Finally, *some* news on RSX

[rog27]

...according to B3D dev...Barbarian:

"'Quote:
Originally Posted by Titanio
I guess RSXs arrived post-December, then..

All I really want to know is if the thing is exactly what was announced at E3 or if it changed subsequently. I mean they did give a reasonable amount of info at E3, if you wanted to take it at face value.'


More like post-January but yes indeed. What was disclosed at E3 is still true but obviously not all inclusive.
And they did improve on a lot of things, but like I said in an evolutionary way. I'm mostly excited about the software side of things, since Sony has decided to provide a down-to-the-metal API which makes me very happy."

[Domination]

I'll be much happier when the curtain opens. :happy:

[xbdestroya]

Well, at this point I never would have expected to get information from a dev on the topic, but indeed 'Barbarian' has provided some information on RSX that I think is worthy of a thread on the matter. I won't make any guesses as to his NDA's, a possible PS3 conference this month, or anything else. I'll just post what he's said, and make some comments after doing so:

All quotes attributed to 'Barbarian'


RSX is an evolutionary processor. I can't really say more right now, but don't expect anything that can be described as 'exotic'.

(the below was in response to a post inferring a post-December RSX dev kit shipping)

More like post-January but yes indeed. What was disclosed at E3 is still true but obviously not all inclusive.
And they did improve on a lot of things, but like I said in an evolutionary way. I'm mostly excited about the software side of things, since Sony has decided to provide a down-to-the-metal API which makes me very happy.

Yes, they have been fabbing Cell for a while, but the question is can you ship with six fans in the box !? I'd be surprised if Sony ships any earlier than June in any territory. Even if the hardware is ready by then, the software side of things is still undergoing radical changes that will need to settle before anyone can ship a game.

The last quote I would not consider RSX related, save that I think the 'radical software changes' he is refering to probably stem from the RSX vs the 7800GTX, and the interaction possible between Cell and the RSX in the final configuration. That's just a guess of course, and I think he was just using the phrase loosely to indicate the work devs still have left to do on their PS3 titles.

But what we can derive from the rest of the comments I would take to be that their is that 1) some form of architectural departure between the G70 and what is contained in RSX, 2) that departure is not 'exotic' a la SPE's attached to the die or anything, and 3) Sony has provided a means of accessing the hardware directly in the vein of the EE and GS.

I'll watch if anything else of note gets said - I for one am not going to bug him on any of these issues, but I think this is as much in the way of hints and clues as any of us could have expected pre-event, where 'event' is left open-ended.

EDIT: Rog you beat me to the post, so you've got the first post credit after the merging, but I'm keeping my thread title. :)

[RzrWire]

ATI just recently released a dual core gpu 'Crossfire' for the masses. I would expect the RSX to be at least this. Although not exotic, I'm sure the way NVidia designed it to work in the PS3 might be revolutionary.

Interest Quote:
but the question is can you ship with six fans in the box !?


Edit:
V Oops! I was thinking of something else and it got stuck in my head =)

[xbdestroya]

ATI just recently released a dual core gpu 'Crossfire' for the masses. I would expect the RSX to be at least this. Although not exotic, I'm sure the way NVidia designed it to work in the PS3 might be revolutionary.

Crossfire isn't a dual core GPU, it's a method of using two video cards in the same system, same as NVidia's SLI.

[Raijin]

ATI just recently released a dual core gpu 'Crossfire' for the masses. I would expect the RSX to be at least this. Although not exotic, I'm sure the way NVidia designed it to work in the PS3 might be revolutionary.

Interest Quote:



Edit:
V Oops! I was thinking of something else and it got stuck in my head =)

And I'm afraid will be disappointed...

[Grandia]

question or two, he said its an evolutionary gpu, compared to what? also what did they improve on?

[Raijin]

question or two, he said its an evolutionary gpu, compared to what? also what did they improve on?

Certainly compared to G70...

What did they improve on?

Maybe clock speed? Bandwith?

We dont know yet...

[Crossbar]

Another question:
What was disclosed at E3 is still true but obviously not all inclusive.


Does this mean that whatever the RSX looks like now, it will be somekind of superset of the functionality presented at E3?
What is your interpretation of this?

I am afraid these comments lead to more questions than answers, still it's great to know the RSX is alive and kicking.

[Crossbar]

question or two, he said its an evolutionary gpu, compared to what? also what did they improve on?
Certainly compared to G70...

...and the GS.

[Handycrap101]

Another question:


Does this mean that whatever the RSX looks like now, it will be somekind of superset of the functionality presented at E3?
What is your interpretation of this?

I am afraid these comments lead to more questions than answers, still it's great to know the RSX is alive and kicking.

Actually I think what he meant by "What was disclosed at E3" mean the slide that was shown. I'm talking about the slide that says "2006 2 PLAYSTATION Conference". Sounds like he is reaffirming that the conference does exist and will take place. Thats what I got out of that atleast.

