Originally Posted by moshpit How did you guys let him say this and NOT confront it when there's TONS of examples of games that use more then one core successfully. Supreme Commander must not exist in their little world...
Actually they are right
Supreme commander, for example, is only capable of using 3 cores
1. Simulation
2. Graphics Driver Management
3. Audio Management
Its hardly multi-threading, at a great level, and this is what valve were talking about when Bit interviewed them all that time ago
What we did [in GRAW 2] was to work very closely with Grin and create this level, which is only available to those who have a physics card. It doesnt technically require the card, but youd have a slideshow without one. Everything in the level is physically simulated.
So, unlock the level so I can see how crappy it runs on my system w/o a Physx card. Honestly - what drives hardware upgrades with us hardcore gamers? Poor game performance. I only upgrade when a new game I've purchased doesn't perform extremely well, if everything's running hunky dorky I hang on to my cash until next years $600 video card is out.
Originally Posted by completemadness Its hardly multi-threading, at a great level, and this is what valve were talking about when Bit interviewed them all that time ago
I think Valve will be using four cores in Episode 2 if they're available... we'll have to see whether that turns out to be the case or not though. SupCom uses several cores, but the performance difference between 2 and 4 isn't that massive to be fair, even at low resolutions where you're CPU limited. A good example of threading is ray-tracing as you see near 100 percent gains - that's something you don't see in SupCom.
Originally Posted by Tim S A good example of threading is ray-tracing as you see near 100 percent gains - that's something you don't see in SupCom.
however, ray tracing isn't actually the best way to do gfx, and its very inefficient
Databases have always been the best way to show threading, something where you can ask for data here, calculate there, print there, etc etc
Say i want to request records 1-100 - if i use one thread, i request record 1, wait, get it, request record 2
With threads i can make 100 threads asking for records 1-100 (1 each) and that way you don't stall
Sadly, just plopping things over extra cores isn't as simple as it looks
Also as to sup comm performance, as it only uses 3 cores max (and 1 is for sound) you really only need dual core, and as the clock better, and overclock better then the quaddies, for the time being dual core > quad for sup comm
you won't see any difference between dual and quad-core in current games at least - they're graphics card limited at the resolutions you're going to play at.
Hey Guys, this is one of my first posts, but I have been a reader for a little over 6 months. But, I digress, lets get back to subject. Where to start with Physx, oh boy this is going to be a doozy.
Alright, what the guy over at Ageia was stating that about multi threading and how it has taken the game developers a very long time to try and develop multi threaded games.
E.G.
Quote:
When it comes to the CPU side, dual-core, quad-core, whatever then the main problem is threading. How are you ever going to thread the two things together? It’s all about timing, when the physic effect hits then how is the second core going to time it and cooperate? At the moment, there’s not a single game that supports multi-threading even at a basic level. I reckon we’re years out with that and it’s already been about for two years. The games that are being developed now only use it a bit, for A.I. and so on where they don’t need extreme threading.”
I think what he is sighting is the trend that game developers and the gaming industry is the slowest to adapt when it comes to multi threading. There are games out there that already have multi threading and games that will be support multi threading. Patches for games like FEAR, COD 2, Quake 4 (If I remember correctly) & Supreme Commander. Its true that games are slow to adapt, but some games that will support and/benefit greatly from multiple cores are Unreal Tournament 3, Crysis, and Alan Wake just to name a few. The difference between a game that was patched and a game that is going to be designed for multi-core is vast. I'm not an expert but if my thoughts are correct, the games designed for multi core will have a better written code and will see better performance versus a game that was patched to support multi-core. He's trying to use the game industry's slow adaptation for adapting to new technology as a type of Achilles for hardware. In reality its not really about how you time things versus write the code properly. Moving on though, the way he says that its we're years from adapting extreme threading is a little out there on the bs side. No offense to Dan who probably has a master's in engineering and decades of experiences with computer technology, but if your development team works well together, anything is pretty much possible, if conditions are ideal.
