Recommendations on 3rd slot physics?

Joined
Oct 10, 2007
Messages
581
I'm upgrading to crossfire (2x x1900xtx's I've got) and I wanted to use a third card for the physics. I was looking at ATI x1650 Pro 512mb 128-bit. I was wondering if that card will be hamstrung by the fact that it's 128-bit and if so how noticably. Should I be considering a 256-bit card instead?
 
Yes, yes it does

There are currently at least 2 boards that support 2 crossfire +1 physics:
~$150 DFI LANParty UT ICFX3200-T2R/G
~$220 INTEL BOXD975XBX2KR

This is on the Intel side and I have no idea about the AMD side, but I know when they demoed RD600 (it became XPRESS 3200 IE) they used 2 cards crossfire and something like a x1600 for the physics. I'm sure there are more expensive boards that also support it in the x38 line, but if you're gonna drop $400 on a board you may as well just get a suped up gpu instead.

So yah, if anyone knows what works best for this please LMK.
 
^^ ...that link is nothing but PR. There's no actual drivers or applications that it will work with. Sorry, useless upgrade right now.
 
Ah yeah, reading more on it right now and seeing that. Thanks for the heads up. It's crazy. I read about Havok/Aegia darn near 1.5 years ago. I figured they would have support by now. But from what I've just read Intel killed it by buying Havok to force Physics onto multi-core cpus instead.

Almost makes me want to throw out my C2D and buy an AMD cpu instead.... almost.
 
Ah yeah, reading more on it right now and seeing that. Thanks for the heads up. It's crazy. I read about Havok/Aegia darn near 1.5 years ago. I figured they would have support by now. But from what I've just read Intel killed it by buying Havok to force Physics onto multi-core cpus instead.

Almost makes me want to throw out my C2D and buy an AMD cpu instead.... almost.

A lot of PR was spinned to steal AGEIA's thunder...
Still no games in sight...or even a working solution for consumers.
And still people advocate this fata morgana solution...

The word "vapourware" comes to mind...
 
From what I can tell, there was actually work done on it until Intel bought Havok and then it was more or less abandoned. Intel's pushing the multi-core approach and since there isn't physics card support wasn't able to draw up enough support before the quad-cores hit, most developers seem to just be going down that route, even when physics support is possible.

It's a shame it died in its tracks, but I guess long run it's probably better we don't have yet another component to be shoving tons of heat and power consumption into cases.

Maybe UT3 will breath new life into PhysX and then maybe physics in general, but I guess we'll have to see. From the preview, it looks like most stuff is done by the PhysX software, but they did say they'll be releasing mods to install for PhysX owners.
 
From what I can tell, there was actually work done on it until Intel bought Havok and then it was more or less abandoned. Intel's pushing the multi-core approach and since there isn't physics card support wasn't able to draw up enough support before the quad-cores hit, most developers seem to just be going down that route, even when physics support is possible.

Could I see your list of quad-able, physics runnning games?

It's a shame it died in its tracks, but I guess long run it's probably better we don't have yet another component to be shoving tons of heat and power consumption into cases.

I won't miss GPU-physics...+100 watt powerusage...compared to the, what, sub 30 watt form a PPU?

Maybe UT3 will breath new life into PhysX and then maybe physics in general, but I guess we'll have to see. From the preview, it looks like most stuff is done by the PhysX software, but they did say they'll be releasing mods to install for PhysX owners.

Ageis PhysX API can run without the PPU, just like Havok does...and then it can do much more with the PPU.
So wether or not you own a PPU, the PhysX API will be running under UT3...
 
A lot of PR was spinned to steal AGEIA's thunder...
Still no games in sight...or even a working solution for consumers.
And still people advocate this fata morgana solution...

The word "vapourware" comes to mind...

WHAT thunder? Ageia had what, 5 games that used their PPU? It was a useless addition back then, and still is.
 
Atech, there are about as many games quad-core as there are using PhysX. Considering they haven't been out that long, that doesn't bode well for PhysX or Havok FX PPU's. Developers will naturally start to code for quad-core since customers will already be buying them in new pc's in a couple years even for the low-end I'd expect.

In contrast, PhysX/3rd slot physics unfortunately seems like a tougherr sell. Looks like PPU's might look great on specific games that use them, but without general acceptance by developers and standardization so that the Havok/PhysX cards are interchangeable (in terms of their respective software being able to unload on them), it doesn't seem like PPU's are going anywhere fast. Quad-cores are a little bit more of a sure thing since people will already be buying them with computers so there isn't anywhere near the same level of uncertainty for developers.

