Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
But a POWER4 is too large, expensive, and energy inefficient to use in current Macs. It also lacks AltiVec. This new PPC is said to be derived from the POWER4. At 2.0GHz, I wouldn't be surprised to see SPEC benchmarks only slightly worse than the 1.3GHz POWER4 - I would actually except to see better CFP results once GCC is improved to automatically make better use of AltiVec. That would mean a 6X+ speed gain in floating-point over the current 1GHz G4, which is very exciting.

Alex
 
Originally posted by alex_ant
Here are two excerpts from this news.com article regarding this next-gen PPC:

A 2GHz G5 or G5-like chip would eat a 2.5GHz P4 for lunch.

Alex [/B]

I am pretty sure it is supposed to eat sparc and sun servers for lunch and also, incidentally be rackable, lower service contract cost, and of course run OSX.

Rocketman

avatar.jpg
 
Originally posted by sneed


Wouldn't there be an issue with 64 bit?

too pissed off for other reasons to explain but, no there wouldnt be any problems.
 
alex_ant:

The Power4 is too lage an expensive, but a single one of it's cores with reduced cache may not be. The cores themselves are not that big. I think IBM could easily hit the size of a P4 yet loose only a few percent of the performance of a Power4 for sinlge-CPU tasks. A good deal of the Power4, including the massive caches, is designed for massive CPU-count scalability and not for single-CPU performance.

In fact, your position that the "G5" will not perform like a Power4 on single-CPU tasks seems quite baseless. What would IBM gain by altering the core itself? When Intel needed a Celeron, they took cache from a Pentium. When AMD needed a Duron, they took cache from an Athlon. Messing with the execution core is just too difficult.

Not even the things Intel did to the Celerons ever crippled them as much as you suggest IBM will do to the Power4.
 
Re: New thought....

Originally posted by Johnny7896
Don't be surprised if Apple makes an agreement with IBM to use their OS to run their servers. Linux sucks eggs. Where Mac OS X is eligant, stable, and not windows. IBM and Apple may be on the same page again. I can see an extremly strong gain (profitable and popular) from this alliance for both companies.

I would be very suprised if IBM provided Mac OS X on its servers. Mac OS X does not support any enterprise applications and does not provide any enterprise class hardware or software support. IBM is in the business of selling servers to moderate to large enterprises. If these companies want a 1U server to run Apache or Samba, it will be x86 based running Linux. Heck, Mac OS X does not even run any of IBM's own enterprise software.

As for Windows server customers moving to Mac OS X server, please. That is ridiculous when they have other options that run on their existing hardware.

IBM is not going to sink tons of bucks into providing Linux on its full line of servers only to embrace Mac OS X. Stevie's servers are toys until they have a real file system (not HFS or ufs).:D When Mac OS X server supports a journaling file system, either Oracle or DB/2 and network backup to a remote tape jukebox, then the situation may change but not until then.
 
Originally posted by ddtlm
alex_ant:

The Power4 is too lage an expensive, but a single one of it's cores with reduced cache may not be. The cores themselves are not that big. I think IBM could easily hit the size of a P4 yet loose only a few percent of the performance of a Power4 for sinlge-CPU tasks. A good deal of the Power4, including the massive caches, is designed for massive CPU-count scalability and not for single-CPU performance.

Yes, welcome to the whole topic of this thread: The next-generation PowerPC derived from the POWER4 architecture. :) And the POWER4 most certainly is designed for single-CPU performance in addition to high-end multiprocessing, unless the fact that it's the fastest chip in the world aside from the Itanium2 is pure coincidence.
In fact, your position that the "G5" will not perform like a Power4 on single-CPU tasks seems quite baseless. What would IBM gain by altering the core itself? When Intel needed a Celeron, they took cache from a Pentium. When AMD needed a Duron, they took cache from an Athlon. Messing with the execution core is just too difficult.

I never said anything about altering the core. I made a guess; I think the POWER4 would be faster per-core than this new PPC because it has the advantage of tons of on-chip cache which would almost certainly be stripped out of a PPC designed for desktops. I still think this new PPC will be a great performer - just not the best performer. Competitive with commodity x86 chips yes, competitive with ultra-high-end server chips probably not.

Alex
 
ok... We know for sure that IBM is not going to make servers that run OS X... That would mean IBM is making a mac 'clone', and I think we all know how Steve Jobs feels about clones...

The only thing IBM ever had to do with Apple is the engineering of processors. IBM won't make a Mac, but they can (and apparently will) make a scaled down desktop processor that is derived from their Power4 processors that will work with Apple computers...

As for air flow, would it be possible for Apple to make a fan that's about 1" thick, and about 1.5-2 feet in diameter, and run it at a slower speed on the side of the computer, sucking air from the bottom and front, and pushing it out the back?
 
Originally posted by Chryx


Um, your MONITOR makes a noise?..

?

It's been a LONG time since I saw a CRT that had a fan in it..
What kind/size monitor are YOU using? Every 19, 20,21, and larger CRT I have seen has had a fan, with the exception of the NEC 29" but those are really just televisions, no matter what NEC says.
 
