Jeff Yates, Havoc, and your GPU

Hardware physics are the way to go but it's gonna take some people extra convincing. Just like hardware accelerated 3d when it first came out. Ironically I've read that the man responsible with convincing the world that 3d acceleration is the future, is the exact same man on board with Ageia telling everyone hardware physics is the next thing. He is an ex 3dFX employee. I believe that hardware physics is the next thing. We have hadware 3d, hardware controlled audio and soon hardware physics.
 
Lord_Exodia said:
Hardware physics are the way to go but it's gonna take some people extra convincing. Just like hardware accelerated 3d when it first came out. Ironically I've read that the man responsible with convincing the world that 3d acceleration is the future, is the exact same man on board with Ageia telling everyone hardware physics is the next thing. He is an ex 3dFX employee. I believe that hardware physics is the next thing. We have hadware 3d, hardware controlled audio and soon hardware physics.
Click to expand...

Bill Gates said 640k is good enough. Well...he was wrong. Same with 2D to 3D. People were wrong. Us as consumers want more with each new purchase. A dedicated physics controller can provide a much more "real" experience in a game. I look foward to this. As part of my support for this evolution, this a good chance i will be an early adopter to help this flegling techonology. We are [H] after all...
 
I think hardware accelerated physics is a good thing, I just don't want to spend the same price as I would for a graphics card, intergrated solutions seem much more viable in my opinion.
 
I think it's all going to come down to what's easier to use.. .and in that regard, Havoc's middleware has the advantage, since it is the current leader that many top developers are used to working with.

I think it's a bit easier to convince people to shell an extra 100 on a BETTER videocard to do graphics/physics than it is to sell a seperate 100 dollar card DEDICATED to physics.
 
Trepidati0n said:
Bill Gates said 640k is good enough. Well...he was wrong. Same with 2D to 3D. People were wrong. Us as consumers want more with each new purchase. A dedicated physics controller can provide a much more "real" experience in a game. I look foward to this. As part of my support for this evolution, this a good chance i will be an early adopter to help this flegling techonology. We are [H] after all...
Click to expand...


Bill Gates never said that, it's a myth.

As far as hardware controlled physics, I'd be fine with the GPU doing it, as long as it didn't require two graphics cards (SLI/Crossfire) or impact 3D performance too much.
 
there CAN be too much of a good thing. How many different add on cards do you want? Sound, Video, why stop at physics? Next lets have a card that ONLY processess shaders, since shaders is the next big thing in graphics they say. AND, lets have add on memory for video cards, that would be great! :rolleyes: oh, good point poster below me, lets not forgot to have an AI only card, THAT WOULD RULE. :eek: :rolleyes:
 
Integration is key, in my opinion. We don't want a million cards for a million different features. We want one card that can do it all for a specific purpose. In the future, I hope to see "gaming cards" that combine video, physics, and really intensive AI processing (if it ever gets to that), since all of it fits under one category: gaming. The sound card should remain separate though, because it's advertised towards a different market as well.
 
I'm all for Hardware Accelerated physics in games. The question for game content developers is, which one is it better/eaiser/more cost effective and less time consuming to program for, Ageia's system or Havok's system.

The aspect I like about Havok's idea is that it will use existing hardware to do it, you won't have to purchase an extra piece of hardware, potentially keeping the price down for the gamer and enthusiast. But for those that can afford it, putting in a second graphics card for SLI or CrossFire (or even just a single card like ATI is proposing) could accelerate the physics and provide a better experience.

What will be interesting will be to test this method (doubling up video cards) directly against an Ageia dedicated accelerator and seeing which one allows a better gaming experience.

Surely ATI is looking at all of this closely as they are pushing GPU's as general purpose accelerators.
 
I would be willing to fork money over for physics cards. I'm more than sure anyone would if games took advantage of it. Dynamic water is amazing to me same with mud and all that kind of stuff. You wouldt just have mud looks but mud feel. Have it beable to fly around when you throw a grenade at it and have chuncks of slimy mud stick to you and everything around you. It just depends how much the games will take advantage of it for me to buy a phyiscs card.

Agerias(or however you spell it) has a better way in the way of pushing advance phyiscs. Havoks probley wont beable to push as extreme physics as agerias but it will cost cheaper for it.
 
