Will Darwin be Ported to the IBM Power 4?

eadint asks: "I have heard rumors thorough the net that Apple plans to port Darwin to the Power 4, 64 bit chip. Currently I work for a university. We are using Apple computers and are considering the platform for our number crunching capabilities. According to this Motorola has no plans on producing a 64-bit chip. Does anyone know if Darwin can or will be ported to a true 64 bit platform."

Comments

Power4

[MonaLisa]

Have you considered Itanium-2 under Linux for
your "number crunching" platform? The McKinley
(Itanium-2) is faster than the Power4, and also
cheaper (although you'll need to buy the Intel
compilers for a few hundred if you want great
performance).

Re:Power4

[Guspaz]

What about waiting for the Hammer to come out? I thought the Hammer's compiler was free.

Regards, Guspaz.

Re:Power4

[Chanc_Gorkon]

Itanium 2 faster then Power4? I doubt it. Again, this is apples to oranges comparisons. Also, has ANYONE ever seen a Itanium system? I thought not. I work on RS/6000 machines and I really doubt an Itantium 2 system could even compare to a IBM RS/6000 (Power4 Based). First off, most RS/6000 machines come equpied with SMP (sometime only having one chip installed, but most are complete). Also, the Regatta p690 machine is THE BEST in my opinion. Only thing that comes close might be the top level Sun machines. Also, this same technology was used in Deep Blue (the machine that beat Gary Kasparov many years ago). All that Deep Blue was was a specially programmed SP system (RS/6000 Super Parallel). Comparing Intel's yet to be produced Itanium 2 which is an evolutionary step of the Itanium which did not really sell all that well. Point is....POINT ME TO THE BENCHMARKS! Since you can't(no silicon kind of stops that), well we shall see. Personally, I would rather run AIX on it. It's proven at least.

Re:Power4

[brejc8]

There are some benchmars out actually. Make of them what you like.
This is the sheet for itanic2 1GHz.
And you can compare it to the others

Its still too early to see what effect the itanics will have but they look quite respectable, if they havent priced them selves out of the parket

Re:Power4

[MonaLisa]

Here are the SPECfp benchmarks from:

http://www.spec.org/osg/cpu2000/results/cfp2000. ht ml

IBM Corporation IBM eServer pSeries 690 Turbo (1300 MHz) 1 1202

Hewlett-Packard Comp hp workstation zx6000 (1000 MHz, Itanium 2) 1 1356

Also, go to: http://www.emsl.pnl.gov:2080/capabs/mscf/?/capabs/ mscf/hardware/results_hpcs2.html
for benchmarks results for some real codes and further synthetic benchmarks.

Re:Power4

[Anonymous Coward]

"Also, has ANYONE ever seen a Itanium system? I thought not." -- Why do people keep saying that? You can buy workstations with Itaniums from HP and Dell right now.

Re:Power4

[CMiYC]

Also, has ANYONE ever seen a Itanium system? I thought not.

I have. Quite often. I have worked with both Merced and McKinley systems. The company I work for sells the logic analyzers and probes for them. I support those products.

Re:Power4

[ericman31]

Also, the Regatta p690 machine is THE BEST in my opinion. Only thing that comes close might be the top level Sun machines.

Have you ever really compared Sun's SunFire servers to the IBM pSeries? If you had you would know that each has their own arena where they are "best". The SunFire x800 and 15K, for example, are much better SMP boxes than Regatta. Their SMP is nearly linear in terms of scalability. This makes them a fantastic box for Data Warehousing.

Regatta, on the other hand, has some very interesting mainframe features, including LPARs and workload management, that should prove extremely useful for cost effective server consolidation and application servers. LPARs, however, are not nearly as "safe" from a hardware perspective as SunFire domains, which partition in hardware rather than software, thereby effectively isolating each domain in the server.

For my money the big iron UNIX ranks Sun, IBM, and then HP/Compaq a distant third. If I had to choose a 64-bit processor it would the UltrSPARC III Cu, although the Power4 is pretty impressive too. BUT, who the heck is going to run a desktop OS on a 64-bit RISC processor that costs $15,000 or more just for the CPU.

