PS3: less RAM needed for the OS = old games run faster?

It is my understanding that the 43MB referred to XDR( "System") usage and does not include GDDR3 ("RSX") usage?

So it's not true that they don't take memory from the RSX anymore? People have been claiming that for awhile now with little proof that I can find. Others are claiming the OS usage is within 10MB of the 360 with no proof. I just wish this wasn't such a big secret.
 
can someone find it for me plz, I'd like to read up on it
Me too. AFAIK no reputable source has said that. I've only seen it passed around as fact by people, but no one can ever seem to locate the source that made the claim.
 
Me too. AFAIK no reputable source has said that. I've only seen it passed around as fact by people, but no one can ever seem to locate the source that made the claim.


Since revealing such information is a breach of NDA, no "reputable" source will want to take credit for that. You'll just have to let it go and be happy with the rumors.
 
Heh, I totally forgot I posted about the 43Mb OS memory before >_<

Am interested to know how well the HDD Virtual Memory works. I believe Insomniac's Blu-ray + HDD streaming system was done independent of it (Too early ?).
 
Since revealing such information is a breach of NDA, no "reputable" source will want to take credit for that. You'll just have to let it go and be happy with the rumors.
I'm referring to reputable websites with anonymous sources, etc. as well not like "Ted Price says so". Or even subtle hints from PS3 devs here.
 
I think it was leaked long ago on a relatively obscure site (www.innerbits.com ? Seems defunct now). Any news/rumors on improvement are likely to be from "gray" sources as well (When it comes to technical domain, regular gaming sites made too many mistakes in the past to be credible anyways).

That 43 Mb reference in the Japanese interview is an exception. The engineers themselves revealed the info, but may be incomplete. ^_^
 
Am interested to know how well the HDD Virtual Memory works. I believe Insomniac's Blu-ray + HDD streaming system was done independent of it (Too early ?).
The VM system works pretty well, but I doubt any streaming system would use it. Much easier to regulate/manage sector reads & memory usage/partitioning (and hence performance) with a system designed for the task.

I would expect that even if VM was available early in development, developers would not consider using a VM system for non-traditional means.

Dean
 
The VM system works pretty well, but I doubt any streaming system would use it. Much easier to regulate/manage sector reads & memory usage/partitioning (and hence performance) with a system designed for the task.

I would expect that even if VM was available early in development, developers would not consider using a VM system for non-traditional means.

Dean

Then what is it used for?

Is it used/usable only by games at the moment?

Does the PS3 OS and its applications use it (maybe to achieve in-game XMB... moving the extra "fat" in HDD memory when game run and restore the whole XMB when game exit)?

If so, why does the web browser still get out of memory errors? Maybe it has to be re-written to make use of the VM system?
 
Not true.. games are told the amount of memory available for the SDK that it was built against. Not the firmware revision.

So, for example, a game built against SDK 1.80 running on a current 2.60 firmware would be given the same free memory information as if it were running on a machine with the 1.80 firmware, ensuring compatibility. If you were to change the memory amount, you change the behaviour of the game (on an internal level), and you invalidate a lot of your testing as a result.

This. But one thing to note is, if you're releasing a patch down the line, it's plausible that you could take advantage of any increased memory allowance freed up in the meantime.

Burnout Paradise was 2.10 (IIRC) at release, but subsequent patches have added features from later SDKs (like custom soundtracks from 2.40), shifting the firmware requirement (and hence memory specs).

Me too. AFAIK no reputable source has said that. I've only seen it passed around as fact by people, but no one can ever seem to locate the source that made the claim.

It's possible to make deductions, whether they are correct or not is another matter ;)
 
This. But one thing to note is, if you're releasing a patch down the line, it's plausible that you could take advantage of any increased memory allowance freed up in the meantime.
Very true! I guess then it comes down to the developers weighing up increased development cost (engineering/testing/content) to utilise any change in this area. Not so likely with bug-fix patches, but more with the larger type of patch (with lots of additional content) such as Burnout Paradise.

Dean
 
Is it used/usable only by games at the moment?
It's certainly usable.. as usable as any other well-engineered HDD VM system I've seen. Whether it's used by any existing titles is something I don't know though. My point was really that while with a VM system you may be able to prod pages and get them to be loaded/unloaded (leaving the management of pages in the hands of the OS), the way applications use memory (and structure their data both in-memory and offline) is typically so custom in order to ensure effective CPU-load, caching and pre-loading of geometry/textures, that teams are more likely to use that than hand over control to the OS. Also a lot of teams have their streaming systems tightly integrated with SPU-based decompression, so wedging that into a pre-existing VM scheme may be more trouble than it's worth.

