Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Re: But we're gaining ground fast

Originally posted by altglbrs
Even if Apple/IBM only reach 3Ghz by late 2004 (which has been promised), we'll be at 75% of the top PC MHz rating. When was the last time we were that close? Sure, its a full GHz back, but percentage differences are much more meaningful in my opinion.

Newbie, just as a 2 GHz G5 is already faster than a 3 GHz P4 so will a 3 GHz G5 be faster than a 4 GHz P4.
 
I've been saying this for a long time, Apple needs to develop a mobile processor.

Just like Intel is having problems fitting a P4 into a laptop, so too is Apple, having trouble with it's G5. It seems like everyone forgets that the PPC 970 is a derivative of the Power 4, a server class chip. It was never designed for a laptop, but engineers are really trying to shoehorn it into one. If it comes at the sake of battery life, who needs it. Why would you want a mobile processor that's not efficient, it truly is one of the best measures of laptop performance.

If this rumor proves to be true, I see the 750VX sometime next year and then on to the PPC300 in laptops. Right now, I just don't think the G5 will ever make it into a Powerbook, although I'll probably eat crow for this comment.
 
Intel is already switching to the lower clocked Centrino chips that run at between 1.0-1.5 Ghz. I am sure their performance is very close to that of any of the P4's.

I think it would be better for Apple and IBM to build a processor for laptops instead of trying to fit a desktop processor into a notebook like intel does.
 
Originally posted by greenstork
I've been saying this for a long time, Apple needs to develop a mobile processor...

If this rumor proves to be true, I see the 750VX sometime next year... Right now, I just don't think the G5 will ever make it into a Powerbook...

You're spot on man. Who even wants a G5 in a laptop? Seen those Dell laptops with desktop-class P4s crammed inside? They're monsters. Forget it.

Now, a low-voltage 2ghz G3 with Altivec and 400mhz FSB? Those will smoke. They should easily compete with Centrino in power and battery life. Heck, the current Moto G4s already do, and the 750GX will only be that much better. I just hope we see them sooner than later.
 
Originally posted by k2k koos
The real fun starts when the OS and applications are all 100% 64bit native, which isn't the case yet, so the G5 still can't show it's true potential.

All together now everyone: 64-bit native is not faster than 32-bit native . Please stop spreading this misinformation that 64-bit applications will be magically faster than 32-bit applications. Most won't, some will be slower, and some will be faster (but only because they were really 64-bit apps shoehorned into a 32-bit runtime). There is nothing implicitly faster about a 64-bit application/OS/runtime.
 
Originally posted by Rincewind42
All together now everyone: 64-bit native is not faster than 32-bit native . Please stop spreading this misinformation that 64-bit applications will be magically faster than 32-bit applications. Most won't, some will be slower, and some will be faster (but only because they were really 64-bit apps shoehorned into a 32-bit runtime). There is nothing implicitly faster about a 64-bit application/OS/runtime.

Untill everything is 64-bit, that's true. 64 bit allows for more info to be crunched per cycle, so it is technically faster, but not much. What it's main focus of 64-bit is that it can cruch more complex information faster. Simply put, you won't notice any difference for the OS, or other things. Now, I can see iTunes ecoding running faster, I can see DVD Burning and creating going faster, but I don't see games or other things that AMD is advertising running faster.
 
Originally posted by greenstork
Apple needs to develop a mobile processor.

Just like Intel is having problems fitting a P4 into a laptop, so too is Apple, having trouble with it's G5. It seems like everyone forgets that the PPC 970 is a derivative of the Power 4, a server class chip. It was never designed for a laptop, but engineers are really trying to shoehorn it into one. If it comes at the sake of battery life, who needs it. Why would you want a mobile processor that's not efficient, it truly is one of the best measures of laptop performance.

But you seem to forget the reason why Intel needed a mobile chip. It's because the Pentium 4 was such a massively inefficient chip designed for only one thing: to crank the Mhz rating as high as possible. The P4 has CPU stages that are meant to do nothing! And for all this, it trades off power like no tomorrow because it was designed for a desktop system exclusively.

At 130nm, the 970 is not a good candidate for a laptop at the speeds that Apple is likely considering. At 90nm, this might be a different story, but we have few if any details to go on that. All we really have is that Apple wants a PBG5 by the end of 2004, so we will almost certainly get one by then, be it the 970 or some other related chip.

So we come to the Centrino, why is it a good laptop chip? Well, mostly because it is power efficient, it only cranks up the Mhz when you need it (albeit slowly from what I've heard from others) and it doesn't use an excessive amount of power in the process. And it is fast, at least fast enough for most things you want to do.

But look at that again - all that really says is fast when you need it, and not power hungry. Ya know what - if I had that on my Desktop, it might save me some money on my light bill too. After all, we've got energy saver refrigerators, water heaters, washers and dryers - why doesn't my Computer try to save me some money too? The damn thing is on all the time, it shouldn't consume anymore power than I need at any one time.

