PowerBook To Get Dual-Core G4?

[AidenShaw]

Hogwash - you've fallen for "The 64-bit Myth"

It's known by now (or atleast you'd think so) that running a 64-bit OS on a 64-bit processor will increase performance no matter what machine it's on, no matter what the usage is for.

This is simply wrong - a fantasy based on the "bigger is better" belief (think "MHz Myth") and encouraged by Apple marketing.

Most programs will run *no faster* on 64-bit OS X, and many will run a bit slower compared to an identical 32-bit program running on a 32-bit OS X.

There are two main cases on OS X where a 64-bit CPU is an advantage over a 32-bit CPU of the same speed.

1. A program does extensive 64-bit integer operations
2. An application can improve its speed by using more memory (> 2 to 4 GiB)

In the first case, note that a 32-bit program on a PPC970 can take advantage of 64-bit integers by using a compiler option - there's no need to have a 64-bit OS. So, 64-bit integers are independent of a 64-bit OS.

The second case is the only true performance advantage of having a 64-bit OS. Here I use "64-bit OS" to mean an OS that gives applications 64-bit virtual address pointers and an individual application can address more than 2 to 4 GiB of RAM in a flat virtual address space. (Due to OS requirements, many 32-bit Os limit the application to something less than 4 GiB - for example Windows limits an app to 3 GiB of user space)

The typical example of the second case is a database application. On a 32-bit machine, the database is limited to caching 2 GiB (or so) of data in memory, and has to go to disk when the cache is full. On a 64-bit machine, the cache can be 4, 8, 10 or more GiB - and the database runs faster because fewer disk accesses are needed.


Why a 64-bit OS can be slower than a 32-bit OS

The increased memory capability comes at a cost - pointers are 64-bits (8 bytes) instead of 32-bits (4 bytes). This has two disadvantages:

1. It takes twice as much memory bandwidth to load and store pointers, which reduces the memory bandwidth available for other data and code
2. Pointers occupy twice as much space in cache, reducing the effective amount of cache available for other data and code

While this effect is usually quite small, some applications have a huge number of relatively small data structures with a number of pointers to other structures (e.g. doubly-linked lists, trees and meshes). Some applications can run significantly slower in 64-bit mode due to the larger pointers.


Summary - a 64-bit OS is of no advantage unless both...

1. You actually have more than 2 to 4 GiB of RAM installed
2. You have individual applications that need more than 2 to 4 GiB of RAM each.

It should be obvious that the database example won't run any faster unless you have the 6, 10 or more GiB of RAM installed. Same for almost any other 64-bit application - if it's not able to allocate more than 32-bits of RAM, it won't have any 64-bit advantage. And I used the word "both" because you don't need 64-bits for the system to support more than 4 GiB of RAM - Windows 32-bit supports up to 64 GiB of physical RAM per system.

Because of these factors, you should not be surprised when you discover that most programs won't be re-written for 64-bit addressing - they'll continue to be 32-bits forever. Why go to the expense and headache of developing and testing and support for 64-bits when your application won't need that much RAM?


Some exceptions....

As with most things in computer science, you can't use the words "all", "none", "always" and "never" when talking about a 64-bit OS. 😉

The above was written with OSX in mind, comparing a G4 to a PPC970. Some fuzzy points or counter-examples are:

• For a system with between 2 GiB and 4 GiB of RAM, there may be some cases where a 64-bit OS lets a program use more memory than a 32-bit program due to operating system overhead issues.
• For x64 systems (Intel's EM64T and Athlon64/Opteron), the processor instruction set is very different in 64-bit mode. There are more general purpose registers and some new instructions. Therefore an x64 program could be faster even if it doesn't need lots of RAM.
• There are some "sparse addressing" algorithms that can use huge amounts of virtual memory, but very little real memory. Such a program could use 64 GiB of virtual memory, yet run happily on a 512 MiB system without paging.
• Some applications simply need more than 4 GiB to hold their data -- but since these won't run on a 32-bit system, it's not an issue of being "faster" on a 64-bit system. 🙂

Fight the 64-bit Myth !!

 

[aus_dave]

AidenShaw you obviously know your stuff - what's the background behind MiB and GiB? I see those terms around a bit but the standard usage is MB and GB so I'm a bit confused 😕.

 

[AidenShaw]

GB = 1,000,000,000 ----- GiB = 1,073,741,824

AidenShaw you obviously know your stuff - what's the background behind MiB and GiB? I see those terms around a bit but the standard usage is MB and GB so I'm a bit confused 😕.

The "small i" in the suffix means that it is a binary multiple, not decimal.

The official international meaning of a prefix like "mega" is decimal ("mega" is 1,000,000 (10^6), not 1,048,576 (2^20)).

