There's been some info, yes. And while it would indeed be of great benefit to a case like this, the Gustafson declined to comment specifically on when systems using the connection design might be rolled out, but said he would expect to see a commercial use of the technology "sooner than five years." part makes me think we'll have plenty of time to watch its progress and see how it turns out--and NOT be getting it in any mainstream 2005-2006 products.
Possible Xbox2 processor configuration?
- Thread starter bbot
- Start date
zidane1strife
Banned
Not sure, but if this is calc like normal area, I think twice the length will give you 4 times the area when squared.
After reading this post,
http://forum.teamxbox.com/showthread.php?postid=3235731#post3235731
this is what I think the Xbox2 cpu will be like. A single cpu with three Power5-based processor cores. The third core will be dedicated to 3D graphics calculations.
http://forum.teamxbox.com/showthread.php?postid=3235731#post3235731
this is what I think the Xbox2 cpu will be like. A single cpu with three Power5-based processor cores. The third core will be dedicated to 3D graphics calculations.
After re-reading what the so-called ms guy said, I realized that it could be done on a single PPC 976. This is what he/she wrote:
"I work for ms let me give the low down...
Basically, it will have a single CPU that will have Hyper-Thread II. Though, the GPU will work in sync with the third thread instead of the first and second thread. The algorithm breaks it into three.
Sort of like network duplexing which you can set two frequencies of lights over a optical cable one for upload one for download or for both upload and have them not interfere.
The first and second will be dedicated to calculations with matrices, vectors, floats, double longs, arrays, data and other stuff.
There's a fourth thread but I can't comment on that."
Fours threads! Exactly what a single PPC 976 is capable of. And that gameindustry.biz article says "four or more processors", so that could mean that Microsoft might add a second CPU, depending on what Sony does.
"I work for ms let me give the low down...
Basically, it will have a single CPU that will have Hyper-Thread II. Though, the GPU will work in sync with the third thread instead of the first and second thread. The algorithm breaks it into three.
Sort of like network duplexing which you can set two frequencies of lights over a optical cable one for upload one for download or for both upload and have them not interfere.
The first and second will be dedicated to calculations with matrices, vectors, floats, double longs, arrays, data and other stuff.
There's a fourth thread but I can't comment on that."
Fours threads! Exactly what a single PPC 976 is capable of. And that gameindustry.biz article says "four or more processors", so that could mean that Microsoft might add a second CPU, depending on what Sony does.
bbot,
That so-called MS guy makes totally no sense at all. First of all, hyper-threading, that's a trademark of intel if I'm not completely mistaken. Second, even if a CPU could do 1000 hardware threads, it would still only execute one of them at a time, so no practical speed gain after a certain point. The more threads you have the more your caches will thrash and the more erratic and unpredictably your code will run, as one thread will push others out of your caches. Relying on all games doing certain things in certain threads is going to mess seriously with software developers, as some of them do not work that way. Remember, the vast majority of all current PC games still do everything in one single thread. 3D engine, physics, AI, sound, etc etc.
It's going to require a total re-think of how games are written, and will smash easy ports from the PC (or other way around I suspect) to bits.
And Quincy,
Yes, Sony did put both major chips on one die, EVENTUALLY, after quite a number of revisions/shrink, ending up with a product where the sum was actually smaller than its two separate components. If you're talking of either three separate chips the size of x or one chip that is the size of 3x, the issue is quite different, wouldn't you say?
Having multiple chips on a board isn't that fricken difficult. Take ye average PC, you have multiple chips with multiple high-speed buses and interfaces connecting them with many hundreds, if not thousands in some cases, of pins distributed amongst them.
Glueless connections such as hypertransport etc makes interfacing multiple CPUs easier than ever before. PCs are much more complicated because they're bogged down with all this legacy crap. A console would be much simpler and cleaner to design.
That so-called MS guy makes totally no sense at all. First of all, hyper-threading, that's a trademark of intel if I'm not completely mistaken. Second, even if a CPU could do 1000 hardware threads, it would still only execute one of them at a time, so no practical speed gain after a certain point. The more threads you have the more your caches will thrash and the more erratic and unpredictably your code will run, as one thread will push others out of your caches. Relying on all games doing certain things in certain threads is going to mess seriously with software developers, as some of them do not work that way. Remember, the vast majority of all current PC games still do everything in one single thread. 3D engine, physics, AI, sound, etc etc.
It's going to require a total re-think of how games are written, and will smash easy ports from the PC (or other way around I suspect) to bits.
And Quincy,
Yes, Sony did put both major chips on one die, EVENTUALLY, after quite a number of revisions/shrink, ending up with a product where the sum was actually smaller than its two separate components. If you're talking of either three separate chips the size of x or one chip that is the size of 3x, the issue is quite different, wouldn't you say?
Having multiple chips on a board isn't that fricken difficult. Take ye average PC, you have multiple chips with multiple high-speed buses and interfaces connecting them with many hundreds, if not thousands in some cases, of pins distributed amongst them.
Glueless connections such as hypertransport etc makes interfacing multiple CPUs easier than ever before. PCs are much more complicated because they're bogged down with all this legacy crap. A console would be much simpler and cleaner to design.
I know PPC 970 but what is this PPC 976 ? I assume is a variant, but can someone telly me what it is exactly. Or PPC 970 has been renamed to PPC 976 ?
"four or more processors" would be capable of eight threads. Unless that command refers to virtual processors capable of a thread each.
originally from macosrumors.com:
http://rds.yahoo.com/S=2766679/K=po...acinsteins.org/newsletters/February_2004.html
"1/18/04
IBM 9xx processor shakeup could benefit the Mac: Multiple sources have now corroborated much of our recent reports about future IBM G5/G6 processors, as well as some of the details we have not yet shared.
* PowerPC 975 - POWER5 core, 90-nanometer/11-layer CMOS process, Hyperthreading, 3GHz+ (previously planned as PPC 980)
* PowerPC 976 - POWER5 Dual Core, 65nm SSOI (Strained Silicon On Insulator) process, VMX2 instructions, 4GHz+
* PowerPC 980 - POWER6 core, 65nm/11-layer FinFET/SSOI process, VMX2, 5GHz+
* PowerPC 985 - POWER7 Dual Core, 45nm, 9GHz+
* PowerPC 990 - POWER8 Multi-Core, 32nm, 15GHz+
There is as yet no finalized delivery date for any of the above processors, as IBM is currently rolling out its initial 90-nanometer 970+ with the new Xserve G5, soon also the PowerMac G5 and later in the year the Powerbook G5. The 975 will follow roundabout September, when Apple follows through on its promise to break the 3GHz barrier within the G5's first year.
1/19/04
More details on future PowerPCs. Some quick updates on the processor front:
The PowerPC 976, the first dual-core 97x chip based on the POWER5 architecture, will probably ship in mid-2005 and will also be the first PowerPC to use the VMX2 vector (“Velocity Engineâ€￾) instruction set; VMX2 will vastly increase the range of applications that will benefit from AltiVec/Velocity Engine-optimized code as well as the performance of that code. In particular, watch for Apple to tout VMX2’s impressive 3D graphics performance.
The single-core PPC 980 will be Apple’s workhorse high-end processor beginning in early 2006 and variants will probably still be powering low-end Macs until nearly 2010.
In late 2006, the PPC 985 will take over the high end of the Mac with a return to dual-core architecture based on POWER7. By this time IBM and Apple project that the cost to performance ratio of the PPC 9xx family will be no less than 5X that of any planned or otherwise likely competitor processor from Intel.
Although processors after the 985 run so far into the future as to be impractical to speculate upon, one notable feature mentioned in internal Apple documents is “the ability to emulate Intel architectures with performance no less than 2X that of native solutions.â€￾ Interesting....
"
http://rds.yahoo.com/S=2766679/K=po...acinsteins.org/newsletters/February_2004.html
"1/18/04
IBM 9xx processor shakeup could benefit the Mac: Multiple sources have now corroborated much of our recent reports about future IBM G5/G6 processors, as well as some of the details we have not yet shared.
* PowerPC 975 - POWER5 core, 90-nanometer/11-layer CMOS process, Hyperthreading, 3GHz+ (previously planned as PPC 980)
* PowerPC 976 - POWER5 Dual Core, 65nm SSOI (Strained Silicon On Insulator) process, VMX2 instructions, 4GHz+
* PowerPC 980 - POWER6 core, 65nm/11-layer FinFET/SSOI process, VMX2, 5GHz+
* PowerPC 985 - POWER7 Dual Core, 45nm, 9GHz+
* PowerPC 990 - POWER8 Multi-Core, 32nm, 15GHz+
There is as yet no finalized delivery date for any of the above processors, as IBM is currently rolling out its initial 90-nanometer 970+ with the new Xserve G5, soon also the PowerMac G5 and later in the year the Powerbook G5. The 975 will follow roundabout September, when Apple follows through on its promise to break the 3GHz barrier within the G5's first year.
1/19/04
More details on future PowerPCs. Some quick updates on the processor front:
The PowerPC 976, the first dual-core 97x chip based on the POWER5 architecture, will probably ship in mid-2005 and will also be the first PowerPC to use the VMX2 vector (“Velocity Engineâ€￾) instruction set; VMX2 will vastly increase the range of applications that will benefit from AltiVec/Velocity Engine-optimized code as well as the performance of that code. In particular, watch for Apple to tout VMX2’s impressive 3D graphics performance.
