Again, I think many people are missing the point. It isn't that we need any of this. It is to improve what we already have for physics in current games. We also didn't need 3D accellerators... But why isn't everyone screaming about the need for those? The CPU is capable of handing the graphics, audio and physics... So, why purchase an Audigy? Why purchase anything except a Triden 4Mb video card? That point is, if we continued what was only needed in games, we would still be playing those text games... Again, I hear no one complaining about APU's or GPU's but all of a sudden people are slamming the PPU's and it just isn't logical from a gamers perspective.
Again, I think many people are missing the point. It isn't that we need any of this. It is to improve what we already have for physics in current games. We also didn't need 3D accellerators... But why isn't everyone screaming about the need for those? The CPU is capable of handing the graphics, audio and physics... So, why purchase an Audigy? Why purchase anything except a Triden 4Mb video card? That point is, if we continued what was only needed in games, we would still be playing those text games... Again, I hear no one complaining about APU's or GPU's but all of a sudden people are slamming the PPU's and it just isn't logical from a gamers perspective.
biostud
Lifer
- Feb 27, 2003
- 19,953
- 7,049
- 136
As far as I remember this are not that great looking, no fancy texture etc.
AI, for one. Lots of folks complain about lame AI, pathing, etc. Freeing up the CPU from having to do physics work allows it to perform better in these areas. More polys, collision detection, etc.
Say a new killer game I want comes out, and my computer isn't up to handling it. If my cpu is the bottleneck, I'm looking at spending an extra $400 or more to get one that can ahead of when I'd planned to upgrade. Or I can get a theoretical $200 physics card that should last far longer and provide a lot better improvement in supported games.
Think about sound cards: Don't know about you, but I sure don't upgrade those very often when compared to cpu/gpu. Physics cards would probably be the same. Just having a Soundblaster 16 vs none at all makes a much huger difference than upgrading that 16 to say a 32.
All it's going to take is a hot title or three to really implement this thing well and give a huge bump to immersion and/or display quality (through cpu cycle savings) and a lot of folks will hop on board. I mean you CAN play Half Life 2 on relatively ancient cards, but most of us choose not to. It's worth spending more for a better gaming experience. Can work the same here. Especially in situations like we've got now where a 6800 x800 will let you play most anything fine. Bumping up to 7800/1800 would be nice of course, but not as much bang for your buck. Done right, a physics card could provide that "difference between night and day" gaming experience. Until they come out with the ScentBlaster add-on card
Originally posted by: fishbits
AI, for one. Lots of folks complain about lame AI, pathing, etc. Freeing up the CPU from having to do physics work allows it to perform better in these areas. More polys, collision detection, etc.
Say a new killer game I want comes out, and my computer isn't up to handling it. If my cpu is the bottleneck, I'm looking at spending an extra $400 or more to get one that can ahead of when I'd planned to upgrade. Or I can get a theoretical $200 physics card that should last far longer and provide a lot better improvement in supported games.
Think about sound cards: Don't know about you, but I sure don't upgrade those very often when compared to cpu/gpu. Physics cards would probably be the same. Just having a Soundblaster 16 vs none at all makes a much huger difference than upgrading that 16 to say a 32.
All it's going to take is a hot title or three to really implement this thing well and give a huge bump to immersion and/or display quality (through cpu cycle savings) and a lot of folks will hop on board. I mean you CAN play Half Life 2 on relatively ancient cards, but most of us choose not to. It's worth spending more for a better gaming experience. Can work the same here. Especially in situations like we've got now where a 6800 x800 will let you play most anything fine. Bumping up to 7800/1800 would be nice of course, but not as much bang for your buck. Done right, a physics card could provide that "difference between night and day" gaming experience. Until they come out with the ScentBlaster add-on card
ROFL
Jeff7181
Lifer
- Aug 21, 2002
- 18,368
- 11
- 81
That's a little rediculous. The reason I say that is because people thought having a video card dedicated to 3D graphics was unnecessary. What would the average person need a computer for, let alone a computer with 3D graphics capability?
