Japanese article about PS3 backwards compatablilty

[leechan25]

Not sure what you mean. Like run some code on the EE/GS while the ps3 cell/rsx is doing its own thing?

If so, if and when Sony decides to take the EE/GS out, wouldnt that cause problems for old games that used the EE/GS for other tasks?

Well, that's if they take it out in the future...

anyway, what if any code could be written for the ee/gs to lighten cell/RSX task?

 

[Bad_Boy]

Well, that's if they take it out in the future...

anyway, what if any code could be written for the ee/gs to lighten cell/RSX task?

well, if they dont plan on taking the EE/GS out, I dont see why that would be a problem. Thats a bit of power to put toward one task. But I dont remember hearing anything about the PS2 using the PS1 hardware for another task.

 

[Murakami]

Hey Guys,

What task could developer off load to the ee/gs within PS3?

I thought that IOP in PS2 was used to offload some tasks to EE or SPU2 in some PS2 games, but if it's true that in PS2 slim the IOP is no more the R3000+GTE+MDEC...

 

[TheChefO]

well, if they dont plan on taking the EE/GS out, I dont see why that would be a problem. Thats a bit of power to put toward one task. But I dont remember hearing anything about the PS2 using the PS1 hardware for another task.

I thought ps2 used ps1 cpu for I/O?

 

[supervegeta]

Chef-O I think you're not reading my post.

Yes, *obviously* the GS bandwidth is the problem area. The question is how will that be resolved. It needn't be with a straight GS; the question is whether it can be done without hardware or not. I think hardware is required, but... I'm just taking a 'wait and see' view to it until more info leaks.

To be honest i don't see any tipe of gfx on the ps2 that the ps3 could not handle...so yeah the ps2 have more vram bandwidth but it is not like every game used all that bandwidth effectively.

It would be cool though to have the gs with it's edram avaible to the devs to be used on ps3 games...

 

[Npl]

How is it possible to emulate the ps2 without having sufficient bandwidth and without rewriting the original code?

1) Having a smarter GPU, which actually uses internal caches, z-compression, hierarchical Z-Buffer... The absolute bandwith is smaller, but used more efficient (there are still cases where these things dont help tough). I guess the most reasonable "customization" for RSX would be bigger caches.

2) defering rendering and detecting what the game does, eg. a PS2-game using multi-texturing would disable z-writes, send the same vertex-data one for each blending-stage, then enable z-writes and send the same vertex-data for the final stage. If such a sequence would be detected RSX could do that in 1 pass.

3) Per-Game patches ( AFAIK called "profiles" on XBox360 ). Could be like 2), but more specific to games, eg. adding detection for sequences used in that game

In the end Sony wont want to add more legacy-hardware each generation, I can see starting them with an added EE+GS, but having a pure-software-emulator too. A PS2 Game would be run with the software-emu, and if it has problems the user could tick to enable a "more compatible" box and use the PS2-Hardware.

 

[Fox5]

How is it possible to emulate the ps2 without having sufficient bandwidth and without rewriting the original code?

The PS2 had pretty much no bandwidth saving techniques, those should help a lot. The PS3 should be fast enough to do hidden surface removal in every game.

 

[Bad_Boy]

I thought ps2 used ps1 cpu for I/O?

ah, that I did not know. makes sense.

 

[Fafalada]

I guess the most reasonable "customization" for RSX would be bigger caches.

I think RSX will have caches well covered. The real problem is that a lot of rendering behaviour is cache unfriendly on PS2 (and even more unfriendly for complex rasterizer pipelines) - some of which I mentioned in this other thread..

Anyway, personally I am not convinced that workarounds for these issues can be done without game "profiles" (even for the simple multitexturing example you gave, I don't think there's one certain way to detect what app is doing - there are just far too many different variations in different apps).

 

[crystalcube]

A PS2 Game would be run with the software-emu, and if it has problems the user could tick to enable a "more compatible" box and use the PS2-Hardware.

If Sony intend to remove PS2 hardware later then having such a checkbox does not make sense. IMO Sony would like to make it as transparent as possible for end user. The only thing it may require is network connectivity so when a user tries a PS2 game , any "profile" for it will be downloaded and executed.

 

[Npl]

If Sony intend to remove PS2 hardware later then having such a checkbox does not make sense. IMO Sony would like to make it as transparent as possible for end user. The only thing it may require is network connectivity so when a user tries a PS2 game , any "profile" for it will be downloaded and executed.

