Plea of a Lurker: Don't let Physics Cards Fail

[velis]

The issue of PPUs is a bit more complex as it seems at first glance. New games will have more and more physics and the way i see it, making it possible to use your gfx card to offload some of the physics calculations will help improve the frame rate despite gfx card already being maxed out. So is the CPU, mind you. Of course, it won't be as powerful as a dedicated PPU, but it will help a bit.
So the ideal scaling would be CPU -> gfx -> PPU where the frame rate would go up from one solution to the next. The problem however is that there is no standard for physics API. So you can't just write drivers for some hardware (gfx or dedicated PPU) that would instantly increase the frame rate when applied. I can only see one player that's capable of doing that (MS), but I fear the resulting salvo of patents instantly pushing out open source libraries. Likewise for linux game development which IMHO is lagging for the same reason (you just can't port directx to Linux).

In the end all of this is possible ONLY if the majort players finally agree on some standard (and hopefully open) API. Otherwise we'll just have games that play mediocre using our gfx cards and games that play great with dedicated PPUs, but both camps will play miserably when their condition (their HW present) is not met.

To sum it all up: I'm all for physics on gfx shaders if that improves my frame rate. But I want to have a choice to plug in a REAL PPU card to increase my performance further. And I'm talking about my favourite game here. I don't want to switch games just to get a few FPS extra. I wonder if Oblivion is done on Novodex?

 

[the Chase]

Thats what I'm wondering too. BFG is supposed to be releasing their PPU card in March. I would think they would want a big title game that uses the thing to make a slash for the release. The only big title I can think of is Oblivion! I guess we will find out soon..(Unless they delay the launch of the card again...)

 

[letdown427]

Originally posted by: the Chase
Wow that was the most eloquent story the Inq has ever posted!! You could actually understand all of it! What happened??

QFT. I was amazed at the quality of that article(for the inquirer). So much so that I fully believe what they're saying, as it makes coherent sense, and it was put so nicely! lol Wow

I'd suggest putting it in the OP. In fact, link to it in the OP of any thread that brings up GPU physics processing.

 

[aka1nas]

Originally posted by: the Chase
Thats what I'm wondering too. BFG is supposed to be releasing their PPU card in March. I would think they would want a big title game that uses the thing to make a slash for the release. The only big title I can think of is Oblivion! I guess we will find out soon..(Unless they delay the launch of the card again...)

Nope, Oblivion uses Havok. The next Unreal game will support PhysX (and might be the reason the card hasn't launched as neither is out yet.).

 

[jiffylube1024]

My two cents is that the increased cost in video cards (and new products every 6 months), combined with the need for things like PPU's now (potentially) are making consoles look more and more attractive. Cost to run a top of the line gaming machine for 5 years is getting ridiculous.

 

[biostud]

Originally posted by: aka1nas
Ultimately, MS needs to add a physics API of some sort to a future version of DirectX, and then game developers can just code to that, and then Havok or Ageia or Nvidia or whoever can just write drivers to offload that code to the GPU or a specialized card.

isn't PhysX an API?, just like Glide from 3Dfx

 

[Drayvn]

Wasnt Ageia getting card vendors and motherboard makers to put their chip on their card.

And also Ageia have 3 different types of PPU, a low, mid, and high end card also?

From what i can remember from long old news is that the mid PPU is being thought of for the motherboard vendors and the likes of ASUS are already partnered with Ageia to utilise this, and also as some of the posters have said Unreal Engine 3 has been said to support the PhysX architecture, though i dont know if this translates into it being able to use the PPU. If i can remember again the PhysX is just like Havok a software based physics engine. But what i cannot remember is if this can be used in conjunction with the PPU when and if it gets released.

I hope the PPU does make it to market and hope it does well. Again it means one more expensive product to buy for only 1 sole reason which is gaming which could be its true downfall as its a very small market compared to consoles. Now again if it can make its way into consoles, then i think it could work out much better and the PC users will get the benefits of this.

The specs on the high end PPU is that it can do up to 50,000 physics objects every second. Which CPUs can only handle a hundred or so objects and they cant be to complex. What i want to know what is the limit of the GPUs power on how many objects it can handle and can this be increased or decreased on the game side or control panel side so the user can balance how much he wants between realism of the graphics or gameplay.