With the enhanced bandwidth of PCI express why not have a Single PCI E 16x with 3 65Nano PU's on them 1 for video, 1 for Physics, and 1 audio processor. Whooooo Baby I would buy that. And then another one and SLi them :D ROFL
 
Physics and graphics should be consolidated, and rename the videocard to gaming card.. or something along those lines. Didn't ATI do something like this?
 
I wasnt a physics card specialized for physics and only that, and i want it to INCREASE framerates, not decrease.
 
When you think about the evolution of computers, it's easy to say who was wrong/right as hindsite is 20/20.
As far as this physics 'card' - why does it need to be seperated, seems to me this is technically part of the whole 'graphics system' ... it should be a intergral part of the next generation of graphics cards...not using the GPU, but it's own chip on the GPU (and it's own memory as well...)
Typically new cards have 5 usable slots - 1 or 2 for graphics, 1 for sound, the rest usually just stay empty unless your mobo doesn't have integrate LAN, and then you have to have a card, so max usage is 4 cards but typically only 2 (graphic/audio) or 3 (dual graphic - sli/audio)...
 
Add-on cards for Physics is not the way to go, it will just further split the PC gaming community. ATi's approach to allowing GPU's to take on the physics load using Havok's software is the best way. This way there is minimum physics support via software using Havok's system and for those that have two cards, they can use one of them for hardware accelerated physics. The interview mentions nVidia and ATi cards both being able to utilize Havok's software in the future so this would be possible to use in Crossfire or SLI. It sure beats requiring customers to go out and get a $150-$200 add-on card just to have physics in their games. Even casual gamers upgrade their GPU's at least once every 1-2 years and if they do, they could always use the older card for physics processing rather than throw it away.

Think about it, with the arrival of 3d accelerated cards, software developers had to (and still do) code for the lowest common denominator which held back the progress of PC games. This was further compounded with ATi and nVidia rebadging old hardware under a new name. If we throw physics add on cards into the mix that cost $150-$200+, then support for hardware physics will never really take off because of the cost. Even as an enthusiast, I'd be reluctant to buy a $500+ card and then have to spend another $200+ on a PPU card. I'd much rather my GPU support it out of the box and then when the card gets old, I could relegate it to PPU duty.
 
Speaking as both as hardware enthusiast and graphics programmer, I think that doing physics on the GPU is a load of crock in terms of practicality, and would MUCH rather prefer having physics cards or CPU driven physics applications.

Firsly, as we strive more and more towards multi-core CPU solutions, if you have physics offloaded onto the already stressed GPU, suddenly you only have AI, game logic and some audio control going on in one CPU. That is a LOT of CPU power being wasted, and it would be far better to have a dedicated thread focused towards physics.

Secondly, they talk about thousands of objects. At that point, we're talking about moving a lot of data stored directly on the video card (render-to-texture data), and having to move that into CPU memory. Not only is the actual texture locking and moving of data itself slow (say all you want about PCI-express x16, it's not as fast as, at the very least, I would like it), but it stalls the GPU AND CPU, eating away precious milliseconds of time per frame. (note: I'm talking about going from an average of ~45fps down to probably 25-30fps, and that's a fairly conservative estimate)

Thirdly, when the hell are games being CPU limited and need to offload their calculations to the GPU? One of the devs of Serious Sam 2 said "Crank your resolution to 2048x1536, turn on 8xAA, HDR, and 16xAF and let's see how CPU limited you are", which is something I TOTALLY agree with. As long as the CPU is able to perform the game logic at >70-80Hz (something which can go out the window if you start thinking about GPU-based physics, and isn't REALLY hard to do if the guys focusing on non-GPU stuff really optimize their code) and possibly reduce the amount of CPU time spent on the graphics side of things, you'll be fine.

Fourthly, how do you think people running dedicated servers will think about this? Last I checked, server motherboards aren't exactly all that graphics focused, and iirc from earlier this summer, they don't even have AGP slots, let alone PCI-express. Not only are there no more graphics cards that can do PCI slots, even if there were the requisite cards, PCI is still pretty slow.

Fifthly, they talk about using different GPU cards for different tasks. Well that's just AWESOME thinking, so let me know when the OGL/D3D API and the hardware vendor drivers even start to THINK about exposing individual GPU control.

So, yeah, this talk about physics on GPU stuff really pisses me off, and if I was working at a dev that decided to do it in such a manner, I'd say they were off their rocker and bail ASAP.

