Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

The 64-Bit Battle

A 64bit chip WILL be faster than a 32bit chip... however, it will not be a significant improvement.

Reason #1
Currently, with 32bit systems, a DOUBLE (64bit floating point) and a LONG LONG (64bit integer) requires several clock cycles to make simple calculations instead of one clock to handle 32bit FLOAT and INT. Any application that is using these larger numbers would benefit from the 64bit system. While 64bit integers are not used very often, DOUBLES are the default when it comes to floating point calculations because the 32bit version does not have the resolution that is needed for most calculations. When 64bit systems are the norm, you will see more applications take advantage of the LONG LONG datatype.

Reason #2
The already exists a 64bit version of the FreeBSD Unix OS which is what Darwin is based on. Having an underlying system that is 64bit capable will allow many system operations take advantage of the benefits of 64bits with existing applications.
 
Re: Re: Windows XP 64 Bit Edition

Originally posted by ktlx


Thanks for posting this.

This squares with what I have heard. If you want to do 64-bits, then your choices from Microsoft are a 64-bit edition of Windows XP Professional or a 64-bit edition of Windows Advanced Server. Just like there will not be a 64-bit version of Windows Server, there would not be a 64-bit version of Windows XP Home.

At least not for the time being. I would assume that if AMD really has significant market penetration in the consumer space with the x86-64, Microsoft would change their mind. But so far everything I read is that there is no 64-bit consumer version of Windows on the horizon.
Click to expand...

I *hate* to do this, and I just wasted an hour at work looking for the link to back me up on this, but I can't find it right now...

anyway... there's basically *no* difference between the various versions of windows. Codebase-wise. There are a few registry settings etc that tell each installation of windows "what" version it is (pro, server, etc.) but beyond that, there is very little difference (in most cases, none) in the software components installed. I've seen a utility (the one I was trying to find a link for) online that allowed you to "switch" versions of Windows 2000 from Pro to Server to Advanced Server. So, a 64-bit "home" version has "already" been written. In Windows XP, the main difference between WinXP Home and WinXP Pro is the ability to join Windows Domains! (why... only *pros* need to do that, right?).

So M$'s idea of different versions is one huge codebase and turn off features for cheapskates. Those features are there, though, just waiting to be turned back on via the registry. With that in mind, I think that we'll see 64-bit Windows as soon as Clawhammer is available in the form of Hobbled versions of the 64-bit codebase. :rolleyes:

Dharvabinky
 
ktlx:

This squares with what I have heard. If you want to do 64-bits, then your choices from Microsoft are a 64-bit edition of Windows XP Professional or a 64-bit edition of Windows Advanced Server. Just like there will not be a 64-bit version of Windows Server, there would not be a 64-bit version of Windows XP Home.
Click to expand...
Why do I get the feeling that it was a waste of my time posting those meager links you? That link that you accepted as proof of what you already are convinced speaks of nothing but the Itanium. That is a totally separate issue from x86-64 Windows, which is very well understood to be on the way. If you are blinded by your apparent need for Apple to be the first to have a 64-bit consumer OS then it's a waste of my time to continue the conversation.
 
Originally posted by DeusOmnis
Although cool, that seems to be a bit odd. Does anyone have one as to look? (i dont know if you could tell anyway.
Click to expand...

Yes, it seemed odd. And I don't follow the nitty-gritty of processor tech that much anyway.

But the price statement grabbed me, under $3k for a T-O-L Mac IS pretty cheap.

As for checking, the processors are marked, but would they conceal that?

Just an odd bit.

Whatever the truth is, the fellow did convince me to delay a purchase with him for about 6 months.

So I figure something is up. And that's why I posted, to see if any of this might be collaborated by some of the more knowledgeable.
 
Re: 970 May be in place.

Originally posted by colonelpanic
Talking with a rep, this is what I believed I heard about the new IBM Chip:

It's already in the dual 1.42. This fellow says you will hear nothing from Apple about it, but if you do enough searching in IBM literature, you'll see the new IBM chip is already in.

He goes on to say this is why the new Dual 1.42s are priced so low. Apparently the new chip is much cheaper to produce. And we may be seeing 2.0 Gig by this fall.
Click to expand...

How soon will it be before we know this for sure. This is exciting if true. This could possibly be the beginning of a price advantage with Windows machines!
 
I once read that OSX is "64 Bits clean", does this have any relevance to anything?
 
Re: 970 May be in place.

Originally posted by colonelpanic
Talking with a rep, this is what I believed I heard about the new IBM Chip:

It's already in the dual 1.42. This fellow says you will hear nothing from Apple about it, but if you do enough searching in IBM literature, you'll see the new IBM chip is already in.

