PPC976: Possible Xbox2 processor?

Status
Not open for further replies.
DM said:
How many Xbox games do you know that makes an extensive use of SSE?
No no... How many games you know that DON't make use of SSE? You are the one that made the claim, you should give evidence to support it.

The difference is that Power5 FPU can be coded in C++ and auto-optimized by compiler, whereas all vector units are asm stuffs.
You can do more in straight C++ then you think - including 'teaching' compiler a thing or two about optimization. But granted SIMD units will still require some amount of asm code to use, no matter how minimal it might be.
 
Certainly the POWER4 project required organizational discipline. The POWER4 implements two complete four-way, superscalar, out-of-order 64-bit PowerPC CPUs in a 0.18 um SOI CMOS process with seven layers of copper interconnect. The 174m transistor device integrates three independent 512 KB L2 caches, controllers for external L3 cache and memory, and inter-processor communication links.

http://www.realworldtech.com/page.cfm?ArticleID=RWT022001001645&p=4



IBM's POWER5 design is rather like the POWER4 design, being dual-core with some L2 cache on-die and a large external L3 cache, in a Multi-Chip Module (MCM). POWER5 systems to replace the current POWER4+ based systems are expected to launch by the end of 2004, 3 years after the original POWER4. All parts of the design have been enhanced to increase performance, while also trying to keep power consumption reasonable. Today, IBM's largest POWER4 servers can take 16 POWER4 processors (32 cores), and for the next generation, IBM are working on a system code-named Squadron that will take 32 POWER5 processors (64 cores, and 128 hardware threads).

The POWER5 will be 389mm² on IBM's 130nm SOI process, run at up to 2GHz and contains 276 million transistors. On the same process, the POWER4+ is 267mm². Though the basic pipeline is the same as the POWER4, the amount of resources has been increased to better optimise for the 2-thread SMT capable cores, which increases the core sizes by about 24%. These resource increases include rename registers, instruction-fetch buffers, and the whole cache design. The SMT threads can also be assigned different priorities, so better optimise the resource allocation - for example, lower the priority of a thread waiting on a lock or I/O. In server tasks, SMT is expected to improve overall performance by around 40% (will vary from benchmark to benchmark of course).

http://www.aceshardware.com/read.jsp?id=60000274

I do not think I need to comment these quotes as they are quite nicely written.

Edit: a nice picture of the POWER5 chip.

power5.jpg
 
Re: ...

Deadmeat said:
What uses would IBM have of CELL? CELL mainframe? CELL AS/400? CELL rackservers?? Get real.
And since they are not in a licensing/royalty deal and not fabbing chips for SCEI directly they wanted to work on CELL to... smile in pride at making a cool new chip?
 
Fafalada said:
DM said:
How many Xbox games do you know that makes an extensive use of SSE?
No no... How many games you know that DON't make use of SSE? You are the one that made the claim, you should give evidence to support it.

Fafalada, you know that logic and Deadmeat are not always good together.
 
Re: ...

cthellis42 said:
Deadmeat said:
What uses would IBM have of CELL? CELL mainframe? CELL AS/400? CELL rackservers?? Get real.
And since they are not in a licensing/royalty deal and not fabbing chips for SCEI directly they wanted to work on CELL to... smile in pride at making a cool new chip?

That might not be far from the truth :lol

Seriously, I think IBM has invested quite a bit of resources into CELL and the only thing they licensed to Sony was their 100 nm SOI manufacturing process around 3-4 years ago.

I think IBM wants to do more than smile at the chip once finished ;)
 
Re: ...

Panajev2001a said:
Seriously, I think IBM has invested quite a bit of resources into CELL and the only thing they licensed to Sony was their 100 nm SOI manufacturing process around 3-4 years ago.
100nm... Is that even used for anything? (And does it apply for any other processes, or did Sony basically lean on some of the tech for improving their own lines?)
 
Re: ...

cthellis42 said:
100nm... Is that even used for anything? (And does it apply for any other processes, or did Sony basically lean on some of the tech for improving their own lines?)

I believe it was part of a technology transfer agreement. In 2002 they further extended the 2001 agreement.
 