In reality I think we should take what he says with an extreme grain of salt. An extreme type of multi threading in this case would probably be Alan wake, when I saw the guys at Intel demoing the game with a Kentsfield processor and an Nvidia Geforce 7900GTX I have to admit I was very impressed. They stated one of the cores was dedicated for physics and the other three were available. That would probably be one of the only "extreme" cases of multi threading there is. But from what it seems like the "extreme" level he's talking about is a game where you can shoot a bullet from a mile away and watch it hopefully hit or miss the fly thats on the wall. Taking into consideration of if the game had wind factors, gravity, accuracy of the gun and the speed of the bullet. That would probably be an "extreme" case where the Physx card would be perfect for and where game developers would not want to pour that much time and resources into try and developing the game in physics. But, I don't think this is the case with most games.
Another point I want to try and bring up is the graphics physics using video cards and the Phys X card. We had a huge discussion about this over at OC forums and well I looked into it and well if the guys over at ATI and Nvidia were able to create a Crossfire/SLI solution to physics, it would be mainly be considered eye candy and not "true physics." This is where the Phys X card would win out. ATI and Nvidia's solutions are not "true physics" rather its more eye candy. The objects would be treated to a degree as physical items, but the possibilities and combinations could be limited to how well the graphics card could process. Added in to how much time, effort and thought was into configuring a game to utilize dual card physics. So in this area the Ageia guys would win out.
Alright, enough of the technical spiel, moving onto the multi-core versus dedicate physics card. From a lot of places I have read, ranging from HardOCP to good old Bit-Tech, it seems that the truth is that a physics card can do many more actual calculations than a single core on a processor can. Now before you jump down my throat and say, "But what if you had an 8 core machine or a quad core and the developers dedicated even more cores to the system," lets try to think about this logically. Would it be more effective and feasible to try and dedicate a unit that can process all the physical calculations versus dedicating 2-3 cores which could be used for processing special effects such as AI, Level, Special effects on the map, etc etc?
All in all the guys at Ageia really want to try and justify the price for the Phys X card, but for $300 I would not spend that much on a calculation unit. The thing is very innovative I'll give it that, but they haven't done so hot ever since they came out back in 05. When I first heard about it, I was all like "Oh holy crap! I have to get me one of those!" But now that I think about it, the price is not very justified. Ranging from the price to the very few developers that want to get on board with them, I say they are kinda dead in the water. Unreal Tournament 3 will help them piggyback and help encourage people to adopt and buy the unit, but until the price comes down I don't think it's very justified. Even for the elite and enthusiast guys, it would not be a very smart buy. All in all, get more developers and try to get more people on board with it and you might see it become more mainstream. But the physics unit might be more valuable for more than just gaming, I mean there is so much untapped potential in that card, I wonder if they can try to look into other industries or other fields that could really use that card. That's my two cents, please critique if you guys have any thoughts on my post.
Originally Posted by InsaneManiac I think what he is sighting is the trend that game developers and the gaming industry is the slowest to adapt when it comes to multi threading. There are games out there that already have multi threading and games that will be support multi threading. Patches for games like FEAR, COD 2, Quake 4 (If I remember correctly) & Supreme Commander. Its true that games are slow to adapt, but some games that will support and/benefit greatly from multiple cores are Unreal Tournament 3, Crysis, and Alan Wake just to name a few. The difference between a game that was patched and a game that is going to be designed for multi-core is vast. I'm not an expert but if my thoughts are correct, the games designed for multi core will have a better written code and will see better performance versus a game that was patched to support multi-core. He's trying to use the game industry's slow adaptation for adapting to new technology as a type of Achilles for hardware. In reality its not really about how you time things versus write the code properly. Moving on though, the way he says that its we're years from adapting extreme threading is a little out there on the bs side. No offense to Dan who probably has a master's in engineering and decades of experiences with computer technology, but if your development team works well together, anything is pretty much possible, if conditions are ideal.
uh i thought supreme commander was a native multicore program
other then that i do agree, ATM i think ageia would do a lot better if they would just move the PPU to a good price range (£40-75) - currently its more important for them to get the things into computers, rather then make new more expensive versions, or whatever
It seems like a good concept and all. But it could probably be integrated into the board somehow, and there really needs to be more programs that use it. It'll probably be a few years before anyone really takes advantage of it.
