1.5ghz G4 Road Map

[Ensign Paris]

MacMinute just posted this, looks like good news for us!

I think we will see G5s San Francisco 2003 which is only 10 months away!

PowerPC G4 roadmap sees speeds of 1.5GHz
February 11 - 17:10 ET: The Register has published a look at Motorola's recently updated G4 roadmap, which calls for three new variations of the G4 to be developed over the coming year. While laden with technological terms, the article notes processor speeds as high as 1.5GHz, and bus speeds of 266MHz, with a new RapidIO technology that could deliver even higher performance.

[Onyxx]

10 months is a little depressing but if thats what it takes to deliver a FLAWLESS chip and a nice palateable price, so be it.

1.5 ghz g4's isn't doing anything for me. I want to see quad processor systems if we are going to be stuck with the g4 for a while. The os is built for it, many high end apps a designed for it. Come on apple, now you design a machine for it!

[Rower_CPU]

What concerns me most is whether or not we're going to see DDR and upgraded USB and Firewire anytime soon.

If we're stuck with the G4, Apple really needs to catch up with faster memory and a speeded up sytem bus.

[Ensign Paris]

Quad machines would be nice, what you could do is buy two dual1ghz and parallel work them.

[Ensign Paris]

Originally posted by Rower_CPU
What concerns me most is whether or not we're going to see DDR and upgraded USB and Firewire anytime soon.

If we're stuck with the G4, Apple really needs to catch up with faster memory and a speeded up sytem bus.

i agree totally! I would like too see an 800 verison of firewire. what the hell is gigawire it never has come to existence!

[Onyxx]

really really really don't have the money for that. some programs wouldn't work with a cluster setup (final cut pro, photoshop) so it wouldn't be that great for my purposes. Also the closeness of the processors allows the processors to work together more eficiently.

[Ensign Paris]

earlier today I had 34 macs (G4 500, d500 and d800s) linked with parallel via firewire networking. it was amazingly fast.

I look forward to work on wednesday.

having tomorrow off for quicktime live stream just for the hell of it!

[Choppaface]

dude where do you get those things? are they firewire hubs or something? where can I get one and how much do they cost?

[Onyxx]

fire wire networking? i don't think i've ever heard of it before. sounds interesting though. put out a link or something so i can get some more info.

[Catfish_Man]

...knew about this "roadmap" (they have a link to one, but it's quite old), I might actually believe this. As it is... maybe. A definite maybe.

[jefhatfield]

Originally posted by Onyxx
10 months is a little depressing but if thats what it takes to deliver a FLAWLESS chip and a nice palateable price, so be it.

1.5 ghz g4's isn't doing anything for me. I want to see quad processor systems if we are going to be stuck with the g4 for a while. The os is built for it, many high end apps a designed for it. Come on apple, now you design a machine for it!

...plus the quad g4 could be a realistic answer to the 3+ ghz chips intel and amd could have by the first of next year

i hope the average, non-techie consumer will not still be judging a computer by its clock speed...it seems to still be a hot issue, but has calmed down that many people have noticed their pentium 4s are not THAT fast

[AlphaTech]

Originally posted by Rower_CPU
What concerns me most is whether or not we're going to see DDR and upgraded USB and Firewire anytime soon.

If we're stuck with the G4, Apple really needs to catch up with faster memory and a speeded up sytem bus.

Ok, according to the start of this thread the 266MHz bus speed is what the systems that use DDR have. Also, a 266MHz bus allows you to run DDR memory as PC2100 (hellafast).

Whatever the 'RapidIO technology' is, it sounds fast (maybe deceptively so).

I can just imagine what a dp 1.5GHz G4 with PC2100 memory would do. I my opinion it would blow the doors off any systems with intel inside (someone mentioned before that 'intel inside' is really a warning lable).

Since I will be waiting for the G5 to get a new system, I will enjoy seeing the advances from the sidelines. I will also probably do something I have yet to do when purchasing a Mac... wait for reviews and real world opinions of them. I might even wait for the first update to the system (not sure if I will be able to wait THAT long).

Granted, I will probably have hands on experience with each revision, even if they are not top speed at the time. One of the benefits of being a tech at a company that buys Mac's for users. Looking back to last year, I think we purchased between 25 and 35 Mac systems (between G4 towers and iMac's). Considering how we didn't bring that many people on staff, many of them went to temps or to replace aged systems. We needed to get the pre-G3 systems off the floor since they were causing productivity isses finally.