He goes on to say this is why the new Dual 1.42s are priced so low. Apparently the new chip is much cheaper to produce. And we may be seeing 2.0 Gig by this fall.
Click to expand...

That smells as bad as it sounds. With all the Mac buffs in existence in these forums and others, how can this NOT be widely known by now?!?

So M$'s idea of different versions is one huge codebase and turn off features for cheapskates. Those features are there, though, just waiting to be turned back on via the registry. With that in mind, I think that we'll see 64-bit Windows as soon as Clawhammer is available in the form of Hobbled versions of the 64-bit codebase.
Click to expand...

That sounds right to me. :)
 
I don't know what IBM chip people are suggesting is in the dual 1.42's, but it sure as heck is not a PPC-970, and it sure seems to have specs matching a Moto 7455.
 
Originally posted by ddtlm
IJ Reilly:


My how funny Apple Rumorvision is. AMD is not actually shipping their 64-bit processor yet and I don't see where anyone said they were.
Click to expand...

Not AMD, Intel. From an August, 2001 article:

"The company announced that its Windows Advanced Server, Limited Edition, is now available for computers based on Intel's 64-bit Itanium chip."
 
Originally posted by ddtlm
I don't know what IBM chip people are suggesting is in the dual 1.42's, but it sure as heck is not a PPC-970, and it sure seems to have specs matching a Moto 7455.
Click to expand...

Thank-you, ddtim!
 
Originally posted by ktlx


You never made a distinction earlier about consumer versus non-consumer.
Click to expand...

Okay, let me make that distinction now. I figured it was a given -- because without that distinction, this would be a rather pointless discussion, since 64-bit processors have been in use in high-end servers and workstations for years.

What I am attempting to determine is if (1) Apple will be able to make the (possibly inconsequential) claim of being the first to market with a desktop OS running 64-bit, and (2) whether it will be of any real importance if they do.

For the record, I don't put much stock in rumors. I prefer to look at what Apple needs to do for survival, and moving quickly to a new microprocessor is one of them. Of the two possibilities (IBM 970 or x86), I happen to prefer the IBM scenario. One thing is for certain: they can't afford to do nothing, and I haven't known them to do nothing. Something is going to happen, it's just a matter of what.
 
Re: Re: Re: Windows XP 64 Bit Edition

Originally posted by DharvaBinky
anyway... there's basically *no* difference between the various versions of windows. Codebase-wise. There are a few registry settings etc that tell each installation of windows "what" version it is (pro, server, etc.) but beyond that, there is very little difference (in most cases, none) in the software components installed. I've seen a utility (the one I was trying to find a link for) online that allowed you to "switch" versions of Windows 2000 from Pro to Server to Advanced Server. So, a 64-bit "home" version has "already" been written. In Windows XP, the main difference between WinXP Home and WinXP Pro is the ability to join Windows Domains! (why... only *pros* need to do that, right?).
Click to expand...

This is not true.

There are three primary differences between Windows XP Home and Windows XP Professional. The first, as you have mentioned is the ability to join a domain. The other two are that Windows XP Home cannot support two processors or have an encrypted file system. To the best of my knowledge, those are not register settings that can be changed by the user.

There are similar differences between Windows XP Professional and Windows 2003 Advanced Server. Advanced Server does not have the number of processor limitation that Windows XP Professional has. If you have more than two or four processors, only Advanced Server will recognize the additional processors.

Having the same codebase does not mean they are the same thing. I work on a project where we use the same codebase to run six different embedded applications on four different hardware platforms. The only difference between them is a handful of small assembler files and some #ifdef's.
 
Originally posted by ddtlm
Why do I get the feeling that it was a waste of my time posting those meager links you? That link that you accepted as proof of what you already are convinced speaks of nothing but the Itanium. That is a totally separate issue from x86-64 Windows, which is very well understood to be on the way. If you are blinded by your apparent need for Apple to be the first to have a 64-bit consumer OS then it's a waste of my time to continue the conversation.
Click to expand...

Geez, don't have a hissy fit. There is no where in my posts where I insist on the need for Apple to be first for a 64-bit consumer OS. Frankly I think the concept of a 64-bit consumer OS in the next year is plain stupid whether from Apple or Microsoft. Just because I don't believe one broken link and another link to the equivalent of a rumor site, doesn't mean I cannot be convinced.

Show me links to some place credible like ArsTechnica, Tom's Hardware, Ace's Hardware, AMD, Microsoft, etc. Hell, I think I might even believe something in /. if it was written by someone credible.