And thus we come to my final point, Apple is not likely to use two different CPUs for Desktop & Laptop because people want the same things in both places. Imagine a CPU that is blazingly fast when it goes all out, and yet when all your doing is writing a paper hums down to the slowest it can get away with. And instantly springs back to life when you switch to a game. Is it possible? You bet, the 970 already does this to a limited extent now when it hums down to 1.3 Ghz and back up to max. That can most certainly be refined to drop down to half speed or less. Additionally there is technology out there to turn off processor units that are not being utilized by the current instruction stream for additional power savings. All this has been rumored for inclusion into the next PPC.

There is no reason to split design resources when the only real difference between what people want can be placed in a switch that says "More Speed or Better Power Efficiency".
 
Originally posted by the_mole1314
Untill everything is 64-bit, that's true. 64 bit allows for more info to be crunched per cycle, so it is technically faster, but not much. What it's main focus of 64-bit is that it can cruch more complex information faster. Simply put, you won't notice any difference for the OS, or other things.


But, all of what you just says hinges on the application falling into the 3rd category that I made: a 64-bit app that has been foisted into a 32-bit runtime. So of course it will get some speed gains because it was already a 64-bit app in design, and on a 64-bit OS it is better able to express that design.

Now, I can see iTunes ecoding running faster, I can see DVD Burning and creating going faster, but I don't see games or other things that AMD is advertising running faster.

iTunes & DVD Burning are both limited by the hardware they are utilizing. In theory a PowerMac G5 can process an entire DVD's worth of data in seconds (granted that isn't much processing). But unless the operations in MPEG encoding are predominately 64-bit integer (which iirc they are not - they are floating point, which we have had in 64-bit flavor since the original PowerMac, and Altivec since the G4) then you won't see a damn bit of difference in encode speed. (and as an aside, iTunes uses mostly 32-bit floating point for it's AAC encoder, roughly the Quicktime Better setting).

So I stand by my original assertion, most won't see any difference in speed (apps that don't care about speed in the first place), some will be slower (apps that rely on the size of a pointer for speed reasons for example) and some will be faster (apps that currently make heavy usage of disk caches or emulated 64-bit integers).
 
Thanks for spreading the truth!

Originally posted by Rincewind42
So I stand by my original assertion, most won't see any difference in speed (apps that don't care about speed in the first place), some will be slower (apps that rely on the size of a pointer for speed reasons for example) and some will be faster (apps that currently make heavy usage of disk caches or emulated 64-bit integers).

Many of these apps won't be any faster unless you have a true 64-bit O/S and application, *AND* more than 2 to 4 GiB of RAM in the machine!

Most of the applications that need 64-bit virtual memory also need more than 31-bits of physical memory to really exploit the machine.


(which iirc they are not - they are floating point, which we have had in 64-bit flavor since the original PowerMac, and Altivec since the G4)

AltiVec does not support 64-bit floating point, only 32-bit floating point (albeit 4 at a time). The Pentium 4's SSE2 does support 64-bit float.
 
Re: Re: But we're gaining ground fast

Originally posted by greenstork
Newbie, just as a 2 GHz G5 is already faster than a 3 GHz P4 so will a 3 GHz G5 be faster than a 4 GHz P4.

I'm not gonna resort to name calling like you, but I assure you, I am not a newbie. I bought my first Mac 12 years ago, and have been employed supporting macs for the last 5 years. While those of us who are "in the know" know that the G5 is already kick-ass at 2.0Ghz, Joe Schmoe looks at Ghz, plain and simple. Also, how many real world benchmarks have you seen where the G5 actually beats top end PCs on something other than Photoshop or some other contrived test? The point I was getting at is we're closing the Mhz gap.. if we're still a Ghz back next year, it'll be a less meaningful Ghz than now. Jackass.. oops.. i forgot, no name calling.
 
Re: Re: Re: But we're gaining ground fast

Originally posted by altglbrs
I'm not gonna resort to name calling like you, but I assure you, I am not a newbie. I bought my first Mac 12 years ago, and have been employed supporting macs for the last 5 years. While those of us who are "in the know" know that the G5 is already kick-ass at 2.0Ghz, Joe Schmoe looks at Ghz, plain and simple. Also, how many real world benchmarks have you seen where the G5 actually beats top end PCs on something other than Photoshop or some other contrived test? The point I was getting at is we're closing the Mhz gap.. if we're still a Ghz back next year, it'll be a less meaningful Ghz than now. Jackass.. oops.. i forgot, no name calling.

Yeah sorry for the newbie comment, but anyone who resorts to comparing MHz has got to get a clue.

To clarify, I never said it could beat any PC, this is far from the truth. I did however say that the Power Mac G5 beats the Intel Pentium 4 and here's a few tests to prove it:

http://www.barefeats.com/pentium4.html

http://www.amdzone.com/articleview.cfm?articleid=1296&page=2

edit: I also think we disagree on what joe shmoe's buying habits are. I think any consumer concerned only about MHz is going to buy whatever the salesperson tells him to buy or whatever is on the cover of the Dell sales magazine that week.
 
Re: Thanks for spreading the truth!

Originally posted by AidenShaw
Many of these apps won't be any faster unless you have a true 64-bit O/S and application, *AND* more than 2 to 4 GiB of RAM in the machine!

