Powerbook and 970 - 7457

hotboxd

Sorry if this has been posted already, search is down.



Anyways, a lot of people have been talking about the likelyhood that we will see a PPC970 Powerbook this year, due to the low wattage of the slowest processors. Others have been talking about how the 7457 would be a pretty awesome laptop processor. This is purely hypothetical, but imagine that the next revision of the Powerbooks marks the introduction of the 7457 running at 1.3ghz. Would this point to the Powerbooks sticking with the G4 for an extended period, and the 970 staying exclusive to the PowerMacs for a while? Or would Apple concievably release a 970 Powerbook in the revision after the 7457?



I've heard that the 7457 is a drop in replacement for the 7455, with no technical issues whatever. If this is true then Apple wouldn't have much problem releasing a 970 Powerbook in the next revision. But if the 7457 does indeed have a 200mhz FSB, wouldn't that require an asynchronous memory controller arrangement to interface with 333mhz RAM? Perhaps the northbridge already has this functionality built in? Does the current Powerbook mobo have a significantly different northbridge from the other DDR Macs? Continuing with that line, could the 970 interface with the current mobo, or would it have to be significantly redesigned (read, totally redesigned)? The fact that a new DDR motherboard was just released for the Powerbooks would seem to point to the G4 staying a while longer, but there always is the chance that it was engineered with the 970 in mind.



Mods feel free to lock this if it's too far into baseless speculation, but I'm just trying to figure out what might happen with the future of the Powerbook. My 500mhz iBook is getting long in the tooth, and I'm looking to get a Powerbook sometime this summer. The 970 would be such a quantum leap in performance that I feel it would be worth waiting up to a year for. Thanks.

fran441

There will be no 970 in the PowerBook this summer. It probably won't even be out in any machine this summer.



I think we'll see the 970 at the very end of this year, but more likely at MWSF next year (10.5 months).

brian green

It's really sad to hear Fran441 say that there won't be a Power Book 970 this summer. That means Apple won't be getting my money this summer. The G4 is dead...it just doesn't know it yet. The 970 is the way to go...it is the future. I was really looking forward to having a PB with a 970 this summer. I'm uninterested in G4 numbers, even if they do manage to bump significantly. We're talking about going from a prop engine to a jet engine here. A huge difference. More than double. Yet Apple wants to drop a G4 in the PB. Whatever. Blah, Blah, Blah. I'll hold on to my cash for a 970 PB.

bacon

Even though it sucks as a desktop chip, the G4 isn't too shabby for notebooks. I wouldn't be surprised if there's a 1.6 ghz 7457 in the powerbook by the end of the year that still gives best in class battery time.

matsu

Look at the PPC again when you look at laptops, it's a completely different game relative to the wintel competition. On the desktop Apple gets hammered, hard, especially from a price performance standpoint. But Mobile X86 is a very far cry from it's firebreathing desktop brethren. Even the lowly G3 still hangs with P4M of twice the speed. A linux journal benched an iBook and got those results. The fastest G4's run neck and neck with the faster P4M's and they really do run away with SIMD tasks. Seen any affordable Wintel laptops that can encode and burn a DVD internally in a reasonable amount of time? Nope, not unless they're using a desktop chip that gets about 90 minutes of battery life and has to be very agressively throttled to avoid melting itself.



Apple is still very much in the game when it comes to mobile performance, and they never really left it. If you had a smaller cooler G4 with more L2 and a health dollop of extra MHz, you'd have a pretty fast laptop, more than the equal of anything Wintel will sell you.



Apple needs to use all the PPC's it can get to release an ALL DUAL PPC970 lineup.

fran441

More logic as to why there won't be a 970 laptop this summer.



People are getting quoted ship times of April for their 17" PowerBooks. That means the channel most likely won't have any stock of these machines until at least April, possibly May.



I know that later in the year, some people will be saying, "the 17" PowerBook came out at MWSF so it's 7 months old" when in reality, it will only be out for 3 or 4 months to the masses.



I'm also being very conservative on when I think we'll see a PPC 970 shipping in any Mac. We also don't know if Apple will drop it into a PowerBook immediately either. Just things to think about.

brian green

Matsu, I certainly don't deny such performance benchmarks when it comes to PPC v.s. P4M. I think the thing that really gets me going is the entirely new architecture in the 970. I realize that Power Mac's need the 970 desperately, and someone even suggested that Apple drop in a Quad for high end just to shock people. I wish I were in a position where I could buy a Power Mac and use it all the time (I've got more than enough cash stashed to get everything I'd need and a huge screen) but I need portability. I currently use a 600 MHz iBook. It is extremely slow and I need a big increase of power to handle Final Cut Pro and the like. I know that the G4 could be throttled and still keep a lower wattage. But the 970 is where the performance is. So I'm going to have to deal with the iBook I currently have in order to wait for the 970 in the PB. When OS X goes to 64bit, I'm not going to want this G3 running the show. Sure it's almost the same as a G4 without Altivec, but a G3 is a G3. The new architecture of the 970 is the way to go, and I'm just going to have to hold out a few revisions of the PB 15 before they finally decide to incorporate the 970 into the PB, even though things are so slow with this iBook. Any idea on when the 970 will actually hit the PB? Hell I'd happily give Apple $1000 extra just to get one now. I don't care at all about the money. All I care about is performance I can have on the road.

