RSX = Stream Processor!?!
- Thread starter "Nerve-Damage"
- Start date
"Nerve-Damage"
Regular
xbdestroya said:
Exotic:
• Transistor count well over the one billion mark
• Dual NVIDIA based G80 cores @800MHz
• Eight Sony based IP sub-cores
• And the GPU able to handle real-time (not approximations) rendering for Nurbs, Global Illumination, Phong shading, ECT…
Non-Exotic
• Nvidia G70 based main core @550Mhz
• Sony based IP sub-cores (2 or 4)
• FlexIO link
• 2 or 4 MB of Cache
Last edited by a moderator:
Nerve-Damage said:
Nerve-Damage...you were the one that posted the patent on re-usable architecture, weren't you? Have you ever thought they might re-use the original toshiba IP but not physically add more SPE like units to the physical GPU itself...rather share them from a pool of SPEs that the PPE core and G7.x core would be load balancing the workstream from?
I've been thinking that the other part of RSX that we haven't figured out yet is programmable DMA logic (like found on Cell), along with the flex I/O bus and customized G7.x core. It would certainly be nice to have that 4 MB of cache as well...but who knows?
Last edited by a moderator:
Are we really back to this stuff again?
Exotic, to anyone whos been paying attention to GPUs for the last 5+ years, is likely to be anything that isn't similarly set up as a R520 or G70 in their architecture (standard architecture with pixel/vertex shaders, texture units, etc). Extra little units on the chip or dual GPUs on a single die is rather exotic, it may not be by the patent-loving group, but in reality basing expectations of anything off patents is not a very good way to do things (infact its a rather horrible way to get information).
I don't see anything previously stated that would make anyone believe that RSX is anything more than a G70 at 550mhz (24pixel/8vertex -- maybe even missing 8 rops) with some tweaking to the shader/texture units and possibly some other small things (tweaks here and there to allow for some things... but no extra SPEs on the die).
It isn't even logical to expect such things. If the "proof" you guys have is based on patents, then it isn't proof at all. The reality of the situation is RSX is likely very much rooted in a design that is very much standard for the PC world.
At best I imagine the evolution of the RSX would be akin to the evolution of NV architecture of the 6800 series to the 7800 series (tweaks mostly, but the architecture is very similar still). There is no logical reason to believe (nor performance reason) that RSX will be anything but a G7X with small tweaks/changes to fit the needs. I'm not even sure I'd want a chip with SPEs and a G70 core and who knows what else... it'd likely be a mess to deal with, and I'm not really sure developers would want that either (they have enough of a mess to deal with without the GPU being some exotic contraption out of some crazy sci-fi movie).
Exotic, to anyone whos been paying attention to GPUs for the last 5+ years, is likely to be anything that isn't similarly set up as a R520 or G70 in their architecture (standard architecture with pixel/vertex shaders, texture units, etc). Extra little units on the chip or dual GPUs on a single die is rather exotic, it may not be by the patent-loving group, but in reality basing expectations of anything off patents is not a very good way to do things (infact its a rather horrible way to get information).
I don't see anything previously stated that would make anyone believe that RSX is anything more than a G70 at 550mhz (24pixel/8vertex -- maybe even missing 8 rops) with some tweaking to the shader/texture units and possibly some other small things (tweaks here and there to allow for some things... but no extra SPEs on the die).
It isn't even logical to expect such things. If the "proof" you guys have is based on patents, then it isn't proof at all. The reality of the situation is RSX is likely very much rooted in a design that is very much standard for the PC world.
At best I imagine the evolution of the RSX would be akin to the evolution of NV architecture of the 6800 series to the 7800 series (tweaks mostly, but the architecture is very similar still). There is no logical reason to believe (nor performance reason) that RSX will be anything but a G7X with small tweaks/changes to fit the needs. I'm not even sure I'd want a chip with SPEs and a G70 core and who knows what else... it'd likely be a mess to deal with, and I'm not really sure developers would want that either (they have enough of a mess to deal with without the GPU being some exotic contraption out of some crazy sci-fi movie).
xbdestroya said:
You would not be asking me to define these patents if the answer was "clear as day". I am still waiting like everyone else for the clear as day answer to this maze of RSX questions. here are some overall thoughts
Sony is trying to build "architecturally" a game console at takes game development closer to rendering like for CGI movies. Everything about cell and the RSX relationship say rendering farm of computers on two chips. I stopped looking at what they are doing from the standpoint of gamer when I saw killzone. That was a cg movie Sony was saying will be playable when PS3 is released.