That would be the best case. Realistically I doubt they will play through each game themselfes, so they let the users "help" them. Similary like you could adjust CD-Speed for PS1 Games and some of them dint like it.

I think RSX will have caches well covered. The real problem is that a lot of rendering behaviour is cache unfriendly on PS2 (and even more unfriendly for complex rasterizer pipelines) - some of which I mentioned in this other thread..

Anyway, personally I am not convinced that workarounds for these issues can be done without game "profiles" (even for the simple multitexturing example you gave, I don't think there's one certain way to detect what app is doing - there are just far too many different variations in different apps).

I was just giving few examples how you could help work around the bandwidth-bottleneck. It wasnt yet mentioned in this thread that the GS can switch states like mad, has 2 simultan renderstates and in short has alot possibilities which current GPUs would struggle at . Neither do I think that emulating the EE will be an easy task, the seperate components wouldnt be hard, but having all of them and the DMA between act together like the real hardware...

 

[TheChefO]

Interesting thoughts on possible workarounds - So without the ps2 guts, they could possibly use "profiles" for each game in the same fashion that MS is doing and achieve there goal but would this require thousands of profiles one for each game or would it be possible to have "themed profiles" for the functions/engines specifically and not necessary per game?

curious - does the slimline ps2 play ps1 games? if so is it using emulation or does it still have the ps1 cpu?

 

[archie4oz]

curious - does the slimline ps2 play ps1 games? if so is it using emulation or does it still have the ps1 cpu?

Yes it plays them (there's about 40 games that don't work on the most recent model). The 75xxx series PStwos don't have the LSI IOP anymore (they're using a different IOP) so I don't know how much "PS1 is left" on it.

 

[leechan25]

well, if they dont plan on taking the EE/GS out, I dont see why that would be a problem. Thats a bit of power to put toward one task. But I dont remember hearing anything about the PS2 using the PS1 hardware for another task.

A bit? I would think that's alot of power if it were used for a simple task. However, I still don't which kinds of task could be done by the ee/gs to lighten cell/rsx loads?

 

[TheChefO]

Yes it plays them (there's about 40 games that don't work on the most recent model). The 75xxx series PStwos don't have the LSI IOP anymore (they're using a different IOP) so I don't know how much "PS1 is left" on it.

So it's using emulation?

 

[ban25]

Yeah I think that EE+GS has been planned for a while, and this emulator project has likely been running concurrent.

Still though there must be a hardware consideration built into the roadmap other than just EE+GS--->emulator, right? Because once that's gone, emulator or no, I just don't see how GS bandwidth could be emulated; will there be a replacement chip for the EE+GS, hardware considerations on RSX, or indeed would total emulation be a feasible achievement?

Don't underestimate what a high-performance set-associative cache can do.

 

[ban25]

So it's using emulation?

No, it's using a different chip that is more favorable from a cost standpoint. I think it is an in-house part owned by Sony, but I'm not sure. However, it's not 100% compatible and the result is that bugs pop up in some games (like the Tekken 1-3 games in T5).

 

[TheChefO]

No, it's using a different chip that is more favorable from a cost standpoint. I think it is an in-house part owned by Sony, but I'm not sure. However, it's not 100% compatible and the result is that bugs pop up in some games (like the Tekken 1-3 games in T5).

gotcha thanks for the info - it must have been significantly cheaper for them to bugger up their bc with this new chip. hmm - just doesn't make sense to me to save a few bucks and ruin a perfectly good bc scheme.

 

[Shifty Geezer]

gotcha thanks for the info - it must have been significantly cheaper for them to bugger up their bc with this new chip. hmm - just doesn't make sense to me to save a few bucks and ruin a perfectly good bc scheme.

$5 * 20 million consoles could be the difference between profit and loss for the division. When you're looking at such large numbers of units, every penny saving amounts to a lot.

 

[Bad_Boy]

A bit? I would think that's alot of power if it were used for a simple task. However, I still don't which kinds of task could be done by the ee/gs to lighten cell/rsx loads?

tomatoes tomahtoes .
What I meant was that I agree thats alot of power for one task. Im not sure what kind of tasks either. physics? particle effects?(which the ps2 was very good at ircc) background gui? who knows.