This I must disagree with. There is an enormous amount of bandwidth between the video card (or any pci-e device) and the northbridge. The bandwidth inefficiencies that people talk about aren't intrinsic to the hardware itself, but rather DirectX. Processing overhead is going to be reduced dramatically in DX10, and people assume this is because of the new DX10 spec hardware, but it is mostly to do with DX.Pyromaneyakk said:
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
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.
Bye bye.. Ageia :)
- Thread starter Flow
- Start date
Unperson
[H]ard|Gawd
- Joined
- Sep 2, 2000
- Messages
- 1,826
jimmyb said:
Why not use OpenGL or somethign else if it's specific to DX?
EDIT: To elbaorate more, I believe that the issue is actually trying to get the data from the shader engine back to the bus efficiently. As of right now I think, and someone please correct me if I'm wrong, the only way to do so is to send the data to the frame buffer and essentially send it back as a screen-shot. Future cards will have a more direct path but current cards to not.
OpenGL has less overhead than DX. Some of the changes in DX10 (like userland drivers) that reduce cpu overhead already exist in opengl.
Note I'm not trying to say that ogl is better or anything like that, but that the inefficiencies are largely the fault of graphics library, not the hardware.
To answer your question, despite its shortcomings, DX is a very good library, and also Microsoft has done a wonderful job of marketing it.
In the case of HavokFX, it can run via glsl, my guess is more efficiently than with hlsl. But that's just complete speculation.
Note I'm not trying to say that ogl is better or anything like that, but that the inefficiencies are largely the fault of graphics library, not the hardware.
To answer your question, despite its shortcomings, DX is a very good library, and also Microsoft has done a wonderful job of marketing it.
In the case of HavokFX, it can run via glsl, my guess is more efficiently than with hlsl. But that's just complete speculation.
Unperson
[H]ard|Gawd
- Joined
- Sep 2, 2000
- Messages
- 1,826
Eulogy said:
How about making video cards modular like motherboards? You get 2 processor slots, and several RAM slots. If you want SLI GPU, you can put in 2 video chips. If you want video and physics, put in one video and one physics chip.
Actually I wonder if it would technically be possible to do something like this on a multi-chip Opteron board? Make a vector co-processor or PPU with Socket 940 / Socket F. The co-processor would use the CPU's memory bank as its own memory, and communicate with the rest of the system via the CPU's hyper-transport link. I know K8L is supposed to have vector co-processor support, should be interesting to see what that brings to the market too.
Pyromaneyakk said:
Even better, how about modular motherboards? A mobo would have around six universal sockets with a memory bank for each. Want more proessing power? Add another CPU and give it a gigabyte of memory. Want ultra-fast gaming performance? Add two GPUs and give them each 512MB of memory. Want realistic sound? Add a DSP and 128MB of memory. Want hardware-accelerated physics? Add a PPU and give it 256MB of memory.
Other things that could fill the sockets include: AIPUs, VPUs, raytracers, video-encoding chips, RAID controllers, NIC processors, and folding chips.
Now THAT would be the ultimate system!
nobody_here
[H]F Junkie
- Joined
- Feb 1, 2006
- Messages
- 8,877
EVIL-SCOTSMAN said:
yeah, but the extra GPU can do physics OR graphics, the PPU can only do Physics, there are what.....2 games supporting Aegia's hardware, and then you need a second GPU to be able to keep performance levels up to where they were when running a single card without the PPU, so with the Aegia card, you still need a second GPU to make it any fun to have, may as well have two GPU's, be able to do Physics (albeit to a lesser degree) when wanted, and have superb framerates in non-physics enhanced games with SLI
i hope Aegia does the right thing and sells out to NV or ATI, at least that way they will get something out of the deal and the technology they developed will go to good use, because as a standalone company, they are going to fail
nobody_here said:
30 by my count. An ever growing list: http://www.ageia.com/physx/titles.html
Ageia have taken the initiative to get this new technology up and running and seem to be doing well, having some big names such as UT2007 adopt their technology and 30 games already.
jimmyb said:
I cannot see any company making PPU hardware not supporting a standard set forth in DirectX or an extension of DirectX, it would be like sound card makers not supporting direct sound, it's just so important and intergrated they don't really have the choice.
That said if Microsoft were to come forth with an API specifically designed for handling physics then we should see all major hardware players adopt support for it, and then we have 1 API to code for and people making games can stick to that.
