RSX Secrets

Then stop posting about it. Didn't you read Shifty Geezer's post above?

You seem to forget you are on a forum with a lot of GPU experts as well as Sony licenced developers. If this forum reaches a consensus on a topic like this, there is no further reason to discuss this topic until you've got some very clear, undisputable proof.

Also, you better believe that moderators do not take kindly to people who ignore their explicit warnings.

 

From here on, there will be zero tolerance of continual bringing up of the RSX clockspeed (infractions etc). If some folks want to believe 18 month pre-console launch specs that have repeatedly been shot down by developers. Fine. Take it elsewhere where someone might care. This board is for intelligent technical discussion, and lately there has only been bickering and tit for tat nonsense.

By the Code, "RSX Secrets" is itself asking people to break NDA. Deal with it and take whatever hints developers have been giving us. It is a privilege that we have developers post on this board, but bringing down the level of quality of discussion here isn't going to make them want to stay.

 

RSX Secret may deadly secret for me? Oh no.

I just only share my information refer to SCEI press release. However I don't know those
press release information can't stand here for what? I've said many times. Does anyone
here was PS3 circuit design engineering? Do they use digital frequency counter to measure it before state that.

If you neead somekind of intelligent technical discussion here.
I need someone proof this by scientist method neither than some rumor from somewhere
and force me to believe that data.

Do Intelligent technicial person method to believe some strange data here are
(1) Shut up all official information from console manufacture. ?
(2) Always accept data from someone whom cliam they're PS3dev as the fact ?
(3) Don't try to proof those fact or you'll be warn and may kick out by moderator ?

Sorry to tell you the truth. I think those kind of three things upper didn't proof anyone
as intelligent technicial anymore.

 

If there are any RSX secrets only some devs would know. And the question is, without asking what these secrets are due to NDA, does it really have any secrets?
Simple yes and no answer.

I think that the RSX may not have any specific secrets, but the whole PS3's "secret" may be how the RSX interacts with the CELL and how the CELL can contribute to the graphics.

I dont know much about technology but I would like to see the few devs that may know something more than others tell us about that CPU-GPU interaction than the GPU itself because thats where I see the PS3's possible unique strengths

 

I understand they position to answer something here are prohibited by NDA.
However in the case of GPU/VRAM clock speed about PS3 console. There are
something very interesting.

(I) SCEI remain E3 2005 Press conference on their official site as you've
seen from my link above.

(II) Developers claim that GPU/VRAM clock was down-grade to some 500/650MHz for save
cost or anything.

(III) The secret is more interesting here that " What's the real GPU/VRAM of PS3 clock rate? "

About Cell and RSX information on 3D pipeline . I think we've some kind of these information
a lot enough from PS3 exclusive developer and more 3rd party around the net. Some said
Cell's SPEs doesn't anything like PS2's vector units. Some said Cell's SPEs does many thing
help RSX in 3D pipeline. I've read far a lot of these like words however we need to see their
games to proof that what kind of thing they've on those words. This like you've not know what the new technic they'll use for cover their words unless you've playing their games.

 

Oh yes.

What you've proven is that, like the dog with the rabbit, you won't drop an argument when asked politely to do so. We know what you think as you've told us 100 times. You are adding nothing to further your argument, just repeating yourself over and over and over. All repeats of this theory of yours are just noise on the forum. It's a waste of typing, and a waste of reading. That's why we asked you politely to be quiet, lots of times, and handed out warnings. Which you still don't get. So a 7 day ban for you to give us all a rest and point out what will happen if when you return you don't make changes. You will be gone.

 

More interestings information about RSX and ps3:

" RSX

FIXME DRAFT

The RSX is the graphics and audio rendering chip of the PS3. It has access to 256MB of dedicated local DDR3 memory. It also has access system memory through the high-speed FlexIO bus of the Cell using DMA operations. The RSX chip is produced by NVIDIA and is similar to a GeForce 7800 GPU. However, no public documentation has been provided yet by either Sony or NVIDIA on how to program this chip.

This page gathers the information obtained so far by experimenting with the Hypervisor API and the Linux framebuffer driver. It summarizes information relevant to the PS3 RSX only. For a better understanding of NVIDIA GPUs in general see the nouveau project. In particular, this page (of the nouveau wiki) introduces many concepts needed to understand NVIDIA GPU programming."


http://wiki.ps2dev.org/ps3:rsx

 

Interesting if correct, I thought the Cell was responsible of processing sound. Or may it be that the recieves processed sounds as part of HDMI?

