What Was The Original Promise of The Itanium Processor?

[tompkin]

I always thought that Intel originally had planned for the Itanium processor to succeed the Pentium. I read somewhere that the problem they had was that they couldn't figure out how to get it cool sufficiently.

 

For some reason, when it didn't work out on the desktop it seemed a bit of a letdown to me. I was wondering if there was any loss of capability in the processors that were eventually used, since the Itanium was deemed unsuitable for the desktop? 

[+virtorio MVC]

I've never heard of Itanium processors being aimed at the desktop market. They're true 64-bit processors and have a totally different instruction set to x86 processors. They're aimed at high performance servers and clusters and they have a bunch of drawbacks when they're not being used for their intended purposes. You wouldn't want one of these in your home machine.

[BajiRav]

IIRC that was intel's original vision for 64bit processors until AMD came up with the x86-64bit stuff.

[DevTech]

I always thought that Intel originally had planned for the Itanium processor to succeed the Pentium. I read somewhere that the problem they had was that they couldn't figure out how to get it cool sufficiently.

 

For some reason, when it didn't work out on the desktop it seemed a bit of a letdown to me. I was wondering if there was any loss of capability in the processors that were eventually used, since the Itanium was deemed unsuitable for the desktop? 

 

I think you might have the Itanium confused with the Pentium 4. The Pentium 4 ran into the current silicon process limits that mainly exist to this day at around the 4 ghz point. Multiple cores and making the instruction set more effient is the main approach and the trend is always to more cores unless some breakthrough comes along. A small group at Intel came up with an energy efficient mobile CPU based on the Pentium 3 and that led to the first Core CPU. The P4 was the wrong branch on the evolutionary tree!

 

The Itanium was designed well before those events as a new 64 bit CPU with a new instruction set that didn't carry forward the legacy crud of the 8086. The problem at the time was the old silicon process required a huge die size which made the Itanium too expensive for a desktop CPU. Intel added lots of cache and sold it to the server market where it performed adequately but not exceptional. That market was clobbered by multi-core 8086 based CPU such as the Intel Xeon and the AMD Opteron. There was a brief period of time where there was some doubt as to whether the Itanium or the AMD 64 bit architecture would win out with AMD's 64 bit extensions to the crufty 8086 architecture being far from elegant. But AMD's design was cheaper by a large amount. Once that battle was over, the cross-license legal agreement court mandated for Intel and AMD permitted Intel to simply adopt the AMD architecture at zero cost. For face saving reasons the Itanium limped on as "Walking Dead"

[smooth_criminal1990]

^That's pretty much it. Thought I'd throw in, Itanium and Itanium 2 were examples of a "Very Long Instruction Word" processor architecture, meaning that it could execute instructions in parallel, and do lots more per clock cycle, than a conventional Pentium 4 or similar.

 

But yeah, the new architecture meant it was completely incompatible with existing x86 processors.

[Lant]

Also programs compiled for x86 had to be run under a slow compatibility layer for IA64. x86-64 had the advantage that any program compiled for x86 could just just fine as it was unaware of the extra extensions.

There probably wasn't the desire to spend time and effort porting existing legacy systems to be support being compiled in 64-bit mode. Much easier and cheaper to do nothing and keep running it without change.

We've seen a slow shift to support 64-bit in code, but it requires a lot of teaching about things like ptr_diff_t and size_t, rather than just assuming that an int would be enough (which it isn't under LLP64).

[PGHammer]

I always thought that Intel originally had planned for the Itanium processor to succeed the Pentium. I read somewhere that the problem they had was that they couldn't figure out how to get it cool sufficiently.

 

For some reason, when it didn't work out on the desktop it seemed a bit of a letdown to me. I was wondering if there was any loss of capability in the processors that were eventually used, since the Itanium was deemed unsuitable for the desktop? 

Itanium was planned originally for the workstation market at the time when RISC (and especially DEC's Alpha) was the "hot ticket".  What wrecked things was, ironically, not only AMD's Opteron, but Intel's own Pentium 4 - both of which offered better bang for buck.  (It certainly didn't help that both Opteron and P4 could also run garden-variety x86 software - which no RISC CPU could do.) Further, Itanium was why Intel did NOT acquire the rights to the Alpha itself when it acquired the rest of DEC Semiconductor (which included DEC's far-from-insignificant Ethernet business).