This is the reason that your new 250 GB disk seems small, the operating system will show you that it's 233 GiB. It's really the full 250 GB, it's just an accounting error in the OS display. (There's a myth that the 17 GB was lost in "formatting", but that's not true.)

The standards bodies are pushing for the "i" to distinguish the correct use of the decimal multiples from the binary forms.

See http://members.optus.net/alexey/prefBin.xhtml for a well written summary of the issue and proposals, with many links to statements from the official bodies like SI, IEC, IEEE, ISO, NIST and others.

The new prefixes are slowly being adopted by Linux and some members of the open source community.

There's currently a class-action suit charging computer companies with "deceiving" customers about the size of their hard drives - hopefully that will spur more operating systems to either shift to decimal displays or at least label their displays with "GiB" instead of "GB"

 

[aus_dave]

Ok, that's what I thought it might be - I've seen new hard drives lose 5% of their "space" straight out of the box before 🙂. Thanks for the comprehensive explanation.

Just out of interest, a 'gib' is a male cat (especially a castrated one) according to Dictionary.com. Watch that capitalization 😀.

 

[AidenShaw]

Just out of interest, a 'gib' is a male cat (especially a castrated one) according to Dictionary.com. Watch that capitalization 😀.

Does that mean if Tiger adopts the binary prefixes we won't have to worry about little tiger kittens all over the place ?? 🙂

 

[CholEoptera36]

Most programs will run *no faster* on 64-bit OS X, and many will run a bit slower compared to an identical 32-bit program running on a 32-bit OS X.

Okay okay, I gotta be a little more careful how I word things.

All of what you said about ram and a 64-bit OS isn't news to me. I didn't mean that every application you run will be at an advantage at all, I meant you can see an increase in performance(which obviously depends on variables). But I guess you can take that to the moon if you wanted because it can be taken a lot of ways, so my mistake. I also wasn't relating what I said to any myth about 64-bit = speed... You said that not me. I said increased performance. And I also didn't say when or where you would find the increase in performance either. I wasn't talking about EVERY single app you run. I definitely understand that some apps actually run slower in 64-bit, and I understand all that stuff about how some companies won't be changing to 64-bit software(for PC software that would be like, updating drivers I think) This is actually why anyone who gets 64-bit Windows won't get jack out of it because there are hardly any 64-bit drivers so far. Correct me if I'm wrong AidenShaw I think that's right? Anyway I didn't say too much about where I was directing my statement so I guess I got no one to blame but myself for how you took it. Just so you know, I was not equating 64-bit with speed at all. I don't even want to be associated with a statement like that unless I said it which I didn't.

Interesting stuff you said though, about Gib and MiB I never knew that. I had always wondered when I got my first 40GB HD why it said something like, 37GB when reformatted. I just figured that a few gigs were taken up in system files that were automatically installed when the OS is installed. So that's interesting 🙂

 

[AidenShaw]

Okay okay, I gotta be a little more careful how I word things.

All of what you said about ram and a 64-bit OS isn't news to me. I didn't mean that every application you run will be at an advantage at all, I meant you can see an increase in performance(which obviously depends on variables). But I guess you can take that to the moon if you wanted because it can be taken a lot of ways, so my mistake. I also wasn't relating what I said to any myth about 64-bit = speed... You said that not me. I said increased performance. And I also didn't say when or where you would find the increase in performance either.)

Now you've confused me.... 😱

Your previous post said "running a 64-bit OS on a 64-bit processor will increase performance no matter what machine it's on, no matter what the usage is for". The "no matter what the usage is" phrase sounds to me like everything will be faster.

What is "increased performance" if it is not "higher speed"? Can you give any concrete examples?

In other words, is it fair to paraphrase your latest post as "no application will run any faster, but you'll have increased performance" ??? ( 😉 Of course it's not fair, but what do you really mean?)

And I'll take a pretty strong stand that if you have 2 GiB or less of RAM you almost certainly will not see any increase in performance or higher speed due to 64-bit virtual addressing support in the OS.

(And again, since 32-bit systems can support more than 4 GiB of RAM, higher performance due to large memory doesn't require a 64-bit OS. I have Windows 32-bit servers that are using most of their 32 GiB of RAM - and they are getting much higher performance than if they had 4 GiB.)

This is actually why anyone who gets 64-bit Windows won't get jack out of it because there are hardly any 64-bit drivers so far.

There are plenty of drivers for new hardware, but compared to the enormous number of legacy drivers available for 32-bit Windows it would seem like there's "hardly any".

Realize, of course, that anyone using 64-bit Windows is buying a new machine with a new CPU and motherboard - so it's not that a big of a problem to make sure that the new hardware that you buy for your new system has drivers available.

