Welcome to the MacNN Forums.

rmendis · Mar 21, 2003, 07:11 AM

John Dvorak predicts that Apple will support an Intel platform in 2004. See:
http://www.insanely-great.com/news.php?id=1853
http://www.pcmag.com/article2/0,4149,939886,00.asp

Some of his comments are very bold.

And i doubt the following:
1. Apple licensing Mac OS X for Intel (even if it were for Itanium)
2. 'Transition' dual processor machines
3. Desktop Mac sporting an Itanium

Despite the lacklustre performance of the Itanium (even in it's second iteration) and the massive heat dissipation of the Itanium 2, my guess is that in the next one maybe two iterations, IA64 architecture will really take off.

The design is a very bold move on Intel's part.
Itanium's IA64 is a VLIW processor.
It's to RISC what RISC was to CISC: it's re-simplifying what again have become very complex RISC-CISC hybrid architectures: SPARC, Pentium, POWER, MIPS and even Alpha.

In fact Transmeta's Crusoe and the Itanium have much in common.
Both are VLIW processors.

The difference is how they achieve backward compatibility with legacy ISA (instruction set architectures).
Intel stuck on a x86 and a PA-RISC core - the stupid, but quick out the door method. (No wonder the things is so slow and hot - temperature wise!)

While Transmeta uses the more superior 'software morphing' technology.
In fact, 'software morphing' should allow for more than one ISA to be supported with much greater ease. It surprises me that Intel haven't thought of using it already?
Or have they?

Transmeta's marcket capitalisation is "small change" for Intel.
I wouldn't be surprised if Intel does aquire Transmeta sometime soon, if only for it's 'software morphing' technology for use with future IA64 processors.

Another thing:
Just because Apple may support Intel, doesn't mean that it will ship an Itanium. Itanium is simply Intel's first IA64 platform.
Mark my words, Intel is sure to ship a IA64 Pentium (or one of its derivatives - Xeon, etc) class processor.

Itanium is a brand it is developing for the enterprise/server market.
It's competitors are SPARC, POWER, MIPS, PA-RISC and Alpha.
Not Pentium. Not Athlon.

Anyway, this is all relevent to the thread, cos, it means that Apple if it were to support Intel would most likely compile Mac OS X for Itanium, most likely for the next revision of the Itanium.
But it would support x86 (IA32) ISA for software.

Next revisions to IA64 be it Itanium or Pentium will turbo charge x86 compatibility using - my guess - software morphing, be it Transmeta's or its own.

So supporting Intel technology wise makes sense.
What about the business sense?

And Apple on Intel strategy would beg the question how Apple would do this without following in the footsteps of NeXT.

The way i would imagine is for Apple to purchase it's own brand and produce Mac OS X on Itanium/IA64 machines. For example SGI is a measly 1/20 th the market capitalization of Apple and should make an easy aquisition.
Apple could then for example ship SGI branded Mac OS X superclusters and servers based on Mac OS X on Itanium.

Or dare is say - ressurect the NeXT brand Apple now owns and sell NeXT branded Itanium servers and workstations based on Mac OS X?

I think going with SGI is probably better.
Cos it's an established brand with sales & marketing and existing customers. NeXT isn't.

Anyway, this would be to dip it's toe in the water.

Already Apple has proven that the market exists with the Xserve and Xserve RAID. However, to penetrate and grow that market more, it really needs its' own server brand.

Apple learned in the soul-searching Scully era that it's not worth trying to be what you are not (a business brand) and realised itself in the candy coloured 'comming out' era of Steve Jobs' second tenure as a commodity/home, creative and education brand.

There would be no point in trying to push big iron with an Apple logo

So this sort of market expansion would allow Apple to continue to sell PowerPC hardware uninterrupted, with innovative new processors like IBM's PowerPC 970 and HyperTransport desktops.

While it can ship SGI branded Itanium superclusters (and possibly workstations) to a whole new market. (The idea at the start would be for those machines to not compete very well with Apple's Xserve and PowerPC 970 PowerMacs...at least for the start

Drizzt · Mar 21, 2003, 07:23 AM

John Dvorak comments are always based on feelings and personnal wishes, never on facts and reality (as a whole). He focuses on one little possibility and expends it until he can black out anything else..

In fact, John Dvorak is one huge troll..

Mac_Nacho · Mar 21, 2003, 07:39 AM

My 2c.
Apple departuring from PPC = suicide
Apple developing OSX for x86 = suicide
Apple leveraging more and more on UNIX = future
that's it.
cheers

codywalton · Mar 21, 2003, 08:16 AM

that's it....

I nominate John Dvorak for the "Biggest Douche in the Universe" Award.... I think he could take John Edwards

turtle777 · Mar 21, 2003, 10:04 AM

Besides the fact that Apple is always good for some surprise, I don't think it's gonna happen.

IMHO it would make more sense to produce dual or even quadruple processor machines based on PPC.
Even if you use rather "outdated" PPC models, if you use 4 of them you still beat any Intel processor by far.

Btw, doesn't Itanium "suffer" from low GHz ranges, too ? The fastest is 1.5 GHZ, right ?
So, what's the big advantage anyway ?