With this physics processor we're talking about something that's 50-100 times as powerful as current CPU's at physics processing. Even if you have a quad core processor, this PPU will STILL be more powerful. If you don't think it's useful... play around with this NovodeX Rocket Demo. And keep in mind a PPU with 50-100 times the physics processing capability would handle those with ease.
Exactily. When a processor can dedicate itself to one function (physics in this case) it can perform 100 times over... I am interested in this technology and I hope it works out.
- Jan 29, 2005
- 5,202
- 216
- 106
Wow ...
I didn't know this subject would start a long discussion here.
Well, thanks to everyone for taking some of your time and post comments and facts or opinios about all this.
So ... to resume all of the general consensus:
1) It's new, and the majority of gamers don't see a logic behind having a PPU.
2) Even if gamers don't see it right now, that we'll want or not, it won't matter, because games are, as a fact, becoming more and more complex, with more things going on at once.
3) The gaming-evolution in itself will, with time, justify the PPU arriving in our computers.
You know what this makes me think about ?
Let's pretend Planes do not exist. A person suddenly thinks about such a capacity, such a technology. Presents it to some sort of "Council", and they consider that as "fiction", or that it's useless because we have ways around that right now, such as walking, cars, public transport and boats, why would we then need the air to transport ourselves, or even protect our aerial "territory".
But, Council decisions or not ... the innevitable evolution of the human race, and its new needs, kind of asked for such a technology.
Eventually, it was created. And is today considered "normal". It's not like people see something in the air that's not a bird, and then thinks it could be a dragon made of iron.
Well ... that's what I believe.
A Physics Processing Unit does make sense to me, even if, at the moment, at the present, we don't need it.
I believe that innevitably, video-gaming evolution and / or revolutions will justify such a technology. And, perhaps, who knows, one day we'll see some sort of new Processors that include the tecnology itself, plus, being able to do "general calculations".
I think we're entering a new ara of entertainment, where the "simulation of realism" will be much more "accessible", and thus will be more exploited, which in turn will "ask" for new technologies, either existing or not, to be there as a "standard", a normality, for regular support of the constantly increasing complexity within 3-D gaming.
I didn't know this subject would start a long discussion here.
Well, thanks to everyone for taking some of your time and post comments and facts or opinios about all this.
So ... to resume all of the general consensus:
1) It's new, and the majority of gamers don't see a logic behind having a PPU.
2) Even if gamers don't see it right now, that we'll want or not, it won't matter, because games are, as a fact, becoming more and more complex, with more things going on at once.
3) The gaming-evolution in itself will, with time, justify the PPU arriving in our computers.
You know what this makes me think about ?
Let's pretend Planes do not exist. A person suddenly thinks about such a capacity, such a technology. Presents it to some sort of "Council", and they consider that as "fiction", or that it's useless because we have ways around that right now, such as walking, cars, public transport and boats, why would we then need the air to transport ourselves, or even protect our aerial "territory".
But, Council decisions or not ... the innevitable evolution of the human race, and its new needs, kind of asked for such a technology.
Eventually, it was created. And is today considered "normal". It's not like people see something in the air that's not a bird, and then thinks it could be a dragon made of iron.
Well ... that's what I believe.
A Physics Processing Unit does make sense to me, even if, at the moment, at the present, we don't need it.
I believe that innevitably, video-gaming evolution and / or revolutions will justify such a technology. And, perhaps, who knows, one day we'll see some sort of new Processors that include the tecnology itself, plus, being able to do "general calculations".
I think we're entering a new ara of entertainment, where the "simulation of realism" will be much more "accessible", and thus will be more exploited, which in turn will "ask" for new technologies, either existing or not, to be there as a "standard", a normality, for regular support of the constantly increasing complexity within 3-D gaming.
While I can understand concerns over pricing, to want something to fail that could possibly improve our overall gaming experience to me is a little short-sighted. Noone wants a 400PPU that needs to be ungraded as often as their GPU. However, a pretty cheap PPU ($125.00 or less) with as much coverage as discrete GPU cards would do alot to provide us with something more then pretty eye candy.
