View Full Version : Physics Graphic card by Nvidia
Graphicmaniac
08-12-05, 10:48 AM
from a news i have read here:
http://www.nvitalia.com/stories.php?story=05/08/12/9648314
It seems like Nvidia's next gen-graphic card will support physics engines in hardware as AGEIA will do. This will start with next genaration games as unreal 2007 and gears of war that will use the new generation physics engine Novodex.
thats the translated version:
http://translate.google.com/translate?u=http%3A%2F%2Fwww.nvitalia.com%2Fstorie s.php%3Fstory%3D05%2F08%2F12%2F9648314&langpair=it%7Cen&hl=it&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools
Interesting. Thanks for the link. I'd much rather see physics processors put onto graphics cards instead of having to buy another card for games.
Interesting. Thanks for the link. I'd much rather see physics processors put onto graphics cards instead of having to buy another card for games.
I dont see the difference, from a pirce stand-point at least. I'm SURE including a phyics processors onto the gpu will mean $200 more for these damn cards.....$799 for a G80 sounds right... :mad:
Pandora's Box
08-12-05, 01:32 PM
I dont see the difference, from a pirce stand-point at least. I'm SURE including a phyics processors onto the gpu will mean $200 more for these damn cards.....$799 for a G80 sounds right... :mad:
im hoping you can buy versions without the physic processor. that way if you already have one your not throwing your cash away.
I'm sure they will sell versions without the physics chip.
Graphicmaniac
08-12-05, 02:06 PM
well maybe it will be integrated into the gpu making it cheaper than gpu+ppu
well maybe it will be integrated into the gpu making it cheaper than gpu+ppu
Right. It might add to the cost of a graphics card by $100, but it makes more sense to integrate it into the gpu as only gamers would take advantage of hardware driven physics.
The GPU would become the GPPU. :D
TierMann
08-12-05, 03:14 PM
Didn't we just get gpu+vpu working correctly.. kinda?
Brimstone7
08-12-05, 03:19 PM
The only problem with this idea is that it's likely that physics processors won't need to be upgraded as often as GPUs. So, if you buy a card that has both features together, you're almost throwing your money away when you upgrade cards because the physics processor is still not dated.
Interesting info .. but today's GPUs are already running hot & requires enormous cooling solutions to run properly , how that would be possible with an extra PPU on the same graphics board ?
frenchy2k1
08-12-05, 05:27 PM
My guess would be that no change is required outside of drivers and software support.
The next generation of GPU will be completly programmable, with unified shader architecture (it was the plan anyway).
Those are almost general purpose processors, or, more to the point, general purpose vector units. Physics processing shares a lot with graphics. Both are vector intensive.
Now, you would just split all those vector processing power between 2 tasks: physics and graphics.
That certainly could be a solution...
If they can make the price reasonable, I'd rather have it on all of them. That way it makes the technology get accepted a lot quicker than if you have to worry about some having the technology, some not having it. Kind of like the old days when the math co-processor was a separate piece of hardware and because not everyone had one, support for it was sometimes lacking in software.
That brings us to another question .. if i bought for instance 2 graphic cards (SLI) & both have there own onboard PPUs , are we going to see PPUs work in SLI ?? :D
AlphaWolf_HK
08-12-05, 05:49 PM
I guess they would eventually start calling these "gaming cards."
Old sound cards used to also be called this. Back in the early early days of PCs, you'd have a card that had a just db15 gaming port which analog joysticks could be plugged into. Then later when soundcards started becoming more common, the game port was commonly integrated into the soundcard (and this port was also used for MIDI instruments.)
IIRC, creative SB16 cards were the last to have game ports.
A physics card wouldn't really have much use beyond gaming, so the business application of these wouldn't be economically practical.
This wouldn't make a lot of sense though if nvidia plans on continuing with their SLI configurations. The only way you could conceivably not waste money in that configuration is if the physics calculations were somehow multithreaded.
Those are almost general purpose processors, or, more to the point, general purpose vector units. Physics processing shares a lot with graphics. Both are vector intensive.
Now, you would just split all those vector processing power between 2 tasks: physics and graphics.
I don't know if that would be a good idea to share them, theres bound to be a bottleneck somewhere in doing that.
this will not cost anything ...except die space... i could see this being made into something like a vs unit..and it could be programed to with custom extensions...i just dont see this costing us anything more..it would be like saying wfg 2.0 cards are gonna cost us more becouse theres more transisters in there....
Interesting info .. but today's GPUs are already running hot & requires enormous cooling solutions to run properly , how that would be possible with an extra PPU on the same graphics board ?
My thoughts aswell..
Why would they want to SHARE the bus, put another heatsource that also needs extra power, and either share the RAM or have its own dedicated RAM (more heat, power and space) on the already stretched video cards we got today..
why not fit them on Mobos that arent exactly overfull these days with 1 chip solutions and all that..
Or as separate cards like they are ment to be initially.
Why would they want to SHARE the bus, put another heatsource that also needs extra power, and either share the RAM or have its own dedicated RAM becouse the bus isnt even being pushed now as it is... and now theres talk of releasing pci-e 32 slots sometime next year ..the heat souce problem would be like adding more vs units to the card... and remeber the next stop is 65 nm for this tech.. (more heat, power and space) on the already stretched video cards we got today.. we have been hearing that video cards are strethed for years...if i told anyone here four years ago that we would have 300 million tranny card you would all laugh at me.. i think having the ppu on the card is a great thing ... this could be the first app that takes full advantage of pci-16 in all of its glory
Interesting info .. but today's GPUs are already running hot & requires enormous cooling solutions to run properly , how that would be possible with an extra PPU on the same graphics board ?
