http://www.xbitlabs.com/news/video/display/20051005202331.html
an early death to the PPU if the GPU's performance will be on par with the PPU?
an early death to the PPU if the GPU's performance will be on par with the PPU?
Graphics Processors Capable of Physics Processing ? ATI.
- Thread starter linkgoron
- Start date
Graphic companies will say anything for sales. Sure it can do some physics, but just how much is some? Right now, not a great deal.
The Ageia card will be dedicated for physics, here is a quote on what we can expect:
Let's assume REAL WORLD performance will be half this, it's still a heck of a lot better than what today's best CPU's and GPU's can offer.
I can't predict the future, but my guess would be that Nvidia and ATI will both integrate some kind of physics chip on to their GPU's. One generation out, two maybe? Who knows.
The Ageia card will be dedicated for physics, here is a quote on what we can expect:
Let's assume REAL WORLD performance will be half this, it's still a heck of a lot better than what today's best CPU's and GPU's can offer.
I can't predict the future, but my guess would be that Nvidia and ATI will both integrate some kind of physics chip on to their GPU's. One generation out, two maybe? Who knows.
great news. I wouldn't need to buy a damn physic card. I am already having hard time upgrading my graphic card , CPU , Ram , Sound card ,Hard drives every year. I think if we keep on adding card for everything single thing then we are going to have very expensive computers ;(. It would be like modding your car or buying a 5K computer because its got crazy physic card , AI card , Sound card , Video card , shader card , card that can do AA with a performance hit blah blah...
I noticed that article also. Although graphics card processors may be capable of operating as a PPU, I wonder if their performance at this task would be enough to make it worthwhile.
It would be a nice way to maintain the usefulness of an older video card as well as increase the overall power of a new card. The physics workload that would ordinarily be computed by the new card would be offloaded to the secondary card, allowing the new video card to maintain higher framerates than it would ordinarily be capable of by itself.
It would be a nice way to maintain the usefulness of an older video card as well as increase the overall power of a new card. The physics workload that would ordinarily be computed by the new card would be offloaded to the secondary card, allowing the new video card to maintain higher framerates than it would ordinarily be capable of by itself.
Err, why would the performance of CPUs at physics relate or correlate in ANY way to the performance of GPUs at physics calculation?
CPU and GPU architecture are totally different.
PPU and CPU architecture are completely different.
PPU and GPU, well I personally don't know, and I doubt you do either, but they could quite feasibly be similar, in which case GPU's WOULD work at doing physics calculations, especially with fully programmable pipelines and unified shader architectures.
But why in the world would I want my Video card handling Physics if it could be better served handling more graphics computations. Once the PhysX card hits $199 (if it comes out) its mine.
I'm not sure some of you guys are getting the picture here. The point is that a PPU is another part that would enhance the game experience just as video accelerators (now GPUs) did for our games when we first were able to use them.
These cards wouldn't be simple hacks to get GPUs as we know it to somehow do physics calculations, they'd be designed to do both, or in a worst case senario there would be a separate PPU chip placed on the video card next to the GPU.
Right now AGEIA has the market all to themselves. Why wouldn't ATI or nVidia step in and produce their own solution? Obviously the easiest way to muscle them out is to include it on the essential part we all need - a video card. Right now we don't need a PPU (we need a GPU in order to play advanced PC games), nor does look like we'll be required to have one any time soon - bad news for AGEIA, good news for ATI or nVidia. If they can successfully incorporate PPU functionality into a video card and have it close to what a separate PPU could do, then AGEIA is going to be SOL, because who is going to buy a separate PPU over a cost effective one included into the video card (probably changed to a "gaming card" or something)?
I can see advantages to both routes, with the PPU being dedicated on its own board with its own memory and the PCI-e bus being more than able to handle it; and then there'd be including it on the video card where you have the PPU and GPU working in tandem, saving on costs, sharing faster memory configurations, and more a more stable unified driver (vs. having a separate one for a different card from a different company).
But I can honostly see the PPU/GPU edging out the lone PPU if it can be done and done correctly.