Graphics-wise games have improved by leaps and bounds, especially when comparing graphic improvement vs. the improvements we have seen in physics or AI. By moving the physics off the CPU, we could possibly see greater improvement in both areas.
Want to know how much improvement is needed in non graphical areas in games like BF2? You can singlehandely destroy any vehicle in BF2. Jet, no problem. Helicopter, no problem. Tank, no problem. Tin shack? Problem.
Graphics-wise games have improved by leaps and bounds, especially when comparing graphic improvement vs. the improvements we have seen in physics or AI. By moving the physics off the CPU, we could possibly see greater improvement in both areas.
Want to know how much improvement is needed in non graphical areas in games like BF2? You can singlehandely destroy any vehicle in BF2. Jet, no problem. Helicopter, no problem. Tank, no problem. Tin shack? Problem.
head, nail, on, hit, you and the.....well said sir
(or madam if not a sir....just in case..heh)
grpahics wise things wont get much better without skipping to photorealism now, so the game producers are going to have to find otherways to make their latest and greatest appeal, while some are capable of dishing out Half-Life type games that just drag you in for almost every reason, the only real thing missing in games is believability in your surroundings, as the guy i quote here says, in a BF2 u can blow the vehicles (tanks, plated armoured metal anyone?) to pieces, yet a ramshackle tin shed is totally oblivious to your C4 or bullets or shells etc.......well, with the PPU that will change, you will (eventually) be able to level a city block by strategic placement of a few C4 charges, and it wont just be a scripted event, you can do it to any block....obviously it will make scriptwriters work harder to make sure this new freedom doesnt detract from the game, but what the hell, they need a challenge, nothing truly new in design has come about since Half Life has it?.....be honest.....yes, Far Cry is good, but its a bloke, in a contained environment trying to get away.....er, Half Life......FEAR (from what i gather) is the same, a contained environment with someone trying to get away.....one problem, you can only get away in a scripted way....even in Far Cry you still need to follow some objectives before you can go further right?......well, what if games came about that truly let you follow your own path, you didnt have to do things in a certain order....now that would be worth a one off PPU chip cost from my point of view....
Ummm...correct me if I'm wrong here, but couldn't the 2nd core of a dual core CPU act as a physics processor?
NVIDIA will build this into their GPUs along with hardware AI acceleration. PhysX add-in cards might be interesting at first, but I expect NVIDIA to jump in with their own PPU as part of their GPU. Dr. Kirk talked about this a while ago, so don't think NVIDIA will be left out. Plus, if the graphics card has all the geometry information anyway, it makes sense for the GPU to do collision dectection, etc.
2nd core will act like a "physics processor", but it's still a "general purpose CPU" and not a "dedicated, hardware physics accelerator". Look at it like a Software Renderer vs. a dedicated Graphics Processing Unit. Basically, if your physics can be done on the PPU then it will be MUCH faster. CPUs can only handle a small fraction of the physics calculations that the PPU can.
2nd core will act like a "physics processor", but it's still a "general purpose CPU" and not a "dedicated, hardware physics accelerator". Look at it like a Software Renderer vs. a dedicated Graphics Processing Unit. Basically, if your physics can be done on the PPU then it will be MUCH faster. CPUs can only handle a small fraction of the physics calculations that the PPU can.
Originally posted by: knyghtbyte
head, nail, on, hit, you and the.....well said sir
(or madam if not a sir....just in case..heh)
grpahics wise things wont get much better without skipping to photorealism now, so the game producers are going to have to find otherways to make their latest and greatest appeal, while some are capable of dishing out Half-Life type games that just drag you in for almost every reason, the only real thing missing in games is believability in your surroundings, as the guy i quote here says, in a BF2 u can blow the vehicles (tanks, plated armoured metal anyone?) to pieces, yet a ramshackle tin shed is totally oblivious to your C4 or bullets or shells etc.......well, with the PPU that will change, you will (eventually) be able to level a city block by strategic placement of a few C4 charges, and it wont just be a scripted event, you can do it to any block....obviously it will make scriptwriters work harder to make sure this new freedom doesnt detract from the game, but what the hell, they need a challenge, nothing truly new in design has come about since Half Life has it?.....be honest.....yes, Far Cry is good, but its a bloke, in a contained environment trying to get away.....er, Half Life......FEAR (from what i gather) is the same, a contained environment with someone trying to get away.....one problem, you can only get away in a scripted way....even in Far Cry you still need to follow some objectives before you can go further right?......well, what if games came about that truly let you follow your own path, you didnt have to do things in a certain order....now that would be worth a one off PPU chip cost from my point of view....
