XBox 360 Specs Released [Triple Core @ 3.2Ghz each]

[Dave the Great]

"My modded $300 Xbox 360 runs Tiger faster than your $3000 powermac" ...

Yeah!!

That's what I was thinking!!

This thing is way more powerful than any PowerMac, yet, costs less than the MacMini. Wow!!

 

[GFLPraxis]

Yeah!!

That's what I was thinking!!

This thing is way more powerful than any PowerMac, yet, costs less than the MacMini. Wow!!

Why don't you people read the first couple pages before you start responding?

This thing wouldn't touch a PowerMac except for the next gen graphics card...

 

[GFLPraxis]

MS owns Virtual PC - they could potentially use that technology for backwards compatibility. Out of order vs in order is irrelevant for emulation - out of order is simply a performance issue, logically the code is executed in order. If most of the graphics code is making DirectX or other OS calls, the emulator can route those calls to the native calls - just like the 680x0 emulator did/doesdoes on the PowerPC with 680x0 apps.

Not only that, they could be using a PowerPC version that supports little endian, which would make the emulation more efficient.

I'll give you the quotes from the debate that convinced me it was not possible.

The PPC in the X360 is NOT a PPC970, nor a POWERx derivative (which have OOOE). In fact, as speculated = here it is the same CPU core as in the PS3, but without the vector units and there are three of them. You may argue against "speculation" but then your argueing that there is another chip exactly identical to the CPU core in PS3 that has never been seen or heard before. I find that exceedingly unlikely, especially considering that only this is the only CPU at IBM has the ability to reach 3Ghz+. In short, it is almost certainly the same, which makes a IOE processor.

EPIC doesn't count as while it is IOE it's also a 6-issue core (compared to 2 for the PPC and 3 for Pentiums) with a huge amount of cache (1.5-9MB L3, 256KB L2), and uses the IA-64 instruction set which was designed to lessen it's IOness, and it still loses to Pentiums and Athlons in integer ops. Integer performance is also the same weakness the much slimmer PPC has to face too. For more realistic CPUs, the 50% claim stands. If you dispute the 50% claim, then you dispute the guy who wrote that, not me. Floating point ops will be much better, which is the PPC's strength, but that's useless in emulation so backwards compatibility is still simply impossible.

The much shorter pipelined out-of-order and wider P3 will likely have a similar if not superior IPC, and in some case will simple OWN the PPC because of it's out-of-order nature. Learn what out-of-order execution means, it's history, etc. It's a very important aspect of a CPU.

An in-order chip has approximately half the performance of an out-of-order CPU (reference) given equivalent technology. This would imply the IPC of the PPC is half that of the P3 assuming perfectly linear scaling to clockspeed. However, there's a lot more to it than that so I'm not making any guesses and it varies too much anyways. However, emulation is pretty straightforward; it all depends on it's ability to convert single-threaded x86 instructions into single-threaded PPC instructions and then process them. Unfortunately, the fact the the PPC is designed for great multithreading makes it terrible candidate for this (It's deeply pipelined and has SMT). Plus PowerPC just suck at the integer apps in most benchmarks I've seen, which are probably going to be the most difficult to emulate, so IMO it'll never work and you'll have to be slightly crazy to even try.

So what? You're trying to emulate a 3-way, out-of-order execution x86 chip with a (relatively) short pipeline (what the P3 in the Xbox is) with a 2-way in-order deeply pipelined PPC chip (what the CPU in the Xbox2 is). Let me explain. The CPU in the Xbox2 is a very simple design and is "skinny," meant for very high clockspeeds, but will have bad IPC (instructions per clock). The P3 on the other had is a much wider design and is more complex, slower in clockspeed (theoretically, since they're different generations of chips) but has good IPC. Somethings will be very suited to the first way but other things will be much more suited on the second CPU. In short they're fundamentally difference designs, and even though the PPC may be moving at 3Ghz and the P3 at 733Mhz, there should still be some cases where the P3 will win. Emulating this will be an ugly, buggy mess I seriously doubt they can do.

If you can debunk this stuff, go ahead, I'd love to hear it. I don't have enough technical knowledge to argue with the original poster. It was an arguement between the two most technical guys on the board and this guy was the winner.

 

[iJWC]

I'll give you the quotes from the debate that convinced me it was not possible.
If you can debunk this stuff, go ahead, I'd love to hear it. I don't have enough technical knowledge to argue with the original poster. It was an arguement between the two most technical guys on the board and this guy was the winner.

The winner at what? Speculation?, because that is all they're doing. No none has even seen the innards yet. We'll probably know more next week.

 

[Dave the Great]

Why don't you people read the first couple pages before you start responding?

This thing wouldn't touch a PowerMac except for the next gen graphics card...

I did read the first couple of pages. And I concurred with at least one of the posts.

This thing wouldn't touch a PowerMac

Why wouldn't it even touch a PM?

except for the next gen graphics card...

What about the processors, too?