And we won't know what if anything PhysX will really be able to do until UT3 actually comes out. The demo only uses the software and doesn't actually dump anything to the PPU. So at present, at least PhysX is still useless.
 
I never understood the reasoning behind people bashing the PPU and calling it a waste of money for not having any games; then going on to talk fondly of Havok FX and how they can't wait to buy a more expensive, greater power eating, more heat generating GPU for physics that has only two announced games that support it. Scratch that, one game, Hellgate London cancelled support.

Not aimed at anyone in particular, just venting at the mentality of some people I have met online.

As for games taking advantage of quad core; there isn't even any dual core games on the market, so what games are being referred to exactly? There is a world of difference between a game spreading tasks between the cores to boost performance and actually taking advantage of the additional cores to do something new like high end physics. The only true multi core game I can think of is Alan Wake and it is dual core based; there is also that problem with processing multiple physics tasks across multiple cores, something to do with the way processors or code is done one task after the other.

If the PPU doesn't get accepted then nothing will take its place, it just means waiting many years until processing power and install base allows better physics like we have always done. Many many years...
 
I've bashed all the solutions. They're all inefficient and/or useless, and provide no noticable GAMEPLAY benefit. Nothing is well supported. However since it seems more CPU cores is the way of the future, I think that's the most logical way to go.
 
Yeah, I would love to believe in Havok FX and/or PhysX PPU's. I posted this topic because I assumed that they would have matured at least a little over the course of the past year and a half, but apparently they haven't. I truly hope UT3 changes things.

In terms of quad-core, there already are a few games:
http://robfisher.net/wiki/wikka.php?wakka=QuadCoreGames

There'll just be more as time passes. I mean, I'm sure they'll come out with an API at some point that automates the threading etc... so that companies don't have to worry about it as much. Same can be said for PPU's, but the problem they face is that people have to first know that they exist and then actually go out and get them in large enough numbers. Otherwise developers won't see any sense in programming for PPU's. If everyone already has quad-core in a couple years, they can count on people having that piece of hardware, but not the PPU's.

Supreme Commander apparently uses extra cores to calculate intelligence so enemies aren't as dumb. I'm sure there's other dual-core games as well. It's true that the gaming industry still has a lot of evolution to get into the multi-threaded, multi-core bit, but in the scheme of things even dual-cores have only been around four years or so. So yeah, it'll probably take a little more time for games to adapt to the new technology, but the thing that gives multi-core cpus the advantage is that they have other uses.

If there aren't many games a PPU can be used on, it'll just be sitting in peoples' machine most of the time so most people won't spend the money on it. Multi-core CPU's, on the other hand, have the advantage that people can use them to do more at once while not playing games, or to play games and do other things at once. So even if multi-core is used extensively in games yet, it's got a place in the market outside of gaming that'll put it in peoples' machines until the capacity is taken advantage of by games.

So I guess my third slot will just go to something else. Some controller card or whatnot, but I can't justify the expense of a card that is hardly used at present.
 
I understand the install base advantages of multi core processors, were as only a few people have a PPU and a secondary GPU to dedicate to physics. But 'everything' I have seen also tells me CPU physics is significantly slower to progress than the experimental alternatives, it is the reason the alternatives were considered in the first place.

Were the GPU and PPU have install base and software support problems, the CPU suffers from average user performance. It is the reason only Ageia supported games utilise the PPU to a point were it is required for smooth game play; fully utilise dual or quad and everyone who doesn't meet the spec cannot play your game, that is a huge incentive for sale concious developers to not take advantage of modern processors. It always annoys me when someone tweaks a PPU effect game to allow additional PPU physics and boast it is proof it is not needed, on their E6850 conroe system. The fact of the matter is not everyone has a processor like that, something steams hardware survey will make clear. People like to think the dual and quad install base is fine because of prices and retail setups shipping with them, over time that will happen but developer decisions on physics level indicate we are not at that point yet.

It is the reason I think either Havok FX or PhysX succeeds, or the next gen physics affects/effects we have witnessed from the technologies are dead. Because that level of physics won't appear in games till software is easily threaded, high end cores are plentiful and everyone has them. The point of the two technologies is to allow those physics today instead of years down the road, and they do that, if people decide they would prefer to just stick with CPUs for physics then fine but they shouldn't expect anything spectacular any time soon.

In the meantime a software optimized piece of high quality damageable cloth eats 60%+ of my 2.8ghz AMD 5600+ x2, not very practical when you also have a game to run on that processor.
 
