• 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.

How fast is your PPU?

sbrown23 said:
In terms of Brent's comment about placing the PPU on the video card PCB, is that even possible? With vid cards already maxing out what power can be delivered to them via an x16 PCIe slot (not to mention AGP), could they stand having another piece of silicon added to the same PCB and requiring 25+ watts more power? It seems like vid cards will only grow in their thirst for power and that it might be more sensible to run this PPU on a separate x1 PCIe or standard PCI slot. I can see nVidia and/or ATi possibly getting into this tech and marketing their own PPUs, but I don't know if I can see them integrating it with vid card architectures.

sure, just throw on another external power connector, heh

power is becoming a huge problem though with hardware, all the latest cpu's, video cards etc.. are drawing huge amounts of power, and putting out large amounts of heat

i would rather see this as part of the video card myself though instead of a seperate add-in board
 
Brent_Justice said:
sure, just throw on another external power connector, heh

power is becoming a huge problem though with hardware, all the latest cpu's, video cards etc.. are drawing huge amounts of power, and putting out large amounts of heat

i would rather see this as part of the video card myself though instead of a seperate add-in board

wish in one hand... shit in the other.. :D

i personally do not mind how things are implemented, as long as it all works.. im ok with it.

the power and heat though is a problem.. somebody is gonna have to figure out how to cut it down!
 
Brent_Justice said:
i would rather see this as part of the video card myself though instead of a seperate add-in board

I still maintain that it's going to wind up in that direction. I think the company promoting this product is itching to be bought by an NVidia or ATI.
 
Computer Games - While we know that the company is developing stand alone cards for the chip, is it also possible that motherboard makers and/or graphics chip board makers can add your processor to their designs and if so will there be any performance hit as opposed to a stand alone board?

Curtis Davis - Most likely, it will be taking advantage of the full PCI express 4X bandwidth so we do not expect that to be the case.

http://www.cgonline.com/content/view/451/2/


am i reading that right? a performance hit will "most likely" incur?
 
Jason711 said:
Computer Games - While we know that the company is developing stand alone cards for the chip, is it also possible that motherboard makers and/or graphics chip board makers can add your processor to their designs and if so will there be any performance hit as opposed to a stand alone board?

Curtis Davis - Most likely, it will be taking advantage of the full PCI express 4X bandwidth so we do not expect that to be the case.

http://www.cgonline.com/content/view/451/2/


am i reading that right? a performance hit will "most likely" incur?

I read that as 'most likely we'll see the PPU integrated into motherboards'.
 
to me, the guy is asking if it is implimented into a video card or a logic board. will there be a performance hit?

the answer says it will be using the full 4x pci-e bandwidth... so they do not expect adding them to a video card or board to be the case.

am i wrong? lol...
 
EvilAngel said:
Easily? I never said that. But for first time it would be possible

Actually, what you say could be valid only if most people dont own it.

This thing [the PPU] only needs to be picked up by 1 console maker, then the kind of stuff the console would be able to do would force the PC games designers to add those features in order to keep the PC at the same level or above such console{s}.

I really hope that happens. True realism is needed more than more FPS right now.

It would be a very nice thing, I just dont see one small add-in board maker putting it all together by themselves. It needs to have that one console marker, that one must have game. Cant blame me for being skeptical...

Edit: Im sure it will happen eventually, the algorithmns should get better with time as well (dont expect a simulation of splitering wood / breaking concrete on the first ppu title) but i really doubt ageia will be the company that does it.

Edit2: Just remembered the problems of sync'n all this craziness server/client wise

Edit3: just saw this dont think its been mentioned http://www.theinquirer.net/?article=21681
 
I wonder if they are gonna have different level of cards, like 100 dollar card, 200 dollar card- 300 dollar card and so on. The better the card the more stuff flies on screen.
 
ludachaz said:
I wonder if they are gonna have different level of cards, like 100 dollar card, 200 dollar card- 300 dollar card and so on. The better the card the more stuff flies on screen.

they are gonna price cards anywhere from 100-400.. would be a safe bet.
 
pettybone said:

There goes my console :D.

