Are quad-processors possible for the Mac at this point or do we have to wait?

Even if the current chipset is capable of more than two processors, they all currently share the same 167-MHz bus.

So the minimal speed benefit would not support the added cost.

 

i am pretty sure that the g4 can work in combinations up to 16. (2,4,8,16) whether we see a benefit or not is another thing, but we could technically have a 16x1.45ghz machine.

 

GeneR:

Quads will not be happening for quite some time, if ever. The PPC-970 could do it, but that would be one heck of an expensive machine with a complex purpose-built system controller, and Apple can't just throw it out there with any expectation of being successful.

 

This is an OLD document.

Keep in mind that the L2 caches have now been integrated into the CPU and another L3 cache is being run as a backside cache. And that the G4s went from MERSI to MESI after the 7400s and 7410s.

So the current G4s can snoop each others caches quite well in dual configurations, but thing get a little messy if you go above duals.

However, there are some Quad G4 blade servers -- but they may still be running 7410s or running the newer CPUs as processing nodes. Been quite awhile since I looked last.

 

I don't know why people want multiple processors...its not needed.

The only reason Apple has duals is so that they can claim that it is faster than the fastest PCs at the moment.

If the 970 is all it's cracked up to be, and can with a single processor be competitve with the intel chip, then I might even see Apple ditch the duals and go back to a single.

But I do not see Apple going beyond dual processors at the moment, and I think quads are nice, but completely unnecessary.

 

It is possible to run Quad G4 Macs, Linus (the creator of Linux) has a Quad G3 Linux Server in his office. But I expect OS X would have to be re-writen to support 4 chips. Different versions of MS Windows are needed to run more than 2 chips. I expect software would also have to be rewriten to take avantage of 4 chips. This is the main problem with it, it is hard enough to code for 2 chips, let alone 4 chips.

 

Try running all this on a machine at once with a single processor.

• Playing MP3s
• Burning a CD or DVD
• Transfering images from Compact Flash
• Editing images in photoshop
• Having Mail, iChat, iCal, and other things running in the background.
• etc

Then you will realize that some people can and will make decent use of multiple processors.

This is only one example of a workload. I can think of many more.

I'll concede that's part of the reason why the previous round of PowerMacs were all dual processors. However, it doesn't invalidate the need/usefulness of multiple CPUs.

I can still see a high end multi CPU system, but the low end their might be a couple of single processor models.

Although, if you only need a single processor, then maybe you only need an iMac?

I disagree about them being completely unecessary, however, only the top model would be more than two CPUs. There are plenty of applications that can suck down CPUs. Animation studios would probably love a Quad CPU Mac right about now.

Just because you don't have a personal need for something, it doesn't mean there aren't valid reasons for other people to need it.

 

Actually, if Apple did it right, OS X can handle more than 2 CPUs. It could've been on an unknown Apple prototype. Or since it's just the core OS, it could've been on an IBM made system. This is pure speculation on where they might have tested it.

Most applications that work for two CPUs are probably already coded to take advantage of an unknown number of CPUs. Also, if you are running multiple applications at the same time, then it doesn';t matter as much if any particular application knows how to take advantage of a multicpu system as other applications would run on the other processors.

 

A 64-bit CPU is FAR more important than a quad G4...

What's the point of wanting a Quad capable G4 now?

Basically for power users the G4 chip is darn near EOL'd, because it's a 32-bit chip.

It would have been helpful to have the option throughout the dual-G4 life, but it's basically to late to worry about it now.

People keep saying, why can I only install 2GB of memory in my Mac? If I use 1GB stick I can put in 4 but the Mac only uses two.

Hello, limitations of a 32-bit chip.

And this 32-but limitation of this are quickly coming for HDs too.

So look to a 64-bit CPU to break through some of the computational walls users are beating their heads against.

So, at this point designing a Quad-G4 would probably be pointless.

 

Re: A 64-bit CPU is FAR more important than a quad G4...

An excellent point. Take my previous posts in this thread to be generic as to the specific processor model. Presume it to be the fastest CPU that Apple is using in a PowerMac at such time Apple might release a Quad processor system.

 

Re: What about people who do rendering? Couldn't they use a Quad?

Rendering is one of the areas that will benefit a lot from more processors.

As for the heat, well, you can keep your feet toasty in winter.

 

Bear:

It sure as heck won't benefit from 4 G4's sharing a 167mhz bus!

 

I don't know about this. The 970 will presumably run on a motherboard supporting ApplePI which, one has to assume, is based on hypertransport. This is a great multiprocessor platform for dualies (which is why AMD dual athlon MPs are so sexy to the uber-linux geeks). AMD will be using this technology for their Opteron servers, which are going to scale big-time with respect to processors.

I could see apple releasing XServes (or possibly very high end powermacs) that go four way with the 970, for use as render nodes and Final Cut Pro stations. They obviously won't be cheap, but workstations aren't.

Not that any (ok, ok, most) of us could ever afford one, but I'll bet that they will show up eventually, perhaps within a year or a year and a half...

Cheers,
prat

 

Considering the architecture they chose for the kernel, I *have* to assume the OSX will scale just fine. The whole point of the Mach microkernel was to abstract as much as possible from the actual hardware itself, which is why it was so easy to port to x86 (Darwin, not Marklar, though I have no problem believing that Marklar exists deep in the bowels of one infinite loop). It gives up speed to accomplish this, though apple has done a lot of work in this area. (Too bad the Window manager is still so dog slow...)

 

praetorian_x:

Hmmm, you should elaborate on how you "have to" assume that "Apple-PI" is based on Hypertransport.

Woa, slow down there... ApplePI is why? Or Hypertransport is why? Athlons as we know them do not use Hypertransport for anything, other than the nForce and nForce2 chipsets for joining the north and south bridges. The FSB's of Athlons implement a protocol developed by Alpha (and in fact Alphas can run on at least some Athlon chipsets).

Hypertransport is not the cause of the ability of Operons to scale, it is the overall topology in which they are connected, using Hypertransport. AMD could have used some other inconnect for that. The PPC-970 is not connected in the same way, and will not scale as the Opterons will.

The thing about a quad-CPU PPC-970 is that each and every CPU needs its own bus to the system controller. Now, that may sound great compared to a single bus, but the cost of making a system conroller supporting 4 900mhz+ 64-bit busses and the quad-channel or higher DDR RAM to feed all that, plus connections for power, for AGP, and for whatever else... whew that's gona cost a bit. It is not going to be a chip Apple just throws into a dual CPU machine. It's gona be purpose-built, and expensive. Very expensive. We're looking at 4 times the pins of Apple's current system controller, or more. Plus those pins are expected to support much higher frequencies.

Add to this the 512k L2, very low for a 4-way system, and it is apparent to me that the PPC-970 will probably not be appearing in quad configurations in Apple systems.

 

GeneR:

In a couple years I bet some video cards will be supporting some sort of real-time rendering.