GPU Physics is a Stupid Idea

Status
Not open for further replies.

xgas

Distinguished
Feb 17, 2006
61
0
18,630
GPU physics is a stupid idea, why? Cause it only gives Eye-Candy.

The PPU from Ageia is somewhat a great idea, but now is not the time to release it, plus they bottle neck its power of 20Giga Instructions per sec with the PCI 3.0/2.2 slot. No GPU can do 20Giga instructions per sec with data going forward and backwards at the same time. Simply said, no data that is feed to the pixel/vertex shader will EVER go backwards(For now, the future, I dont know), so the Physic via a GPU can only bring EYE-CANDY for now.
 

xgas

Distinguished
Feb 17, 2006
61
0
18,630
GPU physics is a stupid idea, why? Cause it only gives Eye-Candy.

The PPU from Ageia is somewhat a great idea, but now is not the time to release it, plus they bottle neck its power of 20Giga Instructions per sec with the PCI 3.0/2.2 slot. No GPU can do 20Giga instructions per sec with data going forward and backwards at the same time. Simply said, no data that is feed to the pixel/vertex shader will EVER go backwards(For now, the future, I dont know), so the Physic via a GPU can only bring EYE-CANDY for now.
Ageia is a waste of time. GPUs will basically do what the Ageia card does. That will happen as time goes on.

errr, no, you did not look at the interactable objects is different from eye candy physics, go do your homework before saying the GPU can do what the PPU can do.
 

FITCamaro

Distinguished
Feb 28, 2006
700
0
18,990
I love it how any of you people feel you're qualified to say what is a stupid idea. People thought VGA cards were a dumb idea. No idea is stupid if it can give tangible results (well in this case tangible in the virtual world). An idea can only be a failed idea if people choose not to accept it due to ignorance or cost.

I think both Ageia's card and GPU physics are both great ideas. Granted, only one can really exist on a system (it would be extremely difficult to implement a way to have both share the physics workload). So the problem is, which way do we go? I like the idea of a standalone card handling it. But I also like ATIs idea of having a 2nd GPU handling the physics because their implementation doesn't call for two identical cards. This lets us recycle a previous generation graphics card instead of selling it for next to nothing, throwing it in a box that doesn't need it, or have it gather dust.

And as far as bottlenecking goes, as long as the average framerate stays around 60 fps, I don't care. Do I want 200 fps and canned physics effects? Or do I want 60 fps and near realistic physics effects? I would rather have the latter.
 

Human1

Distinguished
Jan 3, 2006
306
0
18,780
No kidding, I'm with you. There's nothing wrong with trying to do physics calculations. They really help the immersion of games, but they're super taxing on a system.
I would have loved to play FEAR with all the effects up, but my cpu simply couldn't do better than medium, and that at ~30fps.
I support ATI in this. Fact is that GPU's are very powerful specialized processors and if we can harness that power for physics, then I say do it!
 

linux_0

Splendid
GPU physics is a stupid idea, why? Cause it only gives Eye-Candy.

The PPU from Ageia is somewhat a great idea, but now is not the time to release it, plus they bottle neck its power of 20Giga Instructions per sec with the PCI 3.0/2.2 slot. No GPU can do 20Giga instructions per sec with data going forward and backwards at the same time. Simply said, no data that is feed to the pixel/vertex shader will EVER go backwards(For now, the future, I dont know), so the Physic via a GPU can only bring EYE-CANDY for now.
Ageia is a waste of time. GPUs will basically do what the Ageia card does. That will happen as time goes on.


I would agree :-D

Sadly you almost have to spend $3-5K to be a gamer these days :cry:
 

4745454b

Titan
Moderator
"it only gives Eye-Candy."
Uhmmmm, whats wrong with that? Isn't eyecandy good? I'd much rather look at a screen full of eyecandy than a boring screen full of cr@p. Until they get the problem of lower frames fixed, and the price comes down, I'm not ready to "upgrade"
 

Human1

Distinguished
Jan 3, 2006
306
0
18,780
You certainly can spend $3K or more on a good setup, but if you are content with previous generation equipment (like I am) you only have to spend ~1K to have a decent machine. I don't get the best frame rates on the newer games (~30fps), but I can play anything out right now at high detail. I think FEAR is only game that really kicked my machine in it's glowing red nads, so I had to turn it down to medium detail.
 

sdrawkcaBgoD

Distinguished
Apr 24, 2006
354
0
18,780
GPU physics is a stupid idea, why? Cause it only gives Eye-Candy.

The PPU from Ageia is somewhat a great idea, but now is not the time to release it, plus they bottle neck its power of 20Giga Instructions per sec with the PCI 3.0/2.2 slot. No GPU can do 20Giga instructions per sec with data going forward and backwards at the same time. Simply said, no data that is feed to the pixel/vertex shader will EVER go backwards(For now, the future, I dont know), so the Physic via a GPU can only bring EYE-CANDY for now.
Ageia is a waste of time. GPUs will basically do what the Ageia card does. That will happen as time goes on.

errr, no, you did not look at the interactable objects is different from eye candy physics, go do your homework before saying the GPU can do what the PPU can do.