Here's where we get down to the important part. Itanium and Hammer will be commodity items. They won't be on par with IBM and Sun big iron, although they will probably be very effective at the workgroup/departmental server level. I hope that Apple seriously considers porting to a commodity CPU, it might actually bring competition back to the desktop OS market. Plenty of people out there just want their OS to work. And most people believe that Mac is a better, more user friendly OS than Windows. They just don't want to pay the price for Apple's hardware. Competition and an Intel/AMD platform would make Mac far more affordable and likely to be used.

Re:Power4

[ericman31]

These benchmarks are great. There are problems though, even with SPEC. One of the biggest problems is that we still aren't looking at the real world. The one where you have to integrate Oracle, J2EE, Apache, WebSphere and C++ (as an example, there are so many combinations it's pretty much impossible to list them). Since it's darn near impossible to build a benchmark that measures performance under real world conditions, benchmarks are half the battle, or less.

So we use a variety of benchmarks, our own experience and bidding by the vendors, throw in what our organization is experienced in, and go from there. Based on what I know about SuperDome architecture I seriously doubt that HP's Itanium servers will outperform Regatta or SunFire in "the wild", although it should be a solid performer in general. And the price will probably be very competitive.

Re:Power4

[ericman31]

But, and I'm sure you will admit this, Itanium is NOT in general use in the Enterprise. It will be some time before it is. How many people who work in a typical Enterprise IT environment have seen Itanium systems? Almost none.

Re:Power4

[Anonymous Coward]

So, what you're saying is, is that you're working on your resume... .. you are, right?

Re:Power4

[sharkey]

Again, this is apples to oranges comparisons.

No, it's Apples to Bunny People comparisons.

Re:Power4

[Chanc_Gorkon]

Exactly. Most places aren't going to run a Regatta in a single server situation. They are going to divide it out into LPARS. Most common place would probably buy a Sun if they wanted it to be alll CPU's on one server, although the Sun's have the same or similar thing also. I think they are real close, though I would take the IBM because there service has been very good at our site. I had a drive go bad and they were out the next day and the box never had to be downed either. Plug the new drive in, run cfgmgr and the drive is there and ready to be reallocated. Most times we do down the box though just in case...plus they also run cfgmgr on boot so when you reboot the drive will be there also.

From and Admin standpoint, I like AIX. The Sun has nothing compared to smitty. Admintool doesn't even come close. One upside to sun is there's no ODM so you can admin everything by editing the text files. You can do the same on a AIX box if you don't want to have the settings there on a reboot (if it ain't in a rc script or the ODM, it's gone.). Smitty also makes things easier for a first time admin. You can learn the commandline way by using smitty. Don't get me wrong...sometimes the raw way is good, but if your not sure what your doing, you can always pic your way through smitty.

Also if you want a complete IBM solution, you can get the Shark storage server which scales up to 27TB (TERABYTES!) of fiber channel connected storage! Granted, you could also use pretty much any UNIX®, Windows NT®, Windows® 2000, Novell NetWare®, Linux, IBM iSeries(TM) and AS/400®, IBM zSeries (TM) servers and S/390® too. It's just nice to have everything supported by IBM.

I have never had to call IBM twice to get a tech there. IBM's support has been exemplary which is more then I can say for Xerox, Novell and our ISP. I have no idea what Sun's suport is like because i have never had the opportunity to run a bigger Sun machine. Just ran workstation class machines here.

Re:Power4

[Chanc_Gorkon]

I really can't trust these benchmarks. They are produced by someone that will be SELLING Itanium 2 servers. Also, notice it says at the top hardware available Sept. 2002. That's next month! A site like Anandtech or some other independent reviewer needs to run benchmarks on these before I could trust the results. This is like Microsoft saying NT 7 or whatever is faster then Linux and showing some pretty graph before anyone could go BUY it.

Re:Power4

[ericman31]

Sun Gold and Platinum service is at the same level as IBM's. Our data center (I'm one of the senior infrastructure guys) has both IBM and Sun platforms in it, and the service is outstanding from both. Now the Windows guys in the set of cubes over have to hassle with a nightmare of phone support and then usually their vendors ship the drive (or whatever) to them and they get to install it themselves.