Does the PS3 OS and its applications use it (maybe to achieve in-game XMB... moving the extra "fat" in HDD memory when game run and restore the whole XMB when game exit)?
Consider that while the in-game XMB is running, the game is running.. as far as I can see, there'd be no way for the OS to automagically determine which chunks of memory it may be able to page out to HDD in order to give more room for the in-game XMB. So the memory that's reserved is what the in-game XMB would have to play with. I think the XMB might use the VM system (or something similar) as when I use the in-game XMB at home I do see a fair amount of HDD access going on. This is just my guess though..

Cheers,
Dean
 
Yup, VM for the web browser (in standalone mode) would make a lot of sense. Is it advisable to run such a thing in-game ?

Thanks DeanA for the answers. I was curious whether the VM is a traditional design or enhanced for game specific environments (i.e., allowing the OS to adjust its policies).
 
Thanks DeanA for the answers. I was curious whether the VM is a traditional design or enhanced for game specific environments (i.e., allowing the OS to adjust its policies).
More traditional, with a little bit of PS3 special sauce. Can't really go into details (no matter how mundane) due to the usual NDA shenanigans..

Cheers,
Dean
 
Thank you very much for your answers Dean :),

of course I have more questions :D.

Talking about games... you were saying that basically if you want to get good performance (with or without having streaming implemented in your game) you will have your vm scheme customized to assure optimal CPU load and integrate well with your processing and content pipeline so this OS provided scheme might not be used by those titles pushing for peak efficiency above all.

Might it be a good solution for quick/smaller PSN titles that are not exactly pushing the graphics envelope that much to need those extra CPU cycles, but would enjoy using an OS provided and tested/debugged/certified system (important, less code to test for TRC submissions :)) and implement on top of that a streaming mechanism to allow open world exploration?

Sure, all sub-optimal compared to a custom solution, but it might work ok for small devs (which might already be using something like the Phyre Engine on a Debug Kit).
 
Yup, VM for the web browser (in standalone mode) would make a lot of sense. Is it advisable to run such a thing in-game ?

When was the last time the web browser told you it ran out of memory? This used to happen quite a bit a while ago, but now it never seems to happen. Is your experience the same? This could confirm that the page file is at the very least helping out there.

Also I'm thinking that if you use the VM for general applications on the PS3 (i.e. not performance critical) then that also means you could later on release a PS3 with more memory and benefit from this a little, when it makes sense. I'm not saying it will ever happen, but it's possible - it could keep the PS3 competitive as a multi-media device for a long time.
 
Good point. I seem to recall one months back, but then again I don't browse that long or often on the PS3 (although the frequency is increasing).

Yesterday, I visited the full screen youtube URL (www.youtube.com/tv), singing free karaoke with my family for an hour.

There were brief loading occasionally. I was waiting for an out of memory error but it never come.

Need to fire up some Flash games to try for real.
 
of course I have more questions :D.
Hehehe.. surprise! :D

Talking about games... you were saying that basically if you want to get good performance (with or without having streaming implemented in your game) you will have your vm scheme customized to assure optimal CPU load and integrate well with your processing and content pipeline so this OS provided scheme might not be used by those titles pushing for peak efficiency above all.
Well, you'll have your streaming scheme customized, but again I don't really consider this to be the same thing as an HDD-based VM system. For one thing, VM typically works on large-ish page sizes (the size depends on the underlying OS page support), and the load of a page (of 'n' Kbytes) is usually triggered by an access to the memory associated with that page via hardware page protection mechanisms. Sure, you can hit pages such that they're loaded before they're needed, but the granularity of a page may be very different to a sector-based loading scheme (with sectors being ~2Kbytes).

Actually while thinking about this, it comes to mind that it's worth considering the size of the streamable data if you have thoughts about pushing it onto a VM system. Both PS3 and 360 OS's use a 32-bit ABI (it's not like we have GB's of memory on a console.), so the maximum addressable space is 4GB. Taking into account reserved space (hardware mappings for SPUs, RSX, and other reservations), you've likely got a much smaller space available for to map your VM into than the size of your entire streamable data set. If you see what I mean.

Basically, if you've got a BD-ROM with a single streamable data set of 25GB, then you're not going to be able to fit it into your addressable space, like it or not. So having it all mapped via VM.. well, it wouldn't work. FWIW I can't remember how much you're able to map with the OS VM system - and even if I could remember, I doubt I'd be able to go into any more detail without a lawyer coming up and *smashing me in the face with a hammer*.

My summary still stands that VM is not really anything to do with streaming data systems. With regards to your PSN/smaller title point, these tend to know they're running from HDD, so I doubt that streaming performance (such as you'd get from BD-ROM) is of primary concern. So normal file access may suffice (with a bit of WAD-style file packing where needed).

And using the OS doesn't reduce the testing.. if you feed chuff to the OS, it's going to bail just as if you'd written it yourself. The testing guys won't even be aware of whether you were using OS routines, or your own anyway.. :)

Cheers,
Dean
 
Back
Top