(sadly, this post will probably be ignored because I'm not some high authority in the physics or graphics world.Oh well *sighs*)
 
These bastards are not making me buy yet another card to play games. PC gaming is going to die like a death row inmate in a gas chamber if they make it even more fucking ridiculously expensive a proposition that it already is.

$400 GPU!
$100 Sound card!
$200 CPU and mobo!

And now...

$200 PHYSICS processor!

Give me a fucking break.
 
Cypher19 said:
Speaking as both as hardware enthusiast and graphics programmer, I think that doing physics on the GPU is a load of crock in terms of practicality, and would MUCH rather prefer having physics cards or CPU driven physics applications.

Firsly, as we strive more and more towards multi-core CPU solutions, if you have physics offloaded onto the already stressed GPU, suddenly you only have AI, game logic and some audio control going on in one CPU. That is a LOT of CPU power being wasted, and it would be far better to have a dedicated thread focused towards physics.

Secondly, they talk about thousands of objects. At that point, we're talking about moving a lot of data stored directly on the video card (render-to-texture data), and having to move that into CPU memory. Not only is the actual texture locking and moving of data itself slow (say all you want about PCI-express x16, it's not as fast as, at the very least, I would like it), but it stalls the GPU AND CPU, eating away precious milliseconds of time per frame. (note: I'm talking about going from an average of ~45fps down to probably 25-30fps, and that's a fairly conservative estimate)

Thirdly, when the hell are games being CPU limited and need to offload their calculations to the GPU? One of the devs of Serious Sam 2 said "Crank your resolution to 2048x1536, turn on 8xAA, HDR, and 16xAF and let's see how CPU limited you are", which is something I TOTALLY agree with. As long as the CPU is able to perform the game logic at >70-80Hz (something which can go out the window if you start thinking about GPU-based physics, and isn't REALLY hard to do if the guys focusing on non-GPU stuff really optimize their code) and possibly reduce the amount of CPU time spent on the graphics side of things, you'll be fine.

Fourthly, how do you think people running dedicated servers will think about this? Last I checked, server motherboards aren't exactly all that graphics focused, and iirc from earlier this summer, they don't even have AGP slots, let alone PCI-express. Not only are there no more graphics cards that can do PCI slots, even if there were the requisite cards, PCI is still pretty slow.

Fifthly, they talk about using different GPU cards for different tasks. Well that's just AWESOME thinking, so let me know when the OGL/D3D API and the hardware vendor drivers even start to THINK about exposing individual GPU control.

So, yeah, this talk about physics on GPU stuff really pisses me off, and if I was working at a dev that decided to do it in such a manner, I'd say they were off their rocker and bail ASAP.

(sadly, this post will probably be ignored because I'm not some high authority in the physics or graphics world.Oh well *sighs*)
Click to expand...
There is no games that show true physics yet. Theres no games with dynamic water at all. Just adding dynamic water to a game alone will make one hell of a differance in looks and feel. Half Life 2 pushs pretty card on CPUs due to the ragdolls and all the physics in the game but with PPU there could be actual fluids. There would be dynamic hair also and dynamic clothing. It would add so much life like things to a game. 10 years from now you will regret your post.

Cypher19 said:
Speaking as both as hardware enthusiast and graphics programmer, I think that doing physics on the GPU is a load of crock in terms of practicality, and would MUCH rather prefer having physics cards or CPU driven physics applications.

Firsly, as we strive more and more towards multi-core CPU solutions, if you have physics offloaded onto the already stressed GPU, suddenly you only have AI, game logic and some audio control going on in one CPU. That is a LOT of CPU power being wasted, and it would be far better to have a dedicated thread focused towards physics.

Secondly, they talk about thousands of objects. At that point, we're talking about moving a lot of data stored directly on the video card (render-to-texture data), and having to move that into CPU memory. Not only is the actual texture locking and moving of data itself slow (say all you want about PCI-express x16, it's not as fast as, at the very least, I would like it), but it stalls the GPU AND CPU, eating away precious milliseconds of time per frame. (note: I'm talking about going from an average of ~45fps down to probably 25-30fps, and that's a fairly conservative estimate)

Thirdly, when the hell are games being CPU limited and need to offload their calculations to the GPU? One of the devs of Serious Sam 2 said "Crank your resolution to 2048x1536, turn on 8xAA, HDR, and 16xAF and let's see how CPU limited you are", which is something I TOTALLY agree with. As long as the CPU is able to perform the game logic at >70-80Hz (something which can go out the window if you start thinking about GPU-based physics, and isn't REALLY hard to do if the guys focusing on non-GPU stuff really optimize their code) and possibly reduce the amount of CPU time spent on the graphics side of things, you'll be fine.