-t

turtle777 · Mar 21, 2003, 10:08 AM

Intel declined to comment on pricing, but Insight64's Brookwood expects the cost to stay about the same as the $4,226 price of the current Itanium 2

http://news.com.com/2100-1001-975472.html?tag=rn

For that price, won't I get four G4 processors ?

-t

rmendis · Mar 21, 2003, 10:43 AM

Originally posted by turtle777:
Btw, doesn't Itanium "suffer" from low GHz ranges, too ? The fastest is 1.5 GHZ, right ?
So, what's the big advantage anyway ?

That is largely because of the *two* legacy cores on the chip - one for PA-RISC and the other for x86. (I mean how stupid is that?)

If you want to see a better, smarter and purer design VLIW/EPIC processor consider the Transmeta Crusoe.
It is a smarter, smaller, cooler, more efficient and purer design than the current Itaniums.

I'm fairly confident that Intel will in the next revision or two replace those two on chip cores with 'software morphing' technology.
Having done that, Intel's superiority in chip fabrication and in clock pumping, should *really* turbo charge the IA64 architecture.

VLIW like RISC in its early days, promise much faster clock speed and performance brought about through simplification in design - smaller, cooer, faster, smarter chips.

What Intel has produced now with the Itanium is really a poor offering of VLIW technology. It must have been in order to iron out the VLIW processor technology and architecture and to market the Itanium brand in the server market.

Someone pointed out the high cost (yeah, like anyone is gonna afford a mac with an Itanium *and* a PPC?) makes it prohibitive - only for servers and high end workstations.

But mark my words, i expect IA64 to appear in Pentiums...how soon all depends on Intel...and it's competitors.

NeXTLoop · Mar 21, 2003, 10:55 AM

Dvorak seems to be forgetting one factor... the IBM 970. Why would Apple go thru the horrible process of moving X to Intel, and lose a lot of developers in the process, when there's already, in existence, a chip that is so like the current PPC that it will not even require developers to recompile. Although Cocoa apps would more than likely survive the transfer to Intel ok, a lot of developers worry that Carbon apps would be hopelessly "broke".

In addition, Apple already has a long-standing relationship and partnership with IBM. And IBM has proven that, unlike Motorola who can't seem to get their heads our of the ground, they are a serious chip manufacturer who can keep up with the latest in technology.

And for what it's worth, has anybody ran a high clock speed x86 lately? A buddy of mine has a high speed Athalon. He has a pellet stove in his basement, and as long as its going, he can't even run the PC. It overheats! I mean, come on! Why on earth would I want an oven like that in my laptop?

I'll take the 970 anyday (provided Apple uses it of course).

wataru · Mar 21, 2003, 10:57 AM

WHY WHY WHY won't this stupid thing die already!? There's a thread like this every week!

mitchell_pgh · Mar 21, 2003, 11:06 AM

Apple isn't going to do the Itanium/970 dance.

If... if... if... if... if... if... if... if... if... if... if... if... if... if... Apple ever did port OS X, it would ONLY be an OS X Server port.

Like many of you have said, Apple could go dual or even quad G4 (and perhaps dual or quad 970) for the price of one Itanium. There have been many "Advanced" technologies that just don't take off do to the price. Look at Google etc. (low end solutions...)

The only reason Apple would release a port is so that they could entice sys-admin types over to Apple Hardware...

John Dvorak is a FOOL...

Cadaver · Mar 21, 2003, 11:29 AM

John Dvorak is an idiot.
Regardless of whether or not he likes Apple technology, he's just not even close when it comes to having his finger on the pulse of the industry. And it's not restricted to Apple, either.
He's said FireWire will die. Uhm, my HDTV has a FireWire port on it. He's said Apple's AirPort wasn't worth the expense. Well, now Intel's building it on to their motherboards. Hmmm. Wonder when he'll be wrong again. Answer- probably in his next column.

Leonard · Mar 21, 2003, 12:46 PM

I don't know how anybody could believe a smidge of that article. It's total and utter BS. With the bright and impressive future of the PowerPC 970, and PowerPC 970 Macs coming out within 6 months, how can anyone think Apple is going to use Intel chips. I certainly don't have to wait the 12-18 months he predicts for these to come out to tell that this is BS.

Also, why would Apple build a G4 and Intel hybrid/transition machine. The architecture to support such a machine would probably be too complex and the machine would be too expensive. A G4 PowerMac is already expensive enough.

Dvorak is just a troll trying to get attention.

Superchicken · Mar 21, 2003, 08:03 PM

the worst part is he gets paid to write crap like this...

OpenStep · Mar 21, 2003, 08:49 PM

OH NOS not again

Catfish_Man · Mar 22, 2003, 12:54 AM

Originally posted by rmendis:
John Dvorak predicts that Apple will support an Intel platform in 2004. See:
http://www.insanely-great.com/news.php?id=1853
http://www.pcmag.com/article2/0,4149,939886,00.asp

S

John Dvorak also bashed the mouse at one point. Draw your own conclusions.

beb · Mar 22, 2003, 10:38 AM

One other thing, while speed test now show that the G4 is behind the over-pumped P4 processor in raw speed tests -it's still very impressive that the 1999 introduced G4 processor is on the map.