[teabgs]

I personally hope that this rumor is true. If it is then we'll have DDR, a faster Bus and higher processor speeds. I'm waiting to buy a new computer to replace the 6500 im on right now. I'm a student and have a small budget (which I will spend a lot more) and so I want something really good cause it has to last me a while. If the Dual 1Gig had had DDR and a faster bus I'd have ordered one, but instead I'll wait for these features. Even if its not G5, I still want it. And you should too...these are pretty good upgrades.

[DNA]

...With this G5/G4 discussion. To be honest, I'm not a tech geek, so maybe I've got it all wrong, but it seems to me there are lots of things that they will change in the G4 that make them G5-like. So why moan and groan just because they don't call the processors G5? Is there ONE SINGLE parameter in the processor architecture that determines the "generation" (3, 4, 5...) or what? The impression I've got is that it's the processor as a whole that, with enough changes made, sorta "leaps" to the next generation. See what I mean? If they keep callin' them G4 let them do it! As I said, I'm no techie, so please correct me if I've got it all wrong. I would just love to get this generation terminology sorted out. (And I've been trying for a year now so don't just tell me do do my homework better, please) Cheers!

[OSeXy!]

I think the biggest difference between G4 and G5 is that the G5 will natively process 64-bit instructions, rather than the 32-bit instructions of the G4.

Once software has been recompiled to take advantage of this, it seems like this should really make a difference.

Of course it's hard to do a direct comparison (jast as it is hard to compare the G3 and G4 in 'real world' situations). Especially since no one has seen a G5 in the 'real world' yet.

Could be that other changes to the pipeline, etc, will add to or diminish the benefit of the 64-bit core.

Does seem as though some things highlighted for the G5 are migrating towards the G4, if the rumors are right (Rapid I/O, for example.)

We just have to wait and see. And hope some worms from inside give us a few more clues!

[AlphaTech]

OK, yesterday I went up to one of the Apple retail stores within driving distance (there are three). One of the questions I posted to one of the tech's there was about the bit level of the G4 (current version). His answer was that it is natively a 64 bit processor. The altivec is like an 2x boost for applications written to use it (makes it 128 bit then).

Before anyone starts sending flames my way, and I have to done the fire proof clothing, this confirms what is within Apple's specifications database. They list all the G4 systems that I checked as 64 bit, not 128 (as shown on the information web pages).

This would explain why a dual 1GHz G4 can blow the pants off of a 2GHz+ peecee system. It makes sense, 2x the bit rate=2x the performance per MHz. As well as even more when it is software created to take advantage of the altivec engine.

As always, Mac's are superior. :D

[OSeXy!]

But Motorola sez it's 32-bit:

http://e-www.motorola.com/brdata/PDFDB/docs/MPC7455FACT.pdf

I know the AltiVec is "128-bit", etc. But the chip itself is 32. This is meant to change to 64 with the G5...

[AlphaTech]

Are we going to trust motorola or Apple on this??? I am leaning towards Apple, since they have been posting the information since the day the G4 was released. I remember looking it up when I first purchased my G4 500 (AGP) and then showed it to my cousin who was very impressed (peecee weenie). I also remember something being announced about the G4 being a higher bit rate then the pc chips of the day (those are/were 32 bit).

Also, it makes more sense that the Altivec engine is a 2x implementation of the code, instead of the 4x (if the G4 was only 32bit).

[oldMac]

I'm afraid you've all been "bitten" :) by the marketing folks.

If you think Megahertz is an inaccurate way to measure performance, don't even get started talking about "bit"ness.

Data width has been used by the marketing folks (primarily at the game console companies) as a way of claiming superiority. In reality, it's usually a load of crap. Many of these comparisons aren't even comparing the components of the machines they're attempting to compare.

However, if it matters, *usually* when a chip manufacturer says that a processor is 32-bit or 64-bit, they're referring to how much memory can be addressed by the processor (which is generally determined by the width of the registers). And the most significant thing that matters in this case is how much memory can be addressed by the processor (about 4GB with 32 bit addressing). It could also be used to perform floating point computations with higher accuracy (something that generally has no impact on performance).

However, sometimes "bit"-ness refers to the data path to memory, which is one of the primary factors in how fast you can move information from RAM to the processor (currently 64-bit in the PowerMacs). The other primary factor is the bus speed (133Mhz in the PowerMac).