hotboxd

Basically what I was wondering was how long the Powerbook would stay with the 7457. It sounds like everyone is talking about the 970 going into Powerbooks when IBM has a sufficient number of them. If they start fabbing in June or July, probably spring next year would be a good time for 970 Powerbooks. Or Apple has some sort of engineering problem with putting them in Powerbooks, and we'll have to wait for the 970 to start fabbing with the 0.09 process. That would be probably sometime in the 2nd half of next year. It doesn't sound like there should be a problem putting 970s into Powerbooks though ; the stats on the 1.2ghz part show it to be really power efficient.

drboar

The towers are the line that is suffering most of CPU deficiency so they will get the 970 first. Theoretically Apple could revamp everyting from the Xserver to the iBook to 970 at the same time, but it sound like a logistic night mare. My guess is that they are trying out the CPU in all lines but will introduce it the towers and that when that production is stable say 3-6 months they might introduce them in the portable.



If the 970 towers turn out as well as we all hope, more or less doubling the performance I think there is a large pentup demand for towers and the towers will sell like crazy. These are customers that have not been happy with Apple since the B&W G3 four years ago so Apple will not other 970 boxes if the supplies of some component is strained.

matsu

OSX will not be 64 bit for quite a while AFTER the PP970 arrives. It will load and run a completely 32bit version of OSX for many many months before 64 bit components are added. Relax, there shouldn't be any appreciable performance hit -- PPC was designed and planned with this in mind.

ompus

The G4 target is reasonable performance, cheap, at 6-15 watts.



The 970 is targeted at great performance at 20-40 watts (no word on just how cheap).



Now the 970, at its lowest power config, will certainly be *capable* of notebook duty. But it won't have great battery life. For those people who need to run their notebook for extended periods of time away from a power source, the best solution will still be- ack- the much maligned g4.



So we might see the 970 appear in notebooks that serve as "desktop replacements"- for example the 17", while the 12" powerbook maintains its status as the road warrior king using a low power 7457.



Put another way...I can envision a kick ass tablet with a 7457...I can't do that with a 970.

interface widget

A process shrink of the 970 early next year may make laptops with the 970 feasible only a short while after the 970 ships in desktops.



I believe the motherboard, chipset and surrounding components will prove to be a limiting factor for Apple in getting 970 based computers into the hands of customers. The availability of processors from IBM may not be the most important issue...

amorph

For a single processor 970 board, I don't think the engineering issues are significantly different than they are now: You have the 970, with its really fast (and really short) bus to its companion chip (rather than a G4, with its really fast and really short bus to its L3 cache). You have some other bus to RAM - say, HT or RIO. That should actually be cheaper than the trace-hungry MaxBus. If Apple replaces the AGP bus with an HT/RIO connection to a PCI-X slot, that's even more money saved. RAM doesn't really need to change. The current controller can already handle DDR RAM, async access to RAM, and DMA access to just about everything. If it becomes the "companion chip" in single processor systems, it seems to me that only the part that communicates with the processor bus would need to change. Apple's already got DDR in PowerBooks. The resulting board would be memory bound, but that's still an improvement over the current situation.



A 970-based board only becomes expensive when you bring SMP into the mix (especially more than 2-way SMP), because then you're implementing a NUMA architecture. But in a hypothetical PowerBook I think they can get away with a variation on the architecture they have. The biggest changes would be support for the Giga Bus vs. MaxBus, and HT instead of PCI/AGP.

jrg

[quote]Originally posted by Matsu:

<strong>OSX will not be 64 bit for quite a while AFTER the PP970 arrives. It will load and run a completely 32bit version of OSX for many many months before 64 bit components are added. Relax, there shouldn't be any appreciable performance hit -- PPC was designed and planned with this in mind.</strong><hr></blockquote>



What makes you think this?



Significant portions of pre Mac OS X operating systems ran on Solaris and HP-UX, both 64bit platforms. Getting the kernal to 64-bit will be a n easy step for Apple, supporting 64-bit processes, memory etc. The user visible applications, like Finder, Mail etc (probably) won't ever be a 64-bit app, neither will most of the supplied Applications, but the OS definately will from day one.



