The PS3 OS - Ram Hog or just big boned?

[Barnaby Jones]

I'm hearing murmers that it is a pretty big number, has Sony come out and stated an exact number?

DeanO has aluded to it being pretty high in the past, now I'm hearing a concrete number from a few reliable sources(devs).

That number is....
96 MB's.




The original DeanO quote:

You would think that as we get more and more memory in consoles, that memory management would get easier but alas its not to be…

PS3 has 512Mb of RAM in total, which is a sizeable chunk when compared with the last generation of consoles (64Mb in the Xbox). The problem is that, we want to do much more and we want to do it ‘easier‘.

So where does it all go…

Well first you have to OS thats going to take a bit of the apple. And its a sizeable chunk…

Then we need space for all these lovely HDR AA framebuffers and render targets.

Then we have textures and geometry and least we forget animation data and sound. These content items consume the vast majority of the RAM. Off the top of my head its in the region of about 300Mb in total!

 

[Barnaby Jones]

Since I can't edit my posts, I'll boldface the important quote DeonO made.

Well first you have to OS thats going to take a bit of the apple. And its a sizeable chunk

Any idea what the X360 OS takes up? I heard everywhere from 32 to 48.

 

[Gholbine]

I don't even think my copy of Windows XP takes that much, let alone any Linux distro I've ever used. My FreeBSD box is taking nowhere near that figure.

Somehow I doubt it's 96MB... especially considering it's in a console with limited and very specific features.

 

[Barnaby Jones]

I don't even think my copy of Windows XP takes that much, let alone any Linux distro I've ever used. My FreeBSD box is taking nowhere near that figure.

Somehow I doubt it's 96MB... especially considering it's in a console with limited and very specific features.

well, the guy who made the M2 OS (widely regarded as the reason the M2 was nerfed) is also in charge of the PS3 OS

http://en.wikipedia.org/wiki/Panasonic_M2


BTW, I'm hearing it is 99.99% set in stone, 96 MB of RAM.

 

[Platon]

That sounds to be quite a bit, actually way to much. It depends though what that OS is supposed to do and if the OS functions help out with the game or whether most of those functions are non game related. I have to say that if this is true, then Sony really seem to be putting a lot of effort on functions that sound really good and "futuristic" but in the end might really not be that much used by the majority of PS3 owners...

 

[liverkick]

BTW, I'm hearing it is 99.99% set in stone, 96 MB of RAM.

And Im hearing its 99.9999% set in stone that you wont be providing a link and/or source for this information.

 

[DeanA]

well, the guy who made the M2 OS (widely regarded as the reason the M2 was nerfed) is also in charge of the PS3 OS

And who would that be?

Dean

 

[Shifty Geezer]

That strikes me as excessive. 96 MB will surely go unused almost all the time. If that 96 MB is true and not used by games in any way, it'd be a terrible waste.

 

[patsu]

Heh heh. I think the right way to see it is the PS3 OS is small (embedded OS is always small by definition)... but depending on what kind of data it needs to work with (e.g. HD video), it may take up to 96 Mb ? So to guarantee quality of service as a media server (to perform background jobs), they reserve 96Mb for the "OS". Is this accurate ?

 

[Platon]

Heh heh. I think the right way to see it is the PS3 OS is small (embedded OS is always small by definition)... but depending on what kind of data it needs to work with (e.g. HD video), it may take up to 96 Mb ? So to guarantee quality of service as a media server (to perform background jobs), they reserve 96Mb for the "OS". Is this accurate ?

Yes, of course. The OS in it selfs is really small, or atleast should be, but if you want it to do some more stuff than playing games, then yes, it will take advantage of those 96MB...

 

[kimg]

The number "96 MB" tells us nothing, do it contain libraries? kernel modules who can be loaded to support high level runtimes for languages? who know.

Kutaragi have said the PS3 is to use high level languages for devlopers, and ive read enough about "run time superclustering" in "real time OS" now, im just waiting to see whats popping up when devlopers NDA is lifted, and they can tell us about it

 

[Shifty Geezer]

The number "96 MB" tells us nothing

It tells us something. It tells us, if true, that out of 512 MB available RAM, 416 MBs is availalble for game code and data.

