I'm sure you all read this
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2376
What do you think, good idea or a bad one?
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2376
What do you think, good idea or a bad one?
Originally posted by: cbuchach
I can't see myself buying a separate PPU. The only way I see it to be very useful is if both NVidia and ATI incluce the chip as part of their graphic cards. Initially one will choose to do it as first there seems no end in sight as to what they will charge for a graphics card (and what people will pay) and they will want something more to differentiate performance. Then of course the other company will follow.
Also, the idea of a price and model/performance stratification amongst discrete PPU add-in cards seems ridiculous to me.
Originally posted by: DaveSimmons
Meh. Not worth doing in separate hardware.
This will be a good use for dual-core CPUs, you can code to process the physics in a thread on CPU-2 at least as easily as coding for a wacky new processor type. If it is a single-CPU system your code even still runs, you just disable some of the fancier processing to lighten the load.
Originally posted by: phisrow
The one problem that I'm a bit worried about is how easy it will be to adjust for various levels of hardware performance. With graphics engines, that sort of thing is relatively easy(easy enough for the pros, anyway) as it is pretty obvious to just cut resolution, FSAA, number of polygons and light sources, texture detail, etc. You cannot work miracles; but you can manage a fair bit(e.g. that nutcase who was running Doom3 on dual Voodoo twos a while ago). Physics, on the other hand, seems a lot harder to scale. If you only use it for unimportant stuff, like foliage effects and little waves and stuff, then it will be pretty easy. Everyone with the coprocessor gets real dynamic foliage and waves, everyone else just gets polygons that wiggle, as usual. But what happens when I want to build an engine were accurate physics is necessary to damage modeling, or vehicle handling? Are we going to have to build to totally diffirent physics engines? Will one or the other be a competitive advantage in multiplay? etc, etc. I certainly like the idea, in theory, as the concept of having worlds with real physics is quite attractive; but it seems that it will be difficult to scale between those who have the accelerator and those who don't.
Originally posted by: ribbon13
it should be on the GPU
should add that answer "Yes, only if its part of the Video Card"
Originally posted by: MrPickins
A dedicated ppu sounds awesome for gamers, and depending on price, and market acceptance, I think it can do really well.
I dont like the idea of putting it on the graphics card, as I assume they would be competeing for bandwidth eventually, even with on pci express x16 card. Also seems like there might be power and heat issues.
