Freescale Confirms Dual Core G4

[Analog Kid]

You're wrong.

Depends on how it was meant-- you need a dual CPU aware OS to use both cores, but the app can still be single threaded. Each app will use only one core at a time but will still see the benefit of shifting other apps and system tasks to the other core.

 

[dekator]

im not getting a computer with a g4 in it. g5 has been on the market for more than a year its the way to go

I think one shouldn't get too attached to names. The chips Freescale is proposing are pretty much new chips. Call them G5m (mobile) if you like. They will have most of the benefits of a 'real' G5 anyway, minus the enormous power consumption and heat dissipation. They pretty much are the future 'mobile G5s' if you will.

 

[isgoed]

Now that Freescale has revealed their new chips, it seems obvious what Apple will do with these. The 7448 will go into the Powerbook, iBook, and eMac as soon as possible next year, then the 8641D will go into the Powerbook when it is available.

Unless IBM has a low power G5 planned to be released soon, Apple would be smart to use the 8641D in the Powerbook instead. Since it is a two core 1.5+GHz G4 with two 1MB L2 caches, two slightly improved Altivec units, integrated memory controller and ethernet, up to 667MHz bus, support for DDR and DDR2 memory, PCI Express, and operates at 15-25W. With the elimination of the current slow G4 bus and being dual core it will be faster than a single G5 anyway.

But since these are 90nm processors, will there be issues with actually producing them in sufficient quantity? Freescale/Motorola does not have a good track record and with everyone else having problems with 90nm, its hard to be optimistic.

And for those who feel there is some urgent need for a 64bit laptop because you somehow need and expect to fit more than 4GB RAM in it (yes, I know that 2GB and 4GB dimms exist), Freescale will have its 64bit e700 in the future, maybe by sometime in 2006(?). They are likely to continue their goal of low power chips, which is good news for Apple laptops.

I agree with you completely. It seems that the longer the thread gets, the smarter the people are that are on it. At first there are only the G5 yellers/complainers.

When I read the article at first I was under the impression that they were talking about a near finished product. It amazed me because that would mean that freescale is the first company to make consumer dual cores. Seems waiting is just a thing mac users have to do.

The dual core is really interesting for laptops. It can be really power efficient if they can dynamically shutdown a core.

 

[kotovasii]

PB G5 is the answer!

I sure hope they figure out a way to put a G5 in the powerbook......

-Joe

Totally agree with this – I personally know a lot of people who are waiting for the G5 PB as they would need this processor for bioinformatics calculations. PB is designed for professionals and they need the best, otherwise people will have to “switch back” to a PC platform and fight with dual boot (M$ and Linux) boxes. I certainly would not buy PB G4 even if it is a triple processor. Until then it is Linux laptop and Power Mac desktop… very sad actually.

 

[Analog Kid]

Totally agree with this – I personally know a lot of people who are waiting for the G5 PB as they would need this processor for bioinformatics calculations. PB is designed for professionals and they need the best, otherwise people will have to “switch back” to a PC platform and fight with dual boot (M$ and Linux) boxes. I certainly would not buy PB G4 even if it is a triple processor. Until then it is Linux laptop and Power Mac desktop… very sad actually.

Linux on an Opteron laptop with 4GB of RAM?

 

[thatwendigo]

Totally agree with this – I personally know a lot of people who are waiting for the G5 PB as they would need this processor for bioinformatics calculations. PB is designed for professionals and they need the best, otherwise people will have to “switch back” to a PC platform and fight with dual boot (M$ and Linux) boxes. I certainly would not buy PB G4 even if it is a triple processor. Until then it is Linux laptop and Power Mac desktop… very sad actually.

Just what is it that 'a lot of people' need a portable G5 for that a dual-core with a vector unit couldn't do? For one thing, it's entirely possible to do 64-bit math in a 32-bit processor, but having the inherent ability to use integers of that size tends to speed things up if your code is written to do so. There's absolutely no reason that a G5 is automatically "the best" for a portable application, especially not if the application is threaded and can be fed to more than one processor. The extremely low latency and cache coherency in the 8461D would slaughter a single G5, as is pretty clear when a single 7447A 1.5ghz keeps up with lower clocked single 970s in benchmarks.