Re: ...

cthellis42 said:
Panajev2001a said:
Seriously, I think IBM has invested quite a bit of resources into CELL and the only thing they licensed to Sony was their 100 nm SOI manufacturing process around 3-4 years ago.
100nm... Is that even used for anything? (And does it apply for any other processes, or did Sony basically lean on some of the tech for improving their own lines?)

Sony skipped the 100 nm node: their work with Toshiba has been on CMOS4 ( 90 nm ), CMOS5 ( 65 nm ) and CMOS6 ( 45 nm ).

I would think they needed to improove their own manufacturing ability and needed IBM's secret recipee to improove SOI and low-k materials to not make the manufacturing process a "low-yelds" one ( insider joke which I learned not too long ago ;) ).
 
Re: ...

Deadmeat said:
Still not sure if IBM can produce whatever of CELL it wants for anyone, or if the S/T/I co-operation lets them produce for their own purposes but not farm out.
IBM has no internal use for CELL. It is strictly a "We will license our design to SCEI for money" deal".

Huh???

If I'm remembering correctly, I've read on numerous press releases that IBM intends to fab CELL chips at Fishkill. That depends if they have a customer to purchase them. Sony won't buy any since they want to fab it themselves. But Nintendo, Microsoft, or any other company can purchase them if they've got the money ready to hand over to IBM.
 
It's quite possible and in this case probable that Sony has exclusive rights to use any co-developed technology in a specific arena (say video game systems). IBM would still be able to supply chips for network devices, super computers or very powerful toasters, but not for products seen to compete with PS3. Of course without the exact contents of the legal agreement, there is no way to tell.

I'd be surprised if MS or Nintendo were using Cell per se.
 
Re: ...

Brimstone said:
If I'm remembering correctly, I've read on numerous press releases that IBM intends to fab CELL chips at Fishkill.
Can you find any of those? I've seen Fishkill mentioned in CELL-related releases, but not directly applying to chips they plan to fab there or specifically-mentioned devices that will use them.
 
Re: ...

cthellis42 said:
Brimstone said:
If I'm remembering correctly, I've read on numerous press releases that IBM intends to fab CELL chips at Fishkill.
Can you find any of those? I've seen Fishkill mentioned in CELL-related releases, but not directly applying to chips they plan to fab there or specifically-mentioned devices that will use them.

Link

"We're defining the next era of computing, providing the technology that will bring computer intelligence and network access to a wide array of consumer electronics," said Dr. John Kelly, senior vice president and group executive for the IBM Technology Group. "As a result, IBM's advanced chip technologies are in more demand than ever. We expect a considerable portion of our new, state-of-the-art 300 mm wafer manufacturing facility in Fishkill, N.Y. to be dedicated to this product."
 
Re: ...

IBM engineers expect to try out part of their design in silicon this summer. Final chips are due in 2004. IBM, Sony and Toshiba have committed $400 million to the project over five years. Toshiba and IBM will make the chips, and IBM will push variations of the design into as many applications as it can. "Cell is very core to our processor and networking chip strategy going forward," Kelly says.
 
Re: ...

Deadmeat said:
Power5 already has an excellent FPU and doesn't need the asm coding of Altivec to see a speed gain.

MS isn't going to use power5, that is pretty much a certainty. Power5 is a MCM design intended for SERVER use, it's very large, and it's very expensive.

This is why IBM doesn't do vector SIMD in its line of supercomputers.

To be quite honest, I don't think you have the faintest idea why IBM does, or does not use SIMD in anything.

How many Xbox games do you know that makes an extensive use of SSE?

Poor dodge. Answer my question!

1. Large number of APUs require an APU implementation using bare minimum transistor count.
2. Limited transistor budget means a short pipe design.
3. So EE3 can't clock high because of above two combinatins.

Circular reasoning. Invalid.

Let's say each APU burns 3 watts @ 1 Ghz

Let's say you're just making up. You're grabbing a figure out of the air, do some bogus math, and come up with...NOTHING. Your numbers doesn't mean anything because they're not based on anything! GIGO, you know? Do you even understand this yourself, that your quote up there's just some made-up with no basis in reality?

Talking to you is pointless, you have no knowledge. You have no training or education in the field (or else you'd said so when repeatedly asked about it), and you have no insight! You just make up to try and justify your already made-up, outlandish opinion.