All in all the guys at Ageia really want to try and justify the price for the Phys X card, but for $300 I would not spend that much on a calculation unit.
Quote:
Originally Posted by completemadness
ATM i think ageia would do a lot better if they would just move the PPU to a good price range (£40-75)
I don't know how prices are in the rest of the world, and I'm not too familiar with the VAT's or exchange rates (isn't it about $2 per £1 ?) but at NewEgg, the PhysX card sells for $141.99.
The card made its debut at $300...but its been quite a long time since the price has been that high.
Another thing I find a little amusing is that so many people are *convinced* that a PPU is already useless because a CPU (dual or quad-core) can do all the physics you need....and yet the same people are willing to go out and buy a $150+ sound card, because it frees up the CPU for (very few) extra frames per second, and produces better sound. Granted, games aren't yet making much use of the PPU...but highly involved physics can be much more taxing on a CPU than sound. I, for one, am not yet convinced that a quad-core (or even 8-core) setup will be able to do what a dedicated hardware physics card can do. Of course, we haven't really seen any proof yet either way...we need a neutral party to write some code to test this situation.
Originally Posted by EQC I don't know how prices are in the rest of the world, and I'm not too familiar with the VAT's or exchange rates (isn't it about $2 per £1 ?) but at NewEgg, the PhysX card sells for $141.99.
Yeah, but you can't do a straight conversion. It's easier to just replace the "$" with a "£". :(
We get shafted on prices for pretty much everything.
Originally Posted by atanum141 And we need it to be cheaper. ALOT.
Have you seen the prices as of late? I was gonna make the same comment, but the prices are sub £100 now. Thats ALOT cheaper than the £250-300 they were originally retailing at.
Check OCUK for proof!
Quote:
Originally Posted by capnPedro Multithreading and running processes on a second core are not the same thing.
Agreed.
I think what the PhysX rep was getting at was that even games that use multiple cores, like SupCom, arnt true multi-threaded games, they just farm out one of the engines functions to seperate cores. Which while better than only using the one core, isnt multi-threading by the longest stretch of the imagination.
What also stinks is that their webpage does not provide an email address for general questions or support.
Also, their support website has a knowledge base.. but does not contain any information whatsoever.
I thought I would look up whether or not the card supports the Oblivion game, to what extent and what difficulties others have experienced.
No luck. Not good for a company wanting people to buy their product.
And, I agree. If game developers start utilizing the various other cores for their physics processing, then what will be the point of an add-on card? And there will be no need for multi-threading. Why even bother when you can do parallel processing from multiple cores.
Anyways... Oblivion isn't listed I'm afraid :(
However... they do say in their opening sentence that's just a "partial listing of current and upcoming AGEIA PhysX-accelerated titles"
For me the lack of PCI-E had been a big reason for not getting the card. Other reasons are the heat factor. If you read reviews on places like newegg many of them talk about how the card gets way to hot even when not running anything that would be using the card.
I hope that the new card will support Shader Model 3.0 level. If it will then there will be a big jump in demand for the new card IMHO. Well that is if they can keep the new cards price down to around $150 or less. The reason that support for Shader Model 3.0 level is important to me is because of the following:
Does Havok FX Support AGEIA?
Havok FX supports all hardware that can execute standard OpenGL and Direct3D code at the Shader Model 3.0 level. If the AGEIA card and drivers adopt and support Shader Model 3.0 industry standard, Havok FX support will be possible.
(taken from : http://www.havok.com/content/view/187/77/ )
I really think for the card to be a big hit it will have to be able to work with other engines, or at least more than just AGEIA's engine. Adding havok support would go a long way since it would add many more games that could use the card. Games including HL2, Fear, BioShock and Allan Wake.
Comments 26 to 41 of 41
Supreme commander, for example, is only capable of using 3 cores
1. Simulation
2. Graphics Driver Management
3. Audio Management
Its hardly multi-threading, at a great level, and this is what valve were talking about when Bit interviewed them all that time ago
So, unlock the level so I can see how crappy it runs on my system w/o a Physx card. Honestly - what drives hardware upgrades with us hardcore gamers? Poor game performance. I only upgrade when a new game I've purchased doesn't perform extremely well, if everything's running hunky dorky I hang on to my cash until next years $600 video card is out.