I personally don't use Admintool, or any other graphic interface for administration. If it's a one time deal I do it on the command line. If it's something I do more than once, I have a script to do it for me.

Shark has problems. My personal preference is Hitachi Lightning. The 9960 supports 27 TB raw, about 22 in RAID 5. Since Sun resells Hitachi I get all my storage support from them as well, even for the Hitachi that has IBM and Win2K boxes connected.

We run SunFire 6800's with Sun Platinum support. We have one Sun ES guy permanently onsite (we're a big shop) and any other support we need shows up within 4 hours.

Now back to the Mac OS X on Power 4 idea. Why on earth would I WANT Mac OS X on there? I'd much rather run AIX 5L, a far better and more mature OS. Same with Sun UltraSPARC boxes, why put Mac on it when Solaris 8 or 9 is mature, stable and powerful? However, Mac on a commodity desktop 64-bit CPU (Itanium or Hammer), which should lower the price and make it easier to tinker with? Yeah baby.

Re:Power4

[Chanc_Gorkon]

Good stuff! Glad to here service is good with BOTH companies! Something you hardly ever hear!

Yeah that is exactly what I was saying in another post. Besides, anything that compiles on Darwin would compile on a AIX machine too unless it's OS X and they are Cocoaizing a app. I would never see the point in porting it. Also, a port would be difficult as I am sure that ROS Microcode is different then a Mac's BIOS. Processor wise, it would be easy but there's too much pSeries specific stuff that would need ported to run on the Darwin kernel. Plus the AIX Kernel is way more advanced then Darwin. Darwin would need an equal advance to even make it worth doing. So, why even hope for it? Just run AIX and be happy your running a nice robust system that can also run KDE or GNOME instead of CDE (if you need those GUI things).

Re:Power4

[ivan256]

Also, has ANYONE ever seen a Itanium system? I thought not. I work on RS/6000 machines and I really doubt an Itantium 2 system could even compare to a IBM RS/6000 (Power4 Based). First off, most RS/6000 machines come equpied with SMP (sometime only having one chip installed, but most are complete).

I've seen, used, and programmed for Itanium systems. There is definatly silicon available. You can even buy CPUs from some pricewatch vendors. There aren't any non-SMP capable Itanium systems available right now. They ALL support at least 2 CPUs. There are real benchmarks that have been run on Itanium 2, but there haven't been any independant benchmarks of Itanium 2 yet. The benchmarks show it slightly faster then Power 4, so expect them to be competitive with each other in real world applications. I've personally run benchmarks on the original Itanium, and they were nothing spectacular.

If I were paying the electric bill, I'd buy the Power4. If I needed more then 4 CPUs, I also wouldn't trust Itanium 2 yet, but only because I don't trust the current chipsets that are available. If I wanted a 4-CPU system or less, I didn't care about power or heat, and was only interested in speed, I'd pick the one that cost less.

HERE! I KNOW! I know I know I know!

[Anonymous Coward]

It will be ported. I guess I am not allowed to reveal this kind of strategic information, but since you asked, I just had to tell.

AMD hammer for apple?

[brejc8]

These are just roumors but I heared this from several sources that apple are talking to AMD about a 64bit only OS.
Apple might need to get faster chips to compete but making something so close to a PC will allow clones.
Alternatively apple could just port their OS to the hammer PCs and keep making their own PPC machines.
The OS is sexy enough to make a large wodge of money.
Especially if it comes with M$ office which is the only reason a lot of people dont wish to use anything other than windows.

Re:AMD hammer for apple?

[Sentry21]

Especially if it comes with M$ office which is the only reason a lot of people dont wish to use anything other than windows.

You say that like there isn't an MS Office for OS X.

--Dan

Re:AMD hammer for apple?

[brejc8]

nope im saying that like there is one for osx but not for linux/bsd

This is just software.

[Anonymous Coward]

Does anyone know if Darwin can or will be ported to a true 64 bit platform

This is just software. Anything is possible. Darwin's source is available; if you want it to run on a 64 bit architecture, you can port it or pay someone to do it for you. Just kidding. I'm sure there are other Unix systems that run on POWER4, such as AIX or Linux PPC. Why on earth would you want to use Darwin? Darwin != Mac OS X.

