More On IBM's Next-Gen Xbox Chipset Win

Pieroxy writes "EE Times reports further details on Microsoft's use of IBM chips in its next generation Xbox game and consumer electronics devices, dealing a blow to Intel and providing a much needed boost for IBM's lossmaking chip business." An analyst claims that "IBM is likely to modify its most advanced G5 PowerPC silicon, which is being used in Apple Computer's fastest Macintosh desktops, for the embedded market, reducing the cache and cutting power consumption", and further comments: "This is likely to heat things up at Intel, but it is competition that is healthy for the industry. It's ironic that IBM, with its roots in the computer industry, doesn't supply the processors for the main portion of the personal computer industry. Intel does." We covered IBM's initial announcement as a section-specific story earlier today.

Comments

A question

[FreeBSD Goddess]

Forgive me if I'm being moronic about this, but if it's based off the G5, it has a completely different instruction set. Does this mean that the modified Windows 2000 kernel used with the current XBox will be upgraded and ported to G5, or that we might see a completely different and new kernel?

Re:A question

[grotgrot]

In fact Microsoft developers on the OS side do a lot of their development on non-Intel platforms such as PowerPC. The reason is that it ensures the code is cleaner and architecture neutral. Later on it is "ported" to Intel.

Re:A question

[laird]

Actually, NT 3.x (and, possibly 4.0) ran on many CPU's including the PowerPC. If you remember back when the PPC was being promoted as the CPU of the Common Hardware Reference Platform, you could run MacOS, Linux, and NT on the same hardware. Anyway, I would hope that MS has retained that portability in the core of the OS (the same way Apple makes sure that MacOS X still runs on the x86) in order to keep its options open.

So it'd be easy for new "Xbox 2" games to run natively on PPC. That being said, they'd still need an x86 emulator to run the Xbox games. I bet a 2 GHz G5 could emulate an Xbox pretty well...

Re:A question

[ADRA]

If for instance the code was somehow embedded inside the CPU core, then the following holds:

- The Signature that allows the boot loader to run is stored inside the CPU, but it can only run a boot loader that has been signed by microsoft beforehand. Since the key can't be changed, all it guarantees is that the boot loader is secure. Now, you have the kernel, which one would assume to be upgradable. If the kernel is not upgradable, a major bug / exploit found in the kernel could be exploited and never get fixed until Xbox 3.

If the kernel isn't upgradable, you need to start hackimng the userspace apps to kill the lock-out.

If they make the kernel upgradable, they have to be able to sign the kernel from the boot loader. So, you either use the embedded CPU ID to sign the kernel, or they will have the key somewhere else that normal users won't be able to reach. Of course having the decode key only gives you half the prize. You need to find the encoding key in order to encrypt software that actually runs on the Xbox. Once software has been signed, there is no stopping anyone from running anything as long as that piece of software has kernel-level access to the hardware. If you rely on a userspace exploit, you can only perform a limited set of operations, like running arbitrary commands. Get the kernel, and you could burn in a new kernel in its place tricking everything on the higher levels as to who you are.

So, how do we hack the Xbox 2 if we can't run arbitrary code on the xbox to begin with?
Any external access to the system is exploitable through its IO subssytems. The kernel could have a buffer overrun, the game running could leak precious data, who knows. The more functionality left open, the more potential expoits there are to take advantage of.

I haven't bridged the topic of hardware mods which does get more sticky if the encryption is processes in the CPU. I'd have to see when and how the CPU decrypts a signed working set and how the decoded code is run after the decrypt.

Bodes well for Apple too!

[MarcQuadra]

Who thinks that IBM is going to turn around and sell these things to Aplle as the CPU for the G5 laptop?

This bodes VERY well for IBM, Apple, Microsoft, and anyone else interested in low-power-draw PowerPC systems. It sucks for Motorola, but they lost my favor years ago, and they really charge ludicrous prices for their wares.