There will come a time, when the VGA cards will take sizes of Micro-Atx Motherboards. the nv cards with this size will have GeForce, Soundstorm, PPU, USB, etc etc on it and nv will cal it "The Multimedia Card" or "The Gamers Delight"
:D j/k :D
i think they would bundle a PPU along with the GPU/Card.
I mean they are getting longer and heavier, I could barely fit my 6800GT in my case.
And aside from extra RAM and more routing on future cards, they want to put another chip and possibly more RAM chips for the PPU..
I dont think a PPU would add alot of traffic on the bus, since the bulk of "data" is calculated on the card, and not a whole lot is sent over the bus, and I never said the PCI-E 16x is anywhere near saturated, but still, I can imagine routing and stuff like that to mean some extra work, and shouldnt there be latancies and things with sharing? maybe not..
Well I dunno, if they really plan to do it, It cant be that bad, or they just wouldnt have..
But I would rather have it on a separate card, or on the mobo..
dunno, but I just dont like the idea of putting stuff on my graphics card hehe.
Redeemed
08-12-05, 06:51 PM
That bundle idea sounds better. What will we need to run a vcard that has a PPU on it, a 500watt psu? Not every gamer is the extreme-uber-high-tech-gotta'-have-the-latest-and-greatest-technology kind of gamer. Some just want enough to play the games and no more. For those people, a 500watt psu is kinda' extreme.
But I'm not one of those people... ;) :D :D :D
I do think it is a bad idea to have both the PPU and GPU combined on the same card, at least initially. A second card sounds more realistic OR adding the PPU's core to the GPU die- that MIGHT be a possibility. But what about ram then? There is already enough of a bottleneck with video memory, a processing unit that shares the vcard's memory would only INCREASE that bottleneck.
Then, the whole aspect of the bottleneck with the CPU; if the CPU can't keep up with the modern GPUs, then what do you think will happen when we add a PPU to that GPU? MORE bottleneck.
I feel that the PPU should be a seperate card solution AT FIRST. And THEN it should slowly be integrated onto the mainboard. This way, anybody could enjoy enhanced and advanced physics in their games. The PPU could integrated into both the mainstream and high end mobos leaving the low end mobos for those whom use their computers for nothing more than e-mail, web browsing, and other such simple tasks. Whereas people with a more "mid-range" priced mobo probably play games occasionaly therefore could better utilise the PPU. And for the "high-end" mobos, well, that should be obvious. ;) :D :D :D
That bundle idea sounds better. What will we need to run a vcard that has a PPU on it, a 500watt psu? Not every gamer is the extreme-uber-high-tech-gotta'-have-the-latest-and-greatest-technology kind of gamer. Some just want enough to play the games and no more. For those people, a 500watt psu is kinda' extreme.
and for these ppl, nv makes 5200's 6200's..and will continue makin 7200's n 8200's n so on wich dont need ppu, no psu, no nuthing.
IMO, PPu should/will either be bundled along with the VGA or should be integrated on the mainboard. maybe they are already integrating it, nforce5 should have it.
Just imaginr the length of a GPU + PPU on PCB. reminds me asus dual 6800GT's on same PCB.
Subtestube
08-12-05, 07:31 PM
I dont think a PPU would add alot of traffic on the bus, since the bulk of "data" is calculated on the card, and not a whole lot is sent over the bus, and I never said the PCI-E 16x is anywhere near saturated, but still, I can imagine routing and stuff like that to mean some extra work, and shouldnt there be latancies and things with sharing? maybe not..
My opinion is that it would add a LOT of traffic - just realise that physics isn't like graphics, in that it isn't the 'last stage of the pipeline.' Everything you calculate, unless it's something that just looks cool and has no effect on the game world, has to be fed back to main memory so that everything else can use the new positions of objects in calculating what happens to them.
That said, PCIe IS full duplex, so I strongly doubt that object position and state data (which would be fairly trivial per object) would come even close to 'maxing' it out.
walterman
08-12-05, 08:09 PM
Do we really need physics on chip ?
I mean, cause today, we've 2-3 GHz chips, & multi-core era has started too.
As far as i know, a SSE2 (2x64) multiplication instruction takes 4 cycles in an Athlon64 & 5-6 in the P4. & If you use simple precision (SSE), 4x32.
So, a single core A64 @ 2 GHz, can do around (2000M / 4 cycles) * 2 ops = 1000 Millions of double precission floating point multiplications.
I dunno how many precision is needed for physics, but, 1000M of operations per second, at lets say 50 fps, means, 20M of operations per frame.
Try to imagine a scene where 20M of multiplications are needed for the physics only.
Add dual core & single precision & you get 80M of ops per frame.
Anyway, these are just dumb calculations, real world performance is much worse, due to poor or non existent optimizations, or lots of other things.
& Next year, maybe quad core chips, & higher frequency ....
So, aren't those new physics chips another new hardware/software complot to get more of our money ?
Bah, i'm just anger cause they want to sell those new physics cards for 200-300$
dude you really need to read on how much a cpu can handle as opposed to a dedicated asic... if you statement was true then harware vendors would use nothing but cpu's they would completly avoid using acis. switches and routers use asics becouse they are just faster at certain functions ...period...and this happens to be one of them
sandeep
08-13-05, 01:09 AM
I read somewhere that physics processors will be transient in their physical existence and quad core CPUs will take over the work of physics processing.
Also, motherboards are running out of PCI slots itself. So an extral card would just exacerbate the scarcity. For short term prognosis, there might be ppl who would just prefer an integrated version into Graphics card and some for the alternative til the quad Core CPUs show up.
vBulletin® v3.7.1, Copyright ©2000-2010, Jelsoft Enterprises Ltd.