Why PPU's over generalized computing?

[trickedoutwizard]

Hello, this is an opinion that is only fairly well informed. How’s that for a convincing opening?

The debate over physics acceleration is an interesting one. The argument has a few different sides. Some wish to make PC gaming inexpensive as possible and feel that an additional add-in card just for games is driving the price of gaming on the personal computer too high. Others think that physics cards will not drastically aid in making games more enjoyable and are therefore not worth the price. Still more feel that, with the advent of dual-core and soon quad-core processors, Physics Processing Units (abbreviated PPU) are an unnecessary step towards a problem that does not yet exist.

I will talk about each of these in turn.

The first line of thought, that computer gaming is already expensive and doesn’t need another add-on card is strange. Nobody really ever forced anyone to buy a graphics card until they were extremely common, and even then the game developers usually had some form of software mode for awhile. If gamers want to buy physics cards to make a game more enjoyable to them, what can really be the complaint? If indeed this trend does catch on then by the time a game requires you to actually own a physics card you’ll likely already have one or be able to purchase one for a low amount of money. I would assume the sweet spot for physics cards will be around the selling price of nice gaming sound cards. It is a similar market, people who want to improve their gaming experience, and don’t feel that graphics should consume their entire budget. The one large difference is that a sound card is necessary outside of gaming, while a physics card is not, this will be talked about later.

The second argument that physics will not drastically add to a games fun level is probably completely wrong. People focus on graphics when it comes to immersion because most of the information we take in is from sight. This does not mean that games could not be more enriched by physics. I for one would love games that require you to use your mind more than ones currently on the market. It seems in shooters all of the best things are skipped over with scripted events and use keys. You plant charges on a bridge where the game allows you too and when you’ve used all the charges the bridge blows up, level over. I would find it much more interesting and fun if a game said, “We need you to blow this bridge up”, and then gave me some charges and let me tackle the problem. Perhaps someone would just plant charges on what parts of the bridge seemed important and it would collapse, maybe a clever gamer would be able to save a charge or so and still ruin the bridge. This sort of scenario could be aided greatly with a PPU that could dynamically handle all the rigid body collisions and other calculations leaving the processor/s free to evaluate whether the bridge is indeed still usable. The above example might prove very difficult for a variety of reasons, but that is something I was able to think of in a number of minutes, imagine if a game designer knew the tech was there and had a few years. This argument is just wrong.

The last point, about multi-core processors and sli-cards that handle more than graphics, is probably one the best arguments against PPUs. But really, this argument is not as it seems. This argument is calling for generalized computing, which is the right direction for the industry. Sony does not do a lot of things I feel are innovative or particularly moral, but they were on the right track with the CELL processor. The only problem is that the entire industry is so focused on graphics cards for graphics and processors for game logic etc... That they don’t realize that multiprocessors would be far more useful. Video card companies have realized this and are starting to branch out into other tasks not usually the domain of video cards. For instance, ATI’s program that accelerates video encoding, basically just uses the graphics card as a secondary specialized processor. I would much rather buy generalized processing cards that would just dynamically take away work from my main processor and handle whatever I threw at them. Then a game could say, “you need at least 3 generalized processors for this game and this much hard drive space and ram” instead of “you need at least this processor, video card memory and dx/opengl level, sound card, etc”. Video card companies seem to be primed to take this spot in the near future.

I could be completely wrong, but I just get the feeling that multicore generalized computer hardware is going to happen in some form or another.

I hope someone might find this at least a tad interesting. Worst case scenario everyone just posts “DUH” and I have to disappear forever​

 

[bob]

Anyone want to guess how much the herculese video card in my 1987 DEC cost? No matter what your doing, you will have to buy some kind of video card unless your running a server of some kind.

Im too tired to quote your whole thing and pick it apart, although I agree with you on the most part.

But:

Has anyone compared the prices of PCs from the 1980s, 1990s, and 2000's? If you will notice, PCs have drastically come down in price. Top of the line PC compatible box with some kind of monochrome graphics and a 5-10MB hard drive would've cost more than a car in the 1980s. 1990s... Things were better, I cant recall how much that 386 packard bell cost... but it wasnt cheap, and it could almost play duke nukem.

So here I am with a $740.00 rig I built a year ago, plays all the new games on the low-medium settings. Couldnt pull that off in the mid 1990s . I have a feeling the PPU may end up like the Voodoo card, the original one. When it came out, not much for games could support it. It was pricey and I blew it off, but It was the first 3d-videocard, correct?

 

[pArTy]

I have a feeling the PPU may end up like the Voodoo card, the original one. When it came out, not much for games could support it. It was pricey and I blew it off, but It was the first 3d-videocard, correct?

Thats exaclty my feeling to. Right now the PPU is only useful for mid/low range systems. Systems with not a lot of CPU power. But since PPU is new the prices are high and so high that its worth it just to get a better CPU right now. I really dont think people realise how big the future of PhysX can be if it can pull its self through the beginning without dieing. Right now it can only do 50,000 particles. But once you get to half a million or so theres no way a CPU would handle that and developers could easyly get to half a million particles just with water/blood. I use Cinema4D to make CGIs and I use programs to simulate water sometimes and to get a nice realistic look for just a glass of water I usaly use anywhere between 30,000 to 70,000 particles. Once developers start using PPU for other things as well like calculating bullets going threw certain materials and complex crap like that you will see a need for it.

 

[LuminaryJanitor]

Has anyone compared the prices of PCs from the 1980s, 1990s, and 2000's?

My ex's dad paid $8,500 for his first 5 1/4'' floppy drive...