Also, Could IBM be developing their G3+AltiVec chip for this? It seems to me that if the G3 series was dead IBM would stop working on it, but there are 750GX CPUs due soon (just a 750FX with 1MB on-die cache), and rumors of a G3 with SIMD coming down the pipe. It seems to me that if IBM bastardized some of the SIMD logic from the 970 and strapped it to the 750 they'd have a pretty decent low-power SIMD chip that Apple could market as a 'G3', 'G4' or a 'G5.'

Maybe I'm just a hopeless romantic, but the G3 was the CPU best-suited for what I do, and I hope it doesn't disappear. I have little use for SIMD, and I really appreciate running a CPU without a fan strapped to it, it's just so... elegant.

Backwards Compatible?

[realdpk]

It seems that it would be prudent for Microsoft to announce something about backwards compatibility if they're going to make such a dramatic platform change. That's one of the greatest things for early adopters of the PS2 - they could still play their PS1 games on 'em.

Erm...

[athakur999]

and then adding a pirated disk drive

This is just pure ignorance. Apparently going to Best Buy and buying a hard drive is now "piracy".

PowerPC was *supposed* to become a commodity chip

[KojakBang]

IBM was going to come out with personal powerpc systems which would even have a common motherboard reference design with Macs. The volume of production would drive down costs dramatically. But IBM didn't and Apple basically got screwed on that deal.

Interestly enough, the reason IBM canned the personal powerpc systems was that OS2 for PPC completely blew its schedule several times over. IBM had a personal AIX edition for PPC ready but chose not to go with that. The reason. Unix would never make it as a mainstream operating system for PCs.

Re:What about today's Xbox?

[chill]

Are they not going to have backward compatibility? That seems like a big mistake in the game console market to me.

By the time the X2 comes out, Celeron 733s will probably be cheap enough to put on an add-on card and sell for $49.

That is the price of 1 game. Lot of people with existing X-Box titles would buy that.

Re:Won't MS have to rewrite everything?

[stratjakt]

Yah, MS and IBM are stupid, you better call 'em and set them straight!

Some things to consider:

1) NT ran on PowerPC long ago, porting it wont be hard.

2) Wasn't IBM working on a new mega-chip that can run x86 and PowerPC codes side by side?

3) Perhaps making the Xbox so close to a PC blew up in MSFTs face, and they don't plan to do it again. You can turn your noses at "security through obscurity" all you want, but the Xbox, once initiall cracked, has been blown wide open - it's a no brainer to port PC emulators and apps to it. The GCN or PS2, however, have taken much longer, and are still an obfuscated mess to try and develop custom stuff for.

interesting move for xbox..

[Thrakkerzog]

Wasn't one of the "Benefits" for people making games for xbox the ability to code just like they did for pc games?

perhaps they are trying to lure developers away from GCN by offering a similar cpu architecture?

So is it a fair assumption...

[NanoGator]

... that backwards compatibility isn't part of Microsoft's game here? I suppose they could do like the PS2 did and use the original Intel processor to act as a controller or something, but somehow I doubt that'd be cost effective enough.

Emulator? Eck I hope not. Well.. maybe that wouldn't be so bad. Maybe they could do a combination emulator and wrapper. The emulator would be for the processor instructions, and the wrapper would be to send the graphics commands to the new GPU. Presumably, the difference between the two GPUs wouldn't be big enough as to prevent that from working.

Eh I dunno. Personally, I'm hoping Microsoft does something a little more interesting than just throwing next-gen hardware into a box as an upgrade. Pushing polygons around is nice, but I really like how small and cheap my GameCube is. *Hint hint*

intel/amd/ppc

[Sebastopol]

Intel went after Xbox originally so that AMD wouldn't get the win, but Intel still took it in the shorts, or so it is claimed, by offering a nutrageously low price to outbid AMD. Probably even at a loss given the timeframe.

Intel doesn't give a crap about PPC, as it isn't even a remote threat, what at 4% of the market. Intel could have EASILY played the same power-play and had another design win, but at the cost of lower ASPs for a niche market (compared to its $20b a year market, xbox isn't worth it).

just my $0.02.