Fourthly, how do you think people running dedicated servers will think about this? Last I checked, server motherboards aren't exactly all that graphics focused, and iirc from earlier this summer, they don't even have AGP slots, let alone PCI-express. Not only are there no more graphics cards that can do PCI slots, even if there were the requisite cards, PCI is still pretty slow.

Fifthly, they talk about using different GPU cards for different tasks. Well that's just AWESOME thinking, so let me know when the OGL/D3D API and the hardware vendor drivers even start to THINK about exposing individual GPU control.

So, yeah, this talk about physics on GPU stuff really pisses me off, and if I was working at a dev that decided to do it in such a manner, I'd say they were off their rocker and bail ASAP.

(sadly, this post will probably be ignored because I'm not some high authority in the physics or graphics world.Oh well *sighs*)
Click to expand...

I dont know where you been but ATis newest generation is opening there video cards just so other programs can use the GPU.
 
I dont know where you been but ATis newest generation is opening there video cards just so other programs can use the GPU.
Click to expand...

In my post I was referring directly to games usage, where graphics and physics have to be happily swinging arms together through a field of flowers together. For other apps, like workstations or off-line physics simulations, I could care less about how much the GPU is used, and in fact I'd even encourage it. (Edit: also, iirc, ATi was merely encouraging it. If anything, NV's cards are more capable, since some clever vertex texture usage would save the GPU->CPU transfer of physics results)

There would be dynamic hair also and dynamic clothing. It would add so much life like things to a game. 10 years from now you will regret your post.
Click to expand...

I'll worry about that 10 years from now (at which point PPUs will more than likely be common, not to mention the changes that D3D10 will bring around). Here and now, physics on D3D9 hardware and software is just totally infeasible.
 
Duel Core CPU's are the real way. You cant split up the whole computer, an upgrade will consist of 300 different little things.
 
http://www.realityflux.com/abba/C++/HDR GPU Particles/HDR_GPU_Particles.zip

Check out that demo. The physics (GPGPU) you see there are accelerated by the GPU only. This demo works only with N40 and G70 cards (IIRC).

Also here is an interesting article translated from an italian site:

http://translate.google.com/transla...&hl=it&ie=UTF-8&oe=UTF-8&prev=/language_tools

Don´t mean to diss AGEIA, and I do still think that a dedicated physics card should offer the best performance with hardware accelerated physics, however, that route is also less cost effective and the install-base will be very small. The AGEIA route definitly would take 1 or 2 years only to get a decent install-base. The GPGPU route is much more cost effective and it has already a good install-base.

I do think the Havok guy is obviously trying to talk AGEIA (their main compeditor) down as much as possible because obviously the more successful AGEIA will get, the less profit Havok will make. It would be in the best interest for AGEIA to work also on a physics engine (like current Havok 2) AND add GPGPU support to it. If they concentrate only on their PPUs, they bascially will depend on them and their success, which is still anything but secure.

Compertition is good, so I do hope that AGEIA will establish into the market, that will only benefit us customers. Its good that Havok has compedition.
 
Are we forgetting here that 3D accelleration started the same way? The monster3D, which was the first successfull 3D card, was not a complete video card, but an add-on for your existing card. So was the Voodoo 2. It wasn't until the Voodoo Banshee/Rush that they tried to combine the technologies. And we all know how unstable things were back then. Come on, I know you have selective memories, but I remember keeping my Voodoo2 cards for a long time because the early dedicated 3dfx and Nvidia cards had so-so performance and crappy drivers.

My point is that the physics processing may be forced to start the same way. To get a foothold in the market and guarantee compatibility an additional card may be the only viable chioce Ageia has for the first generation or two. There's no way in the reality of business that ATI AND Nvidia AND the other 3rd tier GPU makers will all willingly adopt Ageia's hardware at the same time, pay Ageia royalties, and risk an unknown return on investment. One of them would rather own Ageia outright, or reverse engineer/r&d and make their own version of the product so they pocket all the money. It's a given that either Nvidia or ATI has the war chest to R&D their own physics chips if they were sure it would pay off. Ageia is a gnat in size compared to either of the graphics giants, they don't make decisions for the industry or push anyone to do anything, they just try to get the best position possible before getting bought out.