[tompkin]

Itanium was planned originally for the workstation market at the time when RISC (and especially DEC's Alpha) was the "hot ticket".  What wrecked things was, ironically, not only AMD's Opteron, but Intel's own Pentium 4 - both of which offered better bang for buck.  (It certainly didn't help that both Opteron and P4 could also run garden-variety x86 software - which no RISC CPU could do.) Further, Itanium was why Intel did NOT acquire the rights to the Alpha itself when it acquired the rest of DEC Semiconductor (which included DEC's far-from-insignificant Ethernet business).

I guess this was what I remembered. It was planned for the workstation market at one time. I knew that it was true 64-bit architecture which was why I was interested in it. Sort of like the AS/400 in that respect. Another reason why i probably thought that was that I remember that Microsoft put out some Itanium software at one time. I'm assuming Windows Server. 

[Aergan]

AMD bringing x86-64 to the desktop market and server markets forced Intel to adopt the same type of approach and helped direct where Microsoft saw fit to invest their time OS & application wise.

AMD have / had a long standing relationship with Microsoft which helped them get support for their hardware whereas Intel (I believe) had to invest significant financial resources into getting IA-64 support from Microsoft (not to mention initially being a niche server-only OS development project with likely no foreseeable desktop options at the time).

[+Fahim S. MVC]

It was a completely different animal from the x86 processors.  Whereas x86-64 processors are broadly backwards compatible with their 32 bit parents (and can run 32 bit code natively), IA64 was not.

 

They weren't CISC or RISC, but a different architecture altogether. Software would need to be specifically compiled to work on it.

[DevTech]

AMD bringing x86-64 to the desktop market and server markets forced Intel to adopt the same type of approach and helped direct where Microsoft saw fit to invest their time OS & application wise.

AMD have / had a long standing relationship with Microsoft which helped them get support for their hardware whereas Intel (I believe) had to invest significant financial resources into getting IA-64 support from Microsoft (not to mention initially being a niche server-only OS development project with likely no foreseeable desktop options at the time).

 

As a small note, "same type of approach" is more like "identical copy" which Intel is allowed to do as a result of the giant court battle between Intel and AMD.

 

As a wildly off-topic addendum to that, it is also little known that Apple and Microsoft have a broad cross-license agreement as a result of Microsoft buying enough Apple stock to save them from destruction. That's why Apple has taken just about everyone to court except Microsoft.

[neufuse Veteran]

As a small note, "same type of approach" is more like "identical copy" which Intel is allowed to do as a result of the giant court battle between Intel and AMD.

 

As a wildly off-topic addendum to that, it is also little known that Apple and Microsoft have a broad cross-license agreement as a result of Microsoft buying enough Apple stock to save them from destruction. That's why Apple has taken just about everyone to court except Microsoft.

 

Hey now, don't point out the MS stock buy to ultra-pro apple people, you might get your face ripped off... they to this day claim it wasn't' about saving the company, but only about cross licensing... this of course was at  a time MS was being considered a monopoly and if apple went under their chief competitor was done.....

[ctebah]

Hey now, don't point out the MS stock buy to ultra-pro apple people, you might get your face ripped off... they to this day claim it wasn't' about saving the company, but only about cross licensing... this of course was at  a time MS was being considered a monopoly and if apple went under their chief competitor was done.....

 

Well obviously Apple would have bankrupted without MS investment even though MS only invested 150 million and Apple had 1.2 billion in cash.  

[MFH]

The idea with the Itanium was to have a new processor design (without the "legacy cruft" of the x86). What most people overlook was the idea of inherent parallelization in the Itanium design - an interesting idea in theory that miserably failed in reality. (The idea was that the compiler could parallelize code for the programmer - something that to this day is not really possible completely automatically.)

[DevTech]

Well obviously Apple would have bankrupted without MS investment even though MS only invested 150 million and Apple had 1.2 billion in cash.  