[Raijin]

...and the GS.

GS?

[VG Aficionado]

Actually I think what he meant by "What was disclosed at E3" mean the slide that was shown. I'm talking about the slide that says "2006 2 PLAYSTATION Conference". Sounds like he is reaffirming that the conference does exist and will take place. Thats what I got out of that atleast.I'd say they're talking about these slides:

http://www.gzeasy.com/ours/edison/rsx_e3_slide_huang.png

I'll try to find more from the E3 conference. Help would be appreciated.

[Crossbar]

GS?
The Graphic Synthesizer of the PS2.

[xbdestroya]

Another question:


Does this mean that whatever the RSX looks like now, it will be somekind of superset of the functionality presented at E3?
What is your interpretation of this?

I am afraid these comments lead to more questions than answers, still it's great to know the RSX is alive and kicking.


I think what he means is that everything stated at E3 is true, ie the numbers haven't changed, but those numbers don't tell the whole story on what RSX really is.

To me, that's better than it is bad because I take 'evolutionary' to be - architecturally - something more ... advanced? ... than what we've seen thus far on the desktop, and an indication of where Barbarian feels things are naturally headed.

I don't know, that's just me guessing and since I can't speak to what that direction might be, it's hard to offer up any solid guesses. Right now though I'm prepared somewhat to be pleasantly surprised by RSX, since I've prepared myself for pretty standard fare up to this point.

Grandia this serves as a default response to you as well. Wish I knew more, but will add it on if I think of anything or learn anything else. :)

[Sephiroth_VII]

Here are the E3 RSX screens, presented by the NVIDIA CEO:

http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_011.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_010.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_009.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_008.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_007.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_006.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_005.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_004.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_003.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_002.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_001.jpg
http://img.photobucket.com/albums/v604/Sephiroth_VII/RSX/PDVD_000.jpg

[Sklaar]

Nice hype me again :clap:

[Pistolero]

Well, nothing new here.
I remember a quote from some Nvidia guys where he stated that the RSX was the completion of an architecture initiated by the G70...Don't understand what he was referring to...and can't get my hand on the interview !

[koldfuzion]

Barbarian ever got his 8th SPE back, just went through most of his old posts, it's the one thing he was actually complaining about. Wonder if the lack of the 8th SPE caused any headaches from a programming perspective (7 being an odd and prime number, may make it harder to program for?). Who knows, just thought it odd, I never saw any dev kvetch about the missing 8th SPE before.

[Nasadus]

Hmm, it says 512MB of graphics render memory.
So am i correct to asume that that number isn't including the 256MB XDR main memory?
The GPU would have 512MB of GDDR3 memory to itself and the Cell would have to do with 256MB XDR memory?

[VG Aficionado]

Hmm, it says 512MB of graphics render memory.
So am i correct to asume that that number isn't including the 256MB XDR main memory?
The GPU would have 512MB of GDDR3 memory to itself and the Cell would have to do with 256MB XDR memory?They did say that both Cell and RSX will be able to access both XDR and GDDR3. Of course, neither Cell nor RSX will ever take advantage of all 512MB of RAM for its own use, and how effectively they will access the other's memory remains to be seen.

[xbdestroya]

Barbarian ever got his 8th SPE back, just went through most of his old posts, it's the one thing he was actually complaining about. Wonder if the lack of the 8th SPE caused any headaches from a programming perspective (7 being an odd and prime number, may make it harder to program for?). Who knows, just thought it odd, I never saw any dev kvetch about the missing 8th SPE before.

I think he's just a little more hardcore than most devs - he can't actually care that much about the 8th SPE. But his excitement at being able to access the hardware directly in this new twist on RSX makes me hopeful that there are some devs out there looking to get the absolute most out of these architectures, regardless of difficulty.

PS - On the side, envying your energy portfolio Kold! :smoke:

[Nerve-Damage]

Could this "Evolutionary" step be more towards a *form* of doing Nurbs rendering?

Modifying a rasterized surface, such as by trimming: File date March 2, 2004 & Issue date December 15, 2005 (http://appft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=%2Fnetahtml%2FPTO%2Fsearch-bool.html&r=3&f=G&l=50&co1=AND&d=PG01&s1=Nvidia&s2=streaming&OS=Nvidia+AND+streaming&RS=Nvidia+AND+streaming)

Embodiments of methods, apparatuses, devices, and/or systems for modifying a rasterized surface, such as by trimming, for graphics and/or video processing, for example, are described.

