Sony's R&D- Linux and PS3

Yousef writes "GameSpy has an interesting article about a presentation given by Sony's head of R&D for Entertainment. It appears that there are some very interesting things in store for the PS3, plus a complete Red Hat Linux installation for the PS2 and many other toys too. An interesting read."

Comments

PS3

[Animats]

It mentions RedHat. It doesn't mention the PS3. It mentions the GSCube, a cluster of PS2 machines set up as a technology demo. That was shown a year ago; it's not a future product.

The PS3 is PowerPC-based, and won't have the wierd vector units of the PS2. Those are generally conceded to have been a mistake. They're hard to program, and required considerable tool development. The competitive effect was that for the first year, PS2 games sucked. (The Xbox is more vanilla; it's basically a PC running Win2K with a GeForce 3, which simplifies development. I know people who had to port a physics engine to the PS2 vector units. Not fun.)

Sony, Toshiba & IBM are creating an OS

[HanzoSan]

I'm betting its a Linux based OS, for the PS3 and other appliances.

IBM Sony & Toshiba

I think consoles are the key to Linux being successful on the desktop, because Consoles are the desktop of tomorrow.

Sony IBM & Toshiba Cell Chip Technology Info

[HanzoSan]

Sony Cell technology
NEWS
IBM, Sony, and Toshiba announced a partnership today in Tokyo to develop new, faster, smaller chips code-named Cell. Over the next 5 years the companies will spend US$400 million to break the 0.10 micron barrier. Cell chips will be targeted for use in high-speed Internet access and network-based computing.

Sony Computer Entertainment (the gaming folks) is the Sony division involved in the partnership, and it already has working arrangements with Toshiba--the two companies formed a joint venture to design and produce the PlayStation 2's chip. Adding IBM to the mix helps all three companies reduce development costs ... and Sony also gets to license IBM's 0.10 micron processing technology, which will probably be used in the PlayStation 3 (PS2 chips are currently at 0.18 and 0.25 microns, though Sony has announced a move to 0.13 micron technology) and other upcoming devices from Sony Computer Entertainment. Development work on the new Cell chips will take place in an Austin, Texas IBM facility and will eventually be produced at a new IBM fab in East Fishkill, New York, slated for completion next year.

IBM will also announce today that it is joining the Extreme Ultra Violet Consortium, another group working to shrink micron processes. Industry watchers think IBM's move may help boost the EUV technology's chances of success.

Interesting facts about official Sony Linux

[Arker]

Here's a good link if you're interested in this - http://www.execpc.com/~halkun/PS2/

It sounds more like it emulates linux in a VM rather than actually loading it as the actual machine OS. This allows them to use all kinds of proprietary code without violating the GPL, and to build in all sorts of copy and "ip" "protection."

Fair use quote:

This is how the Runtime Environment (RTE) works. In order to get Linux running
on your PS2, you must boot the system using the PS2 Linux DVD.
During boot, after all the copy-protection stuff is taken care of, the system
lays down the Runtime Environment. This is basically a layer that hides access
to the SPU2 (Sound Processing Unit), the input/output processor, the hard drive,
the CD/DVD-ROM system, the controllers, memory cards, USB, i.Link and other
peripherals. The RTE does supply hardware looking hooks, an educated guess
being faux-memory address and registers. Then the Linux kernel is loaded on top
of this. There are Linux device drivers that accesses the Runtime Environment
that are open source, but it's just a device driver calling in all actuality,
another device driver that's closed.

What you can and can't do with the system is limited.

You have no ability to read a normal PSX or PS2 memory card directly. For
example you can't open a Final Fantasy X save, edit how much cash you have,
and save it again. Through the RTE you can format a whole memory card(!)
and mount it like an 8 meg hard drive, but that card would be worthless
for saving normal PS2 games. Once you put the Linux formatted card
without Linux running (i.e. you are in the browser) it asks to reformat
the card.

The RTE also not allow audio CDs to be identified.
It also can tell if you have put in a CD-R or not (it can see a wobble track,
which all CD-Rs have ) and likewise not allow the disk to be seen. A PS2 can
read CD-Rs fine, the RTE is just doing copy protection first to make sure you
can't. You will also have no access to the CSS portion of the MPEG decoder,
but you can decode raw MPEG-4. Direct access to the Dolby subsystem is also
denied. Anything dealing with region locks are also restricted.