"So now in your own words, with the massive evidence you have at your disposal, please tell us exactly what the RSX is, since apparently it is clear as day to you"
Dude, you sound like a crying *****! Where all guessing here. The smarty pants need to stop acting like there guess are better than others.
leechan25 said:
The difference is, that you *are* guessing! It's straight guessing from you, based on nothing else but faith and some patents which I don't even think you've read or could explain in your own words. The rest of the 'smarty-pants' are making educated guesses based on what is reasonable. If you don't want to hear what the 'smarty pants' have to say, why even come to B3D? And if you admit that there is a group of people that has a firmer understanding of GPU architecture than you yourself do, why so easily disregard their opinions? It's cognitive dissonance at it's finest.
Listen, I'm not trying to ram my opinion that RSX is not an 'exotic' architecture down anyone's throats; I simply refuse to have the opinion that it *is* some crazy architecture rammed down mine, which is what I perceive your bizzarre 'victory' posts earlier to be.
"Nerve-Damage"
Regular
ROG27 said:
IMO…the SPE presented in the Cell are not going to be presented on the RSX. However, the RSX will have access to them (SPE).
The RSX will contain its own specific sub-cores (Sony IP) for PS2/PS1 emulation and helping with speeding up certain post -processing effects and achieving a higher level of approximation based rendering for Nurbs, Global Illumination, Ray Tracing, Phong Shading, ECT.
I just don’t see the Cell doing everything…or wasting precious resources trying to do Ray-Tracing when that duty should be shared in my opinion along with any other tasking algorithm with the RSX. Leave the Cell for more Physics, sound, AI, game code duties……….
Last edited by a moderator:
leechan25 said:
Such a educated answer...
We there like to justify or answers, and we like that everybody elso do it, so please do it too.
BTW having a patent means nothing as they also have a 3 cells PS3 patets, and a matrix like one, and the list is very long but that dont mean they even exist.
Nerve-Damage said:
It's already been mentioned that nurbs don't work as effectively as polygons with things like physics-based simulation and manipulation, as well as animating appropriately in a realtime environment. I can't see them using nurbs (at least frequently). Ray tracing is too computationally taxing even in today's closed systems that need to balance their resources more effectively anyway. Using other kinds of lighting techniques gives similar results only taxing the available resources to the nth degree less, leaving them available for other processes.
IMO ray tracing and nurbs will not be used much, if at all this time around.
That's why I'm making my case the way I am. Refer again to my post prior.
xbdestroya said:
Oh so your not guessing? Oh your educated guess holds more weight than mine? Your opinions and your reasonable guesses don't mean s*** because you are no closer to answering the question of the RSX GPU than I am. You must have a "guess O meter" that tell you who's guesses are closer to the correct answer!
So maybe we can ask you and all the other smarty pant to put all that high education together and answer what the patents mean? WAIT, it seems we already did that, a the answer most of you smarty pants come up with is - RSX is like the G70 with flex io, more memory, wider data bus... Hell, they told us that at E3, so I think you smarty pants know s***. That's why the guessing continues. So maybe your education failed you. It okay because we're all guessing here!
leechan25 said:
I rest my case.
By the way, you need to adjust the way you post - because you're not winning any points from anyone the way you toss insults around.
Also I want to add to a previous post that Version, eVo Ant-UK and Nerve are all posters I get along with and respect; it doesn't matter if our views differ from time to time. In fact I'm a big fan of Version's 'patent drive-bys.' Certainly neither them nor the average poster on B3D is prone to descending to 'tirade' level as you have, and I think you'd benefit from stepping back and reading over your own posts to see how others might perceive them.
