Official Xbox2 Thread

Not to mention lots of code which could quickly be turned into libraries and tutorials which would just slash game development times down really quick.
 
I still believe the specs i posted would be the most likely for the xbox 2.

The current rumor is dx10 wont come out till longhorn which is going to come out in 2005. So that is most likely what the xbox 2 will have.
 
I still believe the specs i posted would be the most likely for the xbox 2.

On the CPU, front I have to disagree. The A64 you're talking about is a custom part. It falls inbetween a A64 and Opteron 244. The question is why would AMD bother making this since they probably don't stand to make as money off this compared to what they would make selling cheaper A64s in retail. I'm not sure if XBox2 would offer appealing enough volumes that AMD would be willing to commit fab space to such a CPU. Remember, choosing a CPU for a system has more than just performance technical factors. There is also this thing called ability to deliver. Which it doesn't seem like AMD can do. And by deliver I don't mean performance, I mean the product for the likely price.
 
Saem said:
I still believe the specs i posted would be the most likely for the xbox 2.

On the CPU, front I have to disagree. The A64 you're talking about is a custom part. It falls inbetween a A64 and Opteron 244. The question is why would AMD bother making this since they probably don't stand to make as money off this compared to what they would make selling cheaper A64s in retail. I'm not sure if XBox2 would offer appealing enough volumes that AMD would be willing to commit fab space to such a CPU. Remember, choosing a CPU for a system has more than just performance technical factors. There is also this thing called ability to deliver. Which it doesn't seem like AMD can do. And by deliver I don't mean performance, I mean the product for the likely price.


How is it custom ?

Also amd can deliver . Amd would also upgrade a fab of thiers for a fat contract like the xbox 2 or outsource to ibm for chip production if they need help .
 
It is a custom chip. A64 will have a single channel interface with 1 ht link IIRC. It might have two links which cuts down the amount of work.

Again, you say the Xbox2 contract is fat, I say it isn't. Compare this to the money they could get by just selling the processors? IBM is expensive, well if you want some of their good fab technology. Not to mention, CPU designs are tuned to processes, out sourcign isn't as simple as it maybe with less complex designs -- GPUs and so on. IMHO I think AMD made a big mistake going with SOI, it raise their cost for very little gain. I wonder if they'll introduce a revision of the Hammers which don't use SOI.

I'm not sure where you're coming from, but the amount of resource investment on AMD's part for little gain doesn't seem worth it, I'm sure they're more intrested in other OEM design wins which will likely mean larger volumes without any additional investment.

Besides, why would MS want to buy from AMD which is basically one natural disaster away from basically having all their production go boom?
 
Saem said:
Again, you say the Xbox2 contract is fat, I say it isn't. Compare this to the money they could get by just selling the processors?


They wouldn't stop selling processors if they took the Xbox2 contract and they would get a good boost in profits from it.
 
Dural,

You don't seem to understand. AMD as far as I know, has sold everything they have produced in terms of processors for quite a while, not sure if this is true still. Now, the thing is they're already selling stuff. Unless, MS is going to pay them more than their average per unit profit, why would they switch, they don't have any more production capacity.

To get more means porting their CPU over to another process which is non-trivial.
 
Saem:

All Hammer cores are the same chip. 1MB L2 A64s is an Opteron with one half of the memory controller and the additional HT links unconnected/disabled. It wouldn't be a custom chip at all.

And your speculation on AMDs ability to deliver in 2005 using dodgy speculation based on your own perceptions of things as they stand now is flawed to say the least.

Free tip of the day: get a clue, mate. :)


*G*
 
Saem:

All Hammer cores are the same chip. 1MB L2 A64s is an Opteron with one half of the memory controller and the additional HT links unconnected/disabled. It wouldn't be a custom chip at all.

And your speculation on AMDs ability to deliver in 2005 using dodgy speculation based on your own perceptions of things as they stand now is flawed to say the least.

Free tip of the day: get a clue, mate.

All you can come up with is nay saying. I'm making my position clear, my speculation is grounded with facts while others just spit out stuff will little back ground knowledge as demonstrated by your post. BTW, should I start with the ad homenim attacks, it seems you are providing me with lots of fodder.