The first DVD (The boot disk) has a Linux boot loader and the RTE on it.
This disk is not allowed to be copied. It also has the manuals on it too,
which I'm sure are also copyrighted and not allowed to be publicly distributed.
The Linux kernel is on the second disk and also on the hard drive after it's
installed.
In order to use a monitor, you must one that is "Sync on Green". This means
that the refresh rate is only in the green channel. The monitor must use that
sync pulse to sync red and blue channels so they all get painted in the screen
at the same time. The reason why you have to use that is because a PS2 can turn
it's sync on green ability on and off. If you try and use the monitor
adapter for playing PS2 games or watching DVDs, sync on green will be turned
off and only the green channel will show up. Direct video output defeats Macrovision.
Sony doesn't want you making copies of DVDs to tape. Keep in mind that your
network adapter is going to have a MAC address that Sony, no doubt, knows.
Also removing the PS2 hard drive and attempting to mount in a PC will also
likely not work and possibly damage the drive.

Re:bio-playstation

[omega9]

Back of box:
For external use only.
If Playstation 6 contacts skin, flush with water for 15 minutes.
Do not stare directly at Playstation 6.
If Playstation 6 breaks open, run and take cover.

Interesting Concepts, but,,,

[mgrochmal]

The developer responded that the realistic real-time rendering they wanted to do would require 18,000 times more processing power than what was available with the PSOne.

It's difficult to imagine what sort of rendering in a game would require this much power behind it. Then again, I remember when a 1 Gigabyte server was huge. I also wonder that, if the plan for distributed processing is to be successful with multiple PlayStation consoles, how many users would be willing to leave their PlayStation(#)s running all the time for other people to use. On top of that, how would Sony propose to network the consoles togeher? The networking prospects for the current-generation PS2s have occasional problems running the software it already has for networked gaming beyond a jury-rigged LAN.

In regards to why would Sony want Linux on their PSx consoles... An audience of avid PS2 users experimenting with a robust TCP/IP (Internet communication) protocol. Fans will also receive complete documentation with the kit, which includes all the technical details of the PS2 hardware. Normally this info is only available to game developers. Again, Linux for PSx isn't going to rake in millions for Sony, but there may be different benefits for hobbyist developers. It won't be for everyone, but if you're into trying new things, then feel free.

The "Gestural Interface" seems to be the most interesting out of what the article offered. It might be a step closer to a virtual reality interface without needing an expensive headset. It might change gaming for the console. Then again, it depends on the games that use said hardware. It'll be interesting to see how many of these ideas actually come to fruition, and how many of them become Vaporware.

FUD.

[oGMo]

The PS2 uses a CPU that runs at 300 MHz. Would you scoff at an ad in the paper that is selling a "very fast" x86 machine at 300 HMz these days? I would.

Yes, as well you should. However, this is a straw man argument and has absolutely no bearing on your following statement, since a 300MHz x86 is hardly the same as a 300MHz emotion engine.

Console CPUs at lower speeds than PCs used to have higher performance because they were customised for gaming.

Repeate after me: All MHz are not the same. I have a 200MHz StrongARM in my Gameboy Advance. The SNES had a 3MHz processor. The GBA is a bit more powerful than the SNES, but not by the same delta as a 4MHz 8086 and a 200MHz Pentium! All MHz are not the same! In fact, MHz are about as useful for gauging performance these days as BogoMIPS, which is to say not at all.

That's becoming less and less true as newer consoles come out that are closer relatives to PCs.

Gross overgeneralization. The XBOX is the only thing that's a repackaged PC. The PS2, which currently dominates the market by a huge margin, is quite far from a PC architecturally. The GameCube, while using something resembling a PPC, is otherwise architecturally quite remote from a PC. The XBOX may take all its RAM from the same pool (which as has been discussed isn't really a good thing), but that's not much different from what we do now (I've seen cheap SiS motherboards with onboard video that use system RAM for video RAM. Big deal, XBOX.)

Now the hardware to handle graphics runs quite a bit faster. And we can expect the PS3 CPU to be "very fast".

Again, "very fast" is completely relative. The XBOX may have a 700MHz Celery (which is kinda slow anyway, but an OK general-purpose CPU), but take away its graphics accelerator and you'll be lucky to rival a SNES or PSX. The PS2's core is rather tied together, but even though it's 300MHz, it can push a decent amount of polys. Lesson for today: All MHz are not the same!