I think the 970 is the way to go -at least by going with IBM your going with someone who is committed to making faster processors -rather than just concentrating on energy efficiency or whatever Motorola has been doing. And your going with another company who's name is synonymous with computers.

intel may be the way to go eventually, but not today.

rmendis · Mar 22, 2003, 01:23 PM

Originally posted by beb:
I think the 970 is the way to go -at least by going with IBM your going with someone who is committed to making faster processors -rather than just concentrating on energy efficiency or whatever Motorola has been doing. And your going with another company who's name is synonymous with computers.

intel may be the way to go eventually, but not today.

I don't see why it has to be one or the other, when clearly, with Mac OS X, Apple has the technology to support *both*.

Technology is not the issue here...it's the business sense.
The question is: How would Apple support another platform (in this case IA64/x86) without cannibalizing it's own harware sales?

The answer: is by making Mac OS X for Intel a solution for a market Apple currently is not in - that is high end superclusters and servers. Ideally, by purchasing it's own server brand to market such machines, like SGI (or even NeXT :-)

GRAHAMUK · Mar 22, 2003, 04:00 PM

...at least by going with IBM you're going with someone who is committed to making faster processors...and you're going with another company who's name is synonymous with computers.

Ah, seems like only yesterday that IBM were the enemy. How ironic they are now our blessed saviour.

(Quote above edited for grammar)

beb · Mar 22, 2003, 04:26 PM

Heck, I somehow remember Steve saying "Will IBM control it all?!!!"

The eightball answer: Murky, Ask again later.

cwasko · Mar 22, 2003, 08:17 PM

Originally posted by codywalton:
that's it....

I nominate John Dvorak for the "Biggest Douche in the Universe" Award.... I think he could take John Edwards

Hmmm... close call on that one.

To stay on topic here... I actualy seriously doubt that Apple will port the OS soley to Intel anything. They have a huge investment in PPC and AltiVec and anymove away from that would be corporate suicide. Not to mention that Dvorak is a complete and total ignorant fool...

Cipher13 · Mar 23, 2003, 05:44 AM

I can't believe people still listen to Dvorak.

I've said it before and I'll say it again - Apple has an Intel compatible version of OSX locked away kilometers below the earth for a rainy day/tsunami - unless Motorola and IBM suddenly disappear off the face of the earth, along with their fab plants, we'll never see it.

Millennium · Mar 23, 2003, 07:21 AM

Originally posted by Cipher13:
I can't believe people still listen to Dvorak.

Neither can I, Ciph. Dvorak is an unprofessional moron whose reports on Macs are substantiated by little more than his personal vendetta against Macworld. Even if he is an opinion columnist, personal vendettas aren't something which should come into play, but not only does he do it, he's even admitted to it in the past.

Seriously; this guy's a complete jerk. Not worth your time to read.

rmendis · Mar 23, 2003, 07:41 AM

Originally posted by cwasko:
I actualy seriously doubt that Apple will port the OS soley to Intel anything. They have a huge investment in PPC and AltiVec and anymove away from that would be corporate suicide.

Absolutely right.
And Apple will stay with PowerPC for its own hardware.
By that i mean Apple branded hardware.

However, hark back to the NeXT days.

1. When the NeXT computer was first introduced, Steve Jobs, in a typically uncharacteristic gesture and in a great act of diplomacy, forged a strategic alliance with old time Apple arch rival IBM.

(This really was a milestone in tech history, cos i suspect because of that Scully tried to out do Steve and formed the Apple-IBM partnership that blossomed the PowerPC effort.)

What Steve Jobs did was to license NeXTstep to IBM for its POWER workstations and servers.

However, due to various reasons, IBM did not really take advantage of the $25million invested in NeXT to do much with NeXTstep on POWER.

2. Years later, NeXT partnered with HP in the Object*Enterprise alliance in which it subsequently ported NEXTSTEP to HP's PA-RISC workstations. This deal did bear fruit of a shipping version of the OS on HP hardware unlike the IBM deal, though HP lost interest afterwhile probably due to lacklustre demand. Customers must also have been confused whether to go for NeXTstations or high end HP workstations. That in fact may have been the reason why the relationship eventually fizzeled out.

3. A few years on, NeXT abandons harware and NeXT becomes a software only company. It ports NEXTSTEP to Intel. Then it forges an alliance with its old time NeXT arch enemy Sun. (A little bit of history repeating). Together with Sun, NeXT abstracts out an open published set of APIs from NEXTSTEP called OPENSTEP. NeXT licenses OPENSTEP to Sun for Solaris. Eventually, Sun lost interest in OPENSTEP for Solaris while NeXT's focus on OPENSTEP for Windows turns out to be more popular with its clients. The most important product of the Sun-NeXT deal was NEXTSTEP for SPARC which turned out to be more popular with it's customers than OPENSTEP for Solaris.

So Steve tried several times to expand or broaden the NEXTSTEP market by making strategic alliances with high end workstation and server vendors.

Almost always the problem was the conflict of interest and lack of demand.

Apple's Xserve and Xserve RAID have demonstrated that there indeed is a market for Mac OS X servers. Now all it needs to do is to find an alliance/solution where there isn't the kind of conflict of interest that existed in the previous NeXT-IBM, NeXT-HP and NeXT-Sun deals.