 

[DarkRage]

That is related to another concern I have:

After checking all the documentation IBM has made available for optimizing code in Cell, almost every approach came with the same negative effect: more memory used for the code (unrolling loops, reordering instructions, multiple instructions because of the constraints of the SPEs).

Maybe just the "optimized for speed" code uses too much memory?

It would be interesting if someone has some results about how much memory is the same C code (optimized for speed) using in SPE, PPE and xCPU.

 

[Shinjisan]

A few people really can't have a rest

 

[kimg]

Shifty, i just assume, if it can be possible libraries of OS functions is available, i dont just assume an console use same amount of RAM as the operative system i use when i write this (windows XP). winxp use lots of memory, my machine use now, when i write: 90720 KB total, 64220 in swapfile, and 26520kb in memory.

Even if PS3 got support for every function who windows XP and z/OS got, it still dont mean its occupying 96MB in memory, and now we assume the runor is set in stone as truth.

 

[Titanio]

11 posts, red square, questionable post history with references to "rei-rom.com" amongst them. Hmm...screams credibility.

Which isn't to say this couldn't be true, but I wouldn't believe it coming from the OP.

That is related to another concern I have:

It's actually not related..

After checking all the documentation IBM has made available for optimizing code in Cell, almost every approach came with the same negative effect: more memory used for the code (unrolling loops, reordering instructions, multiple instructions because of the constraints of the SPEs).

How is this negative if it achieves better performance? Instruction reordering wouldn't in and of itself result in a larger footprint either, AFAIK.

Maybe just the "optimized for speed" code uses too much memory?

There is often a time/space tradeoff. That's not unique to Cell. Which you trade off depends on what you're doing, your requirements, and the characteristics of your system.

It would be interesting if someone has some results about how much memory is the same C code (optimized for speed) using in SPE, PPE and xCPU.

The same code would take the same amount of space. It's likely you wouldn't use the exact same code, though many of the recommended practices you see used on SPEs (loop unrolling, branch avoidance etc.) would probably often be a win on all the chips you mentioned.

 

[Shifty Geezer]

How is this negative if it achieves better performance?

DarkRage clarified that as the effect of using more memory being negative. Which I'm sure you agree with. Saving memory is a good thing when you have a limited amount to play with. Using up more memory is a negative effect to provide the positive effect of speeding up code, in this example, which I can't comment on whether that's fair assessment or not.

 

[Titanio]

DarkRage clarified that as the effect of using more memory being negative. Which I'm sure you agree with. Saving memory is a good thing when you have a limited amount to play with. Using up more memory is a negative effect to provide the positive effect of speeding up code, in this example, which I can't comment on whether that's fair assessment or not.

I'm not sure if you something that precipitates better overall performance can be construed as a negative If the alternative is to code & compile for space and it runs worse, overall, then that's hardly a more positive alternative. You've got your "positive effect" of saving space, but it's ultimately yielding negative results.

 

[inefficient]

I'm sure that's not the kernel. It has to be when the OS is running a window manager supporting a full GUI for non gaming tasks plus running networking services, peripheral drivers etc.

Anyway, even in the worst case, where the OS persistantly loaded, since the ps3 has a HD you could just page those resources not needed for the game you were playing to the disk.

As an aside... from what I understand, there has never been a gaming console with a virtualized memory system. Not even the xbox used one because of the performance side effects plus the OS overhead required. Maybe the ps3 will be the first?

 

[Hardknock]

11 posts, red square, questionable post history with references to "rei-rom.com" amongst them. Hmm...screams credibility.

Not sure what his credibility or what other forums he posts on has to do with the topic? He has a quote from DeanO a poster on this very forum. Obviously devs know, so this shouldn't take to long to be confirmed or debunked.

If this is true hopefully we can download content while playing games (that's the one thing I hate about the 360 OS).