These cards wouldn't be simple hacks to get GPUs as we know it to somehow do physics calculations, they'd be designed to do both, or in a worst case senario there would be a separate PPU chip placed on the video card next to the GPU.
Right now AGEIA has the market all to themselves. Why wouldn't ATI or nVidia step in and produce their own solution? Obviously the easiest way to muscle them out is to include it on the essential part we all need - a video card. Right now we don't need a PPU (we need a GPU in order to play advanced PC games), nor does look like we'll be required to have one any time soon - bad news for AGEIA, good news for ATI or nVidia. If they can successfully incorporate PPU functionality into a video card and have it close to what a separate PPU could do, then AGEIA is going to be SOL, because who is going to buy a separate PPU over a cost effective one included into the video card (probably changed to a "gaming card" or something)?
I can see advantages to both routes, with the PPU being dedicated on its own board with its own memory and the PCI-e bus being more than able to handle it; and then there'd be including it on the video card where you have the PPU and GPU working in tandem, saving on costs, sharing faster memory configurations, and more a more stable unified driver (vs. having a separate one for a different card from a different company).
But I can honostly see the PPU/GPU edging out the lone PPU if it can be done and done correctly.
And if it can't, and has spare power going to waste because you're CPU limited?
Or if you have Crossfire/SLI and can't tap into the full potential of both cards graphically?
Your the type of buyer they see coming.
I'm flat out refusing to buy another card, it's laughable they want $200 for a card that only does physics.
It will. Am I absolutely sure of this? No, but I believe it will. Also, those initially buying PhysX cards will be the PC enthusiasts, so CPU bottlenecking shouldn't be a huge issue.
Look at the benchmarks for the F.E.A.R. demo benchmarks and tell me where you see excess graphics power going to waste. That's just a low resolution. What happens when I want to play it at 1920x1200? Right now games are coming out that take all that a GPU can give and then some.
And your the type User that will eventually have to make the change or watch people streak away. I on ther other hand enjoy they Idea of playing around with this stuff. I just happen to be poor and set a $150 Limit on a X-Rammed X-FI and $200 on a PhysX processor.
Then those aren't the types of games that will use it, this is going to unlock miles of performance and realism, its not Going to help get better frames on Quake 3 at 1024. Eventually This will also them to totally take CPU physics processing out of the equation, so all of those card capped by the CPU will get another boost in CPU work. Plus I can see one of these cards lasting longer then any graphics card or CPU do to how "fresh" it is and the amount of time it will take for games to use it, let alone take full use of it.
I can say without a doubt more then any graphics card or CPU, this thing interests me most.
Powermoloch
Lifer
- Jul 5, 2005
- 10,084
- 4
- 76
mooncancook
Platinum Member
- May 28, 2003
- 2,874
- 50
- 91
It should be something cheap like the math co-processor in the old days that can be integrated into some mobos
Actually I read in one of the X1800 reviews yesterday something mentioned about running for example an X1800XT for graphics and an X1300 for physics calculations. That way you have an entire gpu dedicated to graphics (x1800xt) and another one for physics (x1300). I think the main disadvantage with this is for sli or crossfire users, these guys would not have an extra available pci-e slot for the 2nd graphics card. Of course I think eventually both will be integrated into one card, as was the case with 2d and 3d (if you recall, in the voodoo days you needed both a 2d and a 3d card in seperate pci slots).
I am really interested to see how this PPU idea works out. Anyone know when we will actually see a game that uses the tech along with a PPU on an actual card?
You wll never be able to tap 2 7800GTX's in SLI at this point. Its electronically impossible with current Processing Bottlenecks.
What with 4x Adaptive (or transparent) AA and 16xAF + HDR and soft shadows and every single goodie maxed out at 1920x1200? You'd be lucky to get 30fps and I don't see the CPU struggling to process physics for 30fps.
thecoolnessrune
Diamond Member
- Jun 8, 2005
- 9,673
- 583
- 126
Lol I remember hearing the same thing when people were saying that they would never pay $600 for a card that only does video. lol
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter