Power5 Processor Notes

[MacRumors]

CBROnline posts some details about the upcoming Power5 processor from IBM.

Bits of information about the Power5 has leaked out over the past few months, with an expected target date of Q2 2004. The Power5 will be the subject of the "Hot Chips" conference at Stanford this week. CBROnline provides some early details of the processor.

New features mentioned include more efficient "best in the industry" simultaneous multithreading (SMT), main memory controller on-chip, and dynamic power management. Initial Power5's are expected to use a 130-nm process with initial clock speeds between 1.5-2GHz.

While the Power5 may not find its way into Apple-branded machines, there have been rumors of Power5-based processors being developed that may be the succesor to the PowerPC 970.

 

[theipodgod16]

Here is to a long standing relationship with Apple and IBM! (Tata, Smotorola!!)

 

[Powerbook G5]

It's always good to hear positive things about Apple's future with a new chip technology.

 

[tazznb]

Power5 derivitive or bust

I'm wondering how the 980 will perform.

 

[yossele]

Better than the 970!

 

[Wonder Boy]

stupid best friend keeps hounding me that i dont have the money to pay for a g5. for some reason, after reading this article, i don't mind waiting for the power5.

 

[Dont Hurt Me]

I wonder why Arn didnt mention the 980 in this article. anyways apple is on its way again ( with IBM's help) and the world will be playing catch up. The best operating system has just met the fastest hardware. Why people still suffer windows is beyond me. I think its great g5 owners will be getting machines this week. yeah im a lil jealous.

 

[simX]

Originally posted by Dont Hurt Me
I wonder why Arn didnt mention the 980 in this article.

Well, he said, "there have been rumors of Power5-based processors being developed that may be the succesor to the PowerPC 970". However, since the PPC 980 hasn't actually been officially confirmed by IBM, I don't think there's any reason to mention it by name.

 

[impierced]

Originally posted by simX
However, since the PPC 980 hasn't actually been officially confirmed by IBM, I don't think there's any reason to mention it by name.

Yeah, and we all know this site doesn't post anything that's unconfirmed...

 

[cb911]

i thought that the 980 was going to be a derivative of the Power5, or is the 980 just a 970 made on a 90nm process?

 

[tex210]

i thought...

that they where being designed concurrently...

 

[Protean Son]

The 980 may not have been mentioned in a press release, but I swear to God I saw it on IBM Germany's website. I don't know how I came to view it there. It was months ago. As far as I know it is being developed at the same time, but they might never release it if for some unforseen reasons 970s dont sell well. Etc.

 

[GregGomer]

Well even if they didn't give it the name, IBM at the announcement made it very clear that BIG things were in the future.

Two things the IBM exec mentioned...

1. We (IBM) plan to have a 3 Ghz Processor with in 1 year of the release of the G5 PPC 970.

2. The replacement to the 970 was already in the works... (he didn't say G6 or 980, but said it was in the works and coming, so to me, that in and of its self is big news....)

 

[GroundLoop]

Originally posted by GregGomer
2. The predecessor to the 970 was already in the works... (he didn't say G6 or 980, but said it was in the works and coming, so to me, that in and of its self is big news....)

The 970's (Apple utilized) predecessor has been in the works for a very very long time...it's called the G4...sorry...I had to do that.....

Hickman

 

[Exponent]

The latest scoop from HotChips Conference....

Hi guys, Exponent reporting here live from the HotChips conference, and IBM is coming up right now with their Power5 presentation....

<P>
Talking about the simultaneos multithreading implementation in Power5

<P>
On target to ship in 04, running in labs now, started in 2000
<P>
What changed since power4 to get SMT in power5
<P>
3 fundamental probs: mem latencies (slow RAM), branch processing, and execution unit utilization (20-25% util usually seen in commercial code)
<P>
more to come.....