2. The method of claim 1, wherein said surface comprises an NURB.

One issue that relates to graphics quality is the rendering of trimmed surfaces. In one approach, trimmed Non-uniform Rational B-spline (NURB) surfaces are rendered with Adaptive Forward Differencing. See "Rendering Trimmed NURBS with Adaptive Forward Differencing," by Shantz and Chang, Computer Graphics, Vol. 22, No. 4, August 1988, pp 189-198. In this approach, adaptive forward differencing is extended to higher order, the basis matrix for each scan is computed, the shading approximation function for rational surfaces is calculated, and the NURB surfaces are trimmed and image mapped. Trimming is accomplished by using AFD to scan convert the trimming curves in parameter space, producing the intersection points between the trim curves and an isoparametric curve along the surface. A winding rule is used to determine the regions bounded by the curve which are then rendered with AFD. In another approach, all trimmed surfaces are converted into individual Bezier patches with trimming regions defined by closed loops of Bezier or piecewise linear curves. Step sizes are calculated in parameter space for each curve and surface which guarantee the size of facets in screen space will not exceed a user specified tolerance. All points on the trimming curves where the tangents are parallel to the u or v axes are discovered, here, the local minima and maxima. Using the extremes, the trimming region of the patch is divided into u,v-monotone regions. Each region is defined by a closes loop of curves. Using the calculated step sizes, each u,v-monotone region is uniformly tessellated into a grid of rectangles connected by triangles to points evaluated along the curves. The polygons defined in u,v parameter space are transformed into facets in object space by evaluating their vertices with the surface factions. Surface normals are also calculated. Each facet is transformed to screen space, clipped, lighted, smooth shaded and z-buffered using 3D graphics hardware. See "Real-Time Rendering of Trimmed Surfaces," by Rockwood, Heaton, and Davis, Computer Graphics, Vol. 23, No. 3, July 1989, pp 107-116.

As illustrated by block 115 of FIG. 1, higher order surface tessellation occurs early in the geometry processing phase of a graphics pipeline. Higher-order surfaces use mathematical formulae and/or functions to represent three-dimensional (3D) surfaces. Examples include Non-uniform Rational B-splines (NURBs), Bezier curves, N-patches, and more. The data transferred is tessellated to generate more complex models. The GPU, therefore, dynamically generates or tessellates the primary model data from the application into much more detailed and complex geometry.

Likewise, a third embodiment or instantiation of dedicated graphics hardware shall be referred to here as a programmable streaming processor. A programmable streaming processor comprises a processor in which a data stream is applied to the processor and the processor executes similar computations or processing on the elements of the data stream. The system may execute, therefore, a program or kernel by applying it to the elements of the stream and by providing the processing results in an output stream. In this context, likewise, a programmable streaming processor which focuses primarily on processing streams of fragments comprises a programmable streaming fragment processor. In such a processor, a complete instruction set and larger data types may be provided. It is noted, however, that even in a streaming processor, loops and conditional branching are typically not capable of being executed without intervention originating external to the dedicated graphics hardware, such as from a CPU, for example. Again, an embodiment of a GPU with this level of capability or a similar level comprises a programmable streaming processor in this context.

[cliffbo]

a question: if nerbs are going to be implimented then that makes it really good at lets say drawing a circle because every pixel can be seperately programmed, but what if you only want a triangle, surely the best way would be to have three points and connect them with a line. has there been anything to suggest that PS3 will have this in hardware and not just in software. if that made sense, i'll be suprised, but i'm trying...

[Sephiroth_VII]

You do love your patents, don't you nerve?

[Nerve-Damage]

You do love your patents, don't you nerve?

:angel:

[Saibo]

NURBS are so 1980's and they are only good for a couple of things, industrial design being one of them. They arent really suited for character work. If anything, they focus on subdivision surfaces. Still im doubtful any type of high order surface is supported in HW on the RSX.

PS i havent given up on the exotic RSX design(cell-ish appoarch), despite this "news" :D. Its almost February!

[jaxmkii]

im kinda bummed about this...
RSX was supposed to be something super!
buy the time i get a PS3 there will be a faster GPU avalible for a PC....

[woundingchaney]

im kinda bummed about this...
RSX was supposed to be something super!
buy the time i get a PS3 there will be a faster GPU avalible for a PC....


From the sounds of it there may likely be a faster gpu available now.:shrug:

[Illmatic]

buy the time i get a PS3 there will be a faster GPU avalible for a PC....


You should've been expecting that from the start. That's the way it's always been, and will always be unforunately :susp:

[liver_kick]

I said this in another thread prior but perhaps it bears repeating in light of this post. Some people were just setting themselves up for a disappointment that isnt really a disappointment at all. Which is a bit of a shame but what can you do. Bottom line, any G70 derivative in a closed platform with Cell is going to equal a monster machine for many years to come. Future proof is a misnomer, especially when it comes to GPU tech and their respective IHVs. You look at a system as a whole, and how its going to be leveraged by the developers writing the actual software during its life span. Im sure here will be 400+ million transitor GPUs in PCs by '07, but will there be any software that takes real advantage of them? No. And will they still be bound by x86 architecture? Yes.

[rpgamer_2k5]

I'm with Saibo. Why the heck would Sony and Nvidia jointly develop a CPU parallel to the G70 if it's just a mere "evolution" (variant, ie. GTX --> GTX-512).

The RSX has to be greater.

