Support NeoGAF

[Panajev2001a]

Each of the cores in the Xbox 2's tri-core CPU is almost for sure not a G5 (latest voice talk about dual-isue cores with a VMX/Altivec enhanced unit each... hey did you really think about three PPC976 running at 3.5 GHz ?), but the fact that the SDKs are based on dual G5s (the early ones were) might be an indication that the ISA of the Xbox 2's PowerPC cores will not be too dissimilar.

[...]

The reason for the incompatibility, according to Microsoft, is that the current version of Virtual PC for the Mac relies on a feature of the PowerPC G3 and G4 processors called "pseudo little-endian mode," which helps boost performance of a Mac when it is trying to emulate a Pentium chip.

"Because the new G5 processor does not support this feature, large portions of the (Virtual PC) for Mac program must be rewritten and carefully tested to work properly on the G5 CPU (central processing unit)," Microsoft said.

[...]

http://www.shahine.com/omar/PermaLink,guid,e706c1bf-87fe-4cd0-aa78-a78ac023cd4a.aspx

 

[HyperionX]

And your point is?

I can see that you think that Xenon is not using a G5 but what's the reasoning behind it?

 

[Ecrofirt]

Wait, why would Microsoft want to use a G5? Wouldn't that be giving money to Apple?

 

[SteveMeister]

Don't you understand? Obviously Microsoft is incapable of altering the Xenon core to run on a big endian system...

 

[emerge]

Don't you understand? Obviously Microsoft is incapable of altering the Xenon core to run on a big endian system...

Now it becomes all clear. MS underestimated the difficulty getting a dedicated OS for Xenon up and running on the big-endian PPC97X cores. Now Virtual PC division is hard at work to provide an extra layer of x86 emulation for the folks at Xenon developement. Genious I tell you, genious.

 

[Blimblim]

Most confusing topic & replies ever ? I think so

 

[Fatghost]

Of course there will be a version of VPC which will work with the G5. But even if MS can get Virtual PC working with G5 in time to make it into the final Xenon hardware, there is still the potential problem/cost of getting Nvidia emulation into the Xenon as well.

 

[Panajev2001a]

Don't you understand? Obviously Microsoft is incapable of altering the Xenon core to run on a big endian system...

What are you saying, Xenon CPUs will be PowerPC based hence B ig Endian, they just probably do not have this special compatibility mode that the G4 and G3 had.

No, but it is extra money (to change VirtualPC for Xenon) that they do not want to spend.

And your point is?

I can see that you think that Xenon is not using a G5 but what's the reasoning behind it?

My point is that Xenon uses IBM PowerPC cores without that special endian mode G3 and G4 had.

VirtualPC can be changed, but it runs slower.

That compatibility mode G3 and G4 had was used to speed up emulation of x86 code which is Little Endian.

Speed and monetary concern (they already prefer to forget backward compatibility because they would have to license the NV2A IP from nVIDIA) are the reasons why they want to skip Xbox 1 BC.

Thsi is just another clue... yet another possible clue.

 

[mr2mike]

"Wait, why would Microsoft want to use a G5? Wouldn't that be giving money to Apple?"

Apple doesn't make the G5 chip, IBM does, and amongst other things other than suppling processors for apple computers, they make wintel PCs

http://www-132.ibm.com/webapp/wcs/s...d=-840&langId=-1&storeId=1&categoryId=2049132

 

[rastex]

...

My point is that Xenon uses IBM PowerPC cores without that special endian mode G3 and G4 had.

VirtualPC can be changed, but it runs slower.

...

Do you know if the speed difference is significant enough such that the extra raw processing power can't make it up?

 

[SteveMeister]

What are you saying, Xenon CPUs will be PowerPC based hence B ig Endian, they just probably do not have this special compatibility mode that the G4 and G3 had.

I was being sarcastic. Endian issues aren't THAT hard to resolve. Binary data files need to be converted, but for code it's just a recompile.

 

[Panajev2001a]

Do you know if the speed difference is significant enough such that the extra raw processing power can't make it up?

Maybe it introduces some nasty timing bugs.

If games go to the low level, intercept system calls, use their own... I do not know... that might be just "yet another reason" why they do not want to go through the hassle (licensing the NV2A IP being one of the most critical reasons).

 

[doncale]

my dream-CPU configuration for Xbox2-Xenon is still a custom POWER5+ MCM with reduced caches.

1 block - the MCM
4 dies
8 cores
16 threads per clock


impossible for Xbox 2? probably

possible for Xbox 3? definitally.

 

[xsarien]

Wait, why would Microsoft want to use a G5? Wouldn't that be giving money to Apple?

Well, for starters, Apple doesn't make the chip. And secondly, Microsoft has a nice chunk of investments in Apple. It's in their best interest to keep Apple alive, their existence is the only [tenuous] argument they can make that they're not a true monopoly.

Anyway, isn't Virtual PC 7 supposed to resolve all of the little-endian (heh, cute) issues? Unless I've heard wrong, VPC7 is G5 ready.

 

[Phoenix]

I was being sarcastic. Endian issues aren't THAT hard to resolve. Binary data files need to be converted, but for code it's just a recompile.

Actually dealing with data files is even easier. Most formats are encoded in such a manner that endian-ness of those files is a non-issue. The world would indeed be a crappy place if websites, for example, had to deal with big and little endian versions of their data

At the CPU level, things are as happy and while PPC processors can handle either mode - they do have a native endian-ness to them.