 

[alandail]

nothing you posted impacts the possibility of emulation, only the performance. And even though MHz don't scale, you are still talking about a 3.2 GHz chip emulating a 700 MHz chip.

And my points before still stand.

- if the programming model has most code go through system calls to do things, they can run natively

- if they are using a power pc chip that supports little endian, things get quite a bit easier - including floating point.

And it's also certainly theoretically possible to translate the entire x86 object code to PPC code - either in advance, or on demand.

But there is no question at all that emulation is technically feasible, especially for a company who writes the OS for both machines and who already owns processor emulation technology.

The questions are
- can it be done with acceptable performance
- would MS rather force people to buy new games

 

[WannabeSQ]

All this talk about which console is the "best", or "fastest" is all moot anyway. What really matters is the games. I didn't buy an xbox because it had the highest specs, I bought it because I wanted to play Halo. That's it. The games that interest me are either xbox only, or multi platform. If Halo 3 is a launch title for the Xbox 360 (doubtful given Bungie's slow progress in making 4 maps for Halo 2) I will definately buy one. However, if that is not the case, I will buy whatever console carries the games I most want to play. HDTV on xbox 360 is intriguing, as I own an HDTV, as is wireless controllers, I will wait to see what kind of software we will have.

 

[d.perel]

OMG!

Yeah!!

That's what I was thinking!!

This thing is way more powerful than any PowerMac, yet, costs less than the MacMini. Wow!!

Now that you mention it, how DID IBM manage to sell those processors to Microsoft at an extreeeemely low price, at a faster speed, and nearly guaranteed availability, while we deal with 2.7 ghz, $3000 computers? Also, this pricing is much lower than a tri-core 3.2 ghz PC! Is there something I am missing?

 

[alandail]

Now that you mention it, how DID IBM manage to sell those processors to Microsoft at an extreeeemely low price, at a faster speed, and nearly guaranteed availability, while we deal with 2.7 ghz, $3000 computers? Also, this pricing is much lower than a tri-core 3.2 ghz PC! Is there something I am missing?

yes, you are missing two things

- the chip isn't available yet - compare it to macs in november, not macs in may
- the chip is stripped down - thus per GHz, the speed isn't the same as a G5.

 

[lalcan]

Unless these chips are many many times more efficient than G5's, they have nowhere near the performance. Just think how much electrical power three 3.2GHz G5s would require?

100 Watts

 

[Dave the Great]

... - the chip isn't available yet - compare it to macs in november, not macs in may
- the chip is stripped down - thus per GHz, the speed isn't the same as a G5.

I think MacWorld had an article stating that they thought there would be PowerMac at 3.2 by November.
I thought they also said the xBox 360 had a 976 in it and supposedly was at 65nm.

However, isn't that is all pure speculation because I think no one has seen the innards and will know for sure exactly what is in there until next week (E3).

 

[GFLPraxis]

I did read the first couple of pages. And I concurred with at least one of the posts.


Why wouldn't it even touch a PM?


What about the processors, too?

It's a stripped down processor designed to be really good for gaming. It would suck for non-gaming tasks. It's in-order, rather than an out-of-order chip. It's NOT a G5.

 

[GFLPraxis]

nothing you posted impacts the possibility of emulation, only the performance. And even though MHz don't scale, you are still talking about a 3.2 GHz chip emulating a 700 MHz chip.

And my points before still stand.

- if the programming model has most code go through system calls to do things, they can run natively

- if they are using a power pc chip that supports little endian, things get quite a bit easier - including floating point.

And it's also certainly theoretically possible to translate the entire x86 object code to PPC code - either in advance, or on demand.

But there is no question at all that emulation is technically feasible, especially for a company who writes the OS for both machines and who already owns processor emulation technology.

The questions are
- can it be done with acceptable performance
- would MS rather force people to buy new games

Yeah, can it be done with acceptable performance is the big if. Those quotes were about emulation, maybe you misunderstood it; he was saying that while that PowerPC will get insanely good floating point calculations, it will suck really bad at integer calculations. The Pentium 3 processor in the XBox 1 is really good at integer calculations, so its a very bad mismatch for emulation.

Like he said, the in-order processor will get about half the performance. So think of it like a 1.6 GHz processor emulating a 733 MHz processor...

 

[GFLPraxis]

I think MacWorld had an article stating that they thought there would be PowerMac at 3.2 by November.
I thought they also said the xBox 360 had a 976 in it and supposedly was at 65nm.

However, isn't that is all pure speculation because I think no one has seen the innards and will know for sure exactly what is in there until next week (E3).

Actually there's no way they could have a PowerPC 970- if I remember right, the G5 (PowerPC 970) was a joint Apple-IBM creation. So they'd have to buy the G5's from IBM..AND Apple.

It makes sense that its not a 970.

 

[alandail]

Yeah, can it be done with acceptable performance is the big if. Those quotes were about emulation, maybe you misunderstood it; he was saying that while that PowerPC will get insanely good floating point calculations, it will suck really bad at integer calculations. The Pentium 3 processor in the XBox 1 is really good at integer calculations, so its a very bad mismatch for emulation.