Something that I see very little in the way of commentary on... The 8461D is rigged to share the caches between processors, allowing any data in multiple threads to remain on chip instead of hitting main memory. That's incredible! The speedup for tasks that use a lot of repetitive calls and instructions ought to be amazing, and that's something that should help any program written to spread out across available cores. It has four times the cache of the current 970 implementation, and can use that for either processor, one processor, or both.

Then there's the fact that they have an on-die DDR and DDR2 memory controller, with support for 667mhz DDR2 RAM out of the box, internally enhanced networking performance with hardware routing for TCP, UDP, VLAN, IPv4 and IPv6, RapidIO, PCI-Express built-in... This thing will likely stomp on the G5, especially in portable formats, since running both cores full-bore draws only 15-25 watts, while the 1.6ghz 90nm 970 pulls as much as the whole 8461D.

 

[Pax]

Engineering effort

Nice to see people starting to believe a dual core G4 is the way to go for laptops. I'd genuinely prefer one of these over a G5 laptop. Remember, Freescale is streets ahead of IBM in the embedded space. Embedded is where performance per watt matters. IBM's "big iron" approach is just about performance. The G5's a great desktop chip, but not a mobile chip.

I wonder whether migrating to a dual core G4 would cost Apple less in engineering effort than building up a G5 laptop? Could they quickly hack the current PowerBook architecture to take the new dual core / on die memory controller chip? It's not a "drop in" replacement but could they quickly design the boards to take the new parts? If so we could see the dual core chip say in mid 2005.

 

[ts1973]

I think dual core is a good path to follow, the only thing that worries me is the time frame : that doesn't look too good...

Meanwhile, an old dual 1,42Ghz PM is about as fast as a iMac G5 today, runnning with slow bus speeds, so a 667Mhz bus and 1Mb cache would allow for much more headroom for a G4. However, by the time the chips come out the iMac will probably be cooled by Water Cooling while running above 2Ghz...

Did anyone notice by the way the dual core chip requires ECC memory to run @ 667Mhz bus. Can anyone explain the reason why ?

 

[G4-power]

The 7445 and higher (excluding the 7448) have support for a 167 MHz frontside bus, but no support for DDR.

I believe the 7400 was introduced at 0.25 microns, though I forget if the die shrink came with the 7410 or the 7450.

Believed accurate- correct me if I'm wrong

The earlier 74xx processors are 180 nanometer (0.18 micron), even the 7400, as is on my Sawtooth machine (350 MHz, see signature.) It is true, that the latter ones that you said are indeed 0.13 micron. The 7410 IIRC is (still in limited production) still 0.18 micron. I don't know about what bus speeds they support, but 167 MHz is the top. Don't know about DDR support either, but remember that new G4 Macs have DDR memory. Apple-History.com (website) and Mactracker (application) have good info about which Mac has which processor and when DDR came to Macs.

Just my opinions, I might be wrong.

 

[G4-power]

Thank you-- finally someone who sees the world like I do... Keep two lines of chips-- best of both worlds.

Yep, I agree with you there. Mobile and desktop proccessors need a different kind of architecture, raw power and power management is not the same thing.

Look at what they did with the "fat binaries" to transition from the 68000 to the PowerPC-- talk about jumping through hoops to support outdated hardware!

Yep, and Tiger will introduce "fat binaries" again, to support both 32- and 64-bit application in one package.

Yeah, from Apple's website Tiger Preview:64-bit

Write Chameleon Code

Tiger simplifies software distribution with support for Fat Binaries, an application that contains both 32-bit and 64-bit binaries within a single file. Using Fat Binaries, network administrators distribute a single version of an application to all users regardless of their system capabilities. Once installed on a user’s system, the Fat Binary automatically selects the appropriate code for the system without user intervention. This greatly simplifies administration, installation and distribution of applications.

 

[MikeBike]

2 is bigger than 1

Totally agree with this – I personally know a lot of people who are waiting for the G5 PB as they would need this processor for bioinformatics calculations. PB is designed for professionals and they need the best, otherwise people will have to “switch back” to a PC platform and fight with dual boot (M$ and Linux) boxes. I certainly would not buy PB G4 even if it is a triple processor. Until then it is Linux laptop and Power Mac desktop… very sad actually.

Isn't the software for bioinformatics calcs using Altivec?
Wouldn't 2 altivec processors, in a Powerbook, be better then 1?
Isn't Motorola's G4 Altivec processor stronger than IBM at the current time?