I find it hard to believe that all Hammers are the same, considering the huge pin count and cache size difference. You do realise that packaging is a factor here, right? We're not talking about trivial cache sizes here, it doesn't seem very economical to have a die and have such a large portion of it just taking up space. Remember, AMD is the one who is having trouble with fab space here. Not to mention they're use low-k waffers which are rather costly, it doesn't make sense to not have different cores.

Also, if you understood MPU manufacturing well, you'd realise that the Opteron is likely produced on a less aggressive process, because they have certain reliability concerns. The A64 is likely manufactured on a different process. You do know that the P4 and Xeons aren't the same die, don't you, there are design differences? If AMD are using the same process, which I'm pretty damn sure they're not, they're going to hose their clock rates in the desktop space which will really hit the sales. Just like P4s and Xeons use different processes, this is one of the reasons why Xeon clocks aren't as aggressive as the P4s. In conclusion, A64 definately differs from the Opterons and not just in cache, HT link and the subsequent pin outs.
 
Saem said:
Dural,

You don't seem to understand. AMD as far as I know, has sold everything they have produced in terms of processors for quite a while, not sure if this is true still. Now, the thing is they're already selling stuff. Unless, MS is going to pay them more than their average per unit profit, why would they switch, they don't have any more production capacity.

To get more means porting their CPU over to another process which is non-trivial.


I seriously doubt AMD sells out of every chip they make, selling an extra few million chips a year at any profit is going to help them more than hurt them. There has also been speculation that Athlon64 would eventually get the dual channel memory controller of the Opteron. The only reason I wouldn't think that MS would want the Athlon64 is because they will probably want to IP license rather than purchase the chips this time around and I doubt AMD would do this.
 
The only reason I wouldn't think that MS would want the Athlon64 is because they will probably want to IP license rather than purchase the chips this time around and I doubt AMD would do this.

Do you think Intel would ?
 
Hmmm mabye ms will use the athlon. Its already been said that the athlons will be outsourced to another company. Mabye they will give ms the athlon liscence and ms will produce the chips some where. That way ms could use an athlon at like 2.5ghz with a fat 400mhz bus. Ms can get them for a song and a dance and amd could make a fe wbucks off each one with no additional cost .
 
Saem said:
All you can come up with is nay saying. I'm making my position clear, my speculation is grounded with facts

...According to yourself, which even IF true, will be two years old come this time of year 2005; half an eternity in the semiconductor business! For crying out loud, AMD hasn't even maxed out their Dresden fab yet, and they've started construction on extending it.

while others just spit out stuff will little back ground knowledge as demonstrated by your post.

Rofl, dude. Whatever! Point at ONE thing that was incorrect in my post please.

One-meg L2 A64s are to all my knowledge EXACTLY the same core as opteron chips, sans HT links and 64 (data) bits of memory interface, I'm not sure about the ECC support. That's why the pincount is less. Check out a die photo of the Opteron, you'll see why this isn't such a stupid idea after all. The HT and memory interface areas of the chip aren't that big after all, and using the same die means they don't need an extra production line dedicated to it, thus saving scheduling issues and overhead costs.

BTW, should I start with the ad homenim attacks, it seems you are providing me with lots of fodder.

Don't get your panties all in a bind here. I didn't attack your PERSON.

I find it hard to believe that all Hammers are the same, considering the huge pin count and cache size difference.

There's no cache size difference between an opteron and a 1-meg A64. There will be a 1/4 meg A64 too, and that will likely be a different die, but that's a different story.

You do realise that packaging is a factor here, right?

Yeeeesssss? So what?

Die and packaging are two different things.

We're not talking about trivial cache sizes here, it doesn't seem very economical to have a die and have such a large portion of it just taking up space.

But it isn't just taking up space.

Remember, AMD is the one who is having trouble with fab space here. Not to mention they're use low-k waffers which are rather costly

...Which will be a non-issue come 2005, since they'll be using a different process then anyway. You see why I said you needed a clue mate? You're using today as a basis for speculating the future.

