PS3 GPU not fast enough.. yet?

Fafalada said:
If you are so starved for CPU XDR bandwith that you can't share it at all, you have bigger problems then trying to assist GPU IMO.

But I disgress. Anyway looking at GDDR to aleviate XDR bandwith seems to be upside down perspective on things to me.
Click to expand...
Though I agree wholeheartedly from a game perspective, what about non-gaming functions? This data suggests PS3 has 256 MB for 'Linux' work. For photo or video editing, working in the GDDR isn't an option it seems, and that RAM will be all but redundant for such tasks :)???:). Which wouldn't be surprising, as these extra applications are something of an afterthought I think, rather than the machine being purposefully designed for them.

Also I still don't get where the 16 MB connection is coming from. Did they just solder a bit of copper wire from pin to pin?! How is that handled over the existing Buses? A tiddly little share of the command processor on RSX? Wouldn't a shared memory channel be 4 GB/s up and down shared, rather than 4 GB down, 0.016 GB up?
 
Fafalada said:
If you are so starved for CPU XDR bandwith that you can't share it at all, you have bigger problems then trying to assist GPU IMO.
Click to expand...

It is true, I think it all comes down to my childhood and how my growing up influenced my relations with others and especially girls... wait a moment maybe I ... am... uhm... I misunderstood something... *whistles away*

Personally I pretty much consider any larger CPU write initiated to external memory (that isn't meant to be highly cloned/reused data instance) a loss, period.
Depending on the machine / memory architecture it may or may not be a loss that you can live with.
Maybe if external memory wasn't so damn slow compared to our computing architectures - I would reconsider my stance, but it doesn't seem like that's likely to happen anytime soon.

But I disgress. Anyway looking at GDDR to aleviate XDR bandwith seems to be upside down perspective on things to me.
Click to expand...

So you think that XDR was chosen instead of a lower bandwidth solution with in mind the need of helping RSX not to be bandwidth starved with its lone 22.4 GB/s link to its VRAM ?

Truth to be told, according to that slide, XDR and its XMC seem to be MUCH more efficient at reaching peak bandwidth when the CBE chip is writing to it rather than reading from it:

XDR Reads initiated by the CBE chip: 16.8 GB/s of 25.6 GB/s ~= 65.625%

XDR Writes initiated by the CBE chip: 24.9 GB/s of 25.6 GB/s ~= 97.27%

It seems that writing to XDR from CELL being so MUCH more efficient than reading from it would favor a LOT the idea of writing to XDR from CELL and initiating a read from RSX rather than writing from CELL over to RSX's GDDR3 memory.

P.S.: Please digress more and more often, I do not think people mind it at all :D.
 
Last edited by a moderator:
Shifty Geezer said:
Also I still don't get where the 16 MB connection is coming from. Did they just solder a bit of copper wire from pin to pin?! How is that handled over the existing Buses? A tiddly little share of the command processor on RSX? Wouldn't a shared memory channel be 4 GB/s up and down shared, rather than 4 GB down, 0.016 GB up?
Click to expand...
I don't think its something physical, I'm beginning to suspect that this is an overall processing rate. If Cell is requesting something from RSX's memory then it may need to go in as a command and be queued, ready for RSX to retrieve and pass out when it can schedule the work.

[Edit] Note the "~" on the Cell read/writes.
 
Magnum PI said:
according to EXIF data the picture was taken the 4th of June, 20:20:59
Click to expand...
Your PI skills are weak.. it seems likely that Sony don't hold DevStation events at 8:20pm on a Sunday night..

:)

Dean
 
Magnum PI said:
according to EXIF data the picture was taken the 4th of June, 20:20:59
Click to expand...
Yes, following Dean's comments, isn't that more likely the time the image was posted/received, seeing as that was yesterday? Unless there was a second DevStation event that we haven't heard about...
 
Shifty Geezer said:
Yes, following Dean's comments, isn't that more likely the time the image was posted/received, seeing as that was yesterday? Unless there was a second DevStation event that we haven't heard about...
Click to expand...

the purpose of exif data is to store in the jpeg some data about the shot like the aperture, the camera model, iso and time the picture was taken.
 
Magnum PI said:
the purpose of exif data is to store in the jpeg some data about the shot like the aperture, the camera model, iso and time the picture was taken.
Click to expand...
sorry, nm.
 
Last edited by a moderator:
DeanA said:
Your PI skills are weak.. it seems likely that Sony don't hold DevStation events at 8:20pm on a Sunday night..

:)

Dean
Click to expand...

I just said what indication gives the EXIF data.. it's a fact

That must be taken with a grain of salt as you can set this value at the value you want by setting the time of the camero or manipulating EXIF headers

And even if this value is true we don't know the timezone, maybe Charlie Demerjian took it (his place of residence would be a clue), maybe someone else..

Why do you think it was taken during a devstation event ? AFAIK it could be a shot of a PDF on some TFT screen, during a flight to Japan..
 
Magnum PI said:
Why do you think it was taken during a devstation event ? AFAIK it could be a shot of a PDF on some TFT screen, during a flight to Japan..
Click to expand...
Actually, good point. Just because it says DevStation, doesn't mean that's when it was being shown for this photo.
 
one thing from this article: the ps3 is my dream pc. monstrous bandwidths in all the right directions.

sans the tri setup - that's a bummer.
 
Okay now this story is really getting interesting.
33.gif
 
So is this info true or false? I myself, never trust the inquire, since they've been wrong many times in the past
 
PSman said:
So is this info true or false? I myself, never trust the inquire, since they've been wrong many times in the past
Click to expand...

It's true, but mostly irrelevent. Which is to say... ignore the entirety of Charlie D's article text, look simply at the slide, and contemplate the implications. There really aren't many.
 
The only real implication is that non-graphics data and graphics data processed on CELL has to be stored in 256MB.

Shock, shock and horror. It's all over.
We're all gonna die tommorow!
... What? Oh, alright, nm then.


Uttar
 
You must log in or register to reply here.

Similar threads

A
Holy Wall of Text: My impressions post Playstation 4 (PS4) unveil
2 3 4
Replies
70
Views
20K
(((interference)))
I
O
PS3 devkit
2 3
Replies
55
Views
12K
ChryZ
C
O
  • Locked
Gamespot infers that Killzone was realtime at 75% power
Replies
17
Views
3K
jvd
J
M
  • Locked
SquareEnix explains why FFXIII is PS3 exclusive - DVD9 not enough
12 13 14
Replies
267
Views
25K
Shifty Geezer
Shifty Geezer
A
  • Locked
Hardware PS2 Emulation + BIG onboard Flash Memory for PS3
Replies
3
Views
3K
Farid
F
Back
Top