By now, I'm sure most of you are familiar with Ageia's new physics processor, the PhysX. But do you really feel that it's pros outweigh spending so much on it?
No, it's not worth it. Unless you're one of the extremely rare breed of people who can afford the best of everything and are happy too, a physics card just ain't worth the dosh. There are many things that money can be better spent on, and considering the extremely limited number of applications that'll actually work with the thing, I really think you'd have to be very wealthy to even bother, because frankly imo for now it's a waste of good paper.
In the future won't the CPU be doing the physics processing as it will have a spare core or two, and current games don't even support the PhysX so in my opinion it is rather pointless.
We don't know for sure that other CPU's will be doing the work in future KingDavies. That fact was argued recently by one of the bit-tech staff(which one it was escapes me). They argued that no co-processor had ever survived on a long term basis. I partially disagree with this, since graphics cards have been with us for around 10 years, and they're likely to be with us in 5 still. Regardless though, untill we can make processors with at least 16 CPU's I don't think we're going to start asking them to do dedicated physics or graphics processing. What a graphics card can make short work off, a CPU would have a hard time chugging through. CPU's are good at lots of basic maths, ask them to do vertex shading and they get a bit tetchy though Whether using CPU's in place of physics or graphics cards will be favourable remains to be seen. I myself feel that for a good while at least 1 graphics card would be preferable to 8 extra cores doing the same work. It just strikes me as more efficient, money and energy wise. Of course if the processor becomes advanced enough then maybe it'll be able to efficiently handle graphics and physics in which case everything will fall into place. But don't write of the physics chips already. They're mostly a waste of money for now, but who knows what the next couple of years will bring?
While you're more than likely correct in assuming that the CPU or GPU will handle physics in the future, however there are games that actively support the PhysX. One of them would be the quite well-known Ghost Recon: Advanced Warfighter (GRAW,) along with City Of Villains and Bet On Soldier: Bloodsport. I'm sure you could look up a list on the games that will support the PhysX on Wikipedia or Google, but there is actually a rather surprising amount of games that support the PhysX that are soon to be released. I consider purchasing the PhysX as soon as Unreal Tournament 2007 is released, and as soon as the price drops (like everyone else is, I'm sure.) [Edit:] Actually, now that you mention it, NVIDIA and ATI have publicly announced that, through their dual-GPU setups, the second core can act as a makeshift physics processor. NVIDIA's most recent patch, 91.31, is able to do such a thing. Whether or not this will run Ageia out of the market, however, remains to be decided.
I can see physX simply licensing their tech to (ati/nvidia/xxxx) and having them included on newer cards, in the same way 3d cards eventually got eaten
No no, not wrong. Just a different opinion - I'm not correcting you, just disagreeing about what I think the future holds, and putting forth my own opinion. As for the dual GPU solution, from what Brett Thomas, one of the bit-tech writers has said, that's not doing it properly. I can't remember the specifics, but it was something about simulating physics using an engine, and not actually calculating the proper physics of the games, IIRC. Regardless, that's a marketing thing, it may sell, but it's not really any technological progression.
I never actually expected NVIDIA or ATI's solution to match or out-perform a dedicated physics processor, but I never expected substituted physics algorithms. I haven't exactly seen a boost or reduction in performance from what I can see on my SLI setup since I've installed the 91.31 patch, nor have I encountered ay sort of options regarding the makeshift physics processing. While I do predict that the physics processor will go the way of the 3D cards (integration or otherwise,) I don't see it happening anytime soon. I'm sure that companies will play with the gaming community for a while by making better, faster, more complex PPU designs, until they feel like suckering more money out of us for G/PPU cards.
IMO PhysX cards are pointless, true, there are a number of games that support it (still very few though) but when you look at what you actually gain from having the card, it's not really that impressive, even the amazing physics demo FPS that came with the card was hacked to run perfectly fine on a system without the card with little in the way of performance, or eye candy decrease. Also, given that the CPU is a maths processor for number crunching, would this be perfectly suited for physics interactions. The eye candy we see go flying everywhere is just basic mechanics really, all it is is just a load of vector maths. Or am I missing something here?