If they need more fab space than they got themselves, they'll have plenty of time after negotiating a deal with M$ to modify their designs for the process of some other fab partner to extend their capacity. A process shrink and the extension to their Dresden facility will automatically give them greater capacity as well. They could also start using larger wafers, there are 12" ones available, they're as big as a dinner plate.

Also, if you understood MPU manufacturing well, you'd realise that the Opteron is likely produced on a less aggressive process, because they have certain reliability concerns.

Oh, so now you're a CMOS manufacturing expert while I'm just some farmer dumbass? LOL. Dude, you're still stuck in the present while 2005 is still the general target we need to be looking at. You think AMD will have process difficulties for a straight two years? They've been making integrated circuits since the sixties. They know these things man.

You do know that the P4 and Xeons aren't the same die, don't you, there are design differences?

CACHE differences, the actual core design is the same, but whatever. Yes, of course I know. Xeons have more cache (these days anyway), but Xeons aren't Opterons. AMD works differently than Intel.

If AMD are using the same process, which I'm pretty damn sure they're not

They'll use the same process mate, you can be sure of that. They might tune it differently, but that doesn't change it's still the same process. Intel might manufacture Xeons in a way that is different enough to be called another process compared to their P4s because they likely make the processors in an entirely different fab anyway (they have a whole bunch of them). However, I'm pretty sure Intel uses the same .13 process for all their P4-based CPUs and just tweak it differently.

In conclusion, A64 definately differs from the Opterons and not just in cache, HT link and the subsequent pin outs.

That's quite a conclusion you come to there, based on nothing but idle speculation and your own suppositions.


*G*
 
Saem:

I'm actually pretty curious as to what microsoft decides to do. In terms of cpus that we know of right now, intel would probably be a better choice with something like banias. Both the P4 and the athlon/opteron processors (along with something like the NV50 or R500) would bake in a case the size of the xbox without some pretty hefty cooling. (would we see vaporphase or water cooling in a console?) On the other hand, microsoft seems to be paying a lot more attention to AMD in the last year or so (which perhaps is simply due to AMD's recent success). Nvidia seems to favor AMD at this point as well, but then Nvidia's role in the xbox2 seems to be fairly uncertain at this point.

Should be interesting to see the design decisions they make. Heat vs speed vs cost vs space vs politics.

Nite_Hawk
 
Why would an athlon , p4 or a64 run that hot . we are talking about 2005. A 2.5ghz athlon in 2005 will be slow . They should be at around 3.5 ghz. So they can easily have an athlon core running at a low vcore . Mabye 1.2 or 1.3 reducing heat. Right now the opteron is running at 1.8ghz at 1.45 and is running pretty cool in a server case. Why not in 2005 a 2.5 running at 1.45 or less vcore ? The reason why ms went with a slower celeron was not just cost but heat. By running them at 700mhz they were able to lower the vcore and produce less heat. The same will be done this time.

Amd is a great company and would easily be able to meet demand for the xbox 2 console . Even if they only make 10 dollars of a cpu in the xbox2 they are going to sell millions of xbox 2s . Not only that but if they use an a64 it would be even better for amd. Even if they barely make any money off the deal they will have a ton of publishers used to coding for thier 64 bit chips . I don't see how amd can loose in a deal with ms .
 
Actually a Pentium M would be fantastic for Xbox 2, now that I think about it... you people should too. It's small. It's quiet. It's cool. It's a P6 core. :)
 
Tagrineth said:
Actually a Pentium M would be fantastic for Xbox 2, now that I think about it... you people should too. It's small. It's quiet. It's cool. It's a P6 core. :)

Yea it wouldn't be a bad choice . Only thing is how does it stack up to the athlons , a64s and p4s....
 
...According to yourself, which even IF true, will be two years old come this time of year 2005; half an eternity in the semiconductor business! For crying out loud, AMD hasn't even maxed out their Dresden fab yet, and they've started construction on extending it.

IIRC, they have maxed out their current Dresden capacity. I think the sales due to the Opterons will further strain production capcity and I'm not the only ONE! AMD for this specific reason in increase Dresden capacity and outsourcing the Athlons to TSMC or some other fab. It seems you don't understand AMD's reasoning.