Oh, and to be somewhat on topic, anything Sony says about the PS3 is likely complete hype at this point. I like Sony (well, no I don't, but I like the PS2 ;-)), but I know just as well as the next guy how the marketting deal works.

But is a 300 MHz PS2 processor really all that amazing as compared to a 200 MHz Dreamcast processor?

All MHz are not the same! All MHz are not the same! All MHz are not the same!

Re:FUD.

[stonecypher]

> I have a 200MHz StrongARM in my Gameboy
> Advance. The SNES had a 3MHz processor. The GBA
> is a bit more powerful than the SNES, but not
> by the same delta as a 4MHz 8086 and a 200MHz
> Pentium!

Whereas the sentiment of the note is correct, I suggest you follow through with some fact checking. The processor in the GameBoy Advance is an ARM7 TDMI, running at 16.78 mHz.

Even so, it should be noted that the Emotion Engine's IPC and so forth (as IPC and mHz aren't everything either, no matter what AMD's webpage told you) aren't as impressive as everyone here seems to think they are.

What's important about the PS2 with regards to speed are a few things: bandwidth (it's just sickening), two reconfigurable vector computation engines (VU0 and VU1 - seems like they might be great for a lot of beowulfish stuff, but then, I have no idea how those clusters really run, so take that with a grain of salt), and the following slap in the face: because all of the comparsions you read were using the DreamCast as the watermark at the time.

The machine, if used aggressively, can still tangle with the GameCube. How it fares against the XBox is the subject of debate; my personal belief is that with careful use it could surpass the XBox, but many people disagree (some feel the higher instruction processing rate is the deciding factor, which would give the issue to the XBox; others feel that the cache problems of an instruction-oriented architecture outweigh the benefits when considered against the bandwidth oriented architecture, which would give it to the PS2, at the cost of being very difficult to write to.)

Moreover, there are facets of the XBox like realtime Dolby 5.1 compression of generated audio which the PS2 has to dedicate a VU to hope to match (this is a significant chunk of the PS2's processing power, making this a Bad Thing).

> [The GameCube ] is otherwise architecturally
> quite remote from a PC.

Not really. Certainly not as far as the PS2, and arguably not even as far as a PS1. Whereas the bus layout and memory maps are completely different, you'll find that things like a normal opcode list make a bigger difference in the long run anyway - I mean, really, nobody in the industry uses magic numbers; it's all a macro called VRAM anyway. On the other hand, you really do use assembly, and quite often.

> Again, "very fast" is completely relative.

Not really. The SNES was "really fast" when it came out, and now it's dog slow. It's relative to what the consumer expects. The judgement made solely on experience, while not being hindered by expectations regarding numbers, operating system concerns, et cetera, is a better measure (in my mind) of "really fast" than anything else.

I mean, we've had Crays which pound your box (whatever it is) into the ground for probably 20 years. Is your box still really fast? Yes: Quake gets three digit frame rates, and even IE doesn't lag.

Now quit doting so much on numbers for subjective judgements. "Fast" doesn't have a number attatched to it in the dictionary, does it?

> The XBOX may have a 700MHz Celery (which is
> kinda slow anyway, but an OK general-purpose
> CPU),

For someone talking about the relativity of speed, you're certainly not thinking about it much. Consider the previous generations of chip, and the current. A celery 700 is more than plenty fast.

> but take away its graphics accelerator and
> you'll be lucky to rival a SNES or PSX.

I don't honestly believe you have any idea what you're talking about, no offense. My 486/33 was able to emulate the SNES in DOS mode, which means it was eating the CPU, graphics cards, sound cards, and so forth all on its lonesome.

If you can't tell the difference between a 486/33 and a Celery 700, you're not qualified to be talking about relative machine speeds. No offense. And, hey: go have alook at Bleem!'s requirements.

> The PS2's core is rather tied together, but
> even though it's 300MHz, it can push a decent
> amount of polys.

The Emotion Engine doesn't push any polys at all, ever. Did you do any research before claiming to know what was going on under the hood? Vertex pushing is almost always solely the province of the Graphics Synthesizer, and sometimes the VU units.

> Oh, and to be somewhat on topic, anything Sony
> says about the PS3 is likely complete hype at
> this point.

It's been in development for almost two years. I bet they have a general idea of how it's going to work.

> I know just as well as the next guy how the
> market[t]ing deal works.

You sure?

(sighs)