I won't be surprised if Steve Jobs tries to get it right this time.
And this time i suspect, or at least it wouldn't surprise me if Apple purchase a small server vendor to release Mac OS X Server on. NeXT certainly did not have that option. But Apple does.

SGI IMHO is the most obvious partner/aquisition. Because of its similar customer base and very small $250 million price tag

SGI in particular (as are other smaller server vendors) have bet the farm on Itanium. So if Apple were to purchase SGI and run Mac OS X on SGI servers, the Itanium Altix servers would make the most likely target. http://www.sgi.com/altix

Let's hope that if and when Steve Jobs does surprise the world again, this time, he will indeed get it right

nsxpower · Mar 24, 2003, 04:50 AM

When are people going to realize that Dvorak is on crack? Dvorak proved to be an idiot when it comes to Macs. According to him Apple should have died several times over already. Not to mention all the great "interpritations" of rumors tunrned into "newsworthy material" on behalf if Dvorak's editor ... who has to be even more clueless then Dvorak himself.

Apple moving to x86/ia64?!? WTF?!? Those are two completely different architectures!
If he actually bothered to read some recent rumor sites he would have realized that IBM's 970 will probably end up in future Macs.
If he actually bother to read up before pulling this "article" out of his ass (I would love to shove it back up there, along with a couple XP boxes and Microsoft mice), he would realize that Apple has a history of supporting 2 architectures at the same time. Apple had "fat" binaries since System 7 days or something that could run on 68K and PPC machines. Same is said to be true of upcoming OS X ...

God, I am so sick of no-nothing-have-a-column writers for PC magazines, like Dvorak. I think I should become an independent columnist. You know sit on my ass for a month downloading mp3 from KaZaA, pull some BS story out of my ass before sending it to the editor and then get praised for writing such utter rubbish. Yeah, and get paid for the whole thing --- so that I can run out and buy more Microsoft software.

Dvorak is "the uber-troll"!

freakboy2 · Mar 26, 2003, 04:49 PM

Originally posted by codywalton:
that's it....

I nominate John Dvorak for the "Biggest Douche in the Universe" Award.... I think he could take John Edwards

If the both died at the same time, which one would rise from the dead first?

and who wuld care.

SpeedRacer · Mar 26, 2003, 07:29 PM

Originally posted by NeXTLoop:
...Why would Apple go thru the horrible process of moving X to Intel, and lose a lot of developers in the process, when there's already, in existence, a chip that is so like the current PPC that it will not even require developers to recompile. Although Cocoa apps would more than likely survive the transfer to Intel ok, a lot of developers worry that Carbon apps would be hopelessly "broke"...

Exactly. But what about down the road when all your flagship apps *are* running in Cocoa, you've built yourself up as a full-fledged consumer software company with the premiere "Office" suite for digital media, you've evolved hardware sales to other hardware offerings (iPod) and you've proven the ability to develop both a formidible market share for your OS platform (OS X) as well as a superior, rapid development alternative to the Windows monopoly via native Cocoa and Java...

What then Intel?

Star Trek... the NeXT Generation anyone?

Eug · Mar 27, 2003, 07:07 AM

I think it's worth repeating:

Dvorak is on crack. His opinions are as bad as David Coursey's.

rmendis · Mar 27, 2003, 10:20 PM

Originally posted by SpeedRacer:
Exactly. But what about down the road when all your flagship apps *are* running in Cocoa, you've built yourself up as a full-fledged consumer software company with the premiere "Office" suite for digital media, you've evolved hardware sales to other hardware offerings (iPod) and you've proven the ability to develop both a formidible market share for your OS platform (OS X) as well as a superior, rapid development alternative to the Windows monopoly via native Cocoa and Java...

What then Intel?

I think it's logical and right to expect Apple to broaden it's market by supporting another platform. Of the two alternative (surviving) chip architectures SPARC and Intel, it would be in Apple's interest to focus it resources on supporting Intel. Soon maybe only Intel and IBM will be players in the field.

I don't expect any announcements to happen this year at WWDC but perhaps next year, once the transition to Mac OS X is complete (I guess to Steve Jobs that means when they no longer ship a Classic environment on Mac OS X?) and when the MHz gap has been closed with newer PowerPC processors.

Itanium is the obvious (first) target for Mac OS X for Intel.
As Itanium machines are priced out of the range of PowerPC boxes.

Also note that if Intel can pump an ancient CICS architecture in excess of 3GHz, imagine what they can do with a VLIW/EPIC implementation. Currently the Itanium speed is hampered by the PA-RISC and Pentium cores. Once they no longer have to ship those cores (by using software morphing to support the two ISAs) on chip, expect the GHz gap to widen even more

freakboy2 · Mar 28, 2003, 03:02 PM

This can't ever happen. Apple makes most of it's money from hardware. If they open up their hardware to the likes of Dell they'd get blown out of the water and slaughtered like sickling pigs.

If OSX could run on intel hardware, I'd go build myself a 400$ machine that was 2x as fast as the apple machine that costs 3,000$ and then I'd spend 120 bucks on jaguar and be done with apple.

Remember what happened when apple tried to have clone makers make macs? they started losing money like crazy.

Bye bye.

Fb2

Originally posted by rmendis:
I think it's logical and right to expect Apple to broaden it's market by supporting another platform. Of the two alternative (surviving) chip architectures SPARC and Intel, it would be in Apple's interest to focus it resources on supporting Intel. Soon maybe only Intel and IBM will be players in the field.