 

[aka1nas]

Originally posted by: biostud

Originally posted by: aka1nas
Ultimately, MS needs to add a physics API of some sort to a future version of DirectX, and then game developers can just code to that, and then Havok or Ageia or Nvidia or whoever can just write drivers to offload that code to the GPU or a specialized card.

isn't PhysX an API?, just like Glide from 3Dfx

Yes, with exactly the same problems in that Ageia's competitors will not want to develop solutions that use that API.

 

[Oemo]

Ageia offers physics support for Cell Playstation 3 chip

Support for consols? Very interesting. The physics saga continues

 

[Nanobaud]

The following is an honest question, not a poorly-disguised rebuttal as is more often the case on this board, but...

What makes the GPU physics a 'stopgap' solution spawned of marketing hype. I know little about the Ageia PPU (certainly sounds cool, but I have yet to get around to looking into it), but I am familiar with the (NVidia) GPU processing capabilities in non-GPU applications, and it is certainly no slouch in the kinds of algorithms needed for physics calculations. There is a little logic, and some firmware (that can be changed within certain limits) that would probably be useless for game physics, but not that much of a fraction - so I find it difficult to believe a more capable PPU would be somehow much cheaper than a GPU being produced in larger volumes. I would guess (admittedly not being familiar with the specific intents of the PPU movement) that with a few die changes (and maybe without, but with clever programming) a GPU would make a fine PPU. Unless there is something substantially more capable in the physical architecture of solutions like the Ageia, the 'rejecting' the marketing hype of GPU promoters is no more enlightened than 'accepting' the marketing hype of the PPU promoters (and if it needs to fall back on something like cell vs cpu, that's the same cop-out). If one is going to purchase two coprocessors, I could certainly forsee the better solution being a single board with two identical processors, possibly connected by a shared memory and/or dedicated on-card bus as well as the PCIe bus, each processor equally capable of GPU or PPU duties on demand. If your gaming already requires full use of 2 GPU's, then you would have been considering 3 co-processors and mulitiple PCIe-G slots anyway, so just extend the reasoning.

I am sure someone here has looked at the proposed Ageia/PPU architectures. If you can summarize why you think the PPU is a fundamentally a better bang-for-the-buck approach, or why it isn't, please post. I am honestly interested.

If there isn't a fundamental difference, perhaps Ageias greatest contribution will be to get the Gorillas to pay attention to some important things they have been neglecting. Unfortunatley, they would not get much financial reward for that.


nBd

 

[bunnyfubbles]

AGEIA was too late to the table, had they come out with the PPU earlier it would have been inside one of our nextgen consoles. With such a major advantage, consoles would actually have a legit chance to inflict much damage to PC gaming. The only practical response would be for PC gaming to adopt PPU support. AGIEA needs to keep it up, if they can survive long enough to get a product into a next gen (the next-next, most likely not this coming gen) then we should see it happen. The problem there is that we're 5 years off before that happens.

I completely agree about the PPU being a good thing. Even if GPUs can process physics to some extent, its rediculous for them to partition part of their power off to another task. The CPU can do it all, but obviously that isn't the most efficient way to do it as we now have GPUs when games in past years only required video cards to display to the monitor the graphics processed by the CPU.

 

[lifeblood]

Depending on price, the nvidia solution is not a good alternative to the PPU. It requires an SLI motherboard and at least an (approximitly) $200 6800GT. If it requires matched video cards then you must get the same card as you have now. Those of you who got high end video cards (7800, etc) must get a second one to enable this. And then if you want to upgrade to a new video card you must upgrade both.

I think I would rather get a seperate PPU that's not tied to the video card.

 

[aka1nas]

Originally posted by: bunnyfubbles
AGEIA was too late to the table, had they come out with the PPU earlier it would have been inside one of our nextgen consoles. With such a major advantage, consoles would actually have a legit chance to inflict much damage to PC gaming. The only practical response would be for PC gaming to adopt PPU support. AGIEA needs to keep it up, if they can survive long enough to get a product into a next gen (the next-next, most likely not this coming gen) then we should see it happen. The problem there is that we're 5 years off before that happens.

I completely agree about the PPU being a good thing. Even if GPUs can process physics to some extent, its rediculous for them to partition part of their power off to another task. The CPU can do it all, but obviously that isn't the most efficient way to do it as we now have GPUs when games in past years only required video cards to display to the monitor the graphics processed by the CPU.

They already are going to have novadex running on the PS3. Keep in mind that it is a physics API that can run in software(and is multithreaded) as well as use the PPU if available.

