Slashdot Mirror
Leaked X-Box 2 Specs Include PPC CPU
Jutebox150 writes "According to the MercuryNews.com, the specifications for Microsoft's successor to the Xbox were revealed. The specs for the next Xbox, at least according to this report, are as follows: Three IBM-designed 64-bit microprocessors, the same chips now used in Apple Computer's high-end G5 PowerMac. This will give the new Xbox 'more computing power than most personal computers.' A graphics chip designed by ATI Technologies that will clock in with speeds faster than the upcoming R400. But what I found most surprising is there are no talks about an internal hard drive, rather suggesting that the next Xbox will instead rely on flash memory, and, depending on hardware cost, backwards compatibility could be out of the question."
From Sunday.
The PPC chips don't have the heat problems of Intel's or AMD's product. So you can use smaller and more importantly more silent fans and cooling.
The only drawback is that they trade power/heat benefits for reduced performance - if main issue with PPC's. This makes me wonder why they don't use mobile processors from scratch.
Over 90 years and counting !
The difference between the past story and this one is that the past story was "rumors" whereas this is "leaked"....
Remember, MS has said they can't make XBox 1 profitable. You can bet they are going to try to drive down their hardware costs with XB2 so that they can actually make some money. The harddrive is a big expense that could be dispensed with without too much pain.
Well, considering MS bought Connix assets (i.e. Virtual PC) I'd have to disagree with you. 733 Mhz emulation should be well within the range in a couple years, if not already today (it runs rather spiffily on a G4, let alone a G5)
A large part of this design is probably lower heat/energy on the part of the CPU. While current G5s are still a beast compared to G4s, they're still cooler and use less power than P4/AMD cpus.
I've heard the reasons for not including the hard drive (I'm in the games industry), and they make sense.
1) People don't care. Believe it or not, it's not actually a big deal to MOST people. Yes, there are lots of people that do want it, but they're a small percentage of the population. It doesn't matter what they save their games on as long as it's fast.
2) With the PPC, backwards compatibility is already broken. Not to mention backwards compatibility is a pain in the ass for developers as well. They don't care about it, either. It's just not worth the money in the end to make a system that's backwards compatible unless it's easy. The PS1 is a single chip in the PS2. The Game Boy is pretty primitive, and is also easy to include in a GBA. For the Xbox 2 to be backwards compatible, it would either a) have to be the same architecture again or b) have an Intel 733 in there again that somehow gets used with XBox 1 games. Interestingly, the majority of the population isn't interested in backwards compatibility as a MAJOR feature anyway. It's just another bullet point to them.
3) Hard drives are expensive. The interesting thing about hard drives is that they never get cheaper, just bigger. Microsoft gets murdered with every hard drive they put in the Xbox.
4) They want this to be part of their digital hub thing. Since the Xbox 2 will likely have a network connection, if you want to store things more permanently, I heard mumblings about being able to do it on your PC.
5) The hard drive does a couple other things: generate heat and take up space. Getting the size down is something that they have to do if they want to make it in the all-important Japanese market.
6) Lastly, they don't want Linux running on Xboxes. If you want a PC, they want you to go out and buy a PC with Windows on it. The margins are better there.
I think this new Xbox sounds exciting. I'm not a big fan of the current model, but the new one will be a huge boon to developers and gamers alike. With 3 general purpose CPUs and a unified memory system, you can do things like generate a single tree and have each processor modify the tree in memory slightly before sending it to the GPU. Voila! Instant forest with quickly generated unique trees.
Have you ever seen VirtualPC run on a Mac? I've seen instances where VPC is able to emulate code pretty close to the x86 equivalent speed. Now if we're talking about a multi-way PPC, (tri? dual?) 970 class processor, even if you penalize one of the 1GHz processors 50%, it should be able to handle the 700MHz P3 that's in the XBox.
I found it fishy when Microsoft purchased VirtualPC. Sure, they can create virtual instances of Windows on top of Windows, but that's not very mass market. On the other hand, if they can use the technologies that VPC perfected to make their software basically architecture independant (backwards compatibility on any reasonably equipped processor), then that really gives them a bargaining chip. Of course, the Mac community felt that Microsoft was going to box VPC up in a small piece of pine and we'd never see it again, but that was not cunning enough.
I've heard that the G5 doesn't have VPC running on it because it's missing one instruction that the G4 had, and although I don't know what that is, I imagine that Microsoft can pay IBM enough to put that instruction in for the XBox2 version of the chip. Heck, Microsoft and IBM can work out a way to make custom logic interface with VPC better instead of it being exclusively modifying VPC (within reason of course).