Am I wrong?
I think Valve will be using four cores in Episode 2 if they're available... we'll have to see whether that turns out to be the case or not though. SupCom uses several cores, but the performance difference between 2 and 4 isn't that massive to be fair, even at low resolutions where you're CPU limited. A good example of threading is ray-tracing as you see near 100 percent gains - that's something you don't see in SupCom.
Databases have always been the best way to show threading, something where you can ask for data here, calculate there, print there, etc etc
Say i want to request records 1-100 - if i use one thread, i request record 1, wait, get it, request record 2
With threads i can make 100 threads asking for records 1-100 (1 each) and that way you don't stall
Sadly, just plopping things over extra cores isn't as simple as it looks
Also as to sup comm performance, as it only uses 3 cores max (and 1 is for sound) you really only need dual core, and as the clock better, and overclock better then the quaddies, for the time being dual core > quad for sup comm
Alright, what the guy over at Ageia was stating that about multi threading and how it has taken the game developers a very long time to try and develop multi threaded games.
E.G.
I think what he is sighting is the trend that game developers and the gaming industry is the slowest to adapt when it comes to multi threading. There are games out there that already have multi threading and games that will be support multi threading. Patches for games like FEAR, COD 2, Quake 4 (If I remember correctly) & Supreme Commander. Its true that games are slow to adapt, but some games that will support and/benefit greatly from multiple cores are Unreal Tournament 3, Crysis, and Alan Wake just to name a few. The difference between a game that was patched and a game that is going to be designed for multi-core is vast. I'm not an expert but if my thoughts are correct, the games designed for multi core will have a better written code and will see better performance versus a game that was patched to support multi-core. He's trying to use the game industry's slow adaptation for adapting to new technology as a type of Achilles for hardware. In reality its not really about how you time things versus write the code properly. Moving on though, the way he says that its we're years from adapting extreme threading is a little out there on the bs side. No offense to Dan who probably has a master's in engineering and decades of experiences with computer technology, but if your development team works well together, anything is pretty much possible, if conditions are ideal.
In reality I think we should take what he says with an extreme grain of salt. An extreme type of multi threading in this case would probably be Alan wake, when I saw the guys at Intel demoing the game with a Kentsfield processor and an Nvidia Geforce 7900GTX I have to admit I was very impressed. They stated one of the cores was dedicated for physics and the other three were available. That would probably be one of the only "extreme" cases of multi threading there is. But from what it seems like the "extreme" level he's talking about is a game where you can shoot a bullet from a mile away and watch it hopefully hit or miss the fly thats on the wall. Taking into consideration of if the game had wind factors, gravity, accuracy of the gun and the speed of the bullet. That would probably be an "extreme" case where the Physx card would be perfect for and where game developers would not want to pour that much time and resources into try and developing the game in physics. But, I don't think this is the case with most games.
Another point I want to try and bring up is the graphics physics using video cards and the Phys X card. We had a huge discussion about this over at OC forums and well I looked into it and well if the guys over at ATI and Nvidia were able to create a Crossfire/SLI solution to physics, it would be mainly be considered eye candy and not "true physics." This is where the Phys X card would win out. ATI and Nvidia's solutions are not "true physics" rather its more eye candy. The objects would be treated to a degree as physical items, but the possibilities and combinations could be limited to how well the graphics card could process. Added in to how much time, effort and thought was into configuring a game to utilize dual card physics. So in this area the Ageia guys would win out.
Alright, enough of the technical spiel, moving onto the multi-core versus dedicate physics card. From a lot of places I have read, ranging from HardOCP to good old Bit-Tech, it seems that the truth is that a physics card can do many more actual calculations than a single core on a processor can. Now before you jump down my throat and say, "But what if you had an 8 core machine or a quad core and the developers dedicated even more cores to the system," lets try to think about this logically. Would it be more effective and feasible to try and dedicate a unit that can process all the physical calculations versus dedicating 2-3 cores which could be used for processing special effects such as AI, Level, Special effects on the map, etc etc?
