Looks a bit more promising than I was expecting, great article, good that you took the time to say wich mobo's are compatable with SLI, a PhysX card and another PCI card.
I think this boxing stuff is getting out of hand, I mean this must be starting to cost quite a bit to come up with these ideas and make them, I'd rather have a rectangular box just big enough to fit everything inside and save a few quid than have a fancy box.
I would've liked it though if you did a third test with a single high end card, and not just high end SLI and a mid range card. But it got the point across anyway :)
Originally Posted by Veles I would've liked it though if you did a third test with a single high end card, and not just high end SLI and a mid range card. But it got the point across anyway :)
That was something that we wanted to do, but unfortunately we ran out of time. We had to choose between ultra high end, high end and mid-range and chose either end of the spectrum.
Originally Posted by Boon The nVidia (and ATI) Havok based onboard solution seems to be the best of both worlds and is quite simply the future for dedicated Physics processing.
That's not quite right. Physics on the graphics card has two major drawbacks:
It can't be used for anything other than eye candy, since the new positions of objects cannot be read back into the main memory from the game.
It requires two graphics cards, of which one must then be dedicated to physics processing.
I'm not sure how permanent either of these drawbacks is. Perhaps someone could shed light on this?
I see two drawbacks with the PhysX card so far:
High price.
Lack of applications.
The price should naturally go down over time. The lack of applications is not permanent, as mentioned in the article.
Personally, I wouldn't buy a PhysX card right now, but I'll be buying one the day UT2007 comes out. :D
They do have a long road ahead though before the properly intergrated into a game engine the cell factor demo is pretty intense but even that slows down when the action gets stupidly intense
the card wasnt designed to only make chunks of debris fly off walls when you shoot stuff. Its a physics processing unit so it manages everything physics-related basically. So if HL2: Episode 2 optimized the game for the PPU, barrels flying,wood breaking, and object throwing would be handled by this PPU, taking a load off your other units. Quad-core CPUs may be an expensive solution, but if youre builing a PC from scratch (like me) its probably a better deal than getting an PhysX card.
HL is based on a joint effort between Valve and Havok iirc (don't hold me to that) so it won't support PhyisX. If GoW was on the PC it would because it's UT3 engine.
Comments 26 to 33 of 33
I think this boxing stuff is getting out of hand, I mean this must be starting to cost quite a bit to come up with these ideas and make them, I'd rather have a rectangular box just big enough to fit everything inside and save a few quid than have a fancy box.
I would've liked it though if you did a third test with a single high end card, and not just high end SLI and a mid range card. But it got the point across anyway :)
- It can't be used for anything other than eye candy, since the new positions of objects cannot be read back into the main memory from the game.
- It requires two graphics cards, of which one must then be dedicated to physics processing.
I'm not sure how permanent either of these drawbacks is. Perhaps someone could shed light on this?I see two drawbacks with the PhysX card so far:
- High price.
- Lack of applications.
The price should naturally go down over time. The lack of applications is not permanent, as mentioned in the article.Personally, I wouldn't buy a PhysX card right now, but I'll be buying one the day UT2007 comes out. :D
i might be the first to watercool one though
They do have a long road ahead though before the properly intergrated into a game engine the cell factor demo is pretty intense but even that slows down when the action gets stupidly intense