I know a Lot of Java programmers waiting for this machine:
dual core: the "G4 Extreme".
I'll buy a new powerbook as soon as this comes out.
A Dual Core Java on an Apple OS X machine would be very very nice.

- Next Powerbook looks like a single 1.8 with 200mhz fsb in 1 to 3 months.

- The Real Battle will be in 8-10 months:
G4Extreme vs. G5-65nm.
If I were Apple I'd build the G4 Extreme.

And I'd be giving IBM guidance that they want to research a dual core G5M
at whatever is after 65nm.

 

[h4ckintosh]

Anybody think there'll be a CPU replacement program, allowing current G4 users to recieve an upgrade?...kinda of a stupid question, I guess

 

[stockscalper]

I for one would love to see this dual core processor in the next Powerbook. The G5 frontside bus had to be severely crippled to contain heat in the imac and I can only imagine how much it would have to be crippled to get it to work in a Powerbook. This chip will give the most power for Powerbook portability and Mac users should welcome it. My only hope is that they offer speeds over the 1.4 GHZ quoted.

 

[shamino]

But since these are 90nm processors, will there be issues with actually producing them in sufficient quantity? Freescale/Motorola does not have a good track record and with everyone else having problems with 90nm, its hard to be optimistic.

IBM and Intel both hit serious problems with 90nm chips (and IBM doesn't seem to have completely solved those problems yet.) On the other hand, Freescale is starting late enough that they can study what IBM and Intel did to fix their problems and possibly avoid hitting them themselves.

 

[shamino]

While I might understand a little G4/G5 confusion from the masses, I really can't understand why so many people who have put in the effort to find these forums and have access to the knowledge of some pretty knowledgeable folks here still seem to think they'd rather have a slower G5 instead of a faster G4... It's like the perception of performance is more important than the real thing.

[SpinalTap]But this one goes up to five![/SpinalTap]

 

[shamino]

64bit workstations and full 64bit OS's are only used in very, very specialized applications: ones that need a lot of memory.

Well, Solaris ships a 64 bit OS that runs on all UltraSparc machines (which are the only kind they've sold for many years - all of which are 64-bit). But it does not require you to run 64-bit apps - and Sun recommends that most apps remain 32-bit unless they have a good reason to go 64-bit (like large memory, large-integer arithmetic, ANSI-C access to large files, or large system constants.)

But I don't know if you consider Solaris to be a full 64-bit OS or not.

My work uses a mix of 32bit and 64bit machines. We've got applications that run for weeks at a time processing data. Whenever possible we run those on 32bit machines because they're faster. We only run on 64bit machines when the data model exceeds the memory limit of the 32bit machines.

Just out of curiosity, are these 64-bit machines Sun workstations? And if they are, have you tried simply compiling the code as 32-bit to run on those workstations? Is 32-bit code on a 64-bit Solaris actually slower than 32-bit code on 32-bit Solaris?

Look at what they did with the "fat binaries" to transition from the 68000 to the PowerPC-- talk about jumping through hoops to support outdated hardware!

Actually, fat binaries were trivial. 68K apps store the code in CODE resources in the resource fork. PPC apps (are supposed to) store the code in the data fork (the way apps on most other operating systems do). To make a fat binary, you just compile the code twice - once for 68K and once for PPC. You store the PPC code in the data fork and the 68K code in the resource fork. At launch time, 68K systems will only look in the resource fork - and they'll find the 68K code; PPC systems will look in both places, trying the data fork first - and they'll find the PPC code.

OS X's package-based application system makes this even simpler, though. In this model, each application is actually a directory tree of files. Among the files is the executable image. It is possible (even simple) to put multiple executable image files in the package. The OS looks for one (in the directory named "Contents/MacOS") and launches it, ignoring the others.

So, if Apple would want to allow 32-/64-bit fat binaries (as they probably should), they can simply define a new folder name (maybe "MacOS64"). A 32-bit system will ignore this folder. A 64-bit system will look there in addition to "MacOS" to find executables. (The choice of which to check first can be made configurable on a system-wide or per-application basis, although it's probably not necessary. Presumably, developers wouldn't release a 64-bit executable image in an application unless the app has a legitimate need to be 64-bit.)

When dual cores are ready, that will be the distinguishing feature between Powerbooks and iBooks.

Until dual-core chips become cheap and abundant. Then the iBooks will get them and some other feature will distinguish them.