I don't expect any announcements to happen this year at WWDC but perhaps next year, once the transition to Mac OS X is complete (I guess to Steve Jobs that means when they no longer ship a Classic environment on Mac OS X?) and when the MHz gap has been closed with newer PowerPC processors.

Itanium is the obvious (first) target for Mac OS X for Intel.
As Itanium machines are priced out of the range of PowerPC boxes.

Also note that if Intel can pump an ancient CICS architecture in excess of 3GHz, imagine what they can do with a VLIW/EPIC implementation. Currently the Itanium speed is hampered by the PA-RISC and Pentium cores. Once they no longer have to ship those cores (by using software morphing to support the two ISAs) on chip, expect the GHz gap to widen even more

chris v · Mar 28, 2003, 03:12 PM

If Apple goes Intel, will Safari still get tabs????

CV

rmendis · Mar 28, 2003, 04:09 PM

Originally posted by freakboy2:
This can't ever happen. Apple makes most of it's money from hardware. If they open up their hardware to the likes of Dell they'd get blown out of the water and slaughtered like sickling pigs.

If OSX could run on intel hardware, I'd go build myself a 400$ machine that was 2x as fast as the apple machine that costs 3,000$ and then I'd spend 120 bucks on jaguar and be done with apple.

Who said Apple would port Mac OS X to run on Pentium (x86)?!

I think it's been repeated several times in this thread that that is not going to happen!!!

Porting the Mac OS X to Itanium (though software/apps be ported to x86 as the native VLIW IA64 isn't really a proper ISA) means that it will run only on high end workstations and servers.

Also nobody mentioned licensing.

My guess is that Apple will follow in the viens of NeXT and either license it to one large Itanium server shop OR purchase a smaller vendor like SGI and produce Mac OS X on Itanium superclusters and servers.

Giv · Mar 29, 2003, 07:52 PM

Originally posted by rmendis:
Porting the Mac OS X to Itanium (though software/apps be ported to x86 as the native VLIW IA64 isn't really a proper ISA) means that it will run only on high end workstations and servers.

IA64 is not a proper ISA ? Ah, that's a good one !

I have been working on IA64/IPF (Itanium Processor Family) for close to 4 years now (originally using simulators). I assure you that IA64 is a proper ISA.

To convince yourself, go look at: IPF manuals
In particular: Application architecture

This misinformation may come from the fact that traditional VLIW processors were not upward compatible (a binary for a given processor may fail when running on a newer processor). However IPF doesn't suffer from that issue, and that's probably one reason Intel/HP don't call it VLIW but use the marketing term EPIC.

A binary compiled for the first Itanium (code-named Merced) will run without modification on Itanium2 (code-named McKinley), despite the fact that Itanium2 is a very different micro-architecture.

You may still want to recompile for Itanium2 to get the best performance, that's exactly the same as ppc601 -> G3 -> G4 transitions. The microarchitecture changes, so there are opportunities for improvements in the code generated.

And running x86 binaries on Itanium processors is not a very smart idea. The x86 compatibility is mostly here for compatibility, not for speed.

I hope this helps clear some confusion. I am not here to hype Itanium but to debunk some myths. Feel free to criticize it, but stick to facts

Giv · Mar 29, 2003, 08:22 PM

Originally posted by rmendis:
That is largely because of the *two* legacy cores on the chip - one for PA-RISC and the other for x86. (I mean how stupid is that?)

It may have been stupid, but thankfully that's not the case.

Itanium and Itanium2 processors contain a single core/ALU/execution unit/... .

In this core there is a unit that converts x86 instructions to equivalent instructions for the "native" core (not necessarily IA64 ISA instructions, they are a level below the ISA and may be using special hardware feature only put there to speed-up emulation).

This "x86" part is really just an add-on to the native core, the one that does the real work. See the Mckinley processor block diagram and the McKinley die to get an idea of how much space that emulation engine takes (about the same space as the 256kB L2 cache).

And there isn't any PA-RISC compatibility in hardware. There is an emulator available on HP-UX, called Aries. This emulator is software-only (almost completely user-space, with minimal kernel involvement). The emulation is made a bit simpler by the fact that the Itanium instruction set has been influenced heavily by PA-RISC, but it's still a fantastic piece of software. Think of Aries as being similar to the 68k emulation that Apple used to transition to PowerPC.

I know some folks who ran a PA version of Oracle under Aries (as they waited to get hold of the native version) ! A bit slow but usable and working (they just used a backup disk of their DB running on PA). Amazing.

I won't comment on your predictions regarding Itanium's future, due to NDAs. I'll just say that I have no idea if Apple will port to Itanium, but I'll send my resume to Apple in the hour if they announce such a port. I'd love to help Apple move into the enterprise space. I certainly don't expect Apple to drop the PowerPC even if they had an Itanium version.

Oh, and until we hear anything from Apple, I'll believe Dvorak's article is really an April's fool column (it appeared in April's issue). Don't panic.