What the consumers want is one product that does everything, we don't want to have to purchas an Agiea PPU to run their physics engine, and then also buy some Nvidia dedicated hardware to accelerate Havock physics, that's just stupid.
The aim of all the companies producing PPU's is to corner the market and make their supported API the one thats dominant, so if they pull it off they have total control over the market, thats no good to the gamers if there's several major players all batteling it out for 4-5 years until all but one gives up.
That is the official list. There's much more though... Here's my unofficial list of developer houses, games, engines, publishers, tools & middleware plus a couple of non-categorized ones too:Matrix said:
http://personal.inet.fi/atk/kjh2348fs/ageia_physx.html#gameslist
Quite a bit more stuff there
Since this thread has completely been derailed again (people are talking about MS DirectX physics API), I figure that I should just chime in:
I hope that MS does NOT make a DX physics API, since I'd much rather see some nice, OS independent API, like OpenGL be used for physics. After all, Microsoft isn't the only company in the world producing software and finally getting some third party involvement would be nice.
I hope that MS does NOT make a DX physics API, since I'd much rather see some nice, OS independent API, like OpenGL be used for physics. After all, Microsoft isn't the only company in the world producing software and finally getting some third party involvement would be nice.
nobody_here said:
This is the worst argument I have ever seen. It takes all of the worst arguments and puts them together. Let's see, where should I begin...
First off, there are what...0 games supporting ATI/NVIDIA's hardware?
Second, there are more than what...2 games supporting AGEIA's hardware (over 30, as someone pointed out).
Third, the extra GPU can do graphics and physics at the expense of graphics performance. Do you really think that an SLI setup is going to be able to do graphics and physics at the same level of performance as an SLI setup doing just graphics? When you use processing power for physics, you take away processing power for graphics. Sorry, but you can't get something for nothing.
Fourth, "and then you need a second GPU to be able to keep performance levels up to where they were when running a single card without the PPU" is the most incorrect statment I've ever seen on this forum! Yeah, you lose a few frames per second. You don't have to buy a second GPU to have an enjoyable gaming experience. Also, GPU physics are going to have the same problem. The extra 1/2 of a GPU that you get for graphics is going to help alleviate the problem, but you are going to pay for that, too.
Microsoft has been the king of standards for years. Things have worked out very well that way. Everyone can be assured that two products will be compatible. If someone learns how to use Windows and Office, they know how to use virtually any other computer they will ever come into contact with. We can be assured that our three main standards (DX, Windows, and Office) will always work nicely together.
I think it's best if we keep things that way. Third-party APIs simply add overhead and raise prices for GPU/PPU developers, since they have to include support for all of the APIs rather than just one. A third-party API would not work together with DirectX or Windows, therefore nullifying the possibility of intertwining graphics and physics processing (i.e. when the GPU isn't being used for graphics, it can be used for physics).It doesn't have to be OS-independent, since virtually everyone (and by "virtually" I mean 99.999% plus one) games with Windows. There is no need to waste time and money supporting dead technologies. Standards are good - they make life easier.
Microsoft standards are only standards on Microsoft platforms. I agree that it would be beneficial for a third party to produce a proper open physics library (or pure spec like OGL). OpenGL is a perfect example of this working well (although perhaps not well enough...).
As always, I know a lot of you don't care about this, but it certainly matters to people like myself who aren't always working on a Windows system.
Oh, and I can't help but respond to this:
As always, I know a lot of you don't care about this, but it certainly matters to people like myself who aren't always working on a Windows system.
Oh, and I can't help but respond to this:
Supporting a non-MS api is no more expensive than supporting a MS api. Greater product support also means more sales. If either Ati or Nvidia dropped support for OGL they would lose an enormous amount of money. I think your comment is totally absurd.HOCP4ME said:
InorganicMatter
[H]F Junkie
- Joined
- Oct 19, 2004
- Messages
- 15,461
Sorry, but have gauranteed hardware support and cheap API implementation for developers is much more important than supporting the dozen Linux users or three Mac users out there. I hope MS does get a good API out the door quick, this hardware war is looking to get expensive!drizzt81 said:
There are some good reasons that platform portability would be good from a purely profit perspective (aside from the obvious altruistic reasons). Console development is certainly much easier when not locked into one platform. Not to mention there would presumably be significant demand for a general purpose vector processor on workstations of varying types.