One-meg L2 A64s are to all my knowledge EXACTLY the same core as opteron chips, sans HT links and 64 (data) bits of memory interface, I'm not sure about the ECC support. That's why the pincount is less. Check out a die photo of the Opteron, you'll see why this isn't such a stupid idea after all. The HT and memory interface areas of the chip aren't that big after all, and using the same die means they don't need an extra production line dedicated to it, thus saving scheduling issues and overhead costs.

I know of two variants of A64s. The 512KB and 256KB cache version. Never heard of any 1MB version. HT, ECC, Memory interface, SMP and a few other things that might not have been mentioned.

Which will be a non-issue come 2005, since they'll be using a different process then anyway. You see why I said you needed a clue mate? You're using today as a basis for speculating the future.

Actually, you need to get a clue. You're a fool if you think AMD would use a bleeding edge process for the Xbox 2 contract. Intel certainly didn't and for good reason, because there are better things that you can produce with that fab capacity which will make you more money. That fab capacity is best used for high margin parts, not low cost moderate-high volume parts.

If they need more fab space than they got themselves, they'll have plenty of time after negotiating a deal with M$ to modify their designs for the process of some other fab partner to extend their capacity. A process shrink and the extension to their Dresden facility will automatically give them greater capacity as well. They could also start using larger wafers, there are 12" ones available, they're as big as a dinner plate.

So lets see, they use the extra space which they're already using for their other lines. Which likely have higher margins, see Opterons and A64s. One thing you have to realise is the new x86-64 family more than anything means AMD's MPUs are closer to price parity with their Intel counter-parts. What you and many others fail to realise is that the XBox2 contract will be for low cost MPUs, which means small margins, which means you don't want to screw the stuff where you're making a lot more. See what Intel did with the Xbox contract, think about what exactly they did this? You might get a clue at that point.

Yes, 12" wafers are so trivial, LoL! We're talking about a more exotic variety, these are the more "run of the mill" wafers that Intel is using. Oh and then we could outsource. Hahaha. You need to get a clue about this stuff, first off this is a very aggressive and exotic process. I'm doubt many except IBM will be able to produce this. If that's the case, which I'm quite sure it is, then AMD would have to go through the non-trivial task of getting this A64 variant to work well in IBM's foundaries and then pay IBM. Did you know that IBM is very expensive on a per unit basis for their more exotic processes?

Oh, so now you're a CMOS manufacturing expert while I'm just some farmer dumbass? LOL. Dude, you're still stuck in the present while 2005 is still the general target we need to be looking at. You think AMD will have process difficulties for a straight two years? They've been making integrated circuits since the sixties. They know these things man.

Do you have a severe inability to comprehend? I'm talking about differences in manufacturing processes. What I'm saying is that Opteron and A64 are very likely going have different processes. A64 is going to be more aggressive. Have you gotten a clue yet? This is done intentionally, not because of problems with a process but reliability concerns! I'm pretty damn sure AMD wants to be taken seriously in the enterpirse sector so they'll do things like this so they're taken more seriously.

CACHE differences, the actual core design is the same, but whatever. Yes, of course I know. Xeons have more cache (these days anyway), but Xeons aren't Opterons. AMD works differently than Intel.

NO, they're NOT the same. There are more differences than the cache and process. If AMD wants to be taken seriously, they'll work the same as Intel. I'm not sure what weird ideas you have about AMD, but AMD has been following the little chipzilla model for a long time. They're trying to be more and more like Intel.

They'll use the same process mate, you can be sure of that. They might tune it differently, but that doesn't change it's still the same process. Intel might manufacture Xeons in a way that is different enough to be called another process compared to their P4s because they likely make the processors in an entirely different fab anyway (they have a whole bunch of them). However, I'm pretty sure Intel uses the same .13 process for all their P4-based CPUs and just tweak it differently.

