

In case your not aware of how SLi physics works, it utilizes one card in SLi or Quad-SLi to process physics, which on paper sounds great because if you have SLi already you won't need anything special. But i don't think there concept mesh's well with traditional SLi and is more suited for Quad-SLi where you can sacrifice a GPU at not cost to performance at all. Technically all the same rules would apply to SLi physics as it would non physics SLi because there physics solution is part of SLi.

ATi has the right approach but i don't think were gonna see either of there techs out till 2008-2009.

SLi Physics is just paper right now and maybe a myth. so don't hold your breath till we see something solid.
Related resources
January 6, 2007 6:25:31 PM

I'm very well aware how the system works.

But the thing, the memory cache' and architecture dont allow for true physics processing. Light particles and reaction yes, but something like a wall falling down, or destruction is still handled by the CPU. and I love how NV says they'll relate the processed data directly to the DX driver without the intervention of the CPU?? :lol:  :lol:  :lol:  I've still yet to hear anything more.
January 6, 2007 6:27:02 PM

Right. As I remember the whole deal - the 7xxx models could be paired up with a, say, G80 or 8xxx series card and used as physics only. Ofcourse one core in a multiple core CPU could be given instructions to process physics also, one or the other will come to be. Most likely the CPU will be chosen, don't you think? To do this as using the other PCI-E slot will shut down SLi graphics (unless we start seeing certain mboards sporting 3+ PCI-E 8O ) imo. From a business point of view :?
January 6, 2007 6:42:03 PM

oh i didnt doubt you knew i just wanted to provide a little background on the theory for the original poster so he understood that because it SLi, the rules for SLi would apply.

i agree with your sig GO BUCKEYES!!!

I think ATi will be Ageia's true physics competitor myself
January 6, 2007 7:19:15 PM

There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.
January 6, 2007 8:51:26 PM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


lanes aren't an issue as graphic cards don't even use 100% of the a single x16 slot... the big thing will be implementation, if ATi can build all the physics engines and SDK API's into all there GPU's like they claim they can then ATi wins the physics battle all the way. But its years off, no ones even seen a working prototype outside the demo ATi showed but wouldn't let people touch and the paper talk nVidia let out.
January 6, 2007 9:09:07 PM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


Thats true. Why not build the physics engine around the 8x slot instead of having the 7xxx using up a 16x slot that could house a videocard, if nV chooses to utilize the 7xxx for physics.
January 6, 2007 9:12:40 PM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


Thats true. Why not build the physics engine around the 8x slot instead of having the 7xxx using up a 16x slot that could house a videocard, if they choose to utilize the 7xxx for physics.


That is ATi's solution, say using current generation hardware.

2 X1950XT's in CrossFire on x16 lanes each
1 X1650Pro running on the x8 slot for physics only

couple there concept with ATis raw calculating power, thats how i conclude they will win if they can implement it right.
January 6, 2007 9:15:35 PM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


Thats true. Why not build the physics engine around the 8x slot instead of having the 7xxx using up a 16x slot that could house a videocard, if they choose to utilize the 7xxx for physics.


That is ATi's solution, say using current generation hardware.

2 X1950XT's in CrossFire on x16 lanes each
1 X1650Pro running on the x8 slot for physics only

couple there concept with ATis raw calculating power, thats how i conclude they will win if they can implement it right.

:)  Heh heh, anyone wanna cook an egg with a 1KW PSU? :) 
January 6, 2007 9:25:57 PM

Yeah, Buckeyes are gonna roll.....

I hope Cleveland picks up Ginn and Smith as free agents, that would be awesome.
January 6, 2007 9:28:23 PM

Could someone post a link to a news source regarding this?
(preferably something more reliable than the inq)
January 6, 2007 9:30:06 PM

Quote:
Yeah, Buckeyes are gonna roll.....

I hope Cleveland picks up Ginn and Smith as free agents, that would be awesome.


i do

but i don't

i want them to be successful

:roll:
January 6, 2007 10:50:00 PM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


lanes aren't an issue as graphic cards don't even use 100% of the a single x16 slot... the big thing will be implementation, if ATi can build all the physics engines and SDK API's into all there GPU's like they claim they can then ATi wins the physics battle all the way. But its years off, no ones even seen a working prototype outside the demo ATi showed but wouldn't let people touch and the paper talk nVidia let out.