TekSomniaK
[H]ard|Gawd
- Joined
- Mar 26, 2003
- Messages
- 1,144
TheGamerZ said:
I'm talking about for use in games that aren't 6 years old. Integrated cards are pieces of shit intended for your grandma
hybrid.09 said:That's great news. I feel sorry for the people who bought Ageia PhysX cards
All but a few that is right?
Now...as for using a spare graphics card for physics, sounds like a good idea because by the time I buy a R600, my x1900xtx will be fair game for physics use...and a tad faster than any ageia card could ever be.
BBA said:All but a few that is right?
Now...as for using a spare graphics card for physics, sounds like a good idea because by the time I buy a R600, my x1900xtx will be fair game for physics use...and a tad faster than any ageia card could ever be.
While I can't comment on the speed, I can say that I'm sure this is how most people feel about using a graphics card to accelerate physics. Nobody wants to get rid of an old card, all of a sudden something comes along giving us something to do with our old stuff once DX10 gets here. Sounds good to me, and to most.
Whether this provides the optimal solution or not is really not at issue here. If nvidia/ati want to create a large user-base for this concept(and they clearly do), they can do this quickly by providing something a lot of people can access quickly and without spending extra money. It doesn't matter much that the physics seen on screen are "effects physics"(when it comes to the spare gpu doing the calculations), what matters is game developers are putting "physics enhanced" games into the hands of gamers.
EVIL-SCOTSMAN said:
EXACTLY! Now can all the ignorant trolls who obviously know nothing about software-hardware interaction please go take some CS classes?
Regarding a GPU doing physics, why use a shovel when you're raking leaves? Why use a rake when your trying to dig a hole? Thanks, but no thanks. I'd rather keep my FPS high and have better physics interactions. And dedicating a graphics card for physics processing, intersting idea, but physics cards are significantly less complex then today's graphics cards. If the market gets large enough for PPUs you would almost certainlly be better off selling your old graphics card on eBay, buying a PPU and pocketing the difference. Unless of course you have a very bad graphics card, in which case your SOL anyway and probably don't post on these forums.
There's also a tinge of irony here. One of the primary arguments of the anti-physics crowd has been that rendering all the extra objects will cause too large a drop in FPS. And now these same people are lauding taking away graphics processing power directly from the GPU and dedicating it to physics. Odd.
Also, Ageia does not need a very large market to survive. Creative X-Fi/Audigy cards comprise less then 5% of the gaming market, yet they are still a huge name. I predict the same for Ageia.
InorganicMatter said:
That's what I'm saying. Gaming on platforms other than Windows is dead and buried. Apple's OS survives only because they can convince some grandmother buying her first computer to buy non-standard software because it is "more secure" without mentioning that it is incompatible with 90% of the software and hardware out there (okay, that was a bit of an exaggeration, but I'm sure it is a pretty high percentage). And as for Linux...who games on Linux?
I am really, really, really tired of hearing this argument from people who don't think about what they are actually saying. There has to be only one OS!!! Can you imagine how it would be if multiple OSs were popular? I can think of two scenarios:
1. Third-parties must make all of their hardware and software compatible with both OSs. They need to hire extra programmers to port it over. Hardware developers have to write two sets of drivers and include extra chips on the cards. Whenever there is an OS update, software developers must issue two patches - one for each OS. The price of everything goes up, and companies have less time and money for R&D and instead use it for porting. Of course, products from the OS companies would never be compatible. Make a text document on your computer, and your friend won't be able to read it.
2. Third-parties don't make their hardware and software compatible with both OSs. Files you create at work won't read on your home PC. You can only play half of the selection of games out there. Partnerships begin to form. ATI cards will only work on one OS, and NVIDIA cards will only work on the other. Intel CPUs work better with one OS, while AMD chips work better with another. One OS will only play blu-ray, and the other will only play HD-DVD, keeping the format war alive.
Or, we can keep it the way it is now, and:
3. You can buy a piece of hardware or software knowing that it will work with your PC. Companies can agree with each other. Costs are lower since hardware and software only has to be coded for one OS. Software becomes more efficient since developers can take advantage of all feature of the standard OS without worrying about those features not being supported by a competing OS. More competition arises in the software and hardware departments, since it is easier for new companies to make profit when they only have to support one OS. You know that all files created on any computer will be readable by any other computer, no matter where you go. In exchange for this, we must settle for increased security hole exploitation, and an OS that has a few bugs and isn't the best product possible, but is at least stable and gets the job done.