 

[Munky]

I dont see the need for dedicated separate physics hardware in games at all. Most modern games are already gpu limited at max settings (on a single card), and the cpu should be more than enopugh to make game physics look realistic. In games like FEAR for example, runing on a dual core opteron and a x1900xtx, I see no difference in performance when I switch the physics options from max to min. Moreover, the game doesnt even use the two cpu cores, and the physics in the game are already decent IMO. Taking advantage of 2 cpu cores should open up even more room for improvement. This is games we're talking about - the physics only have to look real, I dont care whether or not it follows an exact mathematical model of the real world. Hardly any games even use a dual core cpu to its full potential, and here they want me to spend hundreds of dollars more on a separate physics card? Next thing you know, they're gonna try and sell me an AI card, because aparently my cpu is also too weak to handle AI calculations... :roll:. The more useless crap they try and sell me for the PC, the more attactive consoles are starting to look as the alternative.

 

[her34]

how can you make any judgements without either version being released (dedicated ppu or shared gpu). there are many variables that can make either version a better solution for consumers.

1) some people seem to expect the gpu to do this for near free (in terms of fps) to be comparable to the ppu solution. that's not a fair comparison. you will be buying a video card regardless. the comparison to make is how you will spend your extra money towards physics. spend $200 on an extra physics card, or spend $200 on a better video card than you were planning on getting (instead of buying a $300 video card and a $200 ppu, buy a $500 video card). spending money on a better video card would be more versatile because you can still use the extra power on games that don't use physics, whereas a physics card will sit idle for some games.

2) we don't know what the performance difference will be between a ppu or gpu solution for physics. maybe it will be negligible. or maybe a ppu card will be capable of certain effects that gpu's simply can't do well or at all. who knows?


either way, i just want to see both of them out. especially the ppu? wasn't it supposed to be out this month or is it delayed a second time?

 

[aka1nas]

Originally posted by: her34
how can you make any judgements without either version being released (dedicated ppu or shared gpu). there are many variables that can make either version a better solution for consumers.

1) some people seem to expect the gpu to do this for near free (in terms of fps) to be comparable to the ppu solution. that's not a fair comparison. you will be buying a video card regardless. the comparison to make is how you will spend your extra money towards physics. spend $200 on an extra physics card, or spend $200 on a better video card than you were planning on getting (instead of buying a $300 video card and a $200 ppu, buy a $500 video card). spending money on a better video card would be more versatile because you can still use the extra power on games that don't use physics, whereas a physics card will sit idle for some games.

2) we don't know what the performance difference will be between a ppu or gpu solution for physics. maybe it will be negligible. or maybe a ppu card will be capable of certain effects that gpu's simply can't do well or at all. who knows?


either way, i just want to see both of them out. especially the ppu? wasn't it supposed to be out this month or is it delayed a second time?

Actually I would rather have a dedicated physics card over a $500 graphics card if the PPU us anything close to what they are claiming it will be. A $500 top-of-the-line video card is generally not much more than 20% faster than the next model down. If games support Novadex or some other physics API in the future then it is quite possible the most cost effective high-performance solution will be a single $200-$300 graphics card and a PPU. Of course we won't know for sure until they have shipping hardware.

 

[nitromullet]

Originally posted by: her34
how can you make any judgements without either version being released (dedicated ppu or shared gpu). there are many variables that can make either version a better solution for consumers.

1) some people seem to expect the gpu to do this for near free (in terms of fps) to be comparable to the ppu solution. that's not a fair comparison. you will be buying a video card regardless. the comparison to make is how you will spend your extra money towards physics. spend $200 on an extra physics card, or spend $200 on a better video card than you were planning on getting (instead of buying a $300 video card and a $200 ppu, buy a $500 video card). spending money on a better video card would be more versatile because you can still use the extra power on games that don't use physics, whereas a physics card will sit idle for some games.

2) we don't know what the performance difference will be between a ppu or gpu solution for physics. maybe it will be negligible. or maybe a ppu card will be capable of certain effects that gpu's simply can't do well or at all. who knows?


either way, i just want to see both of them out. especially the ppu? wasn't it supposed to be out this month or is it delayed a second time?

You really should try reading some of the articles posted before "adding" to the conversation.

Your buy a better video card option is false since SLI Physics requires the whole second card to do the physics work. You either run dual gpu SLI, as you can now, or you run a single graphics card and your second graphcs card as a dedicated physics card.

What we really would need is the abilty to run any NV graphics card as a PPU so you could use your last gen card to do physics and your new card for graphics. It would be nice for past video cards to have more to look forward to than eBay.