I have no doubt that Microsoft will make sure that Windows XP Home runs on AMD's x86-64 platform when Hammer arrives. But that does not automatically mean that it is a consumer 64-bit OS. AMD has taken enormous pains to make Opteron backwards compatible with the IA32 instruction set. Microsoft may do nothing more than issue a patch to fix places where the backwards compatibility doesn't work.
 
Originally posted by IJ Reilly
Okay, let me make that distinction now. I figured it was a given -- because without that distinction, this would be a rather pointless discussion, since 64-bit processors have been in use in high-end servers and workstations for years.
Click to expand...

Yes, that is true, but there have not been 64-bit versions of Windows for years. Not everyone is aware they exist and I did not know the workstation version was even available yet.

For the record, I don't put much stock in rumors. I prefer to look at what Apple needs to do for survival, and moving quickly to a new microprocessor is one of them. Of the two possibilities (IBM 970 or x86), I happen to prefer the IBM scenario. One thing is for certain: they can't afford to do nothing, and I haven't known them to do nothing. Something is going to happen, it's just a matter of what.
Click to expand...

I agree. I think the x86-64 path is better long term, but Apple has to solve their problems now. Given the urgency for their XServe and PowerMac platforms, the PowerPC 970 seems their only hope.
 
Windows fragmenting?

It appears to me that Windows code bas maybe fragmenting. This is due to AMD and Intel not being code compatible in the 64 bit realm. What I mean by this, is that it doesn't seem to me that a 64 bit application written for AMD will work on a Intel IA64 machine and visa versa.
So ... does one now have to compile for AMD and Intel separately? If so I see this as an advantage for Apple & IBM.

Also someone previously mentioned about the complexity of the "Northbridge" (Intel like term) with the IBM 970. I think it will be simpler that a G4. My understanding is that the memory controller is on the 970. Kind of like having 1 gig of L3 DDR-SDRAM cache. Faster !
Also, I believe that the 970 uses "hypertransport" off the chip to "Northbridge". This is a fast bus. since the main memory is direct to the 970, the hyper transport buss only has to deal with transactions to video and other IO (PCI, disks, USB , Firewire etc.) and other 970s.
This thing should be fast!
Just my $0.02
 
Tim Flynn:

Also someone previously mentioned about the complexity of the "Northbridge" (Intel like term) with the IBM 970. I think it will be simpler that a G4. My understanding is that the memory controller is on the 970. Kind of like having 1 gig of L3 DDR-SDRAM cache. Faster !
Click to expand...
Nope, a PPC-970 system is set up in a rather conventional layout, with a FSB between each CPU and the system controller, and the main memory hanging off of that. The system controller will in fact be significantly more complex than that for a G4 becaue each CPU needs a dedicated FSB and because more than one channel of DDR RAM are going to be needed to feed the FSBs and their processors.

Even if the PPC-970 did have the main RAM connected directly to the CPU, it would have far higher latency than the SRAM currently used in the G4's L3.
 
Originally posted by ddtlm
Tim Flynn:


Nope, a PPC-970 system is set up in a rather conventional layout, with a FSB between each CPU and the system controller, and the main memory hanging off of that. The system controller will in fact be significantly more complex than that for a G4 becaue each CPU needs a dedicated FSB and because more than one channel of DDR RAM are going to be needed to feed the FSBs and their processors.

Even if the PPC-970 did have the main RAM connected directly to the CPU, it would have far higher latency than the SRAM currently used in the G4's L3.
Click to expand...

First of all, this assumes that Apple will be producing Dual-Processor machines. IBM has stated that there will be a 900mhz bus interconnect -- but they have not stated that it would be required for each processor to have an independant bus. More likely it will be a shared bus between the processors. If they are using the shared bus, then the memory controller would be relatively simple... just connecting it to the new 900mhz bus.

This is where it gets interesting... if Apple implements dual channel DDR memory like the nVidia nForce chipset for AMD processors, then memory access will be very fast! It literally doubles the throughput of the memory to about 6.4Gb/s (266 DDR memory) and would be even faster if Apple uses the 333 DDR or even the 400 DDR memory. It would not be quite as fast as the internal cache, but it would allow for a very fast system without requiring the L3 cache in the processor.
 
nighthawk:

IBM has stated that there will be a 900mhz bus interconnect -- but they have not stated that it would be required for each processor to have an independant bus.
Click to expand...

Page 12 of this pdf from IBM clearly states that the PPC-970 FSB is point-to-point, which is to say that it cannot be shared. It goes directly between the CPU and the system controller.

www.simdtech.org/apps/group_public/download.php/23/IBM_PPC970_MPF2002.pdf
 
Re: Re: Re: 3rd Q

Originally posted by FatTony


I think there is also the option of financial quarters. Doesn't Apple's first financial quarter include the holiday retail season? This would shift their subsequent quarters ahead.