64 bit

[Parsec]

Darwin/MacOS X is theoretically 64 bit clean, just waiting for the right (inexpensive) 64 bit CPU. Our campus Apple rep stated that it would soon (with Jaguar, I think) be in sync with one of the *BSDs (IIRC FreeBSD).

I guess if someone really wants to do it, it shouldn't be more difficult than porting to x86, which has already happened. If NetBSD gets ported, that should help.

Chip news sites just make stuff up.

[ivan256]

I was working on a project designing a board that used Motorola's 64bit PPC. They canned the chip in October 2001. They annonuced the cancelation in private meetings to their customers that even knew about it at the first Smart Networks forum in New Orleans. It was NEVER going to be the G5. It wasn't even going to be one of their desktop processors. It was going to be built using their "Book-E" embedded processor spec, and the MMU architecture for it was completely different from the one in the green book. I think that The "we make shit up" Register started the G5 64bit rumour.

Even when the 64bit chip was still in the plans, the G5 was going to come way before it, and was always going to be an evolution of the G4 core. So, the rumors have taken us from the begining, back to the truth, with a whole lot of made up plot in between that never happened.

Re:Chip news sites just make stuff up.

[Hoser McMoose]

I think that The "we make shit up" Register started the G5 64bit rumour.

Actually, I think it's Motorola that made up this rumor, seeing as their publiclly available roadmap shows the G5 with 64-bit support. See this page on their website. Seems like a fairly natural progression of the Gx line, with the next version being the G5 with "32 and 64 bit products, backwards compatibility".

The Register may have blown this out of proportion (they've been known to do that on MANY occasions), but it looks like this time they at least had a tiny smattering of facts to back them up.

If you want it, start writing it.

[dbrutus]

The quickest way to determine if this is a live product would be to start the port using opendarwin.org and see what happens. If you start to get odd or wierd static from Apple, you probably tripped across a secret Apple project and you'll know. If not, then you might just turn it into a live product anyway.

It would certainly be smart both for IBM and Apple to support this as a first step to Mac OS X on RS/6000. Apple could use the increase in its upper end and it would help IBM push some more boxes.

Re:If you want it, start writing it.

[Anonymous Coward]

It would certainly be smart both for IBM and Apple to support this as a first step to Mac OS X on RS/6000.

Mac OS X on an RS/6000... now that's just funny right there. Just doesn't make any sense.

Re:If you want it, start writing it.

[dbrutus]

Why funny? They basically use the same chip family. The PPC was always just a cut down Power.

For IBM, they move boxes so it makes as much sense to sell RS/6000 with Mac OS X as it does to sell it with Linux.

For Apple, they get to keep customers who need higher end computers and are willing to pay 20k for hardware. At least they'll make some money off software rather than losing them to AIX.

Re:If you want it, start writing it.

[norwoodites]

Actually the original Power was a different kind of chip, PowerPC was based on it. The 601 had backwards support for the Power. The 603 and higher got rid of that support. Now the Power series are just PPC processors with Power support in software. The PPC abi and instructions is almost compatible with the original Power's. There are a few removals and few additions to the instruction set. When AIM came together, IBM had the instruction set, Mot had the (funny though) process to make the chips, Apple needed them for their machines.

Re:If you want it, start writing it.

[Anonymous Coward]

What for? So you can finish your Photoshop operations marginally quicker? Oh, you meant as a server OS? Hahahaha! Pretty fucking funny.

BTW they are RS/6000 no longer. IBM calls them the pSeries now.

Yes, it is

[Anonymous Coward]

I can't say too much because of NDA reasons, but you will see more than just Darwin from Apple on Power4. Wait six months.

It would be not that difficult

[Anonymous Coward]

It might already have happened. Whatever CPU Apple might use in future desktop models, either from IBM or Motorola, i wuold be very surprised if a 64 bit port of Darwin would not have been in the works for some time. There are already relatively cheap CHRP Power3 Workstations (like the 44P 170) that one could use for development, and these are not that different (From the software side of things) from Apple products. Even if Apple would not want to release a system based on IBM Power CPU, i'm sure they would want to check wether the OS is 64 bit clean on PowerPC.