I meant to add that in there. Forgot though. I was simply pointing out that Motherboards already have more than two slots.
January 6, 2007 10:51:00 PM

yea market is getting there, but were far from utilizing it to its potential, kinda makes ya wonder
January 6, 2007 10:51:41 PM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


Thats true. Why not build the physics engine around the 8x slot instead of having the 7xxx using up a 16x slot that could house a videocard, if they choose to utilize the 7xxx for physics.


That is ATi's solution, say using current generation hardware.

2 X1950XT's in CrossFire on x16 lanes each
1 X1650Pro running on the x8 slot for physics only

couple there concept with ATis raw calculating power, thats how i conclude they will win if they can implement it right.

I would hope you could do one X1950XT and one X1650pro for graphics and physics respectively.
January 6, 2007 11:02:37 PM

Quote:
I would hope you could do one X1950XT and one X1650pro for graphics and physics respectively.



haven't seen them say anything about it anywhere or i may have and i just don't remember right now (long night of the drinky drinky) but from what i understand the the 3rd PCI-e slot is reserved for x8 only and for a twin card non CF setup, meaning you could run 1/1 Graphics/Physics or Graphics/Graphics for a 4monitor setup or twin 30" setup
January 6, 2007 11:21:35 PM

I still say Ageia wins because, well they have ACTUAL physics card! that does wait for it..... PHYSICS!


All this SLI/CS physics is a just a bunch of Business BS to get people to buy 3 GFX cards.....
January 6, 2007 11:39:01 PM

Quote:
I still say Ageia wins because, well they have ACTUAL physics card! that does wait for it..... PHYSICS!


All this SLI/CS physics is a just a bunch of Business BS to get people to buy 3 GFX cards.....


be reminded Ageia doesn't do physics very well at all so they left the table wide open for ATi and nVidia to try there versions

in a few years about the time DX10 gets adopted we will really see who the physics champ is.
January 7, 2007 12:24:00 AM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


Thats true. Why not build the physics engine around the 8x slot instead of having the 7xxx using up a 16x slot that could house a videocard, if they choose to utilize the 7xxx for physics.


That is ATi's solution, say using current generation hardware.

2 X1950XT's in CrossFire on x16 lanes each
1 X1650Pro running on the x8 slot for physics only

couple there concept with ATis raw calculating power, thats how i conclude they will win if they can implement it right.

:)  Heh heh, anyone wanna cook an egg with a 1KW PSU? :) 

Are you suggesting we put that inside a 4x4?
January 7, 2007 1:24:46 AM

Quote:
I still say Ageia wins because, well they have ACTUAL physics card! that does wait for it..... PHYSICS!


All this SLI/CS physics is a just a bunch of Business BS to get people to buy 3 GFX cards.....


Yeah that's great ! :roll: Nothing takes full use of it yet and I don't think I've heard anything that might. The only thing that comes to mind that uses the PhysX card would be GRAW and that seems to have been slapped on at the last minute.
January 7, 2007 2:46:17 AM

Quote:
There are already motherboards with more than 2 PCI-e slots 680i boards can support full 16x bandwidth with an additional 8x PCI-e slot.


Thats true. Why not build the physics engine around the 8x slot instead of having the 7xxx using up a 16x slot that could house a videocard, if they choose to utilize the 7xxx for physics.


That is ATi's solution, say using current generation hardware.

2 X1950XT's in CrossFire on x16 lanes each
1 X1650Pro running on the x8 slot for physics only

couple there concept with ATis raw calculating power, thats how i conclude they will win if they can implement it right.

:)  Heh heh, anyone wanna cook an egg with a 1KW PSU? :) 

Are you suggesting we put that inside a 4x4?

No need for a furnace.
We forgot a much needed Pentium 4 equivelent processing unit for music; the X-Fi. You may be able to build muscle just lifting the motherboard :D 
January 7, 2007 5:22:14 AM

Quote:
I still say Ageia wins because, well they have ACTUAL physics card! that does wait for it..... PHYSICS!


All this SLI/CS physics is a just a bunch of Business BS to get people to buy 3 GFX cards.....