Multicore GPU
I what I been trying to say also. Cell is multicore. Why not the GPU?
Nerve-Damage said:Why do people feel revolutionary means dual core or multi-core anything?
Don’t get me wrong…I like the Cell processor…but I don’t find it to be a revolutionary CPU.
Rather I find it to be something that was expected or time coming with multiple cores CPU. It’s more evolutionary than anything else……..
So why would the RSX be any different *if* it happens to be a multi-core GPU? I.E. comprised of a NVIDIA based core and a specific Sony IP sub-core.
Man evolving from a banana...that would be revolutionary (and shocking
A multi-core GPU isn’t by no means exotic or revolutionary…IMO!!
I what I been trying to say also. Cell is multicore. Why not the GPU?
leechan - the big difference between the 'smarty pants' and yourself is that the 'smarty pants' present an argument based on 'scientific reasoning' as it were. They consider costs, metrics, feasibility etc. That doesn't mean they're right (we're all guessing!) but their points can be considered on technical merit.
Your argument is 'there are patents, PR talk is of render farms of processors, and there was a CG rendering that they said was going to be produce in realtime on PS3.' The patents evidence holds no weight because patents are not guarentees of present or future technologies ever being produced based on them. The term 'farm of processors' is ambiguous and could readily refer to a shader array. A CG rendering being taken as what will be possible on PS3 doesn't tell us anything about the underlying technology that is supposed to produce those graphics. You've seen an example of what you think PS3 will be producing in realtime and take that as evidence it must be behaving in a strange and exotic way, ignoring the costs and production issues that strange and exotic designs would require.
xbd has asked you to clarify your idea of RSX by explaining some of the essential, unavoidable details that can be considered without needing to much imagination. 'Farm' of processors could meana dozen things, but a transistor of the GPU tells us a relative performance we could expect. Therefore if you give a transitor count to what sort of thing is possible, we can consider on a rational level realism of that. 300 million transistors isn't going to get you a quad core GPU for example. If you believe a multicore GPU is possible, explain how that can be achieved within the transistor budget of a mass-produced part.
This is a technical forum. The principle of consideration is that of technical research and discussion. Claims of 'farms of processors' and multi-core GPUs need to be substantiated by evidence where possible, and thankfully the area of multi-core GPUs is one that can be discussed technically so you should be able to discuss their idea of yours at least on this level. If you can't add technical points to your arguments, you're wasting your time posting in this forum. That someone goes on record saying 'PS2 can render Star Wars Episode 1 in real time' doesn't make it true!
Your argument is 'there are patents, PR talk is of render farms of processors, and there was a CG rendering that they said was going to be produce in realtime on PS3.' The patents evidence holds no weight because patents are not guarentees of present or future technologies ever being produced based on them. The term 'farm of processors' is ambiguous and could readily refer to a shader array. A CG rendering being taken as what will be possible on PS3 doesn't tell us anything about the underlying technology that is supposed to produce those graphics. You've seen an example of what you think PS3 will be producing in realtime and take that as evidence it must be behaving in a strange and exotic way, ignoring the costs and production issues that strange and exotic designs would require.
xbd has asked you to clarify your idea of RSX by explaining some of the essential, unavoidable details that can be considered without needing to much imagination. 'Farm' of processors could meana dozen things, but a transistor of the GPU tells us a relative performance we could expect. Therefore if you give a transitor count to what sort of thing is possible, we can consider on a rational level realism of that. 300 million transistors isn't going to get you a quad core GPU for example. If you believe a multicore GPU is possible, explain how that can be achieved within the transistor budget of a mass-produced part.