System Partitioning from IBM better?

[spike666]

one of the benefits to an IBM RS6000 (not that AIX is a huge benefit IMHO) is that IBM is now sharing all of its low level subsystem code from the Mainframe (os/390 systems) with the other platforms. so things like the way the ibm mainframes allow for systems to be partitioned into multiple machines are now available to the pSeries group.
where Sun partitions across physical CPUs, IBM can now partition w/o regard for the physical CPU. and you can adjust the CPU amount for various machines on the fly...

Re:System Partitioning from IBM better?

[ericman31]

Regatta (or the mainframe for that matter) partitioning is different from SunFire partitioning. You are comparing apples to oranges. On a Sun box if you want to run two servers, one with a data warehouse on it, and the other with an OLTP system on it, and not worry about them stepping on each other, you can. You can't do that on Regatta. The hardware still impacts the whole system. On the other hand, since you can LPAR a Regatta down to 1 CPU, you can consolidate a butt load of servers onto one Regatta box.

Of course, with Sun's Resource Manager you can control workload management at the CPU and process level (kinda like TSO regions). And with Sun's cluster server you get all sorts of nifty virtual server capabilities in a Sun Plex. Basically it is a SysPlex for the UNIX world. For my $$ Sun's approach beats the heck out of IBM's. And I really prefer Solaris to AIX (yes, I've used both, as well as SCO, RedHat, FreeBSD and HP-UX). But, the last couple sentences are just my opinion.

Yay Rumor Mongering

[Sentry21]

I like rumors just as much as the next guy, but this is getting a little silly.

Quoting someone who claims someone he won't name told him Motorola's not making any 64-bit chips as proof ('According to this...') is just silly. I mean, MY source at Motorola says they already have 1024-bit optical chips running at a bajillion petahertz and that OS X 10.a million billion.1 is going to be ported to it Real Soon Now.

If there's one thing Apple does well, it's get people speculating, but the rumors floating around now are pretty baseless. Moving to a system like x86 would be just horrible (x86 assembly is ass-ugly in the first place), not to mention having to support shitty hardware.

Power4/5/whatever is more plausible, though those chips aren't really designed for desktop use, and I don't think IBM loves Apple enough to redesign them. Perhaps Apple would benefit from building a fab plant of their own and doing whatever they like. It would certainly be a perfect compliment to their all-from-us philosophy, and it would give them a lot more freedom. For that matter though, maybe the new IBM .1 micron plant is there to manufacture Power4/5 chips for Apple. Who knows?

All I know is that x86 is a bad, ugly idea, 64-bit or not.

--Dan

Can't say much more than this but..

[Anonymous Coward]

I shouldn't really be posting even this, but it's not particularly important, so here goes..

Apple are currently in negotiations with AMD (as confirmed a few posts above this) so that AMD will eventually produce a special 'locked' x86-64 based chip that Apple can develop its own motherboard for.

This board, along with a ported Darwin (initially it's only on the server end, but eventually for the clients the full OS X bumph will be ported, although once you have Darwin running, that's all pretty easy) will become the top-end Apple server, positioning the current X-Serve as a sort of 'iMac of servers'.

I would like to help in the efforts...

[Anonymous Coward]

I would like to join the efforts in
porting the Darwin OS over to
IBM's Power4 platform.

A few suggestions:
1) I would really suggest the Darwin be ported
to Power4 LPAR as well as the non-LPAR modes.

2) Work with IBM. Put together some convincing
proposals of why IBM would lend support
or assistance to this effort. IBM is already
porting LinuxPPC to POWER4.

3) study the IBM RS/6000 platform.
eg. formerly called CHRP.

4) work on the device drivers.

Hopefully, someday we can run several OSes
( MacOSX, LinuxPPC, and AIX) all at the same time
on POWER4/POWER5. As well as achieving a
much higher performance for MacOSX.
The pSeries 690 is capable of running up to
16 partitions simultaneously. Meaning, we can
run more than 1 OSX, LinuxPPC and/or AIX !!!

CHF

Why bother with Darwin?