But if i understand clearly what you pointed there, it will have support for the AGEIA NovodeX Physics SDK, not the PPU chip unit itself, so it will use one or several of the PowerPC cores to do the physics instead of the PPU chip. Still, a step in the right direction for AGEIA, maybe we will see an addon PPU chip for the xbox2 some day. :eek:
 
Again, (and again forgive me if this is a stupid question), won't multi-core CPUs be able to handle separate tasks like physics and AI? I don't see PPUs seeing a lot of interest if this is the case.
 
eggrock said:
Again, (and again forgive me if this is a stupid question), won't multi-core CPUs be able to handle separate tasks like physics and AI? I don't see PPUs seeing a lot of interest if this is the case.

possibly, however, i still don't think they would be as fast as this dedicated chip for physics

this is a highly optimized, built for that kind of function only kind of thing

CPU's (even with dual cores) will still be all about processing many different types of data, random
 
This still reminds me of many years ago when people debated whether to spend lots of $$ on a math coprocessor (now integrated and called an FPU). Some people thought that they would always remain a niche product. Then, starting with the 80486 it became integrated with the CPU.

I think that PPU's, if they catch on, will follow a similar path.
 
Jason711 said:
i understand that...

however, isnt it plausable that the more you can do with physics.. then you will have to do more graphically.

for example.. more things flying around from explosions... grass, wind blowing through the grass... thats more rendering.

Normally, I would nod my approval and move on without comment here, but I have recent experience to the contrary. In my recent experience (upgrading the vid card on a system virtually identical to my own) large numbers of objects being rendered on screen are often CPU limited due to physics calculations these days. For example, the previously mentioned upgrade: Two identical systems except for the video card. One has a GF FX 5950. The other has a GF 6600GT. The 6600 makes 1200 points more in 3dmark03 but has the same framerate in the stream flyover scene in Game Test 4. Why you may ask? Because of the 500-600 blades of grass and 700-800 leaves displayed. They each require physics calculations (in order to make a more real-world like test).

Thus, a PPU, offloading these physics calculations, would allow the 6600GT to pull out ahead, as it should. I could be mistaken, but I think my logic is pretty sound on this point.
 
you are correct

in geometry/physics bound tests the common bottleneck would be the CPU right now
 
My point in addition to those above is that the PPU will give us existing GeForce 6 users and current Athlon 64 layouts access to this enhanced tech.

Sure in the long run, multicore cell processor setups may not need a physics proc to get great performance. But, I'll be able to upgrade my existing Socket 939 setup to get this enhanced goodness. I don't give a flip over therorectical stuff 2 years out, but what I'll be able to tack onto what I just purchased.

Famous quote from my computer vendor "your system will be obsolete after you leave our store". Any tech that can be added to my existing investment is always an excitement for me.

Note that a superscalar, specific focused proc is always going to be able to do a specific function, at lower power, than a general purpose CPU even if multicore. If multiproc or multicore were all the shiznit that everyone is making out, we'd just have a few powerPC's or other procs on our Graphics cards instead of that superscalar malarky. Heck, we'd just need XVGA cards and all the rendering will be done by those multi core procs. Riighttt. :p
 
Optimus said:
Normally, I would nod my approval and move on without comment here, but I have recent experience to the contrary. In my recent experience (upgrading the vid card on a system virtually identical to my own) large numbers of objects being rendered on screen are often CPU limited due to physics calculations these days. For example, the previously mentioned upgrade: Two identical systems except for the video card. One has a GF FX 5950. The other has a GF 6600GT. The 6600 makes 1200 points more in 3dmark03 but has the same framerate in the stream flyover scene in Game Test 4. Why you may ask? Because of the 500-600 blades of grass and 700-800 leaves displayed. They each require physics calculations (in order to make a more real-world like test).

Thus, a PPU, offloading these physics calculations, would allow the 6600GT to pull out ahead, as it should. I could be mistaken, but I think my logic is pretty sound on this point.


yes. that is true if the number of blades of grass/leaves on trees, etc stays the same. in this case, it is the cpu that is a bottleneck to calculate all of the physics in the scene.

if, however (and this is how i forsee it if PPU's catch on), you ADD more blades of grass/leaves, etc, then it becomes more of a workload on the gpu to render all of the objects. not that they can't handle it, but it is still more of a workload for them.

my guess is that future gpu's will be able to render many more times the number of objects, while the ppu will be able to also calculate movements for many more times the objects, and the cpu will then be able to handle all of the other calculations without any problem.