Like he said, the in-order processor will get about half the performance. So think of it like a 1.6 GHz processor emulating a 733 MHz processor...

again, it'll depend on what percentage of time games spend in system routines - if 80% of the time is spent in system calls and those can all run native, performance on an emulator would be just fine.

 

[myapplseedshurt]

Watching this debate is funny from my point of view

I am bound to a non-disclosure agreement. I work in a group that has its hand directly and indirectly in some processors all of you talk about.

of course I can't say anything about these, but it is fun to watch the debate. With the knowledge I have, I want so badly to straighten this whole thing out, but I can't.

I started watching this site when the first PM were introduced to see if how close the rumors were, and surprisingly they were a bit accurate.

I must say that from what I've seen so far, a few people on this site also have NDA's about these products because they know a little too much info, but they seem to be careful about it.

Others (and this is the majority) don't really have a clue about microprocessor architecture. But I admittedly, I know nothing about system level or OS level designs.

 

[myapplseedshurt]

Actually there's no way they could have a PowerPC 970- if I remember right, the G5 (PowerPC 970) was a joint Apple-IBM creation. So they'd have to buy the G5's from IBM..AND Apple.

It makes sense that its not a 970.

970 is solely an IBM design, but it asked apple what it would like to have in it. IBM owns all rights to 970.

 

[GFLPraxis]

again, it'll depend on what percentage of time games spend in system routines - if 80% of the time is spent in system calls and those can all run native, performance on an emulator would be just fine.

Like I said, the million dollar question. Very difficult to know whether its possible or not.

If it has little-endian mode...then it will be a lot easier, but if it doesn't...

It just seems to me that the processor difficulties combined with the different architecture GPU (you'd think XBox 1 games would be NVidia optimized) would make it incredibly difficult.

 

[GFLPraxis]

I am bound to a non-disclosure agreement. I work in a group that has its hand directly and indirectly in some processors all of you talk about.

of course I can't say anything about these, but it is fun to watch the debate. With the knowledge I have, I want so badly to straighten this whole thing out, but I can't.

I started watching this site when the first PM were introduced to see if how close the rumors were, and surprisingly they were a bit accurate.

I must say that from what I've seen so far, a few people on this site also have NDA's about these products because they know a little too much info, but they seem to be careful about it.

Others (and this is the majority) don't really have a clue about microprocessor architecture. But I admittedly, I know nothing about system level or OS level designs.

Why must you taunt us so????
/me cries

Can you tell us who is completely wrong (without revealing info) at least? Is it me? Is it the people I am quoting? Or am I somewhat on track?

Meanie

 

[GFLPraxis]

970 is solely an IBM design, but it asked apple what it would like to have in it. IBM owns all rights to 970.

Thanks, guess I was misinformed on that.

 

[myapplseedshurt]

but that would ruin my fun!!!

Why must you taunt us so????
/me cries

Can you tell us who is completely wrong (without revealing info) at least? Is it me? Is it the people I am quoting? Or am I somewhat on track?

Meanie

sorry, can't do that. but I must say that I am hooked on this site. I don't currently own an apple machine, but having had my hands in on it, and from the posts I read on this site, I am definitely going to buy one. The question is, when? will I wait till next year or buy one now?


ha ha ha ha ha ha (devilish laugh) ha ha ha ha ha

 

[adamjay]

if it can't run Tiger, i'm not interested.

 

[makey]

I find this quite funny and ironic. Microsoft using processors co-developed with IBM by Apple.

 

[Blackheart]

Well done!!

I thought the claim of 1 Teraflop was a bit sus... Fair enough they are using IBM chips and a dedicated system architecture, but this claim has to be fake. If this 1 Teraflop is true there will be a whole lot of people wanting these things for reasons other than gaming . And considering the price of the 'box' .. peh....

I own an xbox and I would be willing to buy a '360 if there is confirmed backwards software compatibility. Has anyone been able to find some concrete evidence the '360 will have this feature. I have googled for 20 mins and have only found rumors and speculation....


aussie_geek

Yeah. For one, if Microsoft could manage a $400 that could crank out a whole teraflop, I don't see why Microsoft would even continue writing software when they could royally slaughter the competition in hardware. Secondly, I think you'd see a lot of unconventional super computers starting up... and on very low funding at that.

Not all flops are created equal, would be my guess. Someone correct me if I am wrong.

FLoating point OPerations per second

This is a standard by which processors can be benchmarked; so yes, they are all created equal.

 

[muya]

These have to be multi-core G4s

Custom IBM PowerPC-based CPU
• Three symmetrical cores running at 3.2 GHz each
• Two hardware threads per core; six hardware threads total
• VMX-128 vector unit per core; three total
• 128 VMX-128 registers per hardware thread
• 1 MB L2 cache



The G5 does not support L2 cache.