The Altivec is a little bit interesting in the way it works. The Altivec unit does indeed use 128 bit registers. Wider is better because you can fit several smaller groups of information into the same register and then perform a computation on all the information at the same time. However, this is very unique to Altivec. It is not something that happens in general use. If you're using a 64-bit processor (the G4 is a 32-bit processor in the memory-addressing nomenclature), you're really only going to see an advantage if you need to address a *lot* of memory, or you have some other really unique situations.

If it's Sony marketing the Playstation, they're just finding the widest datapath in the machine and quote that number because it's the biggest. IE, if they have special graphics hardware with a 128-bit path to memory, suddenly the Playstation is a "128 bit" machine (regardless of the 20-year-old embedded processor variants used to run the machine.)

[AlphaTech]

So, your basically saying that Apple is lying to everyone about the bit level of the G4 processors.

[oldMac]

Apple is not lying... Maybe just stretching the truth a little. :)

They just choose to quote the higher number, which refers to the data path to memory. Unfortunately, lots of hardware companies are doing this and it's become very confusing to the public. Blame Sony, Nintendo and Sega as they are probably the most flagrant violators.

Motorola is referring to the memory addressing, which is arguably the more classical definition for the "bit"-ness of a processor.

For example, the 6811 is an 8-bit processor, the 68000 was 16-bit, the 68020 was 24 bit and pretty much everything since then has been 32-bit, simply because the memory addressing capability is exponential as you increase the register width.

Not a lot of people are using more than 4GB of memory yet, which is why we are only now seeing 64-bit processors and they are primarily intended for server applications, where you can easily use more than 4GB.

[AlphaTech]

I just popped onto Apple's spec. section which lists information about all systems they have produced. I am attaching the spec.'s as a two jpg files if anyone wants to look. I will have to put the second part as another post, since I am not sure how to do both here in the same post.

It lists the logic boards data path as 128-bit and then further down, software addressing as 32-bit. The only difference from the PCI graphics and QuickSilver models are the bus speed (100 vs 133). The PowerBook systems are listed as having a logic board data path of 64-bit.

I am going to attempt contacting people that are in the know about this to get confirmation either way.

[AlphaTech]

part 2 of the image.

[MacAztec]

I am glad the G4 has a long life. It is a wonderful chip, and fits all of our needs. By the way, if anyone wants to sell me a PM, I am all ears!:)

[Catfish_Man]

...

1) Go to http://www.arstechnica.com and go to the "cpu theory&praxis" section. Read up on processors. It's quite interesting (in a nerdy sort of way) and will prevent you from sounding silly in discussions (to everyone who has the information right, ignore this).

2) The G4 is 32 bit, with a 64 bit wide bus, and 128 bit Altivec. Altivec uses the 128 bitness by doing the same instruction on four 32 bit chunks (32X4 = 128). You can't do this for normal operations. The G5 is theoretically 64 bit. This would have a performance improvement only when you needed variables larger than 32 bit (almost never), or when you needed more than four gigs of memory.

3) The reason the Itanium, Sun's MAJC, etc... are so fast is not that they are 64 bit. They have a lot of interesting new technologies in them (VLIW architecture, thread level parallelism, multi-core, predication, etc...) that are responsible for the insane performance (Also, the Itanium has 4MB of [i]on chip level 3 cache. This is big, expensive, and fast).

4) I think that the MPC 7500 is actually the G5, I've posted my reasons in several other topics, so I won't repeat myself here.

5) To the person who trusted Apple's info over Motorola's: Apple uses the chip, Motorola makes it. Motorola knows what they're talking about.

[ManOfTUnes]

1.5GHz, 266mhz bus, and RapidIO? (whatever that means)
Sounds good enough to
use software synths, samplers, effect plug ins,
eqs, and around 50 audio tracks. WHEN?!#@$#$#%

I really want this machine, so i would probably
be wise to hold off on the new purchase till it
arrives. SOmeone please tell me when!

[b8rtm8nn]

I would wait until MWNY, ManOfTunes, to see if the new mobo design is introduced (which I suspect it will be). I believe that technologically, they could introduce a new PM design in Japan, but Apple is a small company - I do not believe that they can switch on another assembly to handle a new design so soon after dishing out a redesigned iMac.

[OSeXy!]

Originally posted by Catfish_Man
...