RMXO · Mar 29, 2003, 08:43 PM

Originally posted by Giv:
Oh, and until we hear anything from Apple, I'll believe Dvorak's article is really an April's fool column (it appeared in April's issue). Don't panic.

would be a great April Fool's joke......

rmendis · Mar 29, 2003, 11:42 PM

Originally posted by Giv:
In this core there is a unit that converts x86 instructions to equivalent instructions for the "native" core

I see.

I got my information from:
http://www.fool.com/portfolios/rulem...aker000225.htm

Less authorative than the IPF manual or block diagrams.

However, is it possible that the x86 compatibility unit negatively impacts the performance of the Itanium, for one reason or the other?
Also in your opinion how does this compare to Transmeta Crusoe's software morphing technology for achieving x86 ISA compatibility?

rmendis · Mar 30, 2003, 12:01 AM

Originally posted by Giv:
IA64 is not a proper ISA ? Ah, that's a good one !

This misinformation may come from the fact that traditional VLIW processors were not upward compatible (a binary for a given processor may fail when running on a newer processor). However IPF doesn't suffer from that issue, and that's probably one reason Intel/HP don't call it VLIW but use the marketing term EPIC.

Again: I see :-)

This is good news, because upward compatibility is/was an issue.
In which case, if Apple were to support Intel/Itanium, compiling all software to IA64 would make sense instead of x86.

A question: So what in your opinion is the reason Intel is unable to speed pump this theoretically simpler and more elegant processor architecture to the dizzying heights of it'x x86 Pentium? (3Ghz->)

Do you also think that the Itanium EPIC architecture may be used to power future Intel IA32 processors (Pentiums, Xeons) ala the Transmeta Crusoe?

Also how come such a valuable technology/company such as Transmeta (with a tiny market capital and shaky business and future alone) has still not been aquired?

It's software morphing technology is an excellent solution for HP in particular. It could offer HP the opportunity to migrate it's legacy PA-RISC, MIPS and Alpha systems to Itanium...by say an Itanium with software morphing - one processor loaded with different ISA compatibilty software. eg: MIPS + IA64 for HP's NonStop servers, Alpha + IA64 for HP's AlphaServers, etc.

Giv · Mar 30, 2003, 11:59 AM

Originally posted by rmendis:
However, is it possible that the x86 compatibility unit negatively impacts the performance of the Itanium, for one reason or the other?
Also in your opinion how does this compare to Transmeta Crusoe's software morphing technology for achieving x86 ISA compatibility?

Transmeta's software morphing looks fantastic on paper and I had hopes Crusoe would validate the approach. Unfortunately the performance disappointed me (and most observers). I'd be interested to see what this technology could achieve on a better ISA (maybe one specially taylored for easy morphing).

Beyond that I can't comment. Apologies, but I get to see a lot of confidential info as part of my job, and I'd rather keep my job. It would be too easy to give an opinion that would really be hints about what is in plan.

Giv · Mar 30, 2003, 12:56 PM

Originally posted by rmendis:
This is good news, because upward compatibility is/was an issue.
In which case, if Apple were to support Intel/Itanium, compiling all software to IA64 would make sense instead of x86.

Correct.

A question: So what in your opinion is the reason Intel is unable to speed pump this theoretically simpler and more elegant processor architecture to the dizzying heights of it'x x86 Pentium? (3Ghz->)

I don't believe "unable" is the right word. Current Itanium processors have been designed for the server and technical computing markets. As benchmarks (and ports of real applications) show, they do an excellent job there, even if the current Itanium2 doesn't go above 1Ghz.

Let's also keep in mind that when those processors have been designed, x86 frequencies weren't as crazy. The Mhz wars between AMD and Intel hadn't started.

The Itanium2 pipeline is really short (comparable to a G4). If the designers wanted to target very high frequencies they might have gone for a longer pipeline.

Intel recently discussed a processor code-named "Montecito", supposedly appearing in 2005. My understanding is that this will be the first Itanium processor designed since the GHz war started in earnest. We'll see what frequency they achieve then.

Do you also think that the Itanium EPIC architecture may be used to power future Intel IA32 processors (Pentiums, Xeons) ala the Transmeta Crusoe?

I don't think the IPF ISA would look the way it does if speedy execution of IA32 binaries was the primary goal. I wouldn't be surprised if Transmeta had a much better underlying micro-architecture for such a purpose. But Transmeta's technology is only one in a spectrum of hardware/software solutions, so I wouldn't bet on or against anything happening

Also how come such a valuable technology/company such as Transmeta (with a tiny market capital and shaky business and future alone) has still not been aquired?

I don't think they demonstrated enough value yet. They were targeting high-performance, failed miserably and re-purposed their processors as low-power solutions. However they were mostly interesting there because neither Intel nor AMD had a low-power solution. Now those two have woken-up and I am not convinced that Transmeta has a significant advantage anymore. They may still have lower power-consumption, but not by enough, and their performance can't be compared to a Banias processor.

It's software morphing technology is an excellent solution for HP in particular. It could offer HP the opportunity to migrate it's legacy PA-RISC, MIPS and Alpha systems to Itanium...by say an Itanium with software morphing - one processor loaded with different ISA compatibilty software. eg: MIPS + IA64 for HP's NonStop servers, Alpha + IA64 for HP's AlphaServers, etc.

Consider Transmeta's performance and the fact that the Crusoe micro-architecture is specifically designed to support x86 code-morphing. This is not the case for IPF processors and MIPS/Alpha binaries.