Good point, but the things you mentioned have already been implemented to an extent in certain games. Red Faction had destructible environment, and games without a scripted objective - well, that doesnt require a ppu, all you need is to design the game that way, or just play multi-player games. If you want ultimate realism, like placing a bomb in the middle of a city and seeing how the buildings blow up in a life-like fashion, then you might need a separate ppu, or again, you might just need a dual core cpu.
Also, dont forget that even when discrete hardware is faster in some situation, it might be slower in other situations. Like HW T&L on the geforce 1 or 2 - it was fast when there was only 1 light in the scene, but when you had 8 lights, the software T&L on a fast cpu was actually significantly faster.
In addition, there are already a bunch of issues with games that are not realistic. For example, when you get shot and your health decreases to 5%, are you gonna run around and jump all over the place like at full health? Not in real life, and such a thing would be easily fixed in the software without a ppu, and yet most games dont follow the real life model.
The issue with games is that it only has to look convincingly real, it doesnt have to be a totally realistic model of the real world. And even if you had complex calculations involved, I dont see how a separate 200-400 mhz dedicated processor is gonna do any better than a multi-ghz multi-core cpu.
They only licensed their software physics engine (NovodeX), they are NOT using the PPU.Originally posted by: fliguy84
this thing won't fail because sony licensed the technology for the ps3
They have stated that the PPU can handle the simultaneous calculations for 1000x more objects than current CPUs.
I've had a lot of thoughts about this pretty much as soon as I read about it. I'm surprised it's taken so long for the idea to come around, especially when you consider the physics differences between games, etc.
Anyway, had a great number of thoughts and ideas about this so apologies in advance for the long-ish post coming up. I've got a lot of thoughts, hopes. doubts and worries about the problems that such an addition would make.
Such a card would be great to get correct interactions and of course consistant behaviour in a 'world' as well as open up the possibility to maybe maps/levels that span much greater effective distances.
First thing I thought myself was surely wouldn't this be something a 2nd CPU/SMP/ Dual-core could handle but then of course you end up with bus contention and memory bandwidth issues - especially if the environment is complex or large. In such cases it would probably be better to have all of this work off-loaded to another processor where you would only need to pass it information and wait for the result to appear, and an extra / seperated co-processor would be the way to go for that.
Also, using a CPU to do this is not an elegant solution to the problem compared to a custom built processor that would have design advantages (such as the amount of registers and operations that can be done, calculation accuracy, a pre-programmed look-up table for different materials/reactions, etc.) and its own seperate bit of memory so as not to tie-up the PCI/memory bus.
There are other problems though. Such a card would have to compete and gain acceptance from developers, many of whom may already have their own physics engines and get revenue from their use in games - such a card would surely work against them unless they could make use of it somehow but I don't see how without re-writing code to make use of it. And at the end of the day surely the card/processor manufacturer would be in the driving seat as all they need to do is release their own SDK and all other physics software companies would be screwed.....
Another problem is differing specifications: companies like to make different spec products to suit market segments. If they make a 'lite' version of the card how would that affect compatibility with some games, etc.
If they do release a card I doubt it will see wide-spread acceptance. It would just be a fore-runner to adoption onto motherboards or into chipsets themselves and even then I doubt it would be that wide-spread.
And what of multi-player online gaming. Someone who has the physics processor vs. someone who doesn't??? How will it effect gameplay, fairness, cheating, etc.....????
Luckily none of this is my problem.
Anyway, had a great number of thoughts and ideas about this so apologies in advance for the long-ish post coming up. I've got a lot of thoughts, hopes. doubts and worries about the problems that such an addition would make.