Everyone automatically assumes that MS is bad because it is big. But they don't use anti-competitive practices, their product is good, the prices are reasonable, and they ensure that everything works together as it should. What exactly is the problem?
Yes, I was thinking the same thing. Imagine that, for simplicity purposes, a GPU and PPU have raw power measurments in equal units:
Option 1: You have a GPU with 200 units and a PPU with 100 units. Using the PPU's units for physics automatically takes away 50 units from the GPU for rendering. That leaves you with:
Graphics: 150 units
Physics: 100 units
Option 2: You have a GPU with 200 units. To match the performance on the PPU, you use 100 of its units for physics. This automatically takes away 50 additional units for rendering (yeah, you still have to do the rendering regardless of how the physics are done). You are left with:
Graphics: 50 units
Physics: 100 units
Option 1 looks better to me. Also, you have your SLI choices:
Option 1: You have two GPUs with 200 units each, and one PPU with 100 units. Using the PPU automatically removes 50 units from you GPUs for rendering. You have:
Graphics: 350 units
Physics: 100 units
Option 2: You have two GPUs with 200 units each. You use 100 for physics and 50 for physics rendering. You are left with:
Graphics: 250 units
Physics: 100 units
Bottom line: Using the GPU for physics takes more processing power away from the GPU than using a PPU for physics.
BBA said:
That works under two conditions:
1. You have an old GPU (i.e. you don't sell, reuse, or give away your old hardware)
2. NVIDIA's solution allows you to use last-gen cards for physics (can you even do that with DX9 cards?)
In that case, I can see NVIDIA's solution being better if you have an old card. Otherwise, the PPU still seems like a better idea.
- Joined
- Jan 12, 2003
- Messages
- 687
More interesting is the rumoured but as yet unconfirmed Quantum Effects engine, which appears to be Nvidia's pitch for physics processing. Quantum Effects essentially leverages the G80 shaders' processing power for physics calculations, much as PureVideo uses today and tomorrow's GPU shaders for video processing.
Quantum Effects will, we'd guess, tie in to Havok FX's physics API and to Microsoft's anticipated DirectX physics API.
Hints that Nvidia is working on a physics processor of some kind emerged last month when Asus let slip the company's plan in what's presumably a reference to the upcoming 8800s. ®
http://www.reghardware.co.uk/2006/10/05/nvidia_geforce_quantum/
Quantum Effects will, we'd guess, tie in to Havok FX's physics API and to Microsoft's anticipated DirectX physics API.
Hints that Nvidia is working on a physics processor of some kind emerged last month when Asus let slip the company's plan in what's presumably a reference to the upcoming 8800s. ®
http://www.reghardware.co.uk/2006/10/05/nvidia_geforce_quantum/
blitzcraig said:
Isn't that what GPU physics is all about? Using extra graphics headroom (if there is any) for physics?
Is this going to require unified shaders to work? If so, then it isn't going to work with your old cards, at least not first-gen.
Obi_Kwiet said:
You'll need a G80, in other words a Geforce 8 Series GPU, to use the quantum physics engine. So sure, if you happen to right now haven an extra Geforce 8800GTS gathering dust you would be good to go. Let's just forget that inconvienent fact that a 8800 series card costs about three times more then a PPU, and you would always be better selling that extra Geforce 8 series card on eBay, buying a PPU and pocketing the difference.
And everyone who for some reason thinks Nvidia will add features (such as physics acceleration) to cards which they already have on the market? How naïve are you? Nvidia is a for profit comany, and you don't make a profit by spending millions developing features for a product already on the market.
But we wouldn't want silly things like economics or common sense to get in the way of marketing hype, would we?
In reply to a comment earlier regarding whether nvidia "allows you to use last-gen cards for physics", Havok FX interacts with hardware via glsl and hlsl (sm3 compliant), so it is not really up to nvidia what cards are supported or not (unless they have some sort of agreement with Havok where they will specifically disable it).
In regards to,
edit: Am I the only one who finds the name "Quantum Physics engine" a little misleading when in all likelyhood it probably only does Newtonian physics? This seems like a really bad (buzz) word to use of all places.