 

[Exponent]

(From HotChips conference)
SMT is designed to address these three issues

Power5:
130nm, Cu, SOI
Dual processor core
8 way superscalar
SMT core:
-up to 2 virtual procs per real proc
-24% area growth per core for SMT
-"natural" extention to power4 design
-going beyond 2 threads per core diminishes returns

 

[Exponent]

Review of different kinds of Multitreading:
-Single thread has low utilization
-coarse grain multitreading gives a second task to run when utilization is low
-Fine grain threading: problems with long latencies when resources are constrained
-SMT: no problems with constrained resources

(tech stuff follows, coming too fast to type)
How do they do SMT? Add a second Program counter, then register renaming had to be extended.

Group Completion Counter had been extended

(I'm not really sure MacRumors is the best venue for this, and it is coming too fast to type, but I'll try to summarize anything I think is cool....)

 

[Exponent]

Chaches had associativity increased rather than size (more efficient)

Power5 gives dynamic feedback of shared resources, and the machine takes action if resources are constrained - net effect is increased parallelism and increased thruput.

There are situations where inbalanced thread execution of threads is preferred - power5 gives more execution slots to threads with higher priority.

(This kind of runs counter to the idea that with SMT you get a second virtual processor for free due to resource constraints, but they're IBM, and I'm not....)

 

[Exponent]

SMT is a mode that doesn't have to be entered into (I think that's what they're saying)

SMT can make chip-bring-up (i.e. intial chip coming back from the fab debugging) a pain, so Power5 has goodies to help with this)

Dynamic feedback helps things
Chip can dynamically switch between single threading and SMT

40% execution speed improvements for SMT - good return on 24% increase in area.

SMT impacts are pervasive throught the chip.

AIX, Linux and OS/400 have been booted and running applications

End of presentation, taking questions

4 threads per real processor doesn't make any sense. 3 threads per real processor might make sense, but making needed resource changes would be difficult

Didn't see benefit of trace cache.

Speaker "not prepared to comment" on how much power went up, or what the increase of area due to SMT affected frequency.

End of questions

 

[Vlade]

Originally posted by Wonder Boy
stupid best friend keeps hounding me that i dont have the money to pay for a g5. for some reason, after reading this article, i don't mind waiting for the power5.

You can't use the power 5, but you may be able to use a Power 980 if it comes out. By why not wait on a Power 990, it should be at least 3X as fast as the 980.

 

[rjwill246]

Originally posted by Brian Hickman
The 970's (Apple utilized) predecessor has been in the works for a very very long time...it's called the G4...

That was then, MOTO.

This is now, IBM.

 

[MrMacMan]

Originally posted by impierced
Yeah, and we all know this site doesn't post anything that's unconfirmed...

But most of the time the stuff comes with 'uncomfirmed', 'possible' or even 'doubtful'



And thank you Exponent for the blow by blow!

 

[myrdred23]

>>up to 2 virtual procs per real proc

This sounds similar to Intel's Hyperthreading technology. Except this one can have 2 virtual cps for each real one (Intel currently can do 1 virtual per 1 real), meaning a dual 980 (assuming it inherits this technolgy) computer could make itself appear to have 6 processors!

This will speed up multitasking environments (Mac OS X ) quite a bit.

 

[AidenShaw]

terminology...

Originally posted by myrdred23
>>up to 2 virtual procs per real proc

...meaning a dual 980 (assuming it inherits this technolgy) computer could make itself appear to have 6 processors!

No, a dual POWER5 will appear to be four (4) processors.

Intel has a "logical" processor in addition to each real processor. This IBM guy is referring to two (2) "virtual" processors from one physical.

It's the same thing - a doubling of the apparent number of processors.


http://news.com.com/2100-1006_3-5065839.html

"The Power5, which has two processor cores, or brains, looks like four processors."

 

[Plutoniq]

It may be that this chip (or derivitive) will be used by Apple for use in Powerbooks. Considering the 970's lack of Dynamic Power Management, the inclusion of this on the Power 5 seems to hint towards portable application. The speed of the chip (1.5ghz) seems so line up well with where the 7457 G4's will finish off (1.33ghz). As far as the 130nm process, considering that the upcoming PowerPC 750GX & Moto's 7457 utilize the same die size (Both portable CPU's), theres no reason why the G5 has to come down to a 90nm process before it is able to be used in portables.

The lack of dynamic power management onboard the 970's seems to imply that it was never designed, and maybe never will be used for, a portable CPU.

P9