...
2) The G4 is 32 bit, with a 64 bit wide bus, and 128 bit Altivec. Altivec uses the 128 bitness by doing the same instruction on four 32 bit chunks (32X4 = 128). You can't do this for normal operations. The G5 is theoretically 64 bit. This would have a performance improvement only when you needed variables larger than 32 bit (almost never), or when you needed more than four gigs of memory.

3) The reason the Itanium, Sun's MAJC, etc... are so fast is not that they are 64 bit. They have a lot of interesting new technologies in them (VLIW architecture, thread level parallelism, multi-core, predication, etc...) that are responsible for the insane performance (Also, the Itanium has 4MB of [i]on chip level 3 cache. This is big, expensive, and fast).

4) I think that the MPC 7500 is actually the G5, I've posted my reasons in several other topics, so I won't repeat myself here.

...

OK, interesting reasoning. Now I think I get why you call the 7500 the G5. Even though Motorola's PPC Roadmap explicitly calls the G5 an 85xx, Motorola's numbering system has always been a little quirky and ad hoc (it calls the G6 an 86xx...), are you arguing that most of the things projected for the G5 will be present in the rumoured 7500, (possibly minus 64-bit addressing, which Moto might have decided isn't worth the effort at the moment anyway), and therefore the 7500 will in fact "be" the G5 for all our intents and purposes?

This does make sense to me. Motorola then invents a 'new' series number: 75xx. It's not EXACTLY the 85xx, but it is way beyond the 74xx G4s, and has everything we the users wanted from the G5, anyway. Certailny the timing is right if it does come out at MWSF...

[ftaok]

As of right now, Motorola's PPC Road Map shows no existence of a 7500. The only place that I see the 7500 was on The Register website (and then that got picked up by all the other rumor sites).

MOT's map shows a 85xx for the G5 that's meant for communications and consumer applications. I take that to mean routers and Digital DNA stuff. Until I see a 7500 mentioned in an official Motorola Press Release, I refuse to believe that there is a G5 being developed by Motorola for Apple. This does not exclude the possibility of IBM developing the Next-Gen chip for Apple (G5 or whatever you want to call it).

[DNA]

...And just thought I should comment on this subject. As far as I'm concerned they might as well call the processors Al Bundy all over the line. What I think is the big question here is WHEN will Apple consider that enough changes have been made to the G4 to call it a G5. I can't really see that they wouldn't call it a G5 just because some part of it doesn't have "enough" bits or something like that. Ftaok though, has a very interesting point here that may well be the true answer to this riddle. Maybe the G4 is Motorola's last desktop processor and maybe Apple will jump to using an IBM processor in a year or so from now. By the way, is this information that somebody (don't remember the name) provided us with a week or so ago CORRECT??? "All Motorolas processors for embedded applications are called 8XXX and all processors for desktops are called 7XXX. Together with the information in ftaok's post this would mean the Motorola "G5" was never intented to be. If it is correct, then we've been drooling over the wrong candy (the 8500), folks!

[ftaok]

Originally posted by DNA
By the way, is this information that somebody (don't remember the name) provided us with a week or so ago CORRECT??? "All Motorolas processors for embedded applications are called 8XXX and all processors for desktops are called 7XXX. Together with the information in ftaok's post this would mean the Motorola "G5" was never intented to be. If it is correct, then we've been drooling over the wrong candy (the 8500), folks!
DNA,

I'm pretty sure that was me. Follow this LINK (http://e-www.motorola.com/webapp/sps/site/overview.jsp?nodeId=03M943030450467M983989030230) to see Motorola's current road map. It's a gif file and it's kinda tough to read, but you can get the gist of it.

[Onyxx]

Originally posted by b8rtm8nn
I would wait until MWNY, ManOfTunes, to see if the new mobo design is introduced (which I suspect it will be). I believe that technologically, they could introduce a new PM design in Japan, but Apple is a small company - I do not believe that they can switch on another assembly to handle a new design so soon after dishing out a redesigned iMac.

MW Tokyo is traditionally the place where new powerbooks are announced. When you think about it this makes sense too. New iMac, New Powermacs, New ibooks.... The only thing left is a powerbook update.

predictions...

1)New powerbooks in tokyo (apollo chip? new graphics chip? improved screen? improved battery technology [lithium polymer] ? there are a lot of variables.)