Most of the applications that need 64-bit virtual memory also need more than 31-bits of physical memory to really exploit the machine.


Agreed, just didn't feel like writing even more =). Of course, all of Apple's current 64-bit capable machines support more than 4GB of RAM.

AltiVec does not support 64-bit floating point, only 32-bit floating point (albeit 4 at a time). The Pentium 4's SSE2 does support 64-bit float.

I'm aware that Altivec only does 32-bit, guess my post wasn't particularly clear about that. Technically SSE2 doesn't do vectors of 64-bit floats, as SSE2 breaks it's 128-bit vectors in two and processes them back-to-back. This also puts it at a 50% disadvantage to vector processing vs Altivec (and at floating point vs the 970, as the 970 can process two 64-bit floats at once).
 
64 bit and speed..

Oups sorry, I shoulf have been clearer in my post.
It's true that 64 bit does not neccesarily mean fater computing, however I garee with the posting of someone else , of which here the quote:

Untill everything is 64-bit, that's true. 64 bit allows for more info to be crunched per cycle, so it is technically faster, but not much. What it's main focus of 64-bit is that it can cruch more complex information faster. Simply put, you won't notice any difference for the OS, or other things. Now, I can see iTunes ecoding running faster, I can see DVD Burning and creating going faster, but I don't see games or other things that AMD is advertising running faste

End quote....
 
Re: 64 bit and speed..

Originally posted by k2k koos
Until everything is 64-bit, that's true. 64 bit allows for more info to be crunched per cycle, so it is technically faster, but not much. What it's main focus of 64-bit is that it can cruch more complex information faster. Simply put, you won't notice any difference for the OS, or other things. Now, I can see iTunes ecoding running faster, I can see DVD Burning and creating going faster, but I don't see games or other things.

And you would, in general, be wrong.

1) Most computation is not 64-bit integer, and does not need to be 64-bit integer. Please understand that there are only 2 things that you can do with a 64-bit integer that you can't do with a small number 1) Access more than approx. 4 billion consecutive numbers with one register 2) Access more than 4GB of RAM with one register. So unless you need to do either 1 or 2 as your application's predominant reason to be, you will not get ANY benefit from 64-bit computing, and if your application is highly optimized for a 32-bit platform you may in fact lose some performance on a 64-bit platform.

2) Most of the truly computational intensive operations in use today operate on floating point numbers (of which we have 32 and 64 bit floating point native on all PowerPC CPUs) or operate on large numbers of 32-bit or smaller quantities (of which acceleration is generally more significant using Altivec or other SIMD instruction sets). Unless your crunching vast quantities of 64-bit integers or accessing vast quantities of RAM (and have the actual RAM to back it up) you will not see a noticeable performance gain.

3) Most applications run in a very limited working set. It is almost unheard of for most applications to use 4GB of RAM (or even 1GB of RAM for that matter). So unless you need to use more than 4GB of RAM, half of the usages you could have for 64-bit computing are gone. This will certainly change in the future, but by then we will see 32-bit computing as we currently view 16-bit computing.

No matter what you think of 64-bit computing, it is only what it is. Native support for 64-bit integers. That doesn't mean that you can even get double the speed of anything from it. To take advantage of a 64-bit memory space would require an amount of RAM that is currently rather prohibitive (8GB of RAM for a PMG5 is over USD$4000 and that only doubles the physical memory space over a 32-bit system).

So I'm going to be perfectly clear (because I don't feel like coming back to say this again). A 64-bit OS with 64-bit applications will not be faster than a 32-bit OS with 32-bit applications merely due to 64 being greater than 32. That's like saying that you can add numbers faster on a calculator because it's LCD display can show 10 digits instead of 5.
 
Not for a Mac

I'm gonna put one foot outside the box here and say this chip isn't for a Mac.

Rather, I suspect this is the way to a cheaper iPod. Apple wants to get a cheaper iPod out the door, but the thing has 2 StrongARM chips, a Portal Player, an ADC/DAC, and some other chips.

Add to this that its development environment is completely different from the rest of Apple's work, they have 3rd party schedules to coordinate with, PortalPlayer is off doing picture boxes now, and they'd like to get a PDA/Phone/thin client out the door eventually, and be able to add features whenever they want without being constrained by hardware ASIC's.

So, what to do? Build a PowerPC that's really light on energy, whittle down a version of Mac OS X to run on it with some kind of instant-boot technology, and build a system with 1 CPU and 1 custom ASIC, RAM, and some ports.

Put in a small 1" hard disk, and you have an $80 cost iPod. Sell 'em for $199 and you have a nice little market.
 
Thanks Rincewind42: Point taken, and understood, isn't a forum great, we can actually learn things here, cheers for that!

So I'm going to be perfectly clear (because I don't feel like coming back to say this again). A 64-bit OS with 64-bit applications will not be faster than a 32-bit OS with 32-bit applications merely due to 64 being greater than 32. That's like saying that you can add numbers faster on a calculator because it's LCD display can show 10 digits instead of 5.
:) :)
 
In the future, just click the "quote" button at the bottom of the post you want to reply to. It's a lot easier than copying/pasting from the page. :)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.