I think, in true Microsoft fashion, we'll see the new VPC changed slightly and then become the foundation of their (gaming) business.
The problem the G5 has with VPC is that unlike the G4 it cannot accept numbers in both big endian and little endian form. The G4 was able to do this and it saved enourmous amounts of work when emulating x86 and its ass-backwards numbers.
If it's not on fire, it's a software problem.
I have no comment on your comment other than "Power PPC"? A little redundant don't you think?
PPC - Power Personal Computer
So they're putting a Power Power Personal Computer chip in it. P^2PC. EAT THAT G5! (:
No sig for you. YOU GET NO SIG!
From the article:
"Three IBM-designed 64-bit microprocessors. The combined power of these chips means the Xbox Next will have more computing power than most personal computers."
I think "3" G5's can emulate a 733 P3 without too much trouble.
Well, three PPC970s at 2Ghz would be 6Ghz - and IBM and Apple have both mentioned that 3GHz PPC970s will be here in mid 2004 - which would be 9GHz.
I know it's not as simple as 2x 2Ghz = 4Ghz, but generally it's a fair indicator of performance.
I'd be surprised if we had "most" home PCs at 9Ghz by the time the Xbox2 ships.
The fact is that Xbox 2 could be backwards compatible using emulation. Microsoft already owns VirtualPC, which allows PowerPC architectures to run Microsoft Windows applications. However, the problem is that, although Microsoft owns the DirectX API and the Windows kernel, it doesn?t own the nVidia chips found in the Xbox and since it is changing it graphics partner in favor of ATI, there are almost no chances of an agreement between the two companies being reached. This is more of a business problem rather than a technical issue.
Don't think of it as a flame---it's more like an argument that does 3d6 fire damage
Depends on how you have them configured, and what you're doing with them. If you look at multiple CPUs just from the threading level (Like MS has been known to do), you're being really inefficient with the processing power available.
First of all, from an addressing perspective, 3 CPUs is a weird number. In almost all multiprocessor systems (Where the processors share a single board) I know of have a number of processors that is a power of 2. This is because if you're going to build an architecture that addresses a certain number of CPUs, you usually have the full compliment of CPUs to make use of the architecture you've provided for them. Whithout exception, CPU addressing on a single board is done in binary. In a few cases you'll have boards that can have a capacity higher than the number of existing CPUs, but even then the number of existing CPUs uses the full range of addresses given by a certain subset of bits (4 CPUs on an 8 CPU board, using 2 of 3 addressing bits). From a hardware multiprocessing perspective, 3 is a ~really~ weird number.
When programming from a threading perspective, it's quite common to have one thread that will have too much to do, while the other threads aren't particularly active (leaving their CPUs wasting cycles). This is very inefficient behavior.
An alternative to this is to synchronize the CPUs and do a single instruction on multiple sets of data (SIMD multi-processing architype). This is especially useful for things like array transformations (common in 3d gaming and graphics applications) where you can do the work in a fraction of the time, with very little additional overhead. The shared memory model also would help facilitate this kind of work in an efficient manner.
Many tried-and-true parallel programming algorithms are also designed for divide-and-conquor apporaches for solving problems. In these cases, more often than not, the algorithms are designed to deal with a number of CPUs that are a power of 2.
Let's take ordering an unorderd list as an example:
Threaded approach: One CPU orders the list, two CPUs are bored.
Synchronized approach: Have three CPUs ordering the same data. I'm not sure how this algorithm would look/work.
Divide & Conquor: Each CPU gets its own sub-section of the data. Then, each of those sub-sections gets divided recursively... Aside from dividing data by three (much more common for arrays to be multiple of 2 than 3 in my experience), it's not so bad. When you get to merging the sorted results with threes instead of twos, the merges would get complicated and ugly (Differing lengths, differences between origional division and sub-division). So, it would work, but a well-known elegant algorithm for sorting would become much more cumbersome.
As someone who has worked with parallel programming, three is a really strange number of CPUs. In a performance-intense area like gaming, I'd find it hard to believe that they'd really accept many of the inefficiencies of high-level programming. At the same time, it's hard to see the usefulness of a 3 CPU architecture for lower-level programming.
As I stated earlier, they may have 2 in tandem, with one doing other things like sound. However, any divergence in that direction is pure speculation. I'm sure they have some well thought-out plan for the design.
It'll be interesting to see what they finally decide to put out on the market, because the current spec leaves me decidedly curious.
~D
This sig has been enciphered with a one-time pad. It could say almost anything.