PStwo I/O processor, now a PowerPC 440?

paawl

Newcomer
I've seen it mentioned (see below, for example), that Sony may have replaced the R3000A I/O processor (and the SPU2 sound processor?) with an IBM PowerPC 440 in the latest PStwo (slim) design.

http://www.beyond3d.com/forum/showthread.php?p=788153&highlight=r3000a#post788153

Does anyone with first-hand knowledge think that is actually the case? If Sony were to have replaced the IOP with another chip, I assume that it would only be for cost reduction purposes, and then it seems more likely that Sony would have used one of the many R3000A clones, like Toshiba's TX39, rather than a 533 MHz PowerPC 440, which probably has a different bus architecture (and hence different connection to the EE+GS), requires a larger DRAM to hold the software for emulation, etc.

Does anyone know for sure?
 
I've seen it mentioned (see below, for example), that Sony may have replaced the R3000A I/O processor (and the SPU2 sound processor?) with an IBM PowerPC 440 in the latest PStwo (slim) design.

http://www.beyond3d.com/forum/showthread.php?p=788153&highlight=r3000a#post788153

Does anyone with first-hand knowledge think that is actually the case? If Sony were to have replaced the IOP with another chip, I assume that it would only be for cost reduction purposes, and then it seems more likely that Sony would have used one of the many R3000A clones, like Toshiba's TX39, rather than a 533 MHz PowerPC 440, which probably has a different bus architecture (and hence different connection to the EE+GS), requires a larger DRAM to hold the software for emulation, etc.

Does anyone know for sure?

Perhaps its linked to plans to run PS2 emulation on Cell? The PPC could be emulating the I/O chip, then they just have to migrate the code. (This is just a guess)
 
Perhaps its linked to plans to run PS2 emulation on Cell? The PPC could be emulating the I/O chip, then they just have to migrate the code. (This is just a guess)

I suppose it's possible. I was assuming that, this late in the PS2's life cycle, Sony wouldn't do anything to the innards unless it resulted in a console that was cheaper to produce. Even if Sony did develop the emulation software for the PS3, what incentive would they have to use it in the PS2? I don't think they'd use it just as a test of the technology, I assume that there would have to be some cost savings, but the old IOP couldn't have been that expensive to begin with.
 
Makes sense since it would sweeten the deal that they already have going on with PS3 , Also it might open up PC processors from IBM in their VAIO lines in the near Future.
 
paawl said:
Even if Sony did develop the emulation software for the PS3, what incentive would they have to use it in the PS2
Cost reduction - moving the PS2 chipset completely in-house - apparently IOP license fee was costing them more then the chip itself, and new hw(memory was changed as well) is supposedly cheaper to make as well.

And no, I have absolutely no idea what the new chip is - I find it funny that usually people comb over every console's innards with a microscope, but PS2 75000 series got away unscathed even though people KNEW BC compatibility was affected to some degree.

I doubt architecture used is particularly relevant for PS1 emulation - they have clearly perfected this quite far over the years as PSP demonstrates as well (not to mention they own Connectix), and PS3 has plenty power to spare for it.
Maybe some of the resident Sony ppl can shed some light on this - like what chip designs Sony 'owns' that might fit the profile.
 
IBM doesn't manufacture any PC compatible microprocessors.. To my knowledge they never actually have either.
They have, when dinosaurs roamed the earth.
Wikipedia said:
The Cyrix 6x86 is a sixth-generation, 32-bit 80x86-compatible microprocessor designed by Cyrix and manufactured by IBM and SGS-Thomson. It was originally released in 1996.
 
Depending on how you define "manufacture", VIA's x86 CPUs are still produced in IBM fabs. But that's not any different from producing it in a TSMC or UMC fab, it's just a partner choice...
 
Cost reduction - moving the PS2 chipset completely in-house - apparently IOP license fee was costing them more then the chip itself, and new hw(memory was changed as well) is supposedly cheaper to make as well.

I agree that cost reduction is the most plausible explanation for changing/replacing the IOP, it's just the part about replacing it with a PowerPC 440 (mentioned in the other thread I linked) that has me puzzled, because Sony doesn't own the 440 either, they must license it from IBM. And I'd be surprised if the license fee for a high-performance embedded processor like the 440 is less than the license fee for an old dog like the R3000A, unless Sony has some sort of fixed price deal with IBM---maybe Sony pays IBM a flat fee for the 440 that is independent of the number of products shipped using that design. I know IBM has been pretty aggressive promoting PowerPC in the embedded space, so I suppose anything is possible.

I doubt architecture used is particularly relevant for PS1 emulation - they have clearly perfected this quite far over the years as PSP demonstrates as well (not to mention they own Connectix), and PS3 has plenty power to spare for it.

The PS3 I'm not worried about; Cell should be able to fully emulate PS1 by brute force alone. IIRC, the PSP uses a MIPS CPU, so emulation would be pretty easy there as well. In the PS2, though, if you replace the MIPS-based IOP with a PowerPC-based IOP, it will have to run significantly faster just to do the emulation. Again, if cost is the motive, I just can't see how replacing a 40 MHz MIPS-based IOP with a 400 MHz PowerPC-based IOP would be cost-effective.

On second thought, perhaps the problem wasn't the licensing fee, but the manufacturing cost of the old IOP. All the ICs in the PS1 (excluding DRAM) together contained about 2 million transistors. The 440 alone probably contains twice that number, but the PS1 was manufactured on a 500 nm process (at least originally), while the 440 can be manufactured on a 130 nm process . . . hmmm.