2)G5 no later than MWNY (if it is any later apple is in trouble. I wouldn't mind an improved an apollo with 100% symetric multiprocessing capabilities in configuration of 4 [quad processor systems] in the meantime.)

[rastalin94]

Just wanted to say a few things.

64-bit processor does not mean it is any faster than a 32-bit. Within a processor it has to do with how much memory it can access and how big of chunks it can move around within its registers. This does not mean faster.

The way you make a processor faster is by changing how things are done within the chip. You have something called IPC, instructions per clock cycle. Now you take that and multiply it by the speed of the processor, 800 MHz, and than you have how many instructions a processor does per second. The next thing you look at is how many clock cycles does it take to really get anything done. Once you have all that information than you can begin to see if a process is really faster or not. But do not forget no processor runs at 100% efficiency all the time. The PIV has a very high speed, but it takes a lot of clock cycles to really get anything done.

As far as calling a beefed-up G4 a G5, that is not a good idea. In the life-span of a process it often goes through many new steppings. This is most obvious with the Athlon line. You have the
Athlon
Athlon – Thunderbird
Athlon – XP
They are all basically the same chip but with just a few changes and upgrades. You save a change from a G4 to G5 when you have a whole new die come out. The exception to this was the PII upgrade to PIII, the only real difference was SIMD code was added.

What is a big deal right now it to get on DDR-RAM and Rapid-IO. I believe Rapid-IO is an increase in the speed of communication with the north and south bridge chipsets, and might be a replacement for PCI, not 100% sure though. The biggest problem right now is getting data/instructions to a processor, and designing the processor to never be idle when data/instructions are available to it. You would be amazed at how much idle time your processor actually has, there is where on chip multithreading comes in, now that is something that would kick ass on a G5.

[madamimadam]

Originally posted by Onyxx
)G5 no later than MWNY (if it is any later apple is in trouble. I wouldn't mind an improved an apollo with 100% symetric multiprocessing capabilities in configuration of 4 [quad processor systems] in the meantime.)

BEH? What planet you on 'cause I know on this planet there is still no Apple market for a 64-bit processor? Atm, those processors are restricted to specific and tiny fields that have trouble trying to work out which cupboard to use to hold their next truckload of cash. A souped up G4 (not just processor but whole kit and kaboodle) will be fine for now

[ManOfTUnes]

forgive my ignorance, but i don't know when that
is. Also, if it's revealed at that convention,
when would it be available to purchase?

thanks b8rtm8nn

-ManOfTUnes

[DNA]

...I see what you mean, even though I don't understand the language ;-) I hardly know what a "new die" is, but maybe there no such thing as a whole new die for desktops coming from Motorola. ANYONE out there who interpret their road map DIFFERENT from Ftaok? Because if he's right the next generation of processors in our best friends will definitely not be from Motorola, and maybe they will choose to call the (IBM?) processors something completely different.

[madamimadam]

Originally posted by ManOfTUnes
forgive my ignorance, but i don't know when that
is. Also, if it's revealed at that convention,
when would it be available to purchase?

thanks b8rtm8nn

-ManOfTUnes

I believe it is July but I would not be suprised if someone showed me wrong. It is definately around mid-year, though.

[madamimadam]

Originally posted by DNA
...I see what you mean, even though I don't understand the language ;-) I hardly know what a "new die" is, but maybe there no such thing as a whole new die for desktops coming from Motorola. ANYONE out there who interpret their road map DIFFERENT from Ftaok? Because if he's right the next generation of processors in our best friends will definitely not be from Motorola, and maybe they will choose to call the (IBM?) processors something completely different.

I have seen that logic, esp. with the sudden change of number structure of the processors but not something I like to bet on.

[spikey]

These 75xx whatever cpus are unlikely to be the G5. The G5 is likely to be based upon the G4, like the G4 was based upon the 604. But it is also very likely that there are several parts of the G5 which will be completely redesigned from ground up, like the Processors' FPU.

There is a performance gain of 64 bit chips if the data is above 32 bits. The only computing fields i can imagine this to be useful in would be 3d modelling/ Hi Res Graphics design. And possibly for small servers. I also believe that software has to be coded to take any real advantage of this, im wondering how many software developers would want to do that.


Possibly apple wants to use G5 technology in G4 processors just to make sure that consumer lines wont be too far behind the pro machines. Or Apple wants to pioneer the technology on the G4 chip so that the G5 wont have to put up with teething troubles.