Re: Re: New thought....

Originally posted by ktlx


I would be very suprised if IBM provided Mac OS X on its servers. Mac OS X does not support any enterprise applications and does not provide any enterprise class hardware or software support. IBM is in the business of selling servers to moderate to large enterprises. If these companies want a 1U server to run Apache or Samba, it will be x86 based running Linux. Heck, Mac OS X does not even run any of IBM's own enterprise software.

Yeah, I think they would rather use AIX instead. It's trusted by enterprise much more and has proven scalability. Aside from the HPC crowd though, IBMs plans as far as AIX are a little murky. They seem to be pushing Linux on z/OS a lot lately. I'm sure that's because the zSeries (S/390) hardware is damn expensive but still, you don't hear much about Unix from IBM unless you're talking about their SP systems they do. In contrast, vendors like Sun, SGI, and HP market Unix pretty big in the workstation environment though. (Makes me wonder how the RS/6000 line has been doing as far as sales.)

The one thing OS X Server has going for it though is the management console. Granted, there are other solutions out there from other vendors that provide the same amount of functionality, but it is pretty neato cool. :D


IBM is not going to sink tons of bucks into providing Linux on its full line of servers only to embrace Mac OS X. Stevie's servers are toys until they have a real file system (not HFS or ufs).:D When Mac OS X server supports a journaling file system, either Oracle or DB/2 and network backup to a remote tape jukebox, then the situation may change but not until then.

Well, the journaling file system will not happen until Apple either extends HFS+ or gets application developers to stop using metadata features of HFS+ to the point where they could move to an existing journaling filesystem. The trick is that we need to start using the crappy dot extension versus type/creator codes. Too bad we don't have something like BeOS FS with journaling.

Also, the enterprise apps are coming. Sybase is being ported as well as Oracle RAC. Considering the economic climate for tech companies right now, I'll go out on a limb and say I'd assume more enterprise app vendors will port to OS X to make a quick sale.
 
Re: Re: New thought....

Originally posted by ktlx


I would be very suprised if IBM provided Mac OS X on its servers. Mac OS X does not support any enterprise applications and does not provide any enterprise class hardware or software support. IBM is in the business of selling servers to moderate to large enterprises. If these companies want a 1U server to run Apache or Samba, it will be x86 based running Linux. Heck, Mac OS X does not even run any of IBM's own enterprise software.

As for Windows server customers moving to Mac OS X server, please. That is ridiculous when they have other options that run on their existing hardware.

IBM is not going to sink tons of bucks into providing Linux on its full line of servers only to embrace Mac OS X. Stevie's servers are toys until they have a real file system (not HFS or ufs).:D When Mac OS X server supports a journaling file system, either Oracle or DB/2 and network backup to a remote tape jukebox, then the situation may change but not until then.

What enterprise applications that run on Linux will not run on MacOS X Server? From what I've read, porting from one to the other, without any Aqua stuff involved, is fairly easy. I seriously doubt that business decisions regard GUI as a particularly relevant or positive feature in Linux.

I think the more appropo point to make is why MacOS X over Linux? Almost all enterprise apps that do and/or will run on MacOS X undoubtedly run on Linux. Linux is dirt cheap and runs on just about any hardware. MacOS X is 'more expensive' but still a bargain relative to WinDon't. But, it only runs on Apple hardware. So, what compelling features do Apple servers offer the enterprise over that of Linux or WinDon't? Sounds like a good thread, BTW.

A journalling file system is cool, indisputable fact (okay opinion). BUT!!! Is it really that compelling on a server that rarely fails? What kind of server applications benefit from it to the degree that it actually influences enterprise decision-making? IMHO, a journalling file system has considerably greater benefits in client machines that deployed in the jungle and trenches, exposed to nasty external threats as well as relentlessly stupid or inconsiderate users. Bottom line, I don't see journalling as being all that compelling to enterprises today.

As for IBM and Apple partnering in some way, where there is a will there's a way! That said, there are many landmines, such as would you like WebSphere or WebObjects with your server dear customer?

IBM is considered by many industry analysts to be the king of IT. They just announced their intention to acquire PriceWaterhouseCoopers consulting to be combined with IBM"s already prestigious IBM Global Services (or whatever they call it these days). If I were Apple, I'd look for IBM to provide support and implementation services for Apple Xserve customers. IBM could be an enormous knowledge base of problems based by enterprise IT outfits that Apple could mine to discover the problems that Xserve and OS X can solve better for them than do Linux or WinDon't. There are possibilities; I'll leave it at that! Also, I won't hold my breath either. But, if I were SJ, I'd think long and hard about it: instant enterprise credibility enhancement.

Now, back to the subject of this thread:

psych!!!
 
Re: Re: too bad

Originally posted by theranch


Too bad that conference is in October...it would rule out a G5 in August.

Just to elaborate on what I suspect Shrek is thinking:

Would it really?

What is the Microprocessor Forum all about? Is it not to discuss the state of the technology as well as its direction?