I may not be able to handle as many calculations per second as xgas's mighty PhysX paperweight, but I can do this calculation:

4754 posts by prozac26
divided by
44 posts by xgas
= 108


Looks to me like prozac has done more research than xgas by a factor of greater than 100.
I know whose opinion I'm more likely to value...
 

xgas

Distinguished
Feb 17, 2006
61
0
18,630
Well, post count doesnt prove anything, but experience does, I have been using the computer since the age of 4, thats the DOS age when I was using it. Simply said I have an experience of 18 years.
 

xgas

Distinguished
Feb 17, 2006
61
0
18,630
i'm assuming your are either 22 or 18 years of age then. your point. i used to use dos as well when i was young. my first comp was a used one from my dad's friends work. it had a 386 and windows 3.1. does this make me as qualified as you.

you still haven't provided the proof i asked for. it is not optional, either you back up what you say or shut it.

LOL, then I chose to shut up. You know those back then who didnt think GFX cards are the way to go, still bought them now, but I might eat my words. My theory is the GFX and CPU have to wait for the PPU to tranfer data to them(Check out the IO delay between the CPU GFX and standard PCI slot, lol), thats what causing some slow downs, and not pushing the system high enuf in CellFactor, unless you have a supa fast multi-core proc, you will need the PPU, plus the Software version of Ageia can spread the pain to multi-core/multi proc systems, but Havoc havent done this to any resent titles, which is a waste.

You see the Multi-core proc in the Xbox-360? Its a CPU designed to run "CLEAN" codes, mean its good in Graphics geometry prosessing, but the draw back is, its damn stupid in Physics prosessing, the standard CPU in gaming comps we see nowadays have what we call SSE optimization, it rearranges the ramdom data and turn them into "Clean" codes for easy prosessing.
So now, its kind of a smart to design a chip that does this kind of prosessing without the optimization, cause its "DESIGNED" to run datas like that, it does not need optimization, oh yeah, transister amount does not prove anything, the new 7900GT/GTX has 120++ Million less transister than the 7800GT/GTX, but the 7900s run faster, that is all.

In the end, its "Wait And See".
 

BaronMatrix

Splendid
Dec 14, 2005
6,655
0
25,790
While it is true that ATI and nVidia can do physics calculations, they are not HW physics, but SW pumped through the FP on the GPU. A PPU has hard coding for physics. So it is more specialized than a GPU. SInce it is for games I don't know why they didn't use PCIe x4. Most serious gamers have PCIe. No one is really making AGP slots anymore.
 

TheMaster

Distinguished
Jul 4, 2004
488
0
18,780
The physics processing GPU wasn't a stupid idea.
To release a PPU that does practically nothing... now that's a stupid idea.
Great Going Ageia! :lol:
 

dvdpiddy

Splendid
Feb 3, 2006
4,764
0
22,780
The physics processing GPU wasn't a stupid idea.
To release a PPU that does practically nothing... now that's a stupid idea.
Great Going Ageia! :lol:

That pretty much sums up this entire thread... Exactly! btw what happened to your thread about running PhysX games without PhysX?
 

sdrawkcaBgoD

Distinguished
Apr 24, 2006
354
0
18,780
The physics processing GPU wasn't a stupid idea.
To release a PPU that does practically nothing... now that's a stupid idea.
Great Going Ageia! :lol:

That pretty much sums up this entire thread... Exactly! btw what happened to your thread about running PhysX games without PhysX?

I'm sooo glad you asked. :D I think that thread dovetails nicely into this one. If anyone still thinks PhysX is a viable solution as it is right now they should check it out.
 

azrealhk

Distinguished
Apr 28, 2006
122
0
18,680
I support the use of PPUs, but AGEIA screwed up by making the game slower without really improving the overall game. IMHO, the little square bits flying out (GRAW) actually made the game look worse.

For those who did not know, many "ancient" computers used multiple processors. My favorite was the AMIGA with Chips lovingly called

COPPER
BLITTER
AGNES
DENISE
(and the CPU)

which was an Awesome computer when it came out, because the different chips performed different functions offloading the work of the CPU.
 

BaronMatrix

Splendid
Dec 14, 2005
6,655
0
25,790
can you show me where it says they have hard coding and what the hell that means. physics would just be alot of maths. why can't those same equations be run on a PPU and not a GPU. the PPU still need correctly programmed software to work AFAIK.


Both do 3D vectors, but not for the same thing. There is a big differnece bewteen matrix math for physics and texture and vertex mapping. If you setup the ASIC to do a specific mathematical function it will excel. Again I realize that FP wise GPUs are more than capable, but if this were to go to 3rd gen, it would be so far ahead in physics 378 million transistors(X1900) will seem like a few.
 
Status
Not open for further replies.