Yeah that's great ! :roll: Nothing takes full use of it yet and I don't think I've heard anything that might. The only thing that comes to mind that uses the PhysX card would be GRAW and that seems to have been slapped on at the last minute.

Well if dont believe me here, an UPDATED story with Ageia

"First, I want to tell you a little bit about AGEIA's latest advancement in their physics engine. This update is SDK version 2.6 and includes a plethora of new and exciting physics capabilities. These features range from the deformable metals, impact retention, self-collision prevention, Vista and Linux support, and more optimizations of existing features"

http://hardware.gotfrag.com/portal/story/35479/

And we all know GRAW(for PC) is a horrible game you just point out the obvious, Cell factor shows the raw power and what you are able to do with Ageia's Card, o and mind you its PCI so u dont need a 3rd PCI-E slot to run physics if u have SLI/Or CS

You guys who posted here said it yourself nVIDIA and ATI have yet to offer a GAME that boasts use of SLI physics.

"But its years off, no ones even seen a working prototype outside the demo ATi showed but wouldn't let people touch and the paper talk nVidia let out."
January 7, 2007 5:33:31 AM

well theres more than one issue with keeping the PPU on PCI slot
1) most gamers who will buy a PPU has a sound card
2) very few boards or SLi/CF setups allow for two PCI slots to used
3) IF they can use two PCI slots the combined bandwidth usage from a sound card and PPU would start to bottleneck actually hurt peroformance or be wasted money

moving to the PCI-e slot is actually the best idea, so nVidia/ATi will have an edge in this respect, whereas ATis approach is logically better, but yes there only paper now, so yes Ageia is the king of physics till then, but we won't see any real physics competition till DX10 becomes the standard
January 7, 2007 11:32:13 AM

BAM! Score another one for Ageia....

AGEIA has informed me that a PCI-E version will finally be available to those who have limited PCI slots on their motherboards. The majority of motherboard manufacturers are building less and less PCI slots on their motherboards. So anyone using a third party sound card or other PCI card can't add much more. To solve this shortage AGEIA has decided to make use of the 1x PCI-E slot

Source: http://hardware.gotfrag.com/portal/story/34912/

Its going to PCI-E x1 which is what im waiting for my current setup will fit a x1 PCI-E card perfectly, i could get the PCI verison but then i would have to take out my VGA cooler thats right under my 8800GTS
January 7, 2007 12:58:53 PM

It's been said that the Ageia doesn't utilitze more the 15% of it's potential at this point in time. Ageia's got a good thing going even though current games can't touch it's power. ATI's idea on the x8 slot isn't bad either - the only thing to my conern is what the power consumption would look like with 3 ATI cards 8O
January 7, 2007 2:04:57 PM

Mutli-cored CPU's with better FP instruction sets and calculation will effectively kill any other dedicated PPU in due time. If there really is a physics API being developed for a later realese with Vista, this will be the first step towards such a statement.
January 7, 2007 3:08:08 PM

I wonder, what is the point of AMD putting ATi's floating point power on the same die as a general processor? Hmmm...
a b U Graphics card
January 7, 2007 11:06:40 PM

Quote:
I still say Ageia wins because, well they have ACTUAL physics card! that does wait for it..... PHYSICS!


Wait for it... so does the graphics cards. And considering they do most of the visual physics much MUCH faster than Ageia's solution, then they can supplement the CPU when needed, and unlike theAgeia card, when a game doesn't support physics you have the opportunity to use the second or 3rd and fourth card potentially for added visuals while the PhysX card is just an energy sapping heat source doing nothing.

Quote:
All this SLI/CS physics is a just a bunch of Business BS to get people to buy 3 GFX cards.....


You don't need 3 cards, and considering they aren't selling the solution right now, it's not to get people to buy 3 graphics cards, it's to keep people from wasting their money on the Ageia solution which is very limited and has little game support right now. Think of it like a pre-emptive strike to halt Ageia sales from people who don't know squat and just buy based on press.

