I've been reading about the new Physics Processing Unit card that is coming out. However in the future there will be processors with 4 or more cores on them. Wouldn't it be easier (and less cluttered in the case with all those cards) to just code a game to take advantage of one of the cores on the cpu for physics calculations rather than adding in another card?
-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
Multicore vs. PPU
- Thread starter CJP
- Start date
I agree with browntown, a dedicated PPU would be much more efficient than using one core of a CPU, especially considering that using a CPU core requires use of memory bandwidth from the main system, whereas a PPU has its own local memory buffer like a video card. Its one of the reasons descrete graphics are so much faster than IGPs.
That being said, games will not be able to take advantage of the full hardware capabilities of the PPU card for quite a while, since they can simulate physics for enough objects that the CPU and graphics simply cant keep up except at low resolutions. I think there is a thread linking to an article about this somewhere in the forums. In this sense, I think using a CPU core might be more ideal in the short term, since the amount of physics it would be able to generate effectively wouldn't bottleneck the CPU and GPU.
That being said, games will not be able to take advantage of the full hardware capabilities of the PPU card for quite a while, since they can simulate physics for enough objects that the CPU and graphics simply cant keep up except at low resolutions. I think there is a thread linking to an article about this somewhere in the forums. In this sense, I think using a CPU core might be more ideal in the short term, since the amount of physics it would be able to generate effectively wouldn't bottleneck the CPU and GPU.
Ageia claims the Physx can handle 10000 times the complexity that a normal cpu can handle. Thats orders of magnitude more powerful.
Id hate to see it built into a graphics card, because the graphics cards are going to be on a much faster cycle. Much like the "All in wonder" ATi Radeons. Youll be paying for the same chip over and over every time you buy a graphics card.
Ageia claims the PPU is fully programmable and the drivers and API can be edited to add new functionality. They also say the life cycle of the PPU will be more akin to a sound card than a video card.
Id hate to see it built into a graphics card, because the graphics cards are going to be on a much faster cycle. Much like the "All in wonder" ATi Radeons. Youll be paying for the same chip over and over every time you buy a graphics card.
Ageia claims the PPU is fully programmable and the drivers and API can be edited to add new functionality. They also say the life cycle of the PPU will be more akin to a sound card than a video card.
Check out my sig, the video there was captured using a current high end pc.
The game is also being made available for the xbox 360, which makes me think that 2 of those IBM cores and the very high speed memory subsystem we can only dream of on a PC is enough to push that kind of physics around.
What woud be the difficulty in designing a cpu with multicores, which have dedicated cpu/system cores, and then dedicated specialty cores....
I think the idea goes against the more compact mode we have been moving to....Now we will have 2 bay wide vid cards (possibly X2 for SLI or crossfire) then another bay froo PPU....3-5 bays/slots for gaming??? What a waste...I will stick with gaming consoles...
I like my PCs now...I have a mobo, vid card, use the 6.1 sound on the board, use the onboard Lan, and I am set....less clutter allows for better air flow and cooler/quieter running PCs....
I think the idea goes against the more compact mode we have been moving to....Now we will have 2 bay wide vid cards (possibly X2 for SLI or crossfire) then another bay froo PPU....3-5 bays/slots for gaming??? What a waste...I will stick with gaming consoles...
I like my PCs now...I have a mobo, vid card, use the 6.1 sound on the board, use the onboard Lan, and I am set....less clutter allows for better air flow and cooler/quieter running PCs....
The difference between those ppu demo's and real games is that the games already use a lot of approximations on objects to make them easier to render. For exampe, trees use alpha textures for leaves. If you want to accurately model the physics of those leaves on a ppu, you'd have to convert all those leaves to individual polygons, and no sane game developer would do such a thing (at least in the near future). It would be nice to have some less extreme physics effects like destructible envoronments and objects, but in that regard I still havent seen multicore cpu's used to full potential, so it's hard to tell just how much of a practical improvement the ppu will bring.
Both parties are right. That is: the people who say that a dedicated PPU card will be more capable than a general-purpose CPU are correct. And the people who say it is easier to dedicate one core of a multi-core processor to a task like physics, than it is to get people to buy another add-in card are also correct.
My prediction: the PPU will not survive.
Not enough people will have them to make it worth while for game developers to provide support for both hardware paths. Also, as CPUs get increasingly powerful with the addition of extra cores, the physics you can generate on the CPU will be more than adequate for games.
Now they might avert the inevitability of my prognosis by getting mobo manufacturers to embed their hardware.
My prediction: the PPU will not survive.
Not enough people will have them to make it worth while for game developers to provide support for both hardware paths. Also, as CPUs get increasingly powerful with the addition of extra cores, the physics you can generate on the CPU will be more than adequate for games.
Now they might avert the inevitability of my prognosis by getting mobo manufacturers to embed their hardware.
None of the things you said addressed memory bandwidth constraints.
On top of that, the Ageia SDK natively had hardware and software paths for people that dont have hardware accelerated physics. So it doesnt actually increase develpment cost any more than havok or other physics engines of today.
thats 1 RAM chip, not exactly a hige real estate imo, but anyways Ageia is going backrupt in 5 years, and I say good ridance, we already have a piece of hardware whose only purpose is games (gfx cards), any additional hardware like physics or AI calculation can be done there and do not requrie another card. Physics in games are great, and they will continue to increase in the future, but an additional accelerator card is not the solution.
bah, you never said anything about wanting a 128bit bus. but anyways, I can jsut say that would could put it all on one die, deffinitely all in one MCM that you stick into an HT connected socket on a board. But my point was not to argue the validity of putting a chip on a motherboard becasue that is an idiotic idea, my point was to say that this is something that belongs on a graphics card, and not on its own card.
There's a company that is (in a sense) doing just that...
NewsTech Article
It's a programmable coprocessor that drops right into any Opteron CPU socket.
This allows the much higher memory bandwidth as well as coherent HT links to the other processors...
bunnyfubbles
Lifer
Are you saying the current GPU can do it, or are you suggesting the additional hardware can just be added to one card (whether adding parts onto the GPU, or having a 2nd physical chip somewhere else on the card)?
I love the idea of the PPU, but I don't like the idea of CPU (multicore or not) or GPU doing the work if they simply can't do it as well.
Advanced physics is the future of gaming. You can keep making the games prettier and prettier, but they're not going to get anymore fun until you can skip looking for the blue key because you can blow a hole in the blue door and keep moving on...
I am not sure what would be the best idea between having a second chip on the video card (which people already do for SLI on a card solutions), or if you could add some additional hardware on the front end to allow the GPUs pipielines to do the calculations. So, look at ATIs next generation cards with 48 or so unified shader pipelines, and maybe in addition to doind vertex, geometry, and pixel shaders, they can also do physics calculations. Sure you are gonna need a bigger die, and more memmory for everything to work smoothly, but its gotta be cheaper and more efficient than a 2 card solution. In a room with little physics you will have your physics card sitting idle, but if its a combined solution this jsut means that you have more RAM and ALU bandwidth open to the visual calculations. On the flip side, when a big explosion or something happens you can borrow some bandwidth for awhile. If you look at the tech demos they show tons of explosion all at the same time.
Also, since you already have the geometry information being worked on in the GPUs memmory the CPU never even has to get involved.
Also, since you already have the geometry information being worked on in the GPUs memmory the CPU never even has to get involved.
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: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