Such a card would be great to get correct interactions and of course consistant behaviour in a 'world' as well as open up the possibility to maybe maps/levels that span much greater effective distances.
First thing I thought myself was surely wouldn't this be something a 2nd CPU/SMP/ Dual-core could handle but then of course you end up with bus contention and memory bandwidth issues - especially if the environment is complex or large. In such cases it would probably be better to have all of this work off-loaded to another processor where you would only need to pass it information and wait for the result to appear, and an extra / seperated co-processor would be the way to go for that.
Also, using a CPU to do this is not an elegant solution to the problem compared to a custom built processor that would have design advantages (such as the amount of registers and operations that can be done, calculation accuracy, a pre-programmed look-up table for different materials/reactions, etc.) and its own seperate bit of memory so as not to tie-up the PCI/memory bus.
There are other problems though. Such a card would have to compete and gain acceptance from developers, many of whom may already have their own physics engines and get revenue from their use in games - such a card would surely work against them unless they could make use of it somehow but I don't see how without re-writing code to make use of it. And at the end of the day surely the card/processor manufacturer would be in the driving seat as all they need to do is release their own SDK and all other physics software companies would be screwed.....
Another problem is differing specifications: companies like to make different spec products to suit market segments. If they make a 'lite' version of the card how would that affect compatibility with some games, etc.
If they do release a card I doubt it will see wide-spread acceptance. It would just be a fore-runner to adoption onto motherboards or into chipsets themselves and even then I doubt it would be that wide-spread.
And what of multi-player online gaming. Someone who has the physics processor vs. someone who doesn't??? How will it effect gameplay, fairness, cheating, etc.....????
Luckily none of this is my problem.
its been around for a while now but how practicle is it , in my opinion AMD or INTEL will end up with this tech built into a second or third cpu core , it just seems logical to go that way hard ware wise , but the idea behind it sends gamers glimses of brilliance in games and there potential.
l say bring on phyx chip tech n lets see where it goes .....vr maybe......
l say bring on phyx chip tech n lets see where it goes .....vr maybe......
No, AMD and Intel won't adopt this to a new core...why would they? AMD nor Intel care about how well a game performs on a persons computer, that isn't their target audience. It will be included, if with anything, on the video cards themselves...obviously the only people who want our games to perform better are the people who make our graphic cards. say the 8800 Ultra has a PPU and the 8600GT does not....makes sense to me
Bet on soldier will be the first game released to support the physics chip. I wouldnt expect anything revolutionary until unreal 3 engine at the soonest. WTf ever happened to STALKER?
Any game that supports the physics chip will have the benefit of being able to take a major load off the CPU. Now the real question is, how long will it take developers to deliver content that can only bee seen with the physics chip.
Any game that supports the physics chip will have the benefit of being able to take a major load off the CPU. Now the real question is, how long will it take developers to deliver content that can only bee seen with the physics chip.
Persionally i'm interested to see what this can do for physics engines in games. For those that say "why bother with a slower PPU when you have a multi GHz cpu?", If a graphics card that is specifically designed to render 3D video and is clocked at 500-600MHz at the most extreme can far outperform a 2.4+GHz proc very easily on the 3d mark tests, what's to say that a PPU that has an architecture specifically designed to handle insane amounts of physics calculations specifically won't outperform that same proc? As for the argument of "we don't need it now", sure, we might not need it at this instant, but the 3d accelerator card wasn't exactly needed when it came out, and now it's all but impossible to game without it, game companies will program for the technology that is avalible. Once this tech has been out for a year i'm pretty sure we'll see quite a few games that can take advantage of this, or will even require it to unlock the highest lvl of physics detail in the world.
I dont see the gfx card and the ppu as being of equal value to the cpu offloading. The cpu is best suited for number crunching, and graphics is not about crunching numbers for the most part. At the most basic level a gfx card is designed for pushing pixels to the display as fast as possible, it's not something that the cpu was designed for, and even if the cpu could output billions of pixels per second, the actual pixels would eventually have to go through the system memory and the bios, and that would bottleneck the whole process severely. Even with HW T&L, the gpu basically performs a limited, predefined set of math operations on a bunch of matrices, and by limiting the instruction set of the gpu, it can be optimized for those instructions better than a general purpose cpu.