This should be very interesting to watch as it plays out. I'm all for physics processing of any kind, and it looks like we are going to get it. The question is what form will it take?
 
Hardware Physics is too expensive..the PhysX chip will cost like $250 - $300 when it comes out? Who wants to spend so much money on a physics processer when they can buy a X2 for $332? Plus, won't dual core/ SLI systems be able to handle graphics/physics?
...Maybe I am just not a visionary
 
Physx on the GPU or CPU is the way to go.

I don't know for you guys, but I'm gettings bored of better graphx, I want something else, and I would even be happy to turn off some graphic detail to get better physx :)
 
Do physics on a seperate card. Far too much shit happening on video cards these days anyway. I mean, just look how physically large the cards are, even! And the power requirements! No, no, no. Seperate card for now, let them combine them later if they want.

I'm all for it, by the way. I'm no computer engineer, but a dedicated physics processor has got to be better for doing physics than a repurposed CPU or GPU.
 
matt fury said:
Do physics on a seperate card. Far too much shit happening on video cards these days anyway. I mean, just look how physically large the cards are, even! And the power requirements! No, no, no. Seperate card for now, let them combine them later if they want.

I'm all for it, by the way. I'm no computer engineer, but a dedicated physics processor has got to be better for doing physics than a repurposed CPU or GPU.
Click to expand...

Yes, and it also means:

Min Requirenments:
Athlon 64 4800+ or equiv Pentium 4
1280mb ram
Geforce 7800GT or Radeon X1800XL
Physx 120-XT or higher required

Sorry, im not looking forward to another $300 card for my computer. If it does actually come out goodbye PC gaming. This is what duel Core proccessers are for.
 
The same way you can play games with a lower-end card today, you'll be able to play them in the future with a physics card. Just because it's possible, it doesn't mean that you HAVE to play the game on the highest settings. It's not like if they don't use physics cards then they'll be able to have the same performance anyway, so you won't be missing out on anything that you would've had if they never invented them.
 
Remember ATI's proposition, using a seperate video card (GPU) as a general processor.

In that scenerio you wouldn't have to spend $300, you could just buy the cheap X1300 video card spend $100 maybe for a cheap physics upgrade.
 
Cypher19 said:
Speaking as both as hardware enthusiast and graphics programmer, I think that doing physics on the GPU is a load of crock in terms of practicality, and would MUCH rather prefer having physics cards or CPU driven physics applications.

Firsly, as we strive more and more towards multi-core CPU solutions, if you have physics offloaded onto the already stressed GPU, suddenly you only have AI, game logic and some audio control going on in one CPU. That is a LOT of CPU power being wasted, and it would be far better to have a dedicated thread focused towards physics.

Secondly, they talk about thousands of objects. At that point, we're talking about moving a lot of data stored directly on the video card (render-to-texture data), and having to move that into CPU memory. Not only is the actual texture locking and moving of data itself slow (say all you want about PCI-express x16, it's not as fast as, at the very least, I would like it), but it stalls the GPU AND CPU, eating away precious milliseconds of time per frame. (note: I'm talking about going from an average of ~45fps down to probably 25-30fps, and that's a fairly conservative estimate)

Thirdly, when the hell are games being CPU limited and need to offload their calculations to the GPU? One of the devs of Serious Sam 2 said "Crank your resolution to 2048x1536, turn on 8xAA, HDR, and 16xAF and let's see how CPU limited you are", which is something I TOTALLY agree with. As long as the CPU is able to perform the game logic at >70-80Hz (something which can go out the window if you start thinking about GPU-based physics, and isn't REALLY hard to do if the guys focusing on non-GPU stuff really optimize their code) and possibly reduce the amount of CPU time spent on the graphics side of things, you'll be fine.

Fourthly, how do you think people running dedicated servers will think about this? Last I checked, server motherboards aren't exactly all that graphics focused, and iirc from earlier this summer, they don't even have AGP slots, let alone PCI-express. Not only are there no more graphics cards that can do PCI slots, even if there were the requisite cards, PCI is still pretty slow.