Note how the dividing line used to be G3/G4. Then the iBook/G4 came out, and the differences became pretty minor. (Today, there isn't much difference between an iBook and a 12" PowerBook, for instance.)

Powerbooks will continue with the dual G4 until memory limitations become a problem for the majority of users-- not the folks who want bragging rights, but the folks who make smart buying decisions based on all the factors in their portable. When that happens we'll see a move to a dual 64bit core. Maybe from IBM, or maybe the e700 series if it materializes.

Or until IBM gets their yields up (and temperatures down). Sometimes you upgrade a machine simply because it doesn't cost that much to do so.

Of course, by then, the PowerMacs will have dual-core G5's, so there will still be a distinction between the lines.

 

[shamino]

IWhen I read the article at first I was under the impression that they were talking about a near finished product. It amazed me because that would mean that freescale is the first company to make consumer dual cores. Seems waiting is just a thing mac users have to do.

They may well be first.

IBM has had dual- (and 4- and cores for quite some time in their POWER processor modules, but those are anything but consumer.

Intel is working on dual-core, but I don't think they're close to being able to ship anything either.

The way I see it, it's anybody's game. We could see the first consumer dual core processors from Intel, AMD, Freescale or IBM. Everybody's working on it, and everybody says "real soon now".

 

[MikeBike]

Anybody think there'll be a CPU replacement program, allowing current G4 users to recieve an upgrade?...kinda of a stupid question, I guess

- There's never been a cpu replacement program, from anyone that I know of. How do you think they got so rich.

You might be able to get a motherboard / cpu upgrade, if you can find a dealer willing to do it. Would it be legal?
A late model Powerbook 1.5 upgrade to the future 1.8(?) with the 1 mb l2 cache might be feasible, if Apple doesn't change any of the interface locations.

 

[shamino]

Totally agree with this – I personally know a lot of people who are waiting for the G5 PB as they would need this processor for bioinformatics calculations. PB is designed for professionals and they need the best, otherwise people will have to “switch back” to a PC platform and fight with dual boot (M$ and Linux) boxes. I certainly would not buy PB G4 even if it is a triple processor. Until then it is Linux laptop and Power Mac desktop… very sad actually.

How many biotech labs do you know of where the scientists want to (or are even allowed to) do their research on laptops? I would think that corporate security policy would demand that all research be done on machines that don't leave the premesis.

Could you imagine explaining to your boss that the simulation you spent months working on no longer exists because the laptop it was running on got stolen at the bus station?

I certainly would not buy PB G4 even if it is a triple processor.

ROTFLMAO.

So you'd refuse the computer, no matter how fast and powerful it might be, based entirely on whether Apple marketing puts the name "G4" on the box or not?

I hope you're not making any importat decisions where you work.

 

[AidenShaw]

Although G4 can address 32-bit physical addressable memory space (for total of 4 GB), operating systems such as Mac OS X can address up to 31-bit (for total of 2 GB).

Actually, the current G4 chips have 36-bit physical addressing and can support up to 64 GiB of RAM. (Same memory limit as 32-bit Xeon chips.)

The second half of your sentence is true - Apple hasn't supported more than 31-bit addressing, even though Windows and Linux support the 36-bit addressing on the x86 chips.

 

[jbro]

It seems that the longer the thread gets, the smarter the people are that are on it.

I agree.

And, I would love a dual-G4 laptop, especially if it's thin, light and efficient. My G5 tower is a heat machine. What kind of battery life do you want your laptop to have anyway?

It's all about the marketing. Apple would probably have a hard time selling people on a G4 powerbook without disparaging their current G5 lineup. Once they open the G5 can of worms, people start to expect it.

 

[nutmac]

I just don't get the need for the portable render farm. With widely available, high-speed internet connectivity, VPN and remote terminal capability, its a hell of a lot easier to just have a massive set of machines you access remotely. Hell, for *real* video production, you'd have to carry around a G5 sized disk array just to provide the storage and bandwidth. For most mobile warriors, their laptop is for communication and documentation, not in-the-weeds work. Most companies won't let you get a non-issue piece of hardware anywhere near their networks, much less let you put their intellectual property on it...

Just an observation. At some point, enough's enough...

In computing, there is no such thing as a computer being too fast, having too much disk space, or too much memory.

