PlayStation 4 (codename Orbis) technical hardware investigation (news and rumours)

[onQ]

That's an audio decompressor and mixer, which isn't a DSP.

They talk about DSP (as a workload, not as a hardware component - Digital Signal Processing and not Digital Signal Processor) and then how it fits AMD's APU. Surely if PS4 had a DSP, they wouldn't be talking about audio processing on an HSA architecture?

The DSP can be part of the SoC.

 

[Shifty Geezer]

Well, that's a matter of semantics. And audio decoder/encoder block could be a DSP, as could a mix (you are processing signals, after all). However, what PS4 has not got is a programmable DSP for doing audio signal processing a la XB1's. PS4's audio hardware has never been referred to by Sony as a DSP. Ergo, the functional requirements of the audio hardware on PS4 aren't going to need a substantial processor.

 

[Shifty Geezer]

The DSP can be part of the SoC.

Sony doesn't have a DSP in their APU, or they'd have said, "we have an audio DSP that does all sorts of audio processing," instead of, "we have an audio block that does decompression and mixing." Sony have made that distinction for a reason.

I don't ever recall any of these over-optimistic interpretations of hardware (AI processors, physics processors, etc.) that crop up each generation as ever coming true. We always get exactly what the companies tell us. There's zero reason to stretch interpretation to accomodate some new, significant hardware instead of just taking it at face value.

 

[Globalisateur]

On the PS3 there is a chip similar to the one next to the southbridge on PS4, it is the syscon, ARM chip that manage power modes. So I suspect it is the low/normal power mode (famous) ARM chip on PS4.

But the PS4 would need a flash memory also so either it is at the other side (we haven't seen the back side of the motherboard enough yet) and the ARM power manager is next to the southbridge or it is the contrary.

About the audio hardware: Sony never talked about DSP but dedicated hardware, and on leaks: audio custom block, access mediated by OS, not programmable by dev, all leading (IMO) to the assumption that Sony dedicated one OS core (or some percentage) to the audio encoding/decoding and scheduler. It would be the easier solution for them and that would explain why Killzone first demo reveal wasn't using any audio hardware acceleration and we know the current build has it.

Would Sony engineers have designed the first dev-kits without any audio hardware and then suddenly: "WTF guys! Guerrilla just told us we forgot audio hardware, let's just add another chip on the motherboard! that's easy!" or "I am going to call the AMD guys and ask them to add audio chip on the APU" or "let's just add an audio block on the customized southbridge, so trivial".

It is just easier (and cheaper) to update the OS to dedicate some percentage of one OS core to the task.

 

[Brad Grenz]

Well, that's a matter of semantics. And audio decoder/encoder block could be a DSP, as could a mix (you are processing signals, after all). However, what PS4 has not got is a programmable DSP for doing audio signal processing a la XB1's. PS4's audio hardware has never been referred to by Sony as a DSP. Ergo, the functional requirements of the audio hardware on PS4 aren't going to need a substantial processor.

Well, when the DSP was brought up no one implied it was programmable, and point of fact, the Xbox One doesn't expose any of its audio DSPs as programmable to developers either (the audio effects it supports are all fixed function), so I'm not sure what the disagreement actually is here, or why you think it's important to split hairs.

 

[MrFox]

There are some additional questions, but at this point they are trending towards wondering about minor things.


The chip mounted on four sides with wires could be a microcontroller for miscellaneous peripherals.
That kind of mounting doesn't promise a high-performance connection to anything, so I suspect it's not doing anything radical.
That particular mounting, surface markings, and shape are consistent with a number of microcontrollers, although their role doesn't require that they be all that distinctive.

I'm not absolutely certain, though.
When dealing with blurry shots in a system I know little about, my overriding assumption whenever I feel something is wrong is that it's probably me.

I was wondering whether there was a NAND module of some kind somewhere. I figured that would help expedite any drive swapping and keep the OS somewhere more difficult to access, but I'm not sure if there is that kind of storage available.

A micro-controller would have at least a few hundred KB of prom or nand on-die, maybe it would be enough to bootstrap the I/O chip in a secure fashion? Disk swappin requires a 900MB file to reinstall the firmware, so the board firmware could be just the minimum for reinstalling the OS.

Maybe some ARM M3 variant, which is popular in LQFP-100.