Considering that M$ already said they are going the VPU physics route with Direct Physics then it's likely the titles will favour ATi and nV, not Ageia, and announcing anything prior to Vista would make nV and ATi look as stupid as Ageia did with their performance crippling GRAW launch. Sweet $200 to lower your framerate for little benifit, run out and buy one now. 8O

What would be interesting is to see the GF8800 series doing HavokFX physics since it has far more processors with great FP abilities and better i/o communication with the host.

BTW, how is the VPU physics 'business bs' any different than the Ageia BS, both are trying to get you to buy a solution, duh! :roll:
January 8, 2007 12:00:22 AM

^Agreed.
January 8, 2007 10:16:16 AM

"And considering they do most of the visual physics much MUCH faster than Ageia's solution"

awesome, now prove it, The idea of the PPU was to take the load of the CPU and was defined to do better physics, like oozing liquid, cloth and particles, deformable metals. As you can see in Gotfrags updated review, when playing Cellfactor the PPu DOES off load and help the CPU

"it's to keep people from wasting their money on the Ageia solution which is very limited and has little game support right now"

as i recall, i cant run my 8800GTS to render company of heros and then have my old 7900GT put in another x16 slot to run Company of heros Physics which uses the Havok engine.

Here is a list of Videos you can watch that uses the AGEIA solution.
http://www.ageia.com/physx/videos.html

And here is the ATI demo they showed off notice the difference bewteen the two?
http://video.google.com/videoplay?docid=148246142376575...

AGEIA USES GAMES to show off their engine aleast, hell for you know CS/SLI physics might make your FPS DROP and frames stutter, but then again who know seeing as its no released yet, not game announced support for it either.
a b U Graphics card
January 8, 2007 3:51:01 PM

Quote:

awesome, now prove it,


Just look at the floating point numbers, and both companies' official sphere-on-sphere collision numbers, they speak for themselves, the X1600 can calculate 2X the interactions, and the X1900 series ~10x I'd love to see the GF8800 because it's FP prowess likely has boosted it another factor. Also remember, I say 'visual' because they are not depedant interactions the way Ageia does, however those are very limited in most games, and according to nV the stream processors in the GF8800 should be able to handle those much better than before.

Quote:
The idea of the PPU was to take the load of the CPU and was defined to do better physics, like oozing liquid, cloth and particles, deformable metals.


And the VPU physics is perfect for that stuff, in facts that's the list of the ones that are best on VPU physics because they require only limited interactions and very low host/cpu communication.

Quote:
As you can see in Gotfrags updated review, when playing Cellfactor the PPu DOES off load and help the CPU


And the VPU can offload and help the CPU too, and as has been shown many times before (including here) the CELL factor demo can also be done with a second CPU/core to achieve the same effects, it's far from convincing of the 'need' for a single use dedicated PPU. The thing to remember about your statement preceeding this one was that it's to offload some calculations from the CPU, it does need to take the place of the CPU completely as the demos try to do, doing 5 times the physics calcs @ 72% CPU useage using VPU physics is just as impressive as the same effect with 68% CPU useage using a PPU, so the limitation isn't as much of an issue except in games where it all about input physics and not visual like I said. And depending on the calculation (the ones you stated are visual) the VPU based physics.And CELL factor was far from a good view of things like the liquid interactions because they looked unrealitics, which isn't because of the card, but the coding, whcih is the most important part. Will the devs and the API bring out the maximum of each solution.

Quote:
as i recall, i cant run my 8800GTS to render company of heros and then have my old 7900GT put in another x16 slot to run Company of heros Physics which uses the Havok engine.


That's what you think, Havok3 is not Havok FX, Havok 3 is in Oblivion, F.E.A.R. and other games, and it's CPU based physics, any add-on would be in addition to Havok3, and I haven't heard anything that says CoH has VPU based physics added to it.

BTW, my comment was on the lack of suport for Ageia titles, and their only major upcoming title is UT2K7, so it's not like either camp has major compelling reasons right now. Crysis and Halo2-PC are both supposed to support VPU-physics either with their own implementation of DirectPhysics or through HavokFX & DirectPhysics. Crytek's Crysis is a Havok title.

Quote:
Here is a list of Videos you can watch that uses the AGEIA solution.
http://www.ageia.com/physx/videos.html


Of which they put CellFactor in twice, B.O.S. twice, and the only truely noteworthy game is City of Villans. Notice how GRAW is strangely absent from that list. :mrgreen: Like I said, for both companies, there's nothing compeling out there yet, nor for a while, and the game support is both limited in quantity and quality.