[Valdrax]

Seriously. What's wrong with AIX or Linux, the two UNIX operating systems that IBM will have running on them? If you're going to buy a Power4 system from IBM, get the operating systems that it comes with.

In addition, Darwin isn't that great of a UNIX platform. Its thread support isn't all there, its scheduler is terrible, and its missing support for a lot of common advanced UNIX APIs like SysV shared memory and semaphores, and its RPC support isn't fully SunRPC compatible. Trust me. I had to write a multithreaded web server and proxy server that took advantage of all of these UNIX system features, and I kept running into the limitations of Darwin as bundled with Mac OS X 10.1.5.

Stick with AIX or Linux if you want to be doing something serious.

Re:Why bother with Darwin?

[Anonymous Coward]

Ahhhh actually as of August 24 Jaguar OSX 10.2 will be released. The core of 10.2 or Darwin is now in synch with BSD4.4 which is why Jordan Hubbard left the FreeBSD project to work at Apple. Therefore you will soon see Jaguar server released and soon after that a new version of Darwin.

Re:Why bother with Darwin?

[Anonymous Coward]

The main reason IMHO is that we can
get MacOSX running on POWER4 pSeries aka Regatta.
Sure, I grant you that most "unix" software
is already available for linux and/or AIX.
My own personal interest is to see Mac OS X
apps running on the POWER4 in addition to
UNIX apps. Dare to say it?
M$ Office running on MacOSX(Darwin) on
IBM pSeries 630...

Re:Why bother with Darwin?

[Chanc_Gorkon]

This would be a COMPLETE waste of computing power. Granted, a 630 is a deskside system (or can be rack mounted), but I would rather have it in a center as a print server or in some other small server capacity. Noone in my workplace would ever get a 12,000 dollar desktop! This guy could also be used in a HACMP configuration as well so I don't think it would be worth the waste of money it would be running a ported OS/X and a ported MSOffice. It would also be a TOTAL waste to run that on a Regatta (one frame replacing 3 frames). I have 10 servers I could replace very nicely with a Regatta. Only way I would like to see something like this is so that I could serve OS X desktops from a big honkin server for my entire organization. Even then, a Regatta would more then handle it

Will Darwin be ported to power4

[Anonymous Coward]

Isn't Grog (as in the author of The Complete FreeBSD and Porting unix applications)porting FreeBSD to the power 4 platform right now as we speak? And Darwin syncs from time to time with FreeBSD?

GCC

[johnjones]

ok AMD helped SUSE create a GCC backend for x86-64 apple use GCC as the default compiler so thats easy

AMD actually funded ST electronics to do the compiler which does alot better than GCC

in terms of could they move across to a x86-64 technically yes*

*actually they would not do it because backward compatability would be nill and anyone who has done any study of software a backwards compatabilty knows what you should do

so will they move to x86-64 NO
will they move to a 64bit PowerPC ? yes**

** it may not be a Moto part and apple might just buy out motos CPU design licence and use IBM's Fabs like that .10 micron in fishkill

regards

John Jones

get a Linux system

[g4dget]

I use OSX on the desktop and Linux for a compute cluster, so here is my experience.

OSX makes a good desktop system for people who want a no-hassles UNIX system that runs out of the box. It makes it easy for people with modest computer experience to install and maintain their machine. And it runs a bunch of commercial desktop apps.

Once you are talking number crunching with a compute cluster, you are almost certainly better off with Linux-based systems. Linux has extensive cluster administration tools. Cluster installation and maintenance is much easier than OSX installation and maintenance. You can easily run any GUI-based Linux programs remotely using X11, which still beats Apple's remote desktop software in both functionality and performance. You get automatic process migration across a cluster with OpenMOSIX. There is much more numerical and scientific software available for Linux than for OSX, much of it open source; while porting to OSX usually isn't hard, it does require some effort.

Also, you do much better in terms of hardware. While XServe pricing is OK, Pentium and AMD-based servers are still cheaper and offer better performance, and they are offered by many vendors in many different configurations. And, if you like, people already use 64bit Itanium-based machines, or you can still get Alpha-based 64bit compute servers.

Linux better than Mac OS for clustering?

[GPS Pilot]

Check out this article and see if you still feel that way.