X86 emulation under PowerPC

[ahchem]

This is actually a very interesting move on MS's part. Not too long ago they purchased the premier x86 on PPC emulator. Now they are going with IBM's PPC as there next Xbox. Maybe they are ready to have a win-tel divorce and declare their independence from Intel, or maybe just shake the relationship up a bit.

History Repeats Itself

[ewhac]

Not long after the 3DO Interactive Multiplayer came out, we started designing the next-generation platform. The machine was to be PowerPC-based with true 3D rendering capabilities (triangle engine, MIP-mapping, perspective-correct textures, 32-bit rendering, etc.).

The CPU was supplied by IBM. What we ended up with was the PowerPC 602, which was essentially a 603 (?) with a smaller cache and single-precision floating point operations that executed in a single cycle, which were essential for 3D gaming. The part ran at 66MHz.

It was a really nice machine. Sadly, it essentially died on the vine, as Matsushita chose not to exploit its gaming potential, relegating it instead to "kiosk" activities.

IBM also manufactured the triangle engine. It was a five layer chip -- at that time, a rather sophisticated process -- occupying 144 square millimetres.

Schwab

My Prediction

[Goyuix]

Just as with the original XBox, there were plans to use AMD chips, at least that is what most of the speculation on the internet said, up until they announced the specs and low and behold Intel had "won" the contract instead.

I could very easily see this deal with IBM as a backdoor sneaky tatic to get nVidia hardware under the hood again, especially with their cozy arrangements as of late. For that matter, AMD and IBM are also in bed so it wouldn't be surprising at all the see ATI booted and AMD/NV offering becoming the real guts of the console. IBM would be the fab for the chips and assembly for the mainboard... heck they might even just roll the whole thing and take a cool percentage.

Off the wall? Perhaps. But I think this xb0x0r war is far from over. There are no published specs and these talks of industry deals are by no means what the final product will necessarily be.

Re:What about today's Xbox?

[AuMatar]

I do. I own about 6 PS1 games to every PS2 game. Never owned a PS1, I was in college when it came out, had no money. So being able to do both was great, I could play all the games I missed.

In addition, if I had had a PS1, I'd still find it a great feature since it would let me get rid of a piece of hardware and remove clutter.

Re:Hope it's...

[AtaruMoroboshi]

Frankly, what are you talking about? Performance issues?

The Gamecube is way more powerful than the PS2, most importantly with none of the jaggie aliasing problems nor the atrocious load times..

Yes, most of the time those 2 issues are no longer a big deal on the PS2 but it took a lot of developer cleverness to do it.

Re:One wrinkle using the VPC code from Connectix..

[Lars T.]

I have yet to see anything substantiating that claim. Heck, there never was a "virtual little endian mode". When somebody (esp. Microsoft) invents new names for something to prove they don't exist (anymore), I feel uneasy.

3.1.4 PowerPC Byte Ordering

The PowerPC architecture supports both big and little-endian byte ordering. The default byte ordering is bigendian. However, the code sequence used to switch from big to little-endian mode may differ among processors.

The PowerPC architecture defines two bits in the MSR for specifying byte ordering--LE (little-endian mode) and ILE (exception little-endian mode). The LE bit specifies the endian mode in which the processor is currently operating and ILE specifies the mode to be used when an exception handler is invoked. That is, when an exception occurs, the ILE bit (as set for the interrupted process) is copied into MSR[LE] to select the endian mode for the context established by the exception. For both bits, a value of 0 specifies big-endian mode and a value of 1 specifies little-endian mode.

The PowerPC architecture also provides load and store instructions that reverse byte ordering. These instructions have the effect of loading and storing data in the endian mode opposite from that which the processor is operating. See Section 4.2.3.4 Integer Load and Store with Byte-Reverse Instructions for more information on these instructions.

I have yet to see anything by IBM saying that there were changes in that behaviour in the 970.