This is a technical forum. The principle of consideration is that of technical research and discussion. Claims of 'farms of processors' and multi-core GPUs need to be substantiated by evidence where possible, and thankfully the area of multi-core GPUs is one that can be discussed technically so you should be able to discuss their idea of yours at least on this level. If you can't add technical points to your arguments, you're wasting your time posting in this forum. That someone goes on record saying 'PS2 can render Star Wars Episode 1 in real time' doesn't make it true!
xbdestroya said:
insults,
You talk down to me in such a nice way that I could not tell if you were feeding me honey or shit. I give respect to all men that respect me. My points stands. You guys think just because you got some education in a certain field your opinion are gold. That needs to stop because many of us are educated in other areas of business that complete the circle of business at makes the wheels go. Marketing, sales, customer services, economics, distribution and other fields that tell a part of the story. Studying the business model tells me more about and company plans that their PR does. I just because I offer an opinion doesn't mean it only technical. I show you something you said in the past and you got mad and started to talk down to me. that what not cool and should check yourself on that.
To Nurb or...
Hmm, I thought I've been following this thread pretty closely, but it would seem to me that (memory-compact)nurbs+cell --> lots of triangles through high-bandwidth connection to RSX. To my limited knowledge, that seems a most reasonable way to use the cell/rsx bandwidth while preserving the relatively small amount of memory.
Is it really that hard to have a nurb/triangle hybrid? I'm afraid I don't have the math or graphics background, but it would seem that a smart nurb-tesselator could do a lot for geometry. Round shoulders and heads, scopes, barrels, etc.??? And I'm not sure how that would necessarily impact physics/collisions, as that could still happen on the lower-detail triangle-based "base" character.
If you could point me back to the place in the discussion where this was discussed, I'd sure appreciate it. I only see Nerve's patent posting, and not much discussion around it...?
Thanks,
-Dave
ROG27 said:
Hmm, I thought I've been following this thread pretty closely, but it would seem to me that (memory-compact)nurbs+cell --> lots of triangles through high-bandwidth connection to RSX. To my limited knowledge, that seems a most reasonable way to use the cell/rsx bandwidth while preserving the relatively small amount of memory.
Is it really that hard to have a nurb/triangle hybrid? I'm afraid I don't have the math or graphics background, but it would seem that a smart nurb-tesselator could do a lot for geometry. Round shoulders and heads, scopes, barrels, etc.??? And I'm not sure how that would necessarily impact physics/collisions, as that could still happen on the lower-detail triangle-based "base" character.
If you could point me back to the place in the discussion where this was discussed, I'd sure appreciate it. I only see Nerve's patent posting, and not much discussion around it...?
Thanks,
-Dave
Rog but you have to tell us what we're supposed to derive from that.
That's from July 22nd and it was discussed heavily here back in the day, as I'm sure you know. I know your theory of PS3 involves heavy cooperation between Cell and RSX, possibly with the dedicated 'slaving' of SPE's, but no one's disagreeing with you. In fact Cell:RSX communication and assistance has been a central point of what most of us expect from the PS3. It's simply given that theory, there's too little known to either support it or shoot it down. Me personally, I think it sounds viable.
Great post btw Shifty, I'd rep ya but I gotta 'spread more around' first.
That's from July 22nd and it was discussed heavily here back in the day, as I'm sure you know. I know your theory of PS3 involves heavy cooperation between Cell and RSX, possibly with the dedicated 'slaving' of SPE's, but no one's disagreeing with you. In fact Cell:RSX communication and assistance has been a central point of what most of us expect from the PS3. It's simply given that theory, there's too little known to either support it or shoot it down. Me personally, I think it sounds viable.
Great post btw Shifty, I'd rep ya but I gotta 'spread more around' first.
Leechan, whatever - think what you like.
I didn't address you in this thread until you addressed me, and frankly that quote of mine you say I got upset at, I stand by it 100%. It's your tone that upset me. It reminds me of when I argue intelligent design with people.
I didn't address you in this thread until you addressed me, and frankly that quote of mine you say I got upset at, I stand by it 100%. It's your tone that upset me. It reminds me of when I argue intelligent design with people.
Last edited by a moderator:
xbdestroya said:
So you don't like the way I said something...
I was not looking to start a fight but I'll glad to finish one. You may not remember but you and I were talking back then about the same topic. look it up. So I was continuing what was talked about. I don't like your tone. Sounds like my wife when she can't get her way. If you can't talk about things you disagree without talking down to people maybe you should not comment.
Similar threads