[Cookie]

From the sounds of it there may likely be a faster gpu available now.:shrug:
Yeah, that's so weird. Look at the psp? everyone expected psone gphx, but no it delivered. The ps2 delivered too, in its time, impressive, with uber specs, certainly not weaker than gpus released long before it.

At the dawn of this last gen. Dc and ps2 fought on, both were using the .25m process initially, but the ps2 was on a whole nother league. Now we see both xbox360 and ps3 sharing again a similar process, this time 90nm. Once sony trounced their competitors in the gphx dept., and outclassed gpus out even quite near the ps2 launch(in some departments it outclassed pc gpus for years.), I just hope they can do it again. It'd be ironic if for the first time they were actually behind the curve, even if it exceeded the x360, it'd not be good pr if it can't blow it away.

N3 vs Heavenly Sword should be the difference we see in most genres between these two machines, that is if nvidia manages to deliver.

[Sephiroth_VII]

The reason that consoles look so much better than a surperior PC, is that consoles doesn't have all those things running in the background. It of course also has the advantage of being a closed enviorment, meaning that the developers knows exactly what the system can do. The more developers know about the system, the more power they can squeeze out of it.

[Sklaar]

"...so just like Geeforce 6, we had a variety of products, well, we have the 7800 GTX and then the next, future generation of this technology will be the RSX, so kinda what, what we characterise as the parent technology is this second generation shader model 3 engine. And it can be extentiated in a variety of ways, and in the case of the RSX, we have a many pipeline chip which would be directly coupled to the CELL processor..."

http://media.gear.ign.com/articles/628/628947/vids_1.html

:clapping:

[Pistolero]

A G71 (though I do not know if this upcoming card brings anything new to the table except for higher numbers) or another version ?

PS : What we know is that there will be no SPEs or some strange flavour into the mixture. Other than that...will...Anyway, this will be a very powerful part ans that's what matters at the end !

[Revenge]

The special characteristic of the PS3 is the connection between Cell and RSX



The special characteristic of PS3 Graphics is the connection between Cell and RSX. The RSX itself has a similar architecture to the G70, but the host interface for the G70 is meant for the PC and is completely different. The G70 uses PCI Express x16 to connect to the chipset as 8GB/sec (4GB/sec one-way), and it cannot directly access main memory. In contrast, the RSX has a 35GB/sec (20GB/sec down, 15GB/sec up) direct connection to the Cell, and can directly render from the main memory on the Cell side.

This is a big difference, because it allows a completely different way of using the GPU from PC architectures, SCEI explained. First of all, because the bus is wider, the Cell can perform a great amount of geometry operations, then send the vertex data [to the RSX]. Conversely, the RSX side can easily send data back to the Cell.

“The Cell processor can do both pre-processing and post-processing. For example, tessellation, dot filling, etc… Cell can perform physics processing like collision and motion calculations, and transform the vertex array.” said David B. Kirk, Chief Scientist of nVidia.

SCEI basically expects higher abstraction levels to be processed by Cell, and the details (like vertexes and pixels) to be processed by the GPU. The is reasonable – for example, in the case where the CPU side handles geometry transformation, collision detection, which is important in games, is not a problem. In the case where the GPU handles geometry transformation, if the data is not sent back to the CPU, clipping issues may occur. In the case of the PS3, the Cell side can perform transformations, and even if the GPU is used for transformation, it is comparatively easy to send the data back to the CPU side.


In architectures up to now, either the CPU or the GPU have been the bottleneck. It this is not resolved, we cannot go any further. To face this, in PS3 architecture, if the GPU becomes the bottleneck, it can shift work to the Cell, if the Cell becomes the bottleneck it can send work to the GPU, shifting the workload. For example, according to the software, the Cell side can perform more graphics processing, or, oppositely, or easily make an adjustment to leave the graphics work to the GPU, it was explained. In summary, between the CPU and GPU programmable processors, a flexible balance adjustment can be done.


In previous PC architectures, because they were limited by the CPU<->GPU pipe, geometry operations were held to a certain limit, and how rich an environment you can create within that limit became the main technical challenge. In contrast, the PlayStation2-type game consoles created large amounts of polygons, but after that it did not have the expressiveness of PCs. (Trans. note: probably means that PS2 is less capable in applying different effects to polygons than the PC, despite pumping out more polygons.) In the case of the PS3, both are possible, with the flexibility to balance the two.

However, in the case of the currently available PS3 Evaluation System, because of restrictions in the architecture, it is not possible to evaluate the balancing [of Cell and RSX]. This is a difficulty and a weakness, but, if we state it differently, software demos on the current systems still do not demonstrate the full potential of PS3. It is possible that the actual PS3 will have performance greater than current demos.

[Raijin]

Hm where does this come from?

EDIT: I have the impression I already saw this...