What are you going on about? Your tweaking basically changes the process. Yes they'll both be classified as 0.13u or whatever, but that doesn't mean they're the same. Just like Intel's 0.18u and AMD's 0.18u are NOT the same. You do realise to tweak is to change. I certainly hope AMD doesn't use the same aggressive process for the Opterons as they do the A64s, they won't be taken seriously if they do.

That's quite a conclusion you come to there, based on nothing but idle speculation and your own suppositions.

It's a pretty sound conclusion, unless AMD encounters very significant success with Opteron. AMD has been in financial dire straits for sometime. IIRC, they'd been sending their higher ups in coach and declining a lot of air travel for their employees to various conferences. They're hurting bad.

Unless the XBox 2 contract is fairly high on the margins, they likely won't be interested in diverting A64s in that directions, since they'll have higher ASP than the current Athlons and thus margins. Yes the Xbox 2 will mean lots of sales, but over what period of time and currently I don't think the XBox sales are inspiring confidence.

I seriously doubt AMD sells out of every chip they make, selling an extra few million chips a year at any profit is going to help them more than hurt them. There has also been speculation that Athlon64 would eventually get the dual channel memory controller of the Opteron. The only reason I wouldn't think that MS would want the Athlon64 is because they will probably want to IP license rather than purchase the chips this time around and I doubt AMD would do this.

Last I checked they have been. This might have changed recently, but I wouldn't be suprised if they kept on selling out of their stock.

Hmmm mabye ms will use the athlon. Its already been said that the athlons will be outsourced to another company. Mabye they will give ms the athlon liscence and ms will produce the chips some where. That way ms could use an athlon at like 2.5ghz with a fat 400mhz bus. Ms can get them for a song and a dance and amd could make a fe wbucks off each one with no additional cost .

Using an Athlon is more likely. I believe AMD has already started porting the design over to TSMC or some such.

I'm actually pretty curious as to what microsoft decides to do. In terms of cpus that we know of right now, intel would probably be a better choice with something like banias. Both the P4 and the athlon/opteron processors (along with something like the NV50 or R500) would bake in a case the size of the xbox without some pretty hefty cooling. (would we see vaporphase or water cooling in a console?) On the other hand, microsoft seems to be paying a lot more attention to AMD in the last year or so (which perhaps is simply due to AMD's recent success). Nvidia seems to favor AMD at this point as well, but then Nvidia's role in the xbox2 seems to be fairly uncertain at this point.

Yup, Nvidia is very opposed to Intel, since they didn't get a bus license. I wouldn't be suprised that Banais made it into the XBox 2, it's actually quite beefy one thing to note is that it has a significantly upgraded FPU over the PIII, IIRC Paul DeMone and Vincent Dipree (sp?) commented on this. The thing with the P4 is the lack of FPU power, it's clock rate does make up for a fair bit and SSE2 does help quite a lot as well, but it still comes up short when compared to an Athlon. Things would be better if the Athlon L2 cache latency was improved and a faster FSB wouldn't hurt.

Should be interesting to see the design decisions they make. Heat vs speed vs cost vs space vs politics.

Yeah, this is really tricky. I'm not sure what's going to happen here. The heat is the big issue IMHO. I think the PS3 will have a significant advantage in the power/performance ratio which means they can eat the cost which will come down eventually (via improved yields and process shrinks), while the x86 chip in the Xbox 2 will only improve via process shrinks. As for the XGPU, well I think ATI is the logical choice, I believe they're simple far better at releasing faster, smaller, cooler and higher performance chips when compared to Nvidia. NV30 is a mess and unless they really rip out a lot of the legacy it's a lot of wasted power hungry transistors.

Politics is really gonna screw up and techincal decisions. I know a lot of MS guys really like the x86-64, I wouldn't be suprised if this is in the Xbox 2. There is a but however. If MS wants AMD or AMD wants the XBox 2 contract, than MS can hold the Windows 64 knife to AMD's throat to get what it wants. While Intel can play a few cards and screw over AMD, by saying, MS if you screw AMD like so, we'll give you a better deal. Additionally, the AMD chip to go in the XBox 2 if it does go in, will likely be a A64 with 256KB L2 cache.
 
Back
Top