So, not that I'm predicting we'll see this new CPU in the PowerMac this month, it seems plausible to me that IBM would talk about the technology of the new CPU after Apple has introduced it in the PowerMac.

I don't see this as a reason to conclude that this new CPU won't be in this month's new PowerMac. But there are plenty of other reasons.
 
HyperTransport G4 and G5

The HyperTransport issue keep popping up, and the "system interface capable of up to 6.4GB/s" sounds a lot like HyperTransport. If it is, I think this processor was made specifically for Apple, who is on the HyperTransport committee, and is said to be working with nVidia on something. Well, that something would only happen when PowerMac processors get HyperTransport. If it's a HT system interface on the G5, then IBM could be making a G4 with HT too. A G4 would take less design effort than a G5, which uses Power4 technology. IBM would surely use the same vector engine as the G5, which may be a second generation design. It makes sense to have the same system interface on both the G4 and G5, to keep motherboards as much alike as possible.
 
Originally posted by Rocketman


The bottom line is this. When a superchip comes out, it can have a low level kernal written for it to chat with OSX and is immediately useable. I suspect 64 bit Power chips will have 32 bit and perhaps even 16 bit modes for compatibility during transitions.


Here's my one question about 64 bit OSX:

Does OSX need to be fully re-written to take advantage of 64bit processing (or work on a 64 bit chip at all) or can just the mach kernel be rewritten to handle OS X's commands and instructions in 64bit chunks? Could apple (or the open source community) rewrite darwin (and/or mach) to work on a 64bit chip and emulate a 32 bit process quickly?

I'm sure this question was laced with a bit of technical incompetence, so please don't lash back with the normal "jeez, doesn't he know that the 387643813 widget is incompadible with the 984580459-4842-487458484?!?!?!? What year is it? Two-thousand ONE????" posts. I think I have a good grip on what is going on, but this one question lingers in my mind.
 
Power4 64bit processors can run 32bit powerpc code natively with no emulation involved.

Power processors and PowerPC processors have the same number of registers. Difference is that Power 64bit registers are 64bit and PowerPC registers are 32bit, now, running a 32bit app on a power4 would jsut leave half of the register empty, no other real difference. Believe me, thats how they designed PowerPC, remember, PowerPC came from Power processors, just like this new PowerPC processor is being derived from Power4 processors. the PowerPC ISA is backwards compatible unlike that x86 crap :)
 
I love reading comments on this board. You people say things that surely come right out of your ass. If you say something learn to back it up with facts. If you read something always question it. Some of the things said in response to this post are absurd.

1) not really sneed, its PPC compatible, no reason why it cant boot into OS0

First of all I'm assuming that OS0 was OS9. OS 9 is build around Apple's current round of motherboards and 32bit PowerPC processors. If Apple wants it to run on ANYTHING else they have to modify it. Notice a new OS release comes out with every new round of hardware. Even a G4 board with a few minor tweaks requires Apple to release hardware enablers to allow the OS to work with the new components. A 64bit PPC would require quite the massive motherboard redesign and right there without proper maintence of OS 9 code would end OS 9s compatibility with new Mac hardware. That's just the motherboard though. A 64bit processor woud require a complete recompile of all code. Simple as that. Notice MS has different version of XP and .Net server for Itanium. It's not because they like making two boxes. They have to recompile for the different architectures. Apple would have to do the same and without a similar system to the 68k/PPC FAT application system they would have to convince developers to distribute two version of all their apps for this radically different platform.

Nforce with a G4
This one is just crazy. Yes they both use Hypertransport, and yes Apple's current motherboard goals are similar to that of Nforce, but Nforce is a PC chipset build around a PC BIOS. You would have to fundamentally redesign the architecture to make it work with a Mac. Some of you make it sound like they just change the pin out on the ZIF slot to allow a PowerPC and their done. It would be a huge task to create a Mac compatibile Nforce chipset.

Think about what you read.

-Tim
 
Great but...

Great. You answered my question and I now consider myself educated.

My other question: If half of the register is left empty, would all of Mac OS X need to be rewritten in a 64bit native version to take advantage of that extra processing space, or would just darwin / mach be required to be rewritten?
 
Well. SGI, Suns, and HP's desktop market is basically DEAD.

That's why SGI is pushing into the higher end market and not really getting far since Linux is dominating. I expect to see them being bought in the next year or two by big blue or somoene, even terracomputing.

As for AIX VS. etc, it depends on what you are running. AIX is GREAT, and the best after irix, for floating point instruction operations. But for something like databases, you want SUN.

Linux is eating up at the lower to lower/mid range market. But it still can't compete with the super duper mid/high to high end computers. But those super duper servers are only 10% of the market, and that's the place BIG BIG BIG money is, and a place MS can't even think of playing, but they want to.

As for the hammer, and some of the new things, still designed on legacy designs, why do I want a new 2002 car that is still being designed around 30's engineering ideals? That's what's happening there. All the higher end machines, mainly SGI, have no FSB, and are true risc, and each component talks to each other at full speed, elimiating all bottlenecking.

blah blah blah

GPT
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.