Also, with regards to case design and new processors and multiple processors/motherboards in the pro line. Apple did this with the yikes and sawtooth G4s. New case (atleast the color) with the new processore and the low end G4 was just the G3 motherboard with a G4 in it while the higher end models had all new motherboards.

I don't think it is out of the question for the low end PowerMac to have a G4 while the higher ends have IBM's GPUL. That way Apple can purge some inventory and at the same time release the next big thing.
Click to expand...

You are correct about financial quarters FatTony.
I dug this up:
Apple Reports Third Quarter Results


CUPERTINO, California?July 16, 2002?Apple® today announced financial results for its fiscal 2002 third quarter ended June 29, 2002.

So Apple's 3rd quarter ends the last day of June. Ryan said we might see the 970 in the 3rd quarter. But I think he means 3rd calendar quarter not fiscal.
 
Originally posted by nighthawk

First of all, this assumes that Apple will be producing Dual-Processor machines. IBM has stated that there will be a 900mhz bus interconnect -- but they have not stated that it would be required for each processor to have an independant bus. More likely it will be a shared bus between the processors. If they are using the shared bus, then the memory controller would be relatively simple... just connecting it to the new 900mhz bus.
Click to expand...

Can't say where, but I'm pretty sure I've read that the 970 will feature a switch/hub topology to support the CPU. This Apple internconnect or whatever that's on the 970 is actuall 2 point to point connections. Each is 32-bits and 900MHz, yielding 3.6GBs troughput. This gives you 7.2GBs (nominal 6.4GBs according to IBM) with a dedicated bus for each rx and tx. This is different than conventional shared buses (MaxBus and P4) since they can only read *or* write on a clockpulse, not both. Also, it's not a shared bus, but rather a switch topology so that there's no interference between the CPUs.

Needless to say, the 970 will be a bandwidth *monster*, especially if paired Dual DDR Ram..

Has anyone speculated on the possibility that Apple might use DDR2 in their systems? since each processor will be needing about 6.4GBs of data (at maximum burst rate), that kind of need is met by Dual DDR400. But an SMP system here would benefit from even *more* RAM bandwidth since each processor has a point to point connection at the full 6.4GBs (this will make apps that work on Cached data *fly*, especially if there is a "unified" L3 cache in front of the RAM shared by multiple 970s). DDR2-500 will be making a debut soonish featuring 4GBs per channel...

Since the machine is apparantly a "Bandwidth" machine more than a clockcycles machine... do we see DDR2 in line?

:confused:

Dharvabinky
 
With all the talk about super fast 970 processors and ultra high throughput, I wonder what video card Apple would use. HardOCP over-clocked a P3 to 4.44 ghz and found that the video card was the bottle neck. The system also had a serial ata drive and other goodies.

http://hardocp.com/article.html?art=NDIy

Have a nice day!
 
Hey yo... isn't da 970 the same thing as a G-5? Doesn't da "G" stand for "generation"? How long is a 400mhz G3 expected to be a useable and viable processor for real world "household" applications anyways? My meek iMac is still running rings around most of the wintel peecee's that my acquaintances are trying to run. I have no crashes and no need for microsoft, aol, or any of the flavor of the month garbage that everyone in the non OSX world seems to be using as a crutch. I could stand to go a bit faster somedays 9encumbered by 56k modem blues), but I am a happy camper most of the time. How many people with a two year old peecee will you hear saying that?
 
DharvaBinky:

Needless to say, the 970 will be a bandwidth *monster*, especially if paired Dual DDR Ram.
Click to expand...
Better make that only if paired to dual-channel DDR RAM. Single channel DDR is going to be nothing special for a PPC-970.

Has anyone speculated on the possibility that Apple might use DDR2 in their systems?
Click to expand...
The PPC-970 will be available before DDR2 is mainstream, so they will have to start out on DDR1.
 
Yes...

Originally posted by Jimong5
I once read that OSX is "64 Bits clean", does this have any relevance to anything?
Click to expand...

You are right on this, sort of.

I have read that the PowerPC was designed from the beginning to go to 64 bit. The way that the 32 bit instructions are written makes them able to run native on a 64 bit processor, allowing OSes that use the PPC to move from 32 bit chips to 64 bit chips without re-writing the entire thing. So, OSX, a 32 bit OS, can run on the 970 with no emulation loss, becasue it is not emulated. I seem to remember that the oppposite is true as well, that if OSX is adapted and is recoded to be 64 bit native, that it can run on a 32 bit PPC chip with no emulation perfromance loss, since the instruction set can work with the right extention.

I am not an engineer, so if anyone else has a better handle on this, please, chime in. I think that this is a very important point in this discussion.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back