Furthermore code-morphing hasn't proved it can deliver performance even close to a native compilation. I'm still hoping but the evidence is against me. You may want to look at the proceeds of the CGO conference for further insight about software/hardware optimizations. That was an excellent conference.

Another note: HP has announced that True64 was going away, the official migration target being HP-UX on IPF. This makes software emulation almost impossible, due to the differences between both OSes. And True64 customers were very motivated by performance, so emulation makes even less sense for them. Also keep in mind that True64 had a low market-share, much lower than HP-UX, so many applications available on True64 will already be available native on HP-UX IPF (or emulated from HP-UX PA).

So only MIPS/NonStop and Alpha/VMS would really benefit from code-morphing. Did HP provide details on what they are doing to facilitate those migrations ? For all I know they could be repurposing their Aries PA->IPF translator.

If there were any truth to the Apple/Itanium rumor, Apple could do worse than talking to HP about using Aries as a base for a PPC->IPF emulation solution. It's in the best interest of HP to make Itanium succeed.

rmendis · Mar 30, 2003, 10:15 PM

Originally posted by Giv:
If there were any truth to the Apple/Itanium rumor, Apple could do worse than talking to HP about using Aries as a base for a PPC->IPF emulation solution. It's in the best interest of HP to make Itanium succeed.

How does HP's Aries software emulation compare with Transmeta's software morphing?

I imagine the technology is very much in its infancy.
It would be nice to see a company like Intel invest in developing it.

And yes, say if and when Apple does support IA64, it would help to have some sort of PPC emulation or morphing.

rmendis · Mar 30, 2003, 10:48 PM

Originally posted by Giv:
Furthermore code-morphing hasn't proved it can deliver performance even close to a native compilation. I'm still hoping but the evidence is against me.

Well, it is to be expected that a native compilation will perform better.

However, VLIW/EPIC processor architectures are much simpler and take less power + space on die.
In the future (next couple of years) would not the potential for software code morphing not grow by using that extra space for additional parrallism e.g: a second core/pipeline, more execution units or even a larger cache?

In otherwords isn't on chip code morphing software likely to benefit from more powerful micro-architectures available in the future?
I expect there ought to be a convergence (in performance) between native and code-morphed software....IMHO the technology is in it's infancy and it shouldn't be dismissed so quickly. Some technologies take time to develop.

Also, with more complex and parrallel micro-architectures, doesn't on-chip software morphing have the potential to be more sophisticated than hardware morphing? Or even branch prediction/instruction reordering of Intel's x86 offerings? So much so that perhaps soon, even with the performance penalty of code morphing, a VLIW/EPIC powered x86 processor could outperform one based on its native architecture?

rmendis · Mar 31, 2003, 11:25 AM

Originally posted by Giv:
So only MIPS/NonStop and Alpha/VMS would really benefit from code-morphing.

Another point in favor of software code morphing particularly for Itanium has to do with Java.

Java is a big player in the enterprise space and is likely to remain one (there is a certain inertia in the enterprise that keeps technology around. eg: Cobol)

As Itanium is targeted at the enterprise market this may be a feature Intel may want to pay attention to. (It would also have important consequences for Apple if it were at some point to support Itanium).

A VLIW/EPIC could be ideal to implement a Java processor: here of course the code morphing software becomes fairly complex: Java bytecode -> VLIW instructions.

If the Itanium implements software code morphing, could it not fascilitate the loading of additional ISAs? In particular something as complex as Java? Have Sun thought of this already?

sbjordal · Mar 31, 2003, 12:37 PM

Originally posted by beb:
One other thing, while speed test now show that the G4 is behind the over-pumped P4 processor in raw speed tests -it's still very impressive that the 1999 introduced G4 processor is on the map.

It's nothing short of pathetic, that's what it is!!! Apple should have smacked Motorola around a long time ago. Apple is suffering hard core because the processor is essentially the same as it was in 1999.
Come on..how is this impressive????

Drizzt · Mar 31, 2003, 12:42 PM

Originally posted by sbjordal:
It's nothing short of pathetic, that's what it is!!! Apple should have smacked Motorola around a long time ago. Apple is suffering hard core because the processor is essentially the same as it was in 1999.
Come on..how is this impressive????

What is impressive, IMHO, is the G4 design.

What is not is the P4's.

I mean.. look at it this way.. my 400Mhz G4 is about the same speed as a Celeron 800.. with unoptimised code!

I just hope IBM will get the contract.. and that it will outclass anything X86 for a while..

sbjordal · Mar 31, 2003, 01:31 PM

Originally posted by Drizzt:
[B]What is impressive, IMHO, is the G4 design.

What is not is the P4's.

I mean.. look at it this way.. my 400Mhz G4 is about the same speed as a Celeron 800.. with unoptimised code!
B]

Right....just like 4K78 had debug code that Apple forgot to remove...
Your example is dated ca. 1999, and doesn't apply today. The G4/400 was prob. as fast as a celeron 800, but still....The celeron was Pentium's little brother...2ndly, unlike Motorola, Intel put the pedal to the metal and has steadily cranked out faster processors...I mean, Come on!!! It's getting to a point where a dual g4 config is still slower than a Pentium running twice the clockrate.