 

We know RSX handles audio in some way from the discussion a while back where certain people in the know ridiculed the suggestion that a Sony conference mentioning 'RSX Audio' was talking about a years old 3D audio lib . I don't recall if the exact nature of RSX's involvement has been revealed, ut as I understand it, Cell handles audio creation, taking and blending samples, and RSX handles the encoding. I might be wrong on this though.

 

We've moved on from this topic, but for future reference, I realise that we have a pretty clear way of showing non-believers the RSX clock frequency without violating any NDAs, because here's how the Linux driver reports the RSX:

$ dmesg|grep rsx
ps3rsx: PS3 RSX access module, 1.0.0
ps3rsx: reserved XDR memory is @c000000000d00000, len 18874368
ps3rsx: 254MB of DDR video ram at 0x700190000000 mapped at d000080080780000 handle 5a5a5a5b
ps3rsx: context 0x55555554 dma=440000381000 driver=480000684000 reports=4c0000d90000 reports_size=10000
ps3rsx: ctrl=d00008008010c000 drv=d0000800905c0000 reports=d0000800804e0000
ps3rsx: version 2.11 RSX rev17 0MB RAM channel 1 core 500MHz mem 650MHz
ps3rsx: remapped XDR apperture at c000000000d00000 size 18432kB to RSX

Taken from this thread:

http://forums.ps2dev.org/viewtopic.php?p=67476#67476

 

Good find! At least now there is somewhat concrete proof of the numbers. I still find it strange that Sony decided to lower the spec of the Ram. I can't believe doing so would really save them that much in costs? Nor would it really save much on the power and heat requirements. I guess it's something to do with the down clock of the RSX. Any ideas?

 

There's tonnes of reasons, but the one that people gloss over most easily is simple: heat reduction. It's an incredible challenge to keep the device cool enough and still reasonably quiet; as Sony has stated a couple of times publicly, that was one of the toughest challenges when designing the PS3, and it resulted in among others a very innovative design and use of a huge fan that cools outwards. And also interesting to note is that while Sony has just barely managed to make the PS3 both reliable and silent, a certain other company wasn't so successful.

 

Every time I see this thread I'm reminded of this long running skit on Conan O'Brien

 

The cake is a lie

 

Good points but I was thinking in overclock on RSX 65nm

(maybe now with 65nm and future going to 45nm 550MHz is possible in some way)

 

what abo0ut the older models?

Even if its possible I doubt they will do it

 

Maybe guys Homebrews ,linux universe can do it but certainly with some risc (90nm models would need more fan velocity some adjust mods).

(i will like this feature if its possible)

 

Holy crap, I think the same thing!

 

I was wondering something... the RSX and Xenos both have a 4Mp/sec fill-rate yet the memory in the RSX is clocked at 650 MHz and Xenos 700 MHz. This would lead me to believe that what determines the pixel fillrate is the ROP frequency which is 500MHz in both systems. Now since ROP frequency increase would augment pixel fillrate, what would memory frequency do other than provide what I would assume be unusable bandwidth since ROPS in both systems seem to be the bottleneck. In other words was memory frequency dropped 50MHz simply because 8ROPS at 500MHz can only push 4 Mp (with memory at 650 or 700MHz)?

Also, what kind of benefit would have 550MHz ROPS have given RSX? If fillrate average utilization is something like 30% of total bandwdth (major assumption) and RSX has 20GB/sec to GDDR3 memory. That would mean 6GB/sec is being used for fillrate. Assuming the exact same graphics in a game (effects, resolution, textures, etc.) and that the extra 2.4GB/sec was being used for fillrate purposes, what would be the difference in the output of PS3 games today? That would be a 40% increase in fillrate!!!

 

I think it was the interplay between both heat and yields; heat because power draw and heat dissipation are obviously parameters they need to work within, and yields because the RSX likely had a target voltage for 550MHz operation - a target that they probably couldn't reach in sufficient quantities with initial product and for which the resultant voltage bump would have proved too dear. So, they likely went to 500Mhz as the next stable voltage tier that resulted in thermals within their parameters.

Anyway on the side many thanks Arwin for the 'in-writing' affirmation of the 500MHz clockspeed above; maybe finally we will see the end of the madness around here!

I'd say if B3D handed out medals, you would deserve one just for that!