It's something to look out for, but I'd think that any Mac person is well aware of the need to check driver support before buying a piece of hardware !!

 

[CholEoptera36]

Now you've confused me.... 😱

Your previous post said "running a 64-bit OS on a 64-bit processor will increase performance no matter what machine it's on, no matter what the usage is for". The "no matter what the usage is" phrase sounds to me like everything will be faster.

Yeah that's the exact statement I apologized for. I didn't mean "no matter what the usage is for" to be something like, "no matter what app under all circumstances." I concede that was a bad choice of words, definitely. I meant you can use 64-bit to increase or enhance performance no matter if it's at home or office, that's what I meant by usage. I don't like to get into all the "concrete evidence" stuff cause then everybody loves to pick you apart on every technical thing you say. But.. Okay I will show you a few articles I've bookmarked from the past which might shed light on where I was comming from.

Here's something interesting I read about SQL Server 2000:
http://whitepapers.zdnet.co.uk/0,39025945,60092563p-39000372q,00.htm

Didn't you say you run a MS server with 32Mib of ram? That would be nice to get a second oppinion from you on this article since I don't run a server, I just read what it has to say on this site.

Here's another article on how to optimize programing with 64-bit. It's a little old, so I dunno about the validity of the situation as of the moment with programing. But if you want to read it here it is:
http://bmagic.sourceforge.net/bm64opt.html

Now do you have to have anything more than a home computer to be a programer? I'm not a programer, I'm an HVAC tech which is why I'm asking.

Now you can find a lot of benchmarks on the web by now that show examples of how you would be at a greater advantage to staying with 32-bit, this is just to show you an example of how I agree that 64-bit can definitely run slower depending on what apps you're running. In this case for gamers, 32-bit is built up so well that it wouldn't do you much justice at all to go 64-bit:
http://www.amdzone.com/modules.php?...ns&file=index&req=viewarticle&artid=39&page=1

I don't mean to be confusing, but it's hard for me to word my possition without it comming off that way. I tell you what though, I didn't include articles from like, AMD or other companies like that and I'll tell ya why. I gave you some articles and a benchmark from other parties that perhaps might not be bias. I wouldn't want to provide a link on here from like, AMD who makes a 64-bit CPU. There is info out there from companies like that who will state that you will see an increase in system performance per say to them making a sale. Like you said Apple marketing strategy does this too.

I'm not really objecting to anything, my statements were rather nebulous. If I'm just totally wrong about what I'm saying it's because of things I've read or learned when I was in college. If I'm missunderstanding that you can optimize performance by using 64-bit then it's because of the sources I get info like anyone else. Either I am missuderstanding what they mean, or they're not very legit. What they're saying makes sense to me though.

 

[AidenShaw]

I'm not really objecting to anything, my statements were rather nebulous.

OK, I looked at your references, and we're more or less saying the same thing (except for that "all usage" line ;-) ).

Your first two links fall into the "if you have more than 4 GiB, *and* you have a 64-bit application that can use more than 4 GiB, then 64-bit will be a win" category. In fact, your SQL Server link is exactly my "database cache" example.

The third link is an Unreal Tournament review where they looked at how the "extra stuff" that you get with 64-bit on x64 really helps. Again, something I mentioned in my "exceptions" section because it is irrelevant to 64-bit on OS X.

(The PowerPC "book e" architecture is a 64-bit architecture, with a specific subset for a 32-bit implementation that has 32-bit pointers and omits all the 64-bit integer operations. G4 is 32-bit "book e", the PPC970 is a 64-bit "book e" except that it didn't implement the "little-endian" support that Virtual PC has been using).

 

[gekko513]

(The PowerPC "book e" architecture is a 64-bit architecture, with a specific subset for a 32-bit implementation that has 32-bit pointers and omits all the 64-bit integer operations. G4 is 32-bit "book e", the PPC970 is a 64-bit "book e" except that it didn't implement the "little-endian" support that Virtual PC has been using).

Ouch ... how will that affect the performance of Virtual PC on the G5 when the new version comes out?

It sounds like it would slow to an even slower crawl than before.

 

[gekko513]

You're missing the point.

"5" (as in "G5") is bigger than "4" (as in "G4"), therefore the "G5" is obviously better than "G4" - so we must have it.


[/sarcasm]


It doesn't matter if a dual-core G4 makes a lot more sense, marketing is pushing a "G5"....

ah .. noh noh no ...

a dual-core G4 makes it a G8, hence better than G5 😛

 

[AidenShaw]

Good one !!

ah .. noh noh no ...

a dual-core G4 makes it a G8, hence better than G5 😛

🙂

 

[AidenShaw]

Ouch ... how will that affect the performance of Virtual PC on the G5 when the new version comes out?