[VG Aficionado]

In Revenge's post, they talk about "PS3 Evaluation System", while the final development kit is called "PS3 Reference Tool". I'd say the limitations they talk about are no longer an issue since those should be happening just on alpha and beta kits, so that's old news indeed.

[stanDarsh]

Revenge can you post a link to that article?

Thanks.

[Heinrich4]

Revenge can you post a link to that article?

Thanks.

It came from Hiroshige Goto Article :

http://www.babelfish.altavista.com/babelfish/trurl_pagecontent?lp=ja_en&url=http%3A%2F%2Fpc.watch.impress.co.jp%2Fdocs%2F2 005%2F0722%2Fkaigai199.htm

If rsx can certain form auxiliary cell, then it would have "something of cell" in gpu?

[rog27]

"Eureka!

I think I've put the whole puzzle together:

I believe the RSX is this new streaming GPU that we've been seeing talked about in all of these patents, but it's streaming architecture arises not from spe-like units built into the GPU itself. Rather, it is a virtual streaming processor GPU in that the special low level inter-functionality between it and Cell, along with the flex-io bus, allow this specialized G 7.x variant to perform as if it had SPEs of its own.

It would make sense that RSX can lock SPEs (or even load balance them with the CPU/PPE core). This would make sense of the reusable architecture patent Nerve-Damage so kindly posted for us. True, the RSX probably does not have 4 SPEs of its own physically...but it can probably borrow and lock up to 4 (or 6 or whatever) of the CELL's SPEs as a virtual extension of itself.

The archticture Sony had originally imagined probably looked like so:

CPU/PPE core<--->[Pool of 8 SPEs, 1 reserved for redundancy (or something else)]<--->GPU/RSX core (specialized G 7.x variant)

The SPEs act as a processing bridge and can do pre and post graphics work."

http://www.beyond3d.com/forum/showthread.php?t=27884&page=3

[xbdestroya]

I haven't read through all of those patents myself, so I'm no authority here, but I don't see why there is so much emphasis on the word 'streaming' in all of these discussions. SPE's are 'streaming' - sure - but so are the pipes inside GPU's. And when NVidia is the one filing patents for these ideas and such, my first inclination is not to default towards RSX information; rather just view it as ideas for graphics in general. A GPU doesn't need to be a virtual streaming processor - it *is* a streaming processor! :)

By the way, I don't disagree that the SPE's might aid in the graphics department - and in that sense maybe a virtual Visualizer does live on in a sense in the context of the PS3 as a whole (I myself don't know) - but that patent on 'reusable architecture' has to stop coming up; it's a manufacturing and chip-design related patent, not a patent for an actual chip design or anything.

[xbdestroya]

Here's Barbarian's latest addition to the RSX puzzle:

Here's an interesting tidbit - a DX1 texture is always guaranteed to fit in RSX's texture cache. How many megs would that be?

[stanDarsh]

Is that meant to say DX10?

In any even does anyone know the answer?

[xbdestroya]

Nah it doesn't mean DX10, DX1 is what's meant. ROG27 hypothesized on the size, and I'm sure he'll be around sooner or later to provide his own thoughts on that, but what I'm curious to know is where Barbarian is trying to lead us in terms of RSX's architecture.

[cpiasminc]

Actually, it's still a typo -- it's supposed to be "DXT1". DXT1 is 8:1 compression from a 32-bit while dropping alpha content. Which means 4 bits per pixel. Assuming that the largest texture dimensionality supported is say 4096, that makes for 4096x4096x0.5 = 8 MB, which means it's at least that much. Since it doesn't say anything of the sort for a DXT3 or DXT5 texture, that at least suggests that if it's more than 8 MB, it's probably less than 16 MB.

[BlueTsunami]

Vysez and Mintmaster basically gave the same explanation cpiasminc gave but aaaaa00 is saying something different. I'm also with you XB as far as wondering where Barbarian is trying to clue us to. Hopefully its not just an non important RSX cocktease.

[xbdestroya]

*sigh*

Oh well, I figure a good 24 hours before Barbarian clues us in on what he meant; either a typo and indeed it was DXT1 like Cpi and Vysez were venturing, or he's just joking around.

RSX, where art thou?

[section]

Isn't tidbit the same as titbit aka very juicy, firm piece of good tasting breast meat? So that itself is hinting that the texture cache talk is true (as far as Barbarian himself isn't lying all along..) and hence RSX in this context probably walks the path of GS with some nice amount of embedded memory (vram, edram whoknows) with bandwidth to burn.

[overclocked]

*sigh*

Oh well, I figure a good 24 hours before Barbarian clues us in on what he meant; either a typo and indeed it was DXT1 like Cpi and Vysez were venturing, or he's just joking around.

RSX, where art thou?

In case i think nVidia/Sony has taken the BW constraints very seriuosly, and that there will be many little changes adding upp to a potent GPU, interesting to say the least.
Xb just think of how many refer to NV2A as NV27.5 instead of 22.5 because of the apparent Dubble Z potens in the ROPs ie 4ROPs write 8z-samples.