Render farm isn't the only reason to have faster performance and large memory. And the simple fact is, current PowerBook G4s lag seriously behind Pentium-M notebooks.

Consider some of my computing needs, for instance. I do application development by day. Running IDEs such as Eclipse and few other development tools, I easily consume a gigabyte within few minutes. Doing extensive debugging session, the memory footprint quickly skyrocket. Oh, and that's not counting iChat, Yahoo! Messenger, Safari, Mail, and Word that I would like to run on the side.

In the evening, my computing needs is much more consumer-level, but even so, I would be helped by more memory and additional performance. Browsing a message board such as MacRumors, I like to tab-open all the topics I like to read, then proceed to read one-by-one. And running Mail, instant messengers ,and iTunes, I easily consume a gigabyte (my PowerBook is maxed out at 2 GB). If I run iPhoto, iMovie, or GargareBand on top of that, I quickly wish for more memory and faster performance. Yeah, I can close all the application but one, but it can be annoying to re-open the applications later to find where I left off (I wish Safari, etc. remembers the last session).

The point is, PowerBooks need to be at least as fast as Pentium-M based notebooks. I don't care if they use Freescale or IBM CPUs, just as long as PowerBooks run much faster than they are now, with more competitive battery life, high resolution LCD, and can be expanded with more than 2 GB of memory.

 

[Eric_Z]

In computing, there is no such thing as a computer being too fast, having too much disk space, or too much memory.

Exept in mobile computing where you think about power drain vs preformance, do remember that "more RAM==less battery time".

 

[Pedro Estarque]

Exept in mobile computing where you think about power drain vs preformance, do remember that "more RAM==less battery time".

actually it doesn't. Having more RAM, especially in a system as cache hungry as OSX, means much less access to disk which consumes a LOT more then RAM.

 

[Analog Kid]

Just out of curiosity, are these 64-bit machines Sun workstations? And if they are, have you tried simply compiling the code as 32-bit to run on those workstations? Is 32-bit code on a 64-bit Solaris actually slower than 32-bit code on 32-bit Solaris?

Hodgepodge of Sun, Intel and AMD... Yes, we do run 32bit apps on 64bit Solaris which is faster than running on 64bit Solaris, but running on 32bit Solaris is still faster even on a 64bit UltraSparc.

Actually, fat binaries were trivial. 68K apps store the code in CODE resources in the resource fork. PPC apps (are supposed to) store the code in the data fork (the way apps on most other operating systems do). To make a fat binary, you just compile the code twice - once for 68K and once for PPC. You store the PPC code in the data fork and the 68K code in the resource fork. At launch time, 68K systems will only look in the resource fork - and they'll find the 68K code; PPC systems will look in both places, trying the data fork first - and they'll find the PPC code.

You've left out the bit about emulating 68000 code on a PowerPC which is very far from trivial... My point is that Apple has gone to extraordinary lengths to support legacy hardware in the past-- as you point out this will be a much easier task in OSX.

Until dual-core chips become cheap and abundant. Then the iBooks will get them and some other feature will distinguish them.

Note how the dividing line used to be G3/G4. Then the iBook/G4 came out, and the differences became pretty minor. (Today, there isn't much difference between an iBook and a 12" PowerBook, for instance.)

I think it will take more than a supply change to put the duals in the iBook. This may happen if there turns out to be no other route to keep the iBook a functional machine, but I think that will be some time out. I think the iBook went to G4 because the Altivec has really become necessary for media apps, and I think it would take another architectural change like this to bring the two into alignment again-- or simply an inability to increase speed by clock rate alone.

If I were a marketing type, which I'm not, I'd want to differentiate my product lines as much as possible and avoid the current iBook/PB12 confusion.

Of course all of this is assuming we'll see a dual core laptop-- which I think is likely given the announcement, but hasn't happened yet.

Or until IBM gets their yields up (and temperatures down). Sometimes you upgrade a machine simply because it doesn't cost that much to do so.

Of course, by then, the PowerMacs will have dual-core G5's, so there will still be a distinction between the lines.

Again, my thesis here is that different chips are better for different applications. I think 64bits just won't be right for a laptop until the memory is needed, and when it is I think there will be a better portable solution than the G5. The philosophy of the G5 seems to have been performance at the expense of power, which doesn't seem right for a laptop. You can bring the power down by tweaking it, but you could still get a better laptop design with a different architecture while keeping performance degradation limited.