Fifthly, they talk about using different GPU cards for different tasks. Well that's just AWESOME thinking, so let me know when the OGL/D3D API and the hardware vendor drivers even start to THINK about exposing individual GPU control.

So, yeah, this talk about physics on GPU stuff really pisses me off, and if I was working at a dev that decided to do it in such a manner, I'd say they were off their rocker and bail ASAP.

(sadly, this post will probably be ignored because I'm not some high authority in the physics or graphics world.Oh well *sighs*)
Click to expand...

Very good post, in my opinion. Unfortunately, I'm not experienced as you, so I can't comment, but I really like the idea of doing physics on the CPU's extra cores. I'm not sure how much an extra 2.6 GHz Athlon 64 core can do in terms of physics, but if it can do enough, then that would be awesome.
 
5150Joker said:
The interview mentions nVidia and ATi cards both being able to utilize Havok's software in the future...
Click to expand...

Yes, but right now we have no idea how well the Havok system will compare with Ageia's.

http://www.pcper.com/article.php?aid=140 said:
PPU specs:
125 million transistors
20 watts power consumption
128 MB of GDDR3 memory
Click to expand...

As you can see, there's a big difference in horsepower between GPU processing graphics and physics and PPU processing physics alone. Thus, at the minimum you'd need another graphics card to unload the physics work onto.

Now the question is how a GPU stacks up against a PPU in terms of physics processing. I would imagine a PPU would be faster by a big margin since a RISC (Reduced Instruction Set Computer) processor (GPU, PPU, etc) probably does best what it was designed to do.

If a GPU's performance is relatively close to that of a PPU, I'm all for the Havok/GPU solution. If not, I'll pull for the PPU. My vote is for whatever does the job best. But, until we have benchmarks from both, it's all just speculation.

Cypher19 said:
when the hell are games being CPU limited and need to offload their calculations to the GPU?
Click to expand...

They're not yet... because game devs code for what is playable on current hardware. If they had the hardware to code for, they could develop much more complex physics than what's currently out to take advantage of the extra physics horsepower.

-SEAL
 
Sorry for the double post, but I think it's important to address the CPU vs. PPU argument:

CPU = CISC (Complex Instruction Set Computer)
PPU / GPU / etc. = RISC (Reduced Instruction Set Computer)

What's the difference between RISC and CISC processors?

http://en.wikipedia.org/wiki/Risc said:
Yet another part of RISC design came from practical measurements on real-world programs. Andrew Tanenbaum summed up many of these, demonstrating that most processors were vastly overdesigned. For instance, he showed that 98% of all the constants in a program would fit in 13 bits, yet almost every CPU design dedicated some multiple of 8 bits to storing them, typically 8, 16 or 32, one entire word. Taking this fact into account suggests that a machine should allow for constants to be stored in unused bits of the instruction itself, decreasing the number of memory accesses. Instead of loading up numbers from memory or registers, they would be "right there" when the CPU needed them, and therefore much faster. However this required the instruction itself to be very small ("reduced"), otherwise there would not be enough room left over in the 32-bits to hold reasonably sized constants.
Click to expand...

Why is a GPU so much better than a CPU at graphics processing?

http://en.wikipedia.org/wiki/Graphics_processing_unit said:
A GPU implements a number of graphics primitive ("reduced") operations in a way that makes running them much faster than drawing directly to the screen with the host CPU.
Click to expand...

The same principle applies to GPU's. Barring a radically-different implementation of CISC processing from the current x86 standard, RISC's will always outperform them at the specific tasks they were designed for.

-SEAL
 
I agree with everyone else that's it a devil of a lot simpler and cheaper to stick to stick a few more ALUs on a GPU or another core on a CPU to plow through physics calculations than have a dedicated processor do an elegant job of it :p After all, that's been the performance race all along - just add MHz and cache to plow through unoptimized code as fast as possible. Make a developer's life a lot easier when you can just say "eh, they'll have 2GB of RAM and CPUs with 1MB of L2, no need to clean this code up" than "Wow, this new PPU is pretty slick, anybody want to program the thing?"

I think developers would rather multi-thread their games, as that is a guaranteed win/return on investment with the next generation of consoles, than develop an API for a dedicated physics engine which is still very much in infancy.
 