It sounds like it would slow to an even slower crawl than before.

No information, but I suspect that the repeated delays in getting the G5 version out are related to getting the performance back to "slow crawl" on a G5.

Apple shouldn't have let IBM give them "most of a PowerPC" in the 970 - they should have realized that implementing the full spec was important.

 

[thatwendigo]

This is completely unrelated, because you've basically argued every point I would have (and possibly better 😉 ), but it's been bugging me for a while. Is that your cat in your avatar, Aiden?

If so, what's the breed? I have my suspicions, but I don't want to tip my hand just yet.

 

[aswitcher]

This is completely unrelated, because you've basically argued every point I would have (and possibly better 😉 ), but it's been bugging me for a while. Is that your cat in your avatar, Aiden?

If so, what's the breed? I have my suspicions, but I don't want to tip my hand just yet.

I am guessing a Maine Coon, perhaps a brown tabby

 

[AidenShaw]

If so, what's the breed? I have my suspicions, but I don't want to tip my hand just yet.

Yes. Maine Coon, classic tabby with white.

Thanks for the good word....

 

[thatwendigo]

Yes. Maine Coon, classic tabby with white.

Thanks for the good word....

Ah, yes.

I adore Maine Coons, which is why I have three of them. One's a brown mackeral tabby, one a silver-point tabby with white socks, and one a brown calico. The first two came from the same breeder, the last from another when our originaly 'dealer' (because, face it, these rumply monsters are like drugs) moved across the country and took her cattery with her.

 

[AidenShaw]

We stopped at two (so far)

(because, face it, these rumply monsters are like drugs)

Big cats, small voices. And always into any open water....

We've got two, same tom, different moms, six months apart. The other is a solid black. The black's mom is a silver tabby. They're not too big, only 17-18 lbs each (at 2 years).


Here's the dad - Lone Star, a decorated show cat - he has a thing for ladders, which the black one picked up.

Tabby's mom - Dixie Chick

Black's mom - Baci

Enough pictures of the kid's 😉

 

[pigwin32]

OT

Big cats, small voices. And always into any open water....

We've got two, same tom, different moms, six months apart. The other is a solid black. The black's mom is a silver tabby. They're not too big, only 17-18 lbs each.

Man you guys should get a room, or maybe find an appropriate news group. Here we are on the cusp of a new PB, well ok maybe not the cusp but RSN, and you're talking about small furry animals?

 

[AidenShaw]

Man you guys should get a room, or maybe find an appropriate news group. Here we are on the cusp of a new PB, well ok maybe not the cusp but RSN, and you're talking about small furry animals?

Agreed, but this thread is more or less dead now....

 

[CmdrLaForge]

With all that cats, can't help but wonder where has this thread gone ???

Back to topic: I guess with the new iMac the G5 Powerbooks is not that far away 😀 😀 😀 😀 😀 😀

 

[thatwendigo]

Back to topic: I guess with the new iMac the G5 Powerbooks is not that far away 😀 😀 😀 😀 😀 😀

To go back on-topic...

iMac G5 dimensions: 1930.66in^3
PowerBook 17" dimensions: 157.08in^3
PowerBook 15" dimensions: 143.17in^3
PowerBook 12" dimensions: 110.61in^3

There's at least three times the mass and ten times the colume to dissipate heat from the iMac. This means absolutely nothing for a PowerBook.

 

[CmdrLaForge]

To go back on-topic...

iMac G5 dimensions: 1930.66in^3
PowerBook 17" dimensions: 157.08in^3
PowerBook 15" dimensions: 143.17in^3
PowerBook 12" dimensions: 110.61in^3

There's at least three times the mass and ten times the colume to dissipate heat from the iMac. This means absolutely nothing for a PowerBook.

Are you kidding me ? How much place do you think does the display itself need ? And how much space is there for the actual computer.

G5 powerbook - here it comes. I guess we see it in January.

 

[aswitcher]

Are you kidding me ? How much place do you think does the display itself need ? And how much space is there for the actual computer.

G5 powerbook - here it comes. I guess we see it in January.

Well thats a possibility but they'll have to shave another inch off the thickness and still keep it cool... I wouldn't be surprised its still a year away.

 

[CmdrLaForge]

Well thats a possibility but they'll have to shave another inch off the thickness and still keep it cool... I wouldn't be surprised its still a year away.

I don't think its a year away. Now with a that fast iMac (1.8GHz) for little money the powerbooks are lacking a little bit behind performance wise (in a price-performance comparison) If they knew that they couldn't get the G5 in a powerbook for another year it seems unlogical for me that they released the G5 iMac now. They could have speed bumped the iMac first. But they didn't. Thats another reason why I think we see updated G5 Powerbooks sooner then a year.