I dont know if i sad this earlier here but a 6600GT in a console closed box should be able to produce some amazing things!

Heh and know we talking about a chip with a possible shaderperformance of atleast 2 6800U in PC mode. Take that power to enclosed system and wow, just wow. Less BW although

[BlueTsunami]

Deano just commented on the Barbarian comment about "cache" and i'll just quote him...

I'm not going to comment on RSX, just to say current PC chips have caches measured in Kb.

So if PC cards measure cache in the KB and you've got the RSX that can hold MB of data in its cache (if Barbarian is to be believed), what would that implicate? That also seems like a big deviation from a traditional PC part...

[Crossbar]

Actually, it's still a typo -- it's supposed to be "DXT1". DXT1 is 8:1 compression from a 32-bit while dropping alpha content. Which means 4 bits per pixel. Assuming that the largest texture dimensionality supported is say 4096, that makes for 4096x4096x0.5 = 8 MB, which means it's at least that much. Since it doesn't say anything of the sort for a DXT3 or DXT5 texture, that at least suggests that if it's more than 8 MB, it's probably less than 16 MB.
Why would you go with a texture cache of that size instead of a frame buffer of the same size when rendering for let say a 720p screen?
Is it such a big advantage to have a fast texture cache instead of a fast frame buffer?

[Crossbar]

Isn't tidbit the same as titbit aka very juicy, firm piece of good tasting breast meat? So that itself is hinting that the texture cache talk is true (as far as Barbarian himself isn't lying all along..) and hence RSX in this context probably walks the path of GS with some nice amount of embedded memory (vram, edram whoknows) with bandwidth to burn.
If we are talking about a cache of several MB then it is most likely a 2 chips in package solution which opens a lot of options. This is getting interesting!!!!

[section]

prolly shooting at the moon here but again an Nvidia patent from June 28, 2005:

Patent (http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=/netahtml/search-adv.htm&r=1&f=G&l=50&d=PTXT&p=1&p=1&S1='6,911,983'&OS=)


Tile buffers in a graphics processing system are managed use "copy-on-write" semantics, in which tile data stored in a memory location is not transferred to another location until the tile data for one of the buffers is modified. Two memory spaces store tile data, and two logical buffers are used to access the memory spaces. For each tile, a tile association is maintained, indicating which of the two memory spaces is associated with each of the two logical buffers. To copy a tile of the first logical buffer to the second logical buffer, the tile association for the tile being copied is modified. Data for a tile is written to the memory space associated with a target logical buffer after ensuring that the tile association for the tile associates the target logical buffer with a different one of the two memory spaces from the other logical buffer.

guys at B3D talked about this last July:


Jaws:

This patent addresses several topics of recent discussion with PS3 and CELL + RSX. Namely, memory management, memory bandwidth reduction and tiling. These could be modifications/ customisations made to RSX that differ from the current G70/ 7800 PC card...

and another post from Faf:


What would be really interesting is if we were provided with the layout and size of these on-chip buffers, so I can actually force my rendering operations to be tile coherent where it counts.
Of course this is under assumption that the buffer size is not too tiny to be able to exploit effectively.

But I'm worried NVidia will turn a deaf ear to pleas for architectural details though, ruining much of the chances for using the architecture to its full potential Sad


So now after quickly reading the patent and knowing that EIB in PS3 design takes care of cache coherency on hardware level it might be that Nvidia with Sony has actually thought about the tile rendering on "cache" level:


Shifty Geezer:

On-chip buffers? So there's local storage on a GPU in this patent for 'tiled' processing of a sort?

From Jaws' excerpt, I understand it that a screen (buffer) is segmented into areas and an area is only calculated when changed. In a static camera the background will remain the same so not be rendered. In 3D games usually the whole screen is in motion so I can't see any advantage.
But Faf is talking about something else with on-chip buffers.

Anyone care for a rough breakdown/translation of the patentese?


Jaws:

Although the patent was describing double-buffering and frame-buffers, the concept is quite general.

Basically the patent implements logic that decouples 'buffers' from physical memory locations, tiles these 'buffers' and keeps them coherent in a tile 'table'. And the GPU doesn't have to be a tile based GPU.

So essentially you have these tiled 'logical' buffers mapped to 'physical' memory locations available anywhere in the system. And copy/write modifications to the tiled buffers can occur but the 'physical' memory is managed by this 'logic' so that memory bandwidth usage are minimised. The size's of the tiles can also be adjusted to match memory requirements.

This patent would be a good match for RSX + CELL in taking advantage of the NUMA architecture...

Interesting times indeed :D

[Sklaar]

We heard that G80 will be in time for launch in June during Computex and the process technology is likely to be 80nm at TSMC. In the recent statement, NVIDIA has said that they will be backing the 80nm "half-node" process by TSMC where it allows reduction of die size by 19%. We have previously mentioned that G80 is likely to take on the Unified Shader approach and supports Shader Model 4.0. G80 is likely to be paired up with the Samsung GDDR4 memories reaching a speed of 2.5Gbps. As for ATi, the next generation R600 is slated for launch end of this year according to the roadmap we have seen and the process technology is 65nm. It seems that the leaked specs of the R600 that surfaced in June last year is pretty likely.