We'll probably never know for sure, but it would be nice if someone would do a public teardown of the PStwo, just to satisfy my curiosity. :smile:

Does anyone even know which IC, in either the original PS2 or the PStwo, is the IOP? I would guess that it's the package labeled "Korea" in both designs (CXD9618? in PS2 and CXD9799GP in PStwo) as it sits closest to the EE.
 
Haha you got me there I guess!

What I actually meant though they've never developed (and built) one on their own. They didn't own cyrix they just provided 3rd party manufacturing capability.

Peace.

True, but they also sold Cyrix CPUs under the IBM brand, including the IBM 6x86 and IBM Blue Lightning.
 
The PS3 I'm not worried about; Cell should be able to fully emulate PS1 by brute force alone. IIRC, the PSP uses a MIPS CPU, so emulation would be pretty easy there as well. In the PS2, though, if you replace the MIPS-based IOP with a PowerPC-based IOP, it will have to run significantly faster just to do the emulation. Again, if cost is the motive, I just can't see how replacing a 40 MHz MIPS-based IOP with a 400 MHz PowerPC-based IOP would be cost-effective.

The reason whatever they put in there would be cheaper is that the IOP is LSI (I believe) intellectual property, so they have to pay royalties on it and don't control things like process shrinks and integration with other parts.

The PSX already had an emulated IOP so I don't see why this is a surprise.
 
The reason whatever they put in there would be cheaper is that the IOP is LSI (I believe) intellectual property, so they have to pay royalties on it and don't control things like process shrinks and integration with other parts.

Oh, I know that the original PS1 CPU cum PS2 IOP was licensed from LSI. I totally understand Sony replacing it with something that they have more control over, but the PowerPC 440 just doesn't seem to fit that bill. All Sony would have accomplished is switching suppliers from LSI to IBM.

The PSX already had an emulated IOP so I don't see why this is a surprise.

I didn't know that. Do you know what actual hardware was used to emulate the IOP in the PSX? Was it a MIPS-based clone of the IOP, or was it some other architecture (PowerPC?) performing emulation?
 
Wasn't there a topic or article long, long ago, that Sony had purchased some sort of useage rights for the 440? There was some speculation that it might be used in Cell in some manner? Of course, this ended up not being the case, but maybe the use suggested in this topic is what they did have in mind for it?
 
Wasn't there a topic or article long, long ago, that Sony had purchased some sort of useage rights for the 440? There was some speculation that it might be used in Cell in some manner? Of course, this ended up not being the case, but maybe the use suggested in this topic is what they did have in mind for it?

Wikipedia claims (without citing a source) that Sony built a set-top box using the PowerPC 405 core, but that's all I could find on Google.

The simplest explanation might just be that LSI no longer licenses or maintains the R3000A as part of their CoreWare design library on the latest processes. I can't find any reference to the R3000A it on their website, though they do still offer MIPS cores. Sony may have decided to redesign the IOP rather than continue to produce it on an older process. Still, a switch of ISAs seems odd.

ERP, any more info about the PSX's emulated IOP?
 
I think honestly that the move to Power just reflects a path towards greater cooperation with IBM that Kutaragi put the company on when he was still in charge of the Semi division; one does not abandon MIPS-based designs lightly afterall, and the original source article via Reeves indicates that POWER chips were going into Bravias and DVD players.

Electronic News: So does that mean the current Playstation 2 systems have a Cell processor?

Reeves: No, they have a 440 Power processor. It’s a 130-nanometer, single-core ASIC chip. It’s the same technology as if you buy a Sony DVD or a Sony Bravia TV. Sony is replacing all the Mips design points with Power design points.

Between the Cell collaboration, Sony investment in IBM R&D and fab capacity, and a wholesale move to Power from one of the largest CE players in the game, the licensing terms Sony enjoys on these IC's is probably extremely favorable.
 
Between the Cell collaboration, Sony investment in IBM R&D and fab capacity, and a wholesale move to Power from one of the largest CE players in the game, the licensing terms Sony enjoys on these IC's is probably extremely favorable.

There could be other reasons as well. If Sony are using the 440 in other products it'll compatible with their existing silicon process, that means they can reuse the IP without having to port it.
Even if they did need to port it, the 440 is synthesisable - it can be easily ported to other processes. This may not be the case with the MIPs part so switching to the 440 could save them a load of hassle.

It's also compatible with the toolsets they are using for other stuff, that alone could be a viable reason for changing.

The "emulation" may be JIT based so it'll not need to be that much faster than the MIPs core.

Also, the 440 has been replaced by the 460 series which is smaller and uses less power (it's specs are similar to the high end ARM parts).
 
I think honestly that the move to Power just reflects a path towards greater cooperation with IBM that Kutaragi put the company on when he was still in charge of the Semi division; one does not abandon MIPS-based designs lightly afterall, and the original source article via Reeves indicates that POWER chips were going into Bravias and DVD players.
Electronic News: So does that mean the current Playstation 2 systems have a Cell processor?

Reeves: No, they have a 440 Power processor. It’s a 130-nanometer, single-core ASIC chip. It’s the same technology as if you buy a Sony DVD or a Sony Bravia TV. Sony is replacing all the Mips design points with Power design points.
Isn't the PSP based on a fairly fast (compared to the IOP anyway) MIPS R4000 CPU?
That sure would be hard to emulate with a dinky sub-GHz PowerPC. It would appear much saner to me that they keep the MIPS core, and label the Reeves quote as a slight exaggeration.
 
That 440 is by no means a high-end part. It is cheap, so that might be the reason (as well as the SW-issues mentioned above).
 
Back
Top