Drizzt · Mar 31, 2003, 01:37 PM

Originally posted by sbjordal:
Right....just like 4K78 had debug code that Apple forgot to remove...
Your example is dated ca. 1999, and doesn't apply today. The G4/400 was prob. as fast as a celeron 800, but still....The celeron was Pentium's little brother...2ndly, unlike Motorola, Intel put the pedal to the metal and has steadily cranked out faster processors...I mean, Come on!!! It's getting to a point where a dual g4 config is still slower than a Pentium running twice the clockrate.

I'm writing this on a 1.8Ghz P4M.. and believe me.. a Athlon 1800+ feels quite faster!

The processor is brainless, it's like having a 250HP motor.. that is 0.5L

I prefer my G4 400 to the P4.. believe me.

Even if the Celeron is the Pentium's little brother, it's not normal that it gets that slow when compared to a G4.

The problem is Motorola, not the G4.

sbjordal · Mar 31, 2003, 01:54 PM

Originally posted by Drizzt:
I'm writing this on a 1.8Ghz P4M.. and believe me.. a Athlon 1800+ feels quite faster!

I love Athlon, and if I have to purchase processors, I buy Athlon. Agree with you on this one...

I prefer my G4 400 to the P4.. believe me.

There is no way your G4/400 is faster that the P4 you mentioned above. This is bullshit. We're not talking about prefering...we're talking about speed. I too prefer my BW g3/350, but compared to my Athlon 2200+ PC, it's a piece of junk..btw, the g3 is a pretty good server

Even if the Celeron is the Pentium's little brother, it's not normal that it gets that slow when compared to a G4.

Sure, it's completely normal. The Celeron is a slow processor. Which is the reason why you could compare it to a G4 processor running at 1/2 of the mhz speed 4 years ago.

The problem is Motorola, not the G4. [/QUOTE]

AMEN. That's the whole point. Motorola slowed down production and R&D. Apple should have picked up on this two years ago. I love my macs (all 4 of them), I dig OSX wholeheartedly, but the experience would be completely different had the processor been faster.

Drizzt · Mar 31, 2003, 02:04 PM

Originally posted by sbjordal:
I love Athlon, and if I have to purchase processors, I buy Athlon. Agree with you on this one...

I prefer my G4 400 to the P4.. believe me.

There is no way your G4/400 is faster that the P4 you mentioned above. This is bullshit. We're not talking about prefering...we're talking about speed. I too prefer my BW g3/350, but compared to my Athlon 2200+ PC, it's a piece of junk..btw, the g3 is a pretty good server

Even if the Celeron is the Pentium's little brother, it's not normal that it gets that slow when compared to a G4.

Sure, it's completely normal. The Celeron is a slow processor. Which is the reason why you could compare it to a G4 processor running at 1/2 of the mhz speed 4 years ago.

The problem is Motorola, not the G4.

AMEN. That's the whole point. Motorola slowed down production and R&D. Apple should have picked up on this two years ago. I love my macs (all 4 of them), I dig OSX wholeheartedly, but the experience would be completely different had the processor been faster. [/QUOTE]

I prefer my G4 because the balance between the components of the computer is better. I never said it was waster...

A Celeron 800 is a P3 800 with less L2 cache, a 400Mhz G4 with twice the cache shouldn't be as fast

The problem is elsewhere..

The thing is, everybody thinks the most important piece of hardware in the computer is the CPU. I believe it's the hard-drive, the ammount of memory and the memory bandwith. The weakest link is the strenght of the chain.. never forget that..

mamamia · Mar 31, 2003, 04:46 PM

Originally posted by codywalton:
that's it....

I nominate John Dvorak for the "Biggest Douche in the Universe" Award.... I think he could take John Edwards

super sweet!

coolmacdude · Mar 31, 2003, 07:15 PM

That might be a bit of a stretch. It would be close though.

Giv · Mar 31, 2003, 11:38 PM

Originally posted by rmendis:
Another point in favor of software code morphing particularly for Itanium has to do with Java.
...
If the Itanium implements software code morphing, could it not fascilitate the loading of additional ISAs? In particular something as complex as Java? Have Sun thought of this already?

Java on Itanium is fairly new and still has ways to improve, but it is already very competitive. Look at Java benchmark results from HP's version of Hotspot on www.spec.org .

Keep in mind that only 1.3 scores are listed there, while 1.4 is likely to give higher numbers. And Hotspot was designed by Sun for Sparc and x86, it may take a bit longer (or a different VM) to really do a good job on Itanium.

Not all is rosy though. It typically takes more effort to generate decent or good IA64 code than for x86/Risc processors. On short running apps it may put IA64 at a disadvantage unless code is reused/shared between invocations (something Apple has just added to their version of Hotspot 1.4.1). On long-running apps this is balanced by the higher performance.

Oh, don't get mirred by the "code morphing" marketing speak. Hotspot is typically called a JITC (Just In Time Compiler), but as far as Java is concerned it could just as well be called "code morphing".

Transmeta's code morphing is "just" JIT Compilation for a different kind of VM, the x86 VM.

While I am here... Another common myth is that Java is faster on "Java processors". This is wrong, Java is fastest on fast processors, those that aren't specifically designed for Java.

I am sure the 970 will be an excellent processor when running Java.