According to Xpentor, NVIDIA G80 will make ATI stumble on April. Quad SLI itself can be implemented on a single card with two chips solution because it will carry the first dual core GPU ever with the support of DirectX10 and Shader Model 4.0. The development of G80 is also mentioned as being running very intensive since NVIDIA's acquisition over ULi. As for the upcoming G71, there will be 32pipes, increase in ROPs and a little speed bump over the core clock.

65nm
64 Shader pipelines (Vec4+Scalar)
32 TMU's
32 ROPs
128 Shader Operations per Cycle
800MHz Core
102.4 billion shader ops/sec
512GFLOPs for the shaders
2 Billion triangles/sec
25.6 Gpixels/Gtexels/sec
256-bit 512MB 1.8GHz GDDR4 Memory
57.6 GB/sec Bandwidth (at 1.8GHz)
WGF2.0 Unified Shader

http://www.vr-zone.com/?i=3177

[xbdestroya]

VR-Zone is consistently full of it though - not to mention those specs there are for the R600, I just want everyone clear on that. I personally don't see a G71 launch in March with a G80 launch in June. That just seems crowded and stupid. But we'll see what happens... Incidently they got their own information on that from that Xpentor site, of which I've never heard.

Overclocked and Sct-i/on, good points. I feel as if we must be on the verge of an info release or something, either that or Barbarian doesn't operate under the same NDA constraints as other devs do (or he was kidding). But the bandwidth concerns have been what I've wondered about most up until this point, and it seems that if that method were implemented, well that could help considerably in that regard.

[section]

how many refer to NV2A as NV27.5 instead of 22.5 because of the apparent Dubble Z potens in the ROPs ie 4ROPs write 8z-samples.

Few questions here, dubble z potens = double buffering (potential)? :)

So isn't NV2A first xbox's gpu and did it then provide some jazzy programming tricks compared to it's PC siblings?

I'm actually real noob at computer graphics but now I gather that there might be integrated cache on RSX (maybe 2 chips in package solution as Crossbar said) and it could act as coherent cache for "main" memory..


According to one aspect of the invention, a buffer system for storing tile data for tiles of a display includes a first memory space for storing tile data; a second memory space for storing tile data; a memory interface circuit; and a tile table. The memory interface circuit is configured to receive memory access commands referencing a first logical buffer and a second logical buffer, and the memory interface circuit is further configured to respond to the memory access commands by accessing the first and second memory spaces. The tile table configured to store an entry for each of a plurality of tiles, each entry associating each of the first logical buffer and the second logical buffer with one of the first and second memory spaces. The memory interface circuit uses the tile table entries to determine and modify associations of tiles of the first and second logical buffers with the first and second memory spaces.


I wonder how much this type of solution could potentially save bandwidth, I would be happy if Xbd or someone would care to give some academical guesses :)

[rog27]

"102.4 billion shader ops/sec" -G80...interesting.

forget about actual archticture...check out relative power--

The PS3 (system as a whole) will be equivalent in power/performance graphically with a dual core G80 part

...this all stems from the special relationship between CELL and RSX, as RSX on its own is a weaker part than the G80.

And this is why they are equivalent:

PlayStation 3 RSX:
At E3 '05 Sony released a figure of 100 billion shader operations per second (SOPS) for both the RSX and the Cell.
They also told us that the RSX is capable of 136 shader operations per cycle. We are going to use these two numbers to extrapolate the individual SOPS metric for each the RSX and the Cell :
136 (shader operations per cycle) x 550,000,000 (550MHz)=
74,800,000,000.
The RSX alone is capable of 74.8 billion SOPS.
100 billion SOPS (Cell + RSX total) - 74.8 billion SOPS (RSX) = 25.2 billion SOPS for the Cell alone.

The closed system graphically is the equivalent of a dual-core G80 part...

That certainly is reason to smile.

[xbdestroya]

ROG that 102.4 G-ops figure is for R600, not G80. Just want to point that out. :) Even though that VR-Zone article is for G80, those specs listed there are R600.

[xbdestroya]

I wonder how much this type of solution could potentially save bandwidth, I would be happy if Xbd or someone would care to give some academical guesses :)

Sct-i, I have to say that there's not much else I could add to that last quote from Jaws you yourself posted. The logical tiling of course should save bandwidth via the manner implied by that physical mapping scheme - and in PS3 that should assist in the utilization of the XDR pool of memory - but I'm not sure I have any ideas as to what the real-world bandwidth savings/efficiency gains should be expected to be with this.

PS - I'm reading the patent now though to see if there's anything else that might be gleaned.