For further proof the file system is already 64 bit aware (off_t is 64-bit).

hotboxd

[quote]RAM doesn't really need to change. The current controller can already handle DDR RAM, async access to RAM, and DMA access to just about everything. If it becomes the "companion chip" in single processor systems, it seems to me that only the part that communicates with the processor bus would need to change. Apple's already got DDR in PowerBooks. The resulting board would be memory bound, but that's still an improvement over the current situation. <hr></blockquote>

I was under the impression that the controller couldn't handle async access if the RAM speed was lower than the bus speed. For example, I have a 133mhz 512MB DIMM in my 66mhz FSB iBook, but people who overclocked there MDD dual 867mhz bus to 167mhz had to upgrade their RAM to 333mhz to stay synchronous with the FSB, the 266mhz RAM didn't work.



Also, is the bus from the processor to the memory controller MPX all the way, or does it switch over to some other fabric before reaching the controller? I'm trying to figure whether the current controller would be at all compatible with the 970's GX bus.

matsu

There's still 4/5 other model lines to deal with. You won't get 64 bit OSX untill at least one more line moves on up to 64 bits. Adobe, and other's WILL get some lead time first. I'm sure it's there, but Apple won't give it up untill it's neccessary, which given the design of the PPC and the current apps, it isn't, yet.



Some 64 bits at first, and a whole slew later.

telomar

[quote]Originally posted by Matsu:

<strong>There's still 4/5 other model lines to deal with. You won't get 64 bit OSX untill at least one more line moves on up to 64 bits. Adobe, and other's WILL get some lead time first.</strong><hr></blockquote>



You are thinking that it's a bigger issue than it is for older computers. There's no reason Apple can't release a 64 bit operating system at launch and I expect the first public lead time will be at WWDC this year.

programmer

[quote]Originally posted by Matsu:

<strong>There's still 4/5 other model lines to deal with. You won't get 64 bit OSX untill at least one more line moves on up to 64 bits. Adobe, and other's WILL get some lead time first. I'm sure it's there, but Apple won't give it up untill it's neccessary, which given the design of the PPC and the current apps, it isn't, yet.



Some 64 bits at first, and a whole slew later.</strong><hr></blockquote>



This isn't a reason to not ship a 64-bit version of OS X immediately. The only reason not to do so is if its not ready, and while we don't know what state the OS is in these days internal to Apple, I'd be surprised if they were very far off. The OS would load the 32-bit or 64-bit components depending on the hardware it was launched on, so supporting older hardware isn't an issue. The applications will launch into the kind of address space they were written for, so existing and future 32-bit applications will launch just fine under a 64-bit OS. Really, there is no need for panic -- it'll all work fine.

programmer

[quote]Originally posted by hotboxd:

<strong>

I was under the impression that the controller couldn't handle async access if the RAM speed was lower than the bus speed. For example, I have a 133mhz 512MB DIMM in my 66mhz FSB iBook, but people who overclocked there MDD dual 867mhz bus to 167mhz had to upgrade their RAM to 333mhz to stay synchronous with the FSB, the 266mhz RAM didn't work.



Also, is the bus from the processor to the memory controller MPX all the way, or does it switch over to some other fabric before reaching the controller? I'm trying to figure whether the current controller would be at all compatible with the 970's GX bus.</strong><hr></blockquote>



I suspect we'll see a heavily redesigned chipset that works very well with the new FSB. The FSB, memory and I/O buses will probably all operate asynchronously and have enough buffering between them to avoid stall issues. Apple's current chipsets seem pretty well designed, and they've had access to the new FSB specs for some time now.



The current chipsets sit on the G4's MPX bus and connect directly to PCI, AGP, and the memory bus. We'll have to wait and see what the new chipset does but I'd guess that it'll connect the FSB, AGP, memory bus, and some form of connection to a southbridge which will have the PCI and any other I/O connections needed. Or Apple might surprise us all the come right out of the gate with a new "kitchen sink" chipset that does everything internally and interconnects all of the buses and I/O.

jrg

[quote]Originally posted by Matsu:

<strong>There's still 4/5 other model lines to deal with. You won't get 64 bit OSX untill at least one more line moves on up to 64 bits. Adobe, and other's WILL get some lead time first. I'm sure it's there, but Apple won't give it up untill it's neccessary, which given the design of the PPC and the current apps, it isn't, yet.



Some 64 bits at first, and a whole slew later.</strong><hr></blockquote>



I suspect we are saying the same thing but disagreeing over terminology. To me a OS is 64 bit when it supports a 64 bit address spcae (i.e. void * is 64 bits) and allows you to run applications that can execute the 64 bit instructions.



Every component of the operating system required to suppot 64 bit applications, and the kernal will be 64 bit from day one.



Plus what Programmer said

matsu

Yeah, not knowing anything about the techno bits hurts my ability to express meeself.