Quote:
AGEIA USES GAMES to show off their engine aleast,


As for showing ATi's demos sofar, they are early stage demos of a non-commercial product that is relying on yet to be released support from M$. Seriously you're deperate to declare a winner right now when the battle hasn't even warmed up yet. Wait for the commercial release of VPU physics and then compare videos. You're grasping at straws boy, trying to compare a proof of concept demo to something that has had a year to prove it's worth and done no such thing. Are you payed by Ageia to be so obtuse?

Quote:
hell for you know CS/SLI physics might make your FPS DROP and frames stutter,


True, but at that point, then I simply turn of the physics and get the benifit of Xfire/SLi or multi-monitor support, which are two things that Ageia can't do, if it's droping your framerate, then once you disable the PhysX card, then it's just wasting energy and heating your case. So if the default is no effect at least there are other options with dual VPUs,

Quote:
but then again who know seeing as its no released yet, not game announced support for it either.


Well many have announced that they will be based on the Havok engine, but not what components of Havok4 will be used, BTW it's 4 not 3 that involves the HavokFX component that uses VPU physics. And while the companies are pretty secretive, the list of games that use the Havok engine is longer than those that use Ageia's acquired Meqon solution, and Ageia still has Duke Nukem Foerever on that list, which is suspect, and they also include games that use both the Havok and Meqon engines (like GRAW which is still based primarily on the Havok engine with the PhysX engine tacked on).
January 8, 2007 9:27:08 PM

Nicely put, First i seen where it hasn't turned to a flamer war,

I'm not eager to say Ageia wins hands down, i too when it first was released disliked it and thought it was a bogus idea but over time(and some price drops later) it seems a bit more reasonable.

I havent brushed up on HavokFX/havok4 so i think i should i assume it going to go with DX10 hand in hand?

Only till GPGPU is realized and GPU can be utilized for physics we will see reviews comparing Ageia solution VS ATI/nVIDIA's.

What i love to see is Havok and Ageia combine, making one giant physics party where every one can get along, and we can use Either GPU or PPU to do Physics in ANY GAME.
a b U Graphics card
January 8, 2007 11:30:26 PM

While it'd be nice to see this covered by two different approaches in harmony, it seems like the BR/HD-DVD war that the two camps have a vested interest in keeping things separate. I think VPU physics gets a boost from M$' choice of VPU based physics for their direct physics at the expense of the PPU. Had M$ said they would support both I'd think that there'd be a better chance. Right now for Ageia to survive they need to make it to the point at which the killer apps matter, and then prove their worth. VPU based physics doesn't have that same concern (hence little need to rush and botch it) because it's only a secondary or even tertiary benefit of something whose primary function is something else. If Ageia can't get people to appreciate the PPU with the launch of UT2K7 then I'd say it's pretty much over for Ageia in the hardware market (they will still likely have their software engine division), and there will be added pressure if Crysis showcases VPU physics.

At this point in time I just wonder about the financial health of Ageia considering their lackluster sales, and the obvious cost of development. How much are they taking money away from their software engine to pay for this hardware experiment? They need to do something to move cards if they wish to continue, but it can wait if there's a nice launch point like UT2K7, since there's no point to spend the advertising money now if there's little to sustain sales, and little reason to launch a revised PCIe version for the same reason, once that killer app comes by then relaunch as if the PCI/GRAW period didn't exist.

Either way I don't think a winner will be crowned for some time, and right now there's little to get excited about until there's truely more than the lame examples both sides have shown to date. If GRAW or CityOfVillans or something made people truely stand up and say OMFG that's Awesome, then this ould be different, or if Oblivion had been Havok4 not Havok3 then maybe we'd be closer to even seeing this battle start, but right now both companies are pretty much in the demo business.

I wouldn't recommend anyone build specifically for either, but at least with VPU physics there is a nice option of using otherwise leftover/useless cards for another use once the X1600/GF7600 gets old and replaced. However even then, people aren't going to keep their old cards for ever if they can get a good price on eBay for them, so even that benefit has it's limits.
!