RSX patent?

!!! I KNEW IT!!! Makes perfect sense nAo.

 

It will be the fact the EA speech about use 4 SPEs for graphs we could consider a way indirectly to deduce some "Tile cache" (like 4x256KB) as local memory of SPE inside the RSX ?

(or same... it would have something of SPE SIMD/MIMD or cache like in RSX?)

 

I would think that they are using the non-coherent FlexIO lanes for RSX.

 

The cache in MP systems with shared memory has to be coherent or there are problems with a processor working on stale data. This would be important for a 2 Cell system, but I don't think there is a need for the scratch pad type memory / cache in a GPU to be coherent with anything.

If you get a puzzled response from a PS3 developer and it isn't followed by a smiley then you're probably on the wrong track.

 

I suspect that RSX in practice is more likely to fetch data directly from the SPE's local store than from the PPE's cache.

 

I don't think anything can directly access the LS except for the SPE itself.

We know Cell and RSX can communicate but we don't know exactly how that works. Hopefully they'll release more info on it soon.

 

There are quite a few PS3 devs floating around these forums, not just Marco and I...

I will often ask for public docs, cos the NDA's allow me to talk about anything thats in the public domain. So if somebody can point me at a doc, I then know I can talk about it.

As for whether RSX/Cell is coherant or not, I can't say but there is no requirement either way. There are 2 memory controllers in a PS3, one on Cell (essentially the EIB), the other on the GPU. This is the same as PC's today, when you access GPU RAM you have to mark the pages as non-cacheable. This is because the CPU can't snoop the read/writes, so can never know if the the stuff in cache is upto date. Now of course, its possible RSX interfaces so closer with the CPU mem controller that its knows the state of GPU RAM. Feel free to guess away, I can't comment until somebody can provide me with a public statement from Sony...

 

--------------------------------------------------------------------------------

From the Dec. issue of OPM:


Quote:
Nvidia President Jen-Hsun Huang explains why the RSX graphics processor, along with Sony's Cell processor, will help drive graphics into the next generation.

OPM: How does the new processor compare to the GPU in the original Xbox or the ATI GPU's in the Xbox 360 and the Revolution?
Jen-Hsun Huang: The GPU in the original Xbox was phenomenally powerful and flexible for it's time but is quite primitive compared to modern programmable GPU's. GPU performance has been doubling every six to nine months, and since the original Xbox is approximately 5 years old, one would expect modern GPUs, such as the GeForce 7800, to be approximately 50 to 100 times more powerful, and in fact they are.

OPM: What would you say is the key difference between the RSX GPU and previous Nvidia products?
JH: PC graphics and core-logic products must be flexible and work efficiently and reliably in a variety of system configurations. This flexibility forces the GPU designers to be prepared for a wide variety of workloads so that a PC
product can be very good at very many things. The RSX is designed to work closely with [the Playstation 3's Cell processor] in a closed-system environment.

OPM: Can you briefly explain how the Cell processor works in tandem with the RSX GPU? Is the Cell processor a good fit for the RSX?
JH: The Cell processor has multiple processors that can execute different threads and produce geometry, texture, and shader information for the RSX to consume. Since the RSX is so powerful, in some cases it will take almost the full performance of the Cell just to keep the RSX busy. The Cell processor has enormous floating-point processing and throughput capabilities and is an excellent engine for physics, gameplay, and other nongraphical tasks. The RSX is designed to be a powerful, efficient companion processor to the Cell.

OPM: There is some speculation that many of the demos shown at E3 couldn't be done in real time on the PS3 hardware. Do you think this is true?
JH: There's no question the demos are possible on PS3 hardware.<<

 

ok that just proves what ive been saying all the time, lets scrutinize

question 1: He avoids the comparison between RSX and Xenos, he never mentions it when asked straight, but compares G70 and Original Xbox GPU.

question 2: That just nearly confirms that the RSX is a higher clocked, turbocached and flexIO'd G70 to work in a close environment, similiar to the Xbox Geforce3 GPU.

question 3: The EXACT same work process of Xbox 1.

 

CONSUME THIS KNOWLEDGE NOW!!! Before its killed off

 

Care to scrutinise the answer to each question in your opinion?

EDIT: Looking at is "is designed" means its already taped out

 

Well, I believe Jen-Hsun Huang is a handsome man. Also, it primarly points more to the fact that RSX is tailored to a closed environment and that (take it as you will) RSX was made to closley work with the CELL. Which we already knew but...yeah...

 

well that seems like a sarcasmic answer, if a system is closed box, shouldnt it be working closely ? i mean really

 

Well...not..like..physically...not like a hug close but more like teamwork close.

 

Its like Logical topology vs Physical Topology.

 

Hmmm. I make very different observations to you.

1) He doesn't compare RSX to anything. He makes a sweeiping statement that modern processors, like 7800, are 50-100x faster than XB's NV2A. He's wrapped modern processors up into a single group of similar performance, which is what we're expecting, no?

2) He's saying that PC parts have to be versatile to fit the different functions of a PC, whereas RSX is aimed for a specific use with determined hardware. That is, things needed in G70 like video acceleration aren't needed in PS3. RSX is likely to do away witht he extraneous parts needed for PCs.

3) Ummm, isn't that the exact same work process of pretty much every CPU+GPU combination? The only thing I got out of this was a rather scarily ridiculous idea that Cell has to work full-out to power RSX to maximum effect. So either there'll be no actual game and only visuals, or some of Cell's power will be diverted to powering a game and RSX won't be fully used. Thankfully this is just meaningless marketting hype and the real hardware is likely more balanced than that.

Though, as you're saying this article proves what you've been saying all along, and you haven't said anything else in this thread so you're reference things expressed elsewhere, I'm not sure exactly what your points are and you may want to explain them (eg. what is the importance of this being 'The EXACT same work process of Xbox 1'?). Maybe I'd agree with you if I knew what your points were??

 

I don't care what nobody else says from here on out. All I cared to read was this.

 

That similiar to the fact that the Xbox GPU was a souped up Geforce3 GPU, the RSX will be similiar souped up G70 (Nvidia 7800 GTX +Clock +TurboCache +FlexIO -MemBandwidth)

 

he had 3 choices:

1) No those graphics are probably achievable with G80 or beyond, not with RSX or G70

or

2) The game graphics will be even better than the demos

or

3) yes the demo graphics can be achieved.


Question is which of the above would he logically as the president of Nvidia chose and secondly there is a difference between Demo (Unreal 2k6, Killing Day) and Rendered (Killzone 2)