[rog27]

ROG that 102.4 G-ops figure is for R600, not G80. Just want to point that out. :) Even though that VR-Zone article is for G80, those specs listed there are R600.

Well, regardless, performance-wise ps3 will be on par with next-gen pc parts--NVidia and ATI are usually comparable in this area.

[xbdestroya]

I agree that the way things are starting to shape up, RSX seems as if it will be a very powerful part. Of course we always knew that, but it's still a strong feeling of excitement as things slowly... slowly... begin to take shape.

[rog27]

I just have the feeling that, internally, RSX will be best of breed of the tried and true NVidia technologies of recent, tweaked and customized, but Externally...how it communicates and shares its and CELLS assets...will be 'exotic' and revolutionary.

That's just my feeling.

To support my "feeling"...I will point to what EA has been saying about the newest Medal of Honor game that's in the newest Game Informer.

[section]

Obviously they talk a lot about "desktop compositor" in the patent so it might be mostly about driver support for example OpenGL desktop use but there's also the clause While the invention has been described with respect to specific embodiments, one skilled in the art will recognize that numerous modifications are possible. As described above, the tile table can be implemented in a number of ways, so long as an unambiguous association of logical buffers with memory locations on a tile-by-tile basis is provided. The invention is also not limited to the context of a frame buffer; the systems and methods described herein can be adapted to buffering tile data in drawing memories as well.

So while it might seem to primarily cover OS use (PS3 3D accelerated desktop anyone? :D) those tiling ideas could be much more widely used, now we only need to get some numbers straight from the horse's mouth...

[section]

Argh, once again nice speculation created and shot down by B3D pros:


...
Without knowing the texture res, the cache could be,

4k x 4k -> 8MB
2k x 2k -> 2MB
1k x 1k -> 512 KB

The 512 KB cache would match the PPEs L2 cache in CELL. Given consoles limited memory, it could well be these lower res textures...



nAo

Jaws...no
Barbarian smoked something very good..


:(

[xbdestroya]

I want to note though that that paragraph you just quoted in that patent is just for protection, to convey to the patent office that the 'invention' described in the patent is independent of the 'refered embodiment' used to describe it.

Every single modern patent will include almost that exact same language somewhere, simply geared to their own invention.

Basically they're saying: "if you do tiling logical buffers with physical memory mapping, that's us and you're infringing." Be it PC's, cell phones, PDAs, or some A/V device requiring graphics that hasn't even been dreamt up yet.

PS - Yeah I was going to post nAo's comment as well. LOL oh well so we'll move away from this whole cache discussion for a while. The patent we're discussing though is still a valid topic however, and exists independently of all that cache business.

[section]

You can only wonder if those "developer" guys who are supposed to know better get some dirty, twisted kicks out of fooling us mere mortals by their riddles :P

Oh well, I'll drink to that (few nice dark longnecked ones from Czech) :D

[Crossbar]

Shit, I spend some time putting everything in context. I thought a lot of thing started to make sense. Like the recently revealed tool from Web Technology.
The PS3 version of Optix Image Studio adds support for PS3 format 32-bit RGBA8888 images, direct editing and conversion for S3TC compressed textures (DXT1 - 5) http://ps3.ign.com/articles/684/684400p1.html

And a lot of other things, like the requirement for 512 MB graphics memory on the evaluation tool but no number given for the graphics memory of the reference tools etc. etc.

Does someone know where Babarian lives? I would like to give him some serious spanking.

[rog27]

Shit, I spend some time putting everything in context. I thought a lot of thing started to make sense. Like the recently revealed tool from Web Technology.
http://ps3.ign.com/articles/684/684400p1.html

And a lot of other things, like the requirement for 512 MB graphics memory on the evaluation tool but no number given for the graphics memory of the reference tools etc. etc.

Does someone know where Babarian lives? I would like to give him some serious spanking.

Somewhere in California, USA. :)

Good Luck

[Crossbar]

Somewhere in California, USA. :)

Good Luck
Thanks, I'll be on a plane tomorrow.

I'm really pissed, he is not getting away with this.......

[xbdestroya]

Looks like you can save your airfare money Crossbar - Barbarian's explaining his situation:

Umm, yeah sorry I meant DXT1.
Having said that my original post was actually a question. I don't know how many megs of texture cache would provide this kind of guarantee. The thing is, the claim that any DXT1 fits in texture cache was made by an official source at a recent devcon. It sounds a bit absurd to me but for now I don't really have any other information that would contradict this statement.

[Crossbar]

Looks like you can save your airfare money Crossbar - Barbarian's explaining his situation:
Thanks Xb, but he actually wrote that piece after I had spanked him. :asshole: :whip:

Nevertheless, it seems we have some very contradictory information at b3d right now. Barbarian doesn't have the full picture and those who are in the know seems to mock Barbarian.

It's very strange that this kind of information hasn't been made public yet to the developers, it doesn't really make sense. I guess it's some kind of chicken race between Sony and Nintendo.