.02
 
hmm..

most of you are argueing that a CISC processor should be faster than a RISC processor. a CISC will never be as fast as a properly coded RISC.

am i wrong? :)
 
I'd actually like to see a CPU without any FPU MMX or SSE instructions. I yearn for the co-processor days.

Thats not to say I don't think they are necessary and wonderful things, but putting them all into one processor package just seems wrong, especially now with dual cores..

Say you know that a computer will be a desktop machine that does not really need any of the *extras*. You could put a simple CPU or two inside of it and have it run great for simple applications. If you had a video-editing computer, you might want one plain CPU with Two FPU and SSE co-processors. If you play games, you might want one CPU, one FPU, one SSE and one physics processor.

Intel kind of hit this hardware glut, where they hardcoded things like MMX instructions into the chip. When they became obsolete, they still have to leave it in the chip to have backward compatibility with really old software, even though it may be completely obsolete because of a new algorhythm like a MMX2 or SSE/2/3...
 
Brent_Justice said:
eggrock said:
Again, (and again forgive me if this is a stupid question), won't multi-core CPUs be able to handle separate tasks like physics and AI? I don't see PPUs seeing a lot of interest if this is the case.

possibly, however, i still don't think they would be as fast as this dedicated chip for physics

this is a highly optimized, built for that kind of function only kind of thing

CPU's (even with dual cores) will still be all about processing many different types of data, random

Depends on how much more performance you would get. If a PPU adds 10% to your gaming speed would it be worth spending $100-400 on a PPU or an extra $100-400 on a (multicore) CPU? 20%? 50%? There'd be a line somewhere, where it would be more beneficial for a gamer to spend his or her $$$ on a PPU. [H]'ers would obviously buy the best of both worlds (probably at a substantial discount from 'normal' retail prices. :D)

Now if AMD & Intel decide to include a PPU on chip as a 'gamers chip' that would be fantastic.
 
its seems to me like many people are just saying "oh, the CPU can handle it, why should i pay?"

well, the fact of the matter is that NO, it cant handle the things that a PPU will be able to. yea 600 some leaves and blades of grass puts a load onto the GPU, but thats more on the CPU. even with multicore procs and stuff you're still going to get a huge slowdown if you try to run the things that a PPU is coded for.
\
right now im running software emulation on my vid card becuase im in the middle of an upgrade... it blows because the CPU has to use a bunch of instuctions where a GPU only needs 1... its like using onboard sound, where the CPU is doing all the processing effects. thats why the Audigy increases FPS and stuff.


anyway, got a bit off topic, but it seems to me like people arent willing to give in to the fact that with a dedicated PPU physics will be way better than they are now. sure, a CPU is fine NOW, but add those 50,000 "bones" and all of a sudden things look more real and awesome, but your processor is limiting you to 1 fps
 
yeah, people shouldn't underestimate the power of a dedicated chip for a certain task

that's why we have video cards now that do all the 3d rendering, a cpu simply cannot do it fast enough

video cards are fast at pixel pushing/texture filling, but slow at geometry/physics, and so is the cpu

we either need the video cards to kick it into gear and offload the physics from the cpu to them, or use a dedicated chip
 
Just as a side note:

For "true" physical "realism" every "bone" in the environment must have physics associated with it (though not necessarily actual polygons). This would include leaves, multiple points (technically infinite points) on a blade of grass, an entire tree trunk, the frame of a car, each of the various body shells (door covering, trunk lid, hood, etc.) of the car, about 265 (?) bones in a human body (for each character on screen), etc. Of these, only the tree trunk, leaves, and car body shells need to be rendered in 3d, because the others are not actually visible; they are merely points of reference.

Of course, with 50,000 bones, you would be able to render the physics for a pile of roughly 188 people with no physically modifyable aspects in the environment.
 
Back
Top