I'm sure there's a number of machines that have disproved the PhysX's usefulness, but what sort of system requirements would one need to see littlee or no hinderance on gameplay in a physics-heavy situation, otherwise handled by the PhysX? I heard that the demo (the name eludes me as of now) was hacked to run without a PhysX, but I was never given any information on the PCs that ran the hacked demos.
I don't think they were incredibly powerful, if you've got enough money to buy a PPU card then you'll probably have enough to buy a machine capable of handling that demo game without it. My system below might be able to handle it, a new conroe rig would probably have little problems. You have to bear in mind also, the extra effects made by the PPU (even though they are a little anticlimatic for a £200 add-on card) puts extra load on you're GPU, so you really need to have your computer up to scratch anyway. Bit did an article on what the PhysX card adds to GRAW, the results wern't really that impressive, all you get is extra particles, and existing particles reacted using physics, not just an animation. IMO, thats not worth the money they want for the card.
For the price it costs to buy the PhysX now, and the amount of games it currently supports, it would be impossible to get your money's worth out of it. When more games start to flow out, probably around the fall or winter seasons, we may see improvements on the card's design and improvement on system performance. Honestly, the only reason I'd consider purchasing it is for Unreal Tournament 2007, which looks like it'd need two 7950 GX2's in Quad-SLI to run smoothly.
lol, you will if you add all the eye candy a physX processor adds, AFAIK, the PPU only makes the game harder for your system to run, since most of the work is done by the GPU these days, a physX processor only adds to work it needs to do and takes a bit of load off the CPU that it could probably deal with anyway.
Well UT engines are known for their amazing compatability with older hardware, thus I really do not think that you'll need amazing hardware for UT 2007. Their use of shaders to create stunning environments with considerably small polygon count would probably just require a decent CPU and a good video card. For now and for 2007 I think Aegia Physx (fu*** this flickr and razr spelling crap) will not be worth a damn. I'd maybe buy it for $50... It is the future, and it will be awesome, but right now all it does is add some nice flapping flags, few more flying tires and more "stuff" coming out from the trees when you shoot them. Game developers need to take Ageia seriously, but to take them seriously, gamers need to have Physx card. In order for gamers to have these cards, either the price has to be $100 or below, OR there has to be a huge advantage to having one. In order to have huge advantage, the game developers have to make games that support these cards. So it's a loop. And it will take time. But when it becomes popular among gamers and developers, we'll see some pretty cool sh*t.
I suppose you'd need something powerful to run the game well even if you didn't have the PhysX. The average polygon count of player models in UT2007 is approximately 2,000,000. That aside, you're probably right about it adding more objects to render -- but if you've got a card that can handle it, why not?
I thought I read somewhere that Havok (Company providing Physics for all the next-gen consoles and already have lots of PC games such as HL2) had an agreement with nVidia to develop a physics processor on the graphics card itself. Can anyone jusitfy spending £200 on a something that makes an explosion look prettier? I'd want Smell-o-vision on my PC for that price. (Somebody will now make a Smell-o-vision PCI Card )
Havok! That's the name. Anyhow, I never heard that NVIDIA and Havok would team together to create a G/PPU card. I know NVIDIA has tried / is trying to create a physics processor by using an SLI setup's additional GPU, but like previously mentioned, it uses substituted and static physics alrogrithms. Would you happen to have the link to the article you read? I can agree with this. If Ageia plans on grabbing the gaming market by the neck, they need to appeal to a larger portion of it. While I find Ageia's premature release to be foolish, I still believe that the PhysX will remain as it gains more support both from game developers and gamers themselves. In order for larger gamer support, though, prices need to be lowered. Not everyone can afford to spend $250, and there's little sense in doing so at it's current stage. Down the road, though, who knows?
Especially when we already need to fork out a few hundred for a single graphics card, sure if it was about £80ish I might consider it. We also have the problem with the fact there are very few motherboards that can support SLI or crossfire, a sound card like an X-fi AND a physX card, there just simply isn't enough room on most boards to squeeze in the high end equipment. Hopefully this will change though as mobo companies realise thier boards are a bit pokey for the bleeding edge enthusiasts.