Don't get me wrong, I totally think that PPU is a more effective and better solution, at least in the foreseeable future and I would love it if it enjoyed a broader support base. It just sucks that Intel bought out Havok and they are gonna push their inferior multi-cpu model while crippling Havok FX (by not developing it further). I think Havok et al should release their PPU API's for the same price as the normal price ones. I think that single fact would make more developers consider it - if they weren't paying more to get it in their game. That's one of the big things. But anyway, the ATI PPU seems dead as of now so I won't be tossing a card in for physics.

Maybe I'll get PhysX if it picks up a broader following. I know Nvidia also is working on some stuff, but I've got 2 x1900xtx's so I'm in no rush to switch gears at this point.
 
agiea keeps calling their solution "true physics" and I gotta tell ya, they really don’t seem to have much down. The agiea API supports hundreds or even thousands of parallel simple kinetic equations, but it still doesn't take into account just how elastic the collision should be (as if it did we could theoretically have an accurate representation of just how much and what kind of sound would be created). Nor does it take into account density, heat, any form of sound... it's only got a really basic fluid dynamics sub program... overall I was un impressed.

Depending of course on how the program was written, an API that supported all the features above would probably be better served on a GP PU then one of agiea's dedicated PPUs.

anyways, the tri-crossfire drivers arn't out yet but there are grumblings from both Nvidias camp and ATIs camp about future tri-gpu configurations, granted Nvidias grumblings seem to be louder and more eminent.

edit: and in regards to that steam survey, this is a shocking number (to me at least)

[name]................................[total #].......[%]
NVIDIA GeForce 6600..........75,541........6.91
NVIDIA GeForce 7600..........71,429........6.54
NVIDIA GeForce FX 5200.....51,612.........4.72
NVIDIA GeForce 8800.........49,850........4.56 %
 
WHAT thunder? Ageia had what, 5 games that used their PPU? It was a useless addition back then, and still is.

You talk like AGEIA isn't around anymore, and no games are commming out with the PhysX API any more?
BTW, the list is a bit longer than 5 games ;)
 
edit: and in regards to that steam survey, this is a shocking number (to me at least)

[name]................................[total #].......[%]
NVIDIA GeForce 6600..........75,541........6.91
NVIDIA GeForce 7600..........71,429........6.54
NVIDIA GeForce FX 5200.....51,612.........4.72
NVIDIA GeForce 8800.........49,850........4.56 %
Not really, the 8800 is a high end card which sells substantially less than the mainstream selection. People forget that what we consider normal hardware is actually from the enthusiast selection, the 8600GT and upcoming 8800GT will be the high performance cards for a good portion of the market, while many gamers are happy to turn down the res on their older cards.

This is why scalability is very important, if you game requires the best of today's hardware to run then only a small percentage can play it.

Where can I get that demo?!?! Very interested :)
Just go to ageia's website and grab the latest drivers, they should be in the control pannel.
 
LOL physics is such a waste of money look at all the jackasses who buy the cards so they can take advantage of demos and like 1 game. By the time it's really put into use their cards are going to be outdated and just wait I'm sure you'll see it integrated into video cards. Much more cost effective than SLI anyways.
 
I was looking around and apparently Crysis is going to use their own physics engine, which will scale according to cpu's, so more cpu's = better physics FOR THIS GAME. I'm not trying to say anything in general. But if dev's keep developing physics on the cpu, I don't see much future for PhysX or 3rd slot physics in general.
 
LOL physics is such a waste of money look at all the jackasses who buy the cards so they can take advantage of demos and like 1 game. By the time it's really put into use their cards are going to be outdated and just wait I'm sure you'll see it integrated into video cards. Much more cost effective than SLI anyways.

jackass's is a little extreme dont you think?

Agiea really cant seem to shake this belief. There are multiple games that support agiea physics. You can find the list if you were so inclined, go educate yourself kiddo.

your sure its going to be integrated onto a video card eh? do you work for nvidia or ati? You are partially correct, both ati and nvidia have found that there general purpose shaders are suprisingly good at general calculations. No where near powerful enough to power a real API, but certainly better at cranking the collision detection and reaction calculations better then anyone thought they would be. And as you may recall, at the 8800 launch, they stated that the G80 will support some form of physics.

in regards to sli, i would have to agree, because of the drivers in their current state. the SLI drivers need a real rebuild. Something that Nvidia seemed to have in mind with SLI2.0 but, with vista on there plate, I guess they figured it was just too much.
 
Nor does it take into account density, heat, any form of sound... it's only got a really basic fluid dynamics sub program... overall I was un impressed.

Oh come on, be realistic.
 
Back
Top