All in all the guys at Ageia really want to try and justify the price for the Phys X card, but for $300 I would not spend that much on a calculation unit. The thing is very innovative I'll give it that, but they haven't done so hot ever since they came out back in 05. When I first heard about it, I was all like "Oh holy crap! I have to get me one of those!" But now that I think about it, the price is not very justified. Ranging from the price to the very few developers that want to get on board with them, I say they are kinda dead in the water. Unreal Tournament 3 will help them piggyback and help encourage people to adopt and buy the unit, but until the price comes down I don't think it's very justified. Even for the elite and enthusiast guys, it would not be a very smart buy. All in all, get more developers and try to get more people on board with it and you might see it become more mainstream. But the physics unit might be more valuable for more than just gaming, I mean there is so much untapped potential in that card, I wonder if they can try to look into other industries or other fields that could really use that card. That's my two cents, please critique if you guys have any thoughts on my post.
other then that i do agree, ATM i think ageia would do a lot better if they would just move the PPU to a good price range (£40-75) - currently its more important for them to get the things into computers, rather then make new more expensive versions, or whatever
I don't know how prices are in the rest of the world, and I'm not too familiar with the VAT's or exchange rates (isn't it about $2 per £1 ?) but at NewEgg, the PhysX card sells for $141.99.
The card made its debut at $300...but its been quite a long time since the price has been that high.
Another thing I find a little amusing is that so many people are *convinced* that a PPU is already useless because a CPU (dual or quad-core) can do all the physics you need....and yet the same people are willing to go out and buy a $150+ sound card, because it frees up the CPU for (very few) extra frames per second, and produces better sound. Granted, games aren't yet making much use of the PPU...but highly involved physics can be much more taxing on a CPU than sound. I, for one, am not yet convinced that a quad-core (or even 8-core) setup will be able to do what a dedicated hardware physics card can do. Of course, we haven't really seen any proof yet either way...we need a neutral party to write some code to test this situation.
Yeah, but you can't do a straight conversion. It's easier to just replace the "$" with a "£". :(
We get shafted on prices for pretty much everything.
Have you seen the prices as of late? I was gonna make the same comment, but the prices are sub £100 now. Thats ALOT cheaper than the £250-300 they were originally retailing at.
Check OCUK for proof!
I think what the PhysX rep was getting at was that even games that use multiple cores, like SupCom, arnt true multi-threaded games, they just farm out one of the engines functions to seperate cores. Which while better than only using the one core, isnt multi-threading by the longest stretch of the imagination.
Also, their support website has a knowledge base.. but does not contain any information whatsoever.
I thought I would look up whether or not the card supports the Oblivion game, to what extent and what difficulties others have experienced.
No luck. Not good for a company wanting people to buy their product.
And, I agree. If game developers start utilizing the various other cores for their physics processing, then what will be the point of an add-on card? And there will be no need for multi-threading. Why even bother when you can do parallel processing from multiple cores.
You didn't come across this page then :?
Anyways... Oblivion isn't listed I'm afraid :(
However... they do say in their opening sentence that's just a "partial listing of current and upcoming AGEIA PhysX-accelerated titles"
Reports today of the new PCI-E version:
http://www.xbitlabs.com/news/multimedia/display/20070830161921.html
For me the lack of PCI-E had been a big reason for not getting the card. Other reasons are the heat factor. If you read reviews on places like newegg many of them talk about how the card gets way to hot even when not running anything that would be using the card.
I hope that the new card will support Shader Model 3.0 level. If it will then there will be a big jump in demand for the new card IMHO. Well that is if they can keep the new cards price down to around $150 or less. The reason that support for Shader Model 3.0 level is important to me is because of the following:
Does Havok FX Support AGEIA?
Havok FX supports all hardware that can execute standard OpenGL and Direct3D code at the Shader Model 3.0 level. If the AGEIA card and drivers adopt and support Shader Model 3.0 industry standard, Havok FX support will be possible.
(taken from : http://www.havok.com/content/view/187/77/ )
I really think for the card to be a big hit it will have to be able to work with other engines, or at least more than just AGEIA's engine. Adding havok support would go a long way since it would add many more games that could use the card. Games including HL2, Fear, BioShock and Allan Wake.