I think to really get this thing off the ground we need a standard API for it. Now that video cards have the capability to do these things and act in this way there needs to be an API. I wouldn't be against Microsoft creating a Physics API built into DirectX to allow developers to directly access the hardware for those functions.
 
Brent_Justice said:
I think to really get this thing off the ground we need a standard API for it. Now that video cards have the capability to do these things and act in this way there needs to be an API. I wouldn't be against Microsoft creating a Physics API built into DirectX to allow developers to directly access the hardware for those functions.
Click to expand...

I agree Brent. However, I think they are doing that already. I just think it will take somebody like MS though, as you said, to build it into DirectX or whatever it is going to be called now.

Why do people here seem to always complain how much things cost and then they go out and buy it anyways? :confused:

Also putting physics on a graphics card does not make sense in the terms of data flow. The graphics card is the last stop to make things look pretty. The CPU has already done 99% of the work when it comes to the wireframes, motion, etc.

As for people who complain about a $$$ soundcard...that is because you made a poor investment choice. A dedicated soundcard compared to onboard sound may get you a few more FPS. A dedicated physics card will give you more visual punch for the $ than any CPU every could.

-tReP
 
Brent_Justice said:
I think to really get this thing off the ground we need a standard API for it. Now that video cards have the capability to do these things and act in this way there needs to be an API. I wouldn't be against Microsoft creating a Physics API built into DirectX to allow developers to directly access the hardware for those functions.
Click to expand...

*cough* The PhysX SDK on the site (aka Novodex), which is essentially an API, is completely free and works with the physics card. However, they ask that if you are to use it for a shipped product that you purchase a license. I haven't worked with it at all, but I know a couple people who could provide further details.
 
Hello! While I hope that all alternatives will be available(using a dual core CPU OR a physics card OR another GPU), and I think that would be ideal, I really don't think people should be complaining about price. Its pretty obvious that people around here are willing to spend what it takes to have the best. Would a few hundred(or less) dollars really affect those who can already afford a dual core/FX CPU or a 7800? I really feel sorry for people like myself who are still running w/ Athlon XP 2400s and such. Were the ones who are struggling to afford anything, even the games themselves. I realize that by next year(UT2007) my system will be hopeless, but I have two choices--either sell my wife and kid or give up PC gaming. Im not going to complain about the price, ill just have to either pay or quit PC gaming(at least if I want to stay current). But I would rather be able to get a modest system, and either use an add on card or a cheap vid card that be forced to spend 500-800 on a processor, and keep my budget under 1000.
 
Cypher19 said:
*cough* The PhysX SDK on the site (aka Novodex), which is essentially an API, is completely free and works with the physics card. However, they ask that if you are to use it for a shipped product that you purchase a license. I haven't worked with it at all, but I know a couple people who could provide further details.
Click to expand...

Yep, I'm aware of that. My primary point is how much clout will they have with the developers to use purchase the license, use the API and program it for their dedicated PPU. Will smaller game developers go that route? What about the big ones?

IMO this will remain a niche thing until it's backed by a big ole company like MS who has it directly built into DirectX for free so anyone can use it with any SM 2.0/3.0/4.0 capable video card.
 
Standpoint said:
Hello! While I hope that all alternatives will be available(using a dual core CPU OR a physics card OR another GPU), and I think that would be ideal, I really don't think people should be complaining about price. Its pretty obvious that people around here are willing to spend what it takes to have the best. Would a few hundred(or less) dollars really affect those who can already afford a dual core/FX CPU or a 7800? I really feel sorry for people like myself who are still running w/ Athlon XP 2400s and such. Were the ones who are struggling to afford anything, even the games themselves. I realize that by next year(UT2007) my system will be hopeless, but I have two choices--either sell my wife and kid or give up PC gaming. Im not going to complain about the price, ill just have to either pay or quit PC gaming(at least if I want to stay current). But I would rather be able to get a modest system, and either use an add on card or a cheap vid card that be forced to spend 500-800 on a processor, and keep my budget under 1000.
Click to expand...

It isnt really just that though. There is no reason to introduce a product that isnt needed. Multi-threaded games if coded correctly should do fine on newer CPU's.
 
CPU's still aren't as powerful as a dedicated GPU or PPU for doing specific things, back to the RISC/CISC comparison.
 
You must log in or register to reply here.
Back
Top