Physics, on the other hand, can be boiled down to a bunch of math equations, but those equations can get quite complex, and thus are adequately suitable for a cpu. Only in limited cases would you need a dedicated ppu, when you need to calculate a simple interaction of forces between millions of objects. But this is only true if the math is simple enough to be broken down into a SIMD model (single instruction, multiple data), and can be fully performed using the simple instructions of dedicated hardware. If the math turns out to be complex, then only a cluster of cpu's working in parallel (aka a supercomputer) would be fast enough to do the math.
You do have a good point there, however i still believe that there is potential for the hardware to improve the physics of modern games. With the advent of the PPU the next logical step becomes a few standard physics engines that are used the same way game engines are used, which are ultimately designed to take specific advantage of an indipendent PPU. If nothing else it will drive programmers to new levels of realism in the physics engines, to the point where in a few years the havok engine will look like nothing at all, because the whole system will become standardized. As for the math operations getting complex it would be possible to pre-define the solution methods to all of the various equations that are likely to be encountered in everyday gaming, which will greatly accelerate the physics. I persionally think that it will take at least a year before we truly see the advantage of the PPU, but at the same time i fully expect that they will be there.
It will never get integrated into a CPU - a PPU would require its own bit of memory and probably a fair bit of it if it has to store a particularly large environment packed full of different objects all with different properties attached.
For example, take a game like Serious Sam where the levels in that were huge - if we started modelling every element in the game to the point of where every building was constructed of blocks we could destory until it fell down, a whole lot of memory will be needed. With a CPU the memory would either have to be on-chip which is unlikely or use the system memory which eventually means that an integrated PPU on a CPU would be of little benefit especially in this era of SMP on a single chip.
The graphics card idea actually holds a bit of promise. For example, as DirectX ramps up in versions so do the graphics architectures. If the physics requirements go up also then as you upgrade a graphics card from say a DX10 level to a DX12 level you get the improvements in physics also. Also graphics cards usually have a lot of RAM on anyway so a PPU with its own little bit of memory isn't gonna make a massive difference in some cases especially if it doesn't need lightning fast RAM - it could just take up a small space o have the PPU chip itself and a single memory chip (after all high-density memory is not too hard to get anymore). I'm sure ATI or nVidia can be persuaded to add a chip to the boards if it proves even mildly popular or beneficial. Besides, it wouldn't be the first time we've seen 3D cards with extra chips on offering sound output, etc., all on the same board.
Integrating the PPU into the graphics chip would be good but then you have 2 things using up the graphics RAM bandwidth which would hurt performance so that wouldn't be in their interest.
For example, take a game like Serious Sam where the levels in that were huge - if we started modelling every element in the game to the point of where every building was constructed of blocks we could destory until it fell down, a whole lot of memory will be needed. With a CPU the memory would either have to be on-chip which is unlikely or use the system memory which eventually means that an integrated PPU on a CPU would be of little benefit especially in this era of SMP on a single chip.
The graphics card idea actually holds a bit of promise. For example, as DirectX ramps up in versions so do the graphics architectures. If the physics requirements go up also then as you upgrade a graphics card from say a DX10 level to a DX12 level you get the improvements in physics also. Also graphics cards usually have a lot of RAM on anyway so a PPU with its own little bit of memory isn't gonna make a massive difference in some cases especially if it doesn't need lightning fast RAM - it could just take up a small space o have the PPU chip itself and a single memory chip (after all high-density memory is not too hard to get anymore). I'm sure ATI or nVidia can be persuaded to add a chip to the boards if it proves even mildly popular or beneficial. Besides, it wouldn't be the first time we've seen 3D cards with extra chips on offering sound output, etc., all on the same board.
Integrating the PPU into the graphics chip would be good but then you have 2 things using up the graphics RAM bandwidth which would hurt performance so that wouldn't be in their interest.
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