EDIT: There's a Macronix chip connected to the SB, 16-sop means it's a serial flash. Maybe their secure flash product?
http://www.macronix.com/QuickPlace/...58F25146E9B6237482574170009A62A/?OpenDocument

 

[patsu]

On the PS3 there is a chip similar to the one next to the southbridge on PS4, it is the syscon, ARM chip that manage power modes. So I suspect it is the low/normal power mode (famous) ARM chip on PS4.

But the PS4 would need a flash memory also so either it is at the other side (we haven't seen the back side of the motherboard enough yet) and the ARM power manager is next to the southbridge or it is the contrary.

About the audio hardware: Sony never talked about DSP but dedicated hardware, and on leaks: audio custom block, access mediated by OS, not programmable by dev, all leading (IMO) to the assumption that Sony dedicated one OS core (or some percentage) to the audio encoding/decoding and scheduler. It would be the easier solution for them and that would explain why Killzone first demo reveal wasn't using any audio hardware acceleration and we know the current build has it.

Would Sony engineers have designed the first dev-kits without any audio hardware and then suddenly: "WTF guys! Guerrilla just told us we forgot audio hardware, let's just add another chip on the motherboard! that's easy!" or "I am going to call the AMD guys and ask them to add audio chip on the APU" or "let's just add an audio block on the customized southbridge, so trivial".

It is just easier (and cheaper) to update the OS to dedicate some percentage of one OS core to the task.

Vita developers also do not have direct access to the programmable audio hardware (made by Wolfson Microelectronics). Perhaps Sony want absolute control of the audio paths so that they can manage sound from different apps and system software centrally.

I suppose they may be able to evolve the audio h/w further later.

 

[scf]

Some bits about remote play implementation in AC4 http://www.polygon.com/2013/11/11/5091294/assassins-creed-4-black-flag-ps4-ps-vita-remote-play

 

[patsu]

Some bits about remote play implementation in AC4 http://www.polygon.com/2013/11/11/5091294/assassins-creed-4-black-flag-ps4-ps-vita-remote-play

Article reiterates that RP support is done at the h/w level, which makes it odd that Blizzard tries to block it.

 

[scf]

Article reiterates that RP support is done at the h/w level, which makes it odd that Blizzard tries to block it.

Well... they don't http://www.gamespot.com/articles/diablo-3-reaper-of-souls-on-ps4-supports-remote-play/1100-6416109/

 

[taisui]

yea, the red line of pure hotness.

 

[patsu]

Well... they don't http://www.gamespot.com/articles/diablo-3-reaper-of-souls-on-ps4-supports-remote-play/1100-6416109/

Cool, so the developers really don't have to worry about RemotePlay at all, except for button mapping.

 

[MrFox]

But the PS4 would need a flash memory also so either it is at the other side (we haven't seen the back side of the motherboard enough yet)

Macronix flash memory chip (the MXIC logo), lower right from the south bridge.

 

[scf]

Some positive feedback about RP

 

[patsu]

Phhhfppt, should be spectating in my book. Already knew RemotePlay would work well over LAN. Show me universal spectating.

 

[onQ]

Macronix flash memory chip (the MXIC logo), lower right from the south bridge.

Good eye!

 

[Grall]

You don't need to attach photos or powerpoint slides to every other of your posts. It just makes you look obsessive-compulsive.

 

[onQ]

You don't need to attach photos or powerpoint slides to every other of your posts. It just makes you look obsessive-compulsive.

Just making it easier for other people to see.

 

[morphinapg]

Personally, I still strongly suspect it's a video encoder.
- the video encoder for the PS4 needs to be 'very low latency' for remote-play.
- ideally the PS4 would produce 2 streams - one at 1080p for the record feature, and one scaled for the Vita's screen.

Not long to wait until we find out .

I'm pretty sure the record feature only records video at 720p. Screenshots are 1080p, but not video.

 

[MrFox]

I'm now 99% sure that the the sata bridge for the HDD is the Fujitsu MB86C311B.

Unless I'm missing something (probable), it should be able to saturate the sata2 port close to 300MB/s. It supports UASP on the client side so I see no reason it can't reach close to 300MB/s there either. There's been plenty of tests with SSD used on such bridges and they can definitely do 270MB/s. Whatever the hardware bottleneck is, I guess I'm still looking for it.

That would require some effort on the driver for UASP, and they probably didn't if the stock drive cannot do more than 100MB/s.