 

I am aware that there is a popular recent fashion to create a retrospective history of that time using various approaches and it puzzles me that it could be worth the effort. The permanent "Get Out Of Jail Free Card" that Apple handed out to Microsoft is proof enough of the value to Apple of that deal at that time (for whatever reason) and it is the only reality that remains today and is a valuable asset for Microsoft.

 

If Apple would have died or not is only an abstract alternative history speculation that always seems to stop with different evaluations on the probability instead of the more interesting S.F. type of speculation on what would a world without Apple be like? Would Microsoft have been split up into separate companies? Would the publc be less interested in battery life and fashion, and more interested in productivity? We will never know.

 

Whatever happenned, happened and Microsoft got a juicy prize out of it that I think they mostly wasted, probably due to the usual corporate NIH syndrome...

[PGHammer]

Well obviously Apple would have bankrupted without MS investment even though MS only invested 150 million and Apple had 1.2 billion in cash.  

It was the timing of the investment - not the investment itself, that was critical for Apple.

Remember, at the time of the investment, there was no OS X (it was still in development), no iDevices at all, and Apple was in the process of acquiring NEXT.  Worse (for Apple) Adobe (then the most critical of Apple's stable of MacOS publishers behind Microsoft) had started making their MacOS and Windows products (especially Photoshop) functionally identical.  Result - Windows was no longer a poor second to MacOS - for content creators; instead, it (and especially the multiprocessor-aware Windows NT, which really gained ground) was better (due entirely to bang-for-buck).

 

As far as NIH, Microsoft is not the only company to have suffered from that; if anything, Apple is even more NIH than it was when Microsoft made that critical investment.  Also, do you REALLY think that "if you can't beat them, BUY them" is any safer a stratagem for Apple than it would be for Microsoft, or anyone else?  (Still, it is IBM - more than any other company - that really showed the destructive affects of NIH; they are, in fact, still paying rather large penalties for that today.)

 

Lastly, there is STILL a significant part of Microsoft's userbase that seeks to preserve NIH - this core group is now turning against Satya Nadella, just as they turned against Sinofsky, and even Ballmer.

 

And that one billion dollars in cash was, in fact stranded, and for the SAME reason that an even larger hoard of cash is stranded today - the silliness of the Internal Revenue Code of the United States.  (Then, it was sales of Apple EMEA, and Macs in particular, that resulted in the Stranded Hoard - today, it's iTunes and the App Stores, along with hardware sales in the same regions.

[tompkin]

Lastly, there is STILL a significant part of Microsoft's userbase that seeks to preserve NIH - this core group is now turning against Satya Nadella, just as they turned against Sinofsky, and even Ballmer.

 

 

The NIH is a real issue. But there are two other points to be mentioned. 

 

First, historically it's been difficult for companies to change course when they have successful current revenue streams. It's hard to transition from desktop Windows to something else when it still is bringing in a significant amount of revenue. Companies have always struggled with this. Apple has to be struggling with this right now.

 

The other things is, though. It's not just the user base that's in disagreement. There are pockets of disagreement within Microsoft itself that would rather things go one way rather than the other. There are always turf-battles within companies and you can bet that has been the case the last 5-6 years at Microsoft.

 

When you looked at where Nadella came from within Microsoft, they are doing excellent. Azure is rolling right along. If you were watching the first part of Build 2015 and weren't too bored, it was a pretty impressive demonstration as to where they are, and are headed in their cloud offering. 

 

My own view of Microsoft is that their business divisions are doing quite well. There are nit picks here and there (OneDrive for Business as an example) but for the most part things are going well.

 

The consumer side is where  the failings have been. (XBox being an exception). I will spare you what I think they should have done but needless to say, their current course is Universal apps. We'll see how that goes. 

[wiretap]

Itanium was planned originally for the workstation market at the time when RISC (and especially DEC's Alpha) was the "hot ticket".  What wrecked things was, ironically, not only AMD's Opteron, but Intel's own Pentium 4 - both of which offered better bang for buck.  (It certainly didn't help that both Opteron and P4 could also run garden-variety x86 software - which no RISC CPU could do.) Further, Itanium was why Intel did NOT acquire the rights to the Alpha itself when it acquired the rest of DEC Semiconductor (which included DEC