CELL simply can't clock high as Pentium4s and superpipelined PPCs.

..."Because I say so." Like *I* said: pointless. You're making up to fit your opinion. Nothing more.

IBM has no internal use for CELL.

...Because you said so? LOLFL, this is utterly pathetic. You're most easily summed up in one word: HUMBUG.
 
why are we talking about cell in a thread about xbox ?


Lets drop the cell talk and stick with the topic at hand .

If you can not do that I will lock this thread.


As for those asking why this thread exists . Well we allowed many rumor threads bout the ps3 cpu and we will allow them with any other system that has rumors attached to them.

So if you try and post again like that your posts will be deleted.
 
ERP said:
It's quite possible and in this case probable that Sony has exclusive rights to use any co-developed technology in a specific arena (say video game systems). IBM would still be able to supply chips for network devices, super computers or very powerful toasters, but not for products seen to compete with PS3. Of course without the exact contents of the legal agreement, there is no way to tell.

I'd be surprised if MS or Nintendo were using Cell per se.


Well the Team Xbox article stated that work on the Xbox 2 CPU just started when the contract was announced, so I don't think it would be CELL as Sony knows it.

Maybe they could do something like Sun did with MAJC, but instead of Java in mind, they design a CPU with .net as the intended target. Some CELL ideas could be included if needed along with inspiration from MAJC itself.
 
Massive heat generation from 16 APUs and short pipe design makes it difficult for EE3 to clock high.

1. Noone can make the claim of 16 APU's, not even you. As it stands, Sony plans for BE to have 32APU's.

2. The name of the PS3 Cellular IC is the Broadband Engine as outlayed in the Toshiba, Rambus, SCEI contract not EE3.
 
jvd said:
why are we talking about cell in a thread about xbox ?
Because when talking about chip speculation and comparing many IBM chips--especially advanced process and what's gearing up for 2005 and beyond--one may wander a lot. (Various PPC's, Power 4-6, CELL, etc... All of which came up.)
As for those asking why this thread exists . Well we allowed many rumor threads bout the ps3 cpu and we will allow them with any other system that has rumors attached to them.
There was all of one person mentioning that, and if you notice he was summarily ignored. ;)
 
MS isn't going to use power5, that is pretty much a certainty. Power5 is a MCM design intended for SERVER use, it's very large, and it's very expensive.
Lose the cache and the core size drops to around 120 mm2 @ 90 nm.

To be quite honest, I don't think you have the faintest idea why IBM does, or does not use SIMD in anything.
That's because you don't have any idea what the requirement for FPU heavy computing is.

Circular reasoning. Invalid.
Why don't you come up with a better reasoning if you are unhappy with mine?

Let's say you're just making up.
You are right, I gave you a hypothetical number.

You're grabbing a figure out of the air, do some bogus math, and come up with...NOTHING. Your numbers doesn't mean anything because they're not based on anything!
OK, Let's base my calculation on something real.

PSX2OAC = 8 watts @ 300 Mhz
PSX2OAC = probably ~30 watts @ 1 Ghz
PSX2OAC x 8 times the execution units = ~240 watts.

See, my 3 watts per APU was a very optimistic guess, because I don't believe the actual CELL APU will hit such number.


Do you even understand this yourself, that your quote up there's just some made-up with no basis in reality?
YOu sound so much like Grall. Have you returned under a different name after getting banned from here??? In that case, mods have their hands busy.
 
cthellis42 said:
jvd said:
why are we talking about cell in a thread about xbox ?
Because when talking about chip speculation and comparing many IBM chips--especially advanced process and what's gearing up for 2005 and beyond--one may wander a lot. (Various PPC's, Power 4-6, CELL, etc... All of which came up.)
As for those asking why this thread exists . Well we allowed many rumor threads bout the ps3 cpu and we will allow them with any other system that has rumors attached to them.
There was all of one person mentioning that, and if you notice he was summarily ignored. ;)

I should have said why are we fighting about cell things in a thread that has nothing to do with it . Why don't you save the argueing for a thread that focuses on cell . Not on the chips in the xbox 2 .
This thread is quckly becoming a flame thread.
 
Status
Not open for further replies.
Back
Top