In regards to,
This is a naive comment. Developing competing products to take market share (and ultimately profit) is a fundamental aspect to the free market. Notice how Ford is not the only automobile manufacturer.Hvatum said:
edit: Am I the only one who finds the name "Quantum Physics engine" a little misleading when in all likelyhood it probably only does Newtonian physics? This seems like a really bad (buzz) word to use of all places.
so, what can a physix card do for me now? i've seen small proof of concept (is that right?) demos, and now upcoming games that just dedicate a core on kentsfield cpus to handle all the physics.
does this mean the card is not really useful to you if you have a quad core cpu, because one of the cores can handle all the physics? or are there still benefits above and beyond what a cpu can render?
does this mean the card is not really useful to you if you have a quad core cpu, because one of the cores can handle all the physics? or are there still benefits above and beyond what a cpu can render?
iroc409 said:
You look like you're just asking a question, so I will answer it:
Right now, there is not much reason to buy a PPU. A few games are out that make limited use of it, but they aren't considered to be very good by most of the community here.
Over 100 games are upcoming for it. Most will not utillize it to its full potential, but UT2007 and CF:R are a few notable exceptions. Still, you shouldn't buy a PPU right now. By the time those games come out, the price may have gone down, or a new model may be out that is more powerful.
NVIDIA and ATI are developing a technology that uses an extra GPU (or an existing one, at the expense of graphics performance of course) for physics. This technology may end up winning in the long run, which is another reason why you should not buy a PPU right now.
Havok is developing something that will uitllize extra CPU cores for physics. It will not replace PPU or GPU physics, however. Extra CPU cores still cannot come close to the power of a PPU or GPU for physics. You would probably need at least 16 cores dedicated to physics to equal the PPU. That estimate does not take into account the overhead that is caused when you must get all of those cores to work together, either.
If you are an early-adopter of technology, or want to support new technology to help it be utillized faster, go ahead and buy one. After all, if everyone went out and bought one, it would be utillized within a few months. Unfortunately, its usually the developers who have to act before the consumers will buy.
Hvatum said:
Slow down there, Turbo.
People have said that being able to use an old card would be *nice*, but not that it's the only reason the option is more attractive.
If I'm looking at getting a new video card AND a ppu in the future(say, after the G80 is available and processing physics in games), here are my options:
1) Buy one graphics card and one physics card and be stuck with a specific level of performance.
2) Buy two graphics cards, leaving me with a versatile setup.
The fact is simple; Aegia's PPU will NEVER be able to enhance my graphics performance, while nvidia's G80 appears very likely to be able to enhance both graphics and physics. Why buy a product that can only do one thing when you can buy a product that does both?
Since there isn't any reasonable way to compare the performance of Aegia's option against any of nvidia or ati's solutions yet, it's really difficult to come to a concrete "The ppu is better" conclusion. If a video card can process physics just as well as a PPU, why on earth would you buy a PPU instead of a video card? (Unless you already have two cards in sli, a third video card would be somewhat pointless unless it offered better performance than aegia's option)
EQTakeOffense said:
I know today's kinda different from 1998, but eight yrs ago ppl were doing well over that $300 mark just so they could play Q2/HL/Unreal with all the bells and whistles...
Besides, I thought the reason they were saying you could use a third GPU for a PPU was so that ppl who are upgrading their vid cards don't have to toss/Ebay their old GPU, it can now just become their PPU?
(Nvidia or ATI said that anyway...)
which wouldn't be unheard of: see SLi and Intel board.jimmyb said:
jimmyb said:
Tell me why nVidia would want to enhance the value of the last gen cards, unless it is sitting on stockpiles of unsold GPUs. Their interest is to sell as many as possible high-end cards. Yes, if ATi was adding backwards support, then nVidia may be forced to as well, but there is not certainty about whether it even makes sense -from a computational standpoint- to enable old(er) cards.
Also, let's take a page from the Intel Performance/ watt playbook. IIRC the PhysX card takes somewhere in the "low double digits" in terms of watts. A 8800 GTX will most certainly require triple digits.
no you are not the only one.
legendz411
2[H]4U
- Joined
- Jul 3, 2006
- Messages
- 3,534
jimmyb said:
Lol. Glad to see I wasn't the only one thinking this