The single-core PPC 980 will be Apple’s workhorse high-end processor beginning in early 2006 and variants will probably still be powering low-end Macs until nearly 2010.
In late 2006, the PPC 985 will take over the high end of the Mac with a return to dual-core architecture based on POWER7. By this time IBM and Apple project that the cost to performance ratio of the PPC 9xx family will be no less than 5X that of any planned or otherwise likely competitor processor from Intel.
Although processors after the 985 run so far into the future as to be impractical to speculate upon, one notable feature mentioned in internal Apple documents is “the ability to emulate Intel architectures with performance no less than 2X that of native solutions.â€￾ Interesting....
"
Also known as the 970FX! (Oooh! Aaaah!) I'm really hoping companies discover a new labelling system soon... :?bbot said:
Its Dual Power5 Cores, so its 4 virtual processors.
So PPC 976 is what becomes 130nm Dual Power5 core chip that was shown last year, shrunk to 65nm for production. So at 4 GHz that would give 32 GFLOPS, not too bad indeed.
Mean while PPC 970FX is going to be used for the G5 server, which mean its 90nm Power4 core.
90nm PPC 970FX has size of ~60mm2 and power of 24.5 W with
clockspeed 2.0 GHz.
bbot said:
That was perhaps the biggest load of pipe-dreams I've seen in quite a while. Especially the last paragraph. That future PPC chip's going to emulate x86 chips at twice the speed of native solutions? Wow, gimme some of that shit now! My life needs some bright, flashy lies!
Jesus christ, I feel sorry for anyone actually believing that trash. It's one thing with a site dealing with RUMORS, but obvious UNTRUTHS, that only Deadmeat's allowed to spread around!
Guden,
It's not jsut the difficulty in implementing a motherboard with multiple sockets and chips, it's also the difficulty in keeping the cost down for tha tmultiple chip motherboard. As you know dual processor mother boards are expensive. Even more processor sockets would raise the cost of that board quite a bit.
It's not jsut the difficulty in implementing a motherboard with multiple sockets and chips, it's also the difficulty in keeping the cost down for tha tmultiple chip motherboard. As you know dual processor mother boards are expensive. Even more processor sockets would raise the cost of that board quite a bit.
Quincy,
Why is it so 'difficult' to have multiple chips? If you look at the Saturn for example, or early PS2s for that matter, you'll find consumer hardware with quite a number of (large) ICs on them sold at low prices.
Also, we know consoles often start out more complicated than later on in their life, that a machine might have multiple chips at launch doesn't mean it'll stay that way forever.
Btw, you brushed off (or more correctly, ignored) the point I brought up in my previous point where I said a situation with 2 chips added and ending up smaller than their two separate counterparts does not compare with a chip that is 3x larger than its individual components.
Why is it so 'difficult' to have multiple chips? If you look at the Saturn for example, or early PS2s for that matter, you'll find consumer hardware with quite a number of (large) ICs on them sold at low prices.
Also, we know consoles often start out more complicated than later on in their life, that a machine might have multiple chips at launch doesn't mean it'll stay that way forever.
Btw, you brushed off (or more correctly, ignored) the point I brought up in my previous point where I said a situation with 2 chips added and ending up smaller than their two separate counterparts does not compare with a chip that is 3x larger than its individual components.
Once again you're missing the point. I didn't say it was difficult to "have" I said it's difficult to keep the cost down ona mothe rboard that supports multiple chips, whcih is the exact reason those motherboards for PC are more expensive than regular boards. The sega saturn didn't have a choice as a process for multiple cores wasn't avaialbe at that time.
I know that. That doesn't mean however that if the process is available from the start and you can ruduce all those chips into one, that you wouldn't take that process. We all know IBm already has a process for using mutliple cores on a single chip, so I don't even know why you are arguing against that.
Who says it has to always be three times larger? I really don't even know why you are continuing to argue this as the xbox 2 will NOT have multiple chips, it will have one cpu with mutltiple cores on it.
I agree with Quincy on this one. IC's costs will [theoretically] decrease at a rate roughly bounded by Moore's Law - Sony has done this to great effect with the PlayStation(s).
Multiple IC's require added fixed costs (eg. Packages, PCB, cooling, etc) that don't scale. This is bad when you're in a price war. Actually, it's analogous on the surface to the current XBox model which has fixed costs as opposed to PlayStation's scaling in time.
Regarding multiple chips on the mobo, sometimes it isn't always possible to have everything you want on a single chip.
If this was true, Sony would have combined EE and GS and everything else on a single chip since the launched of Playstation2.
They'll just go which ever cheaper or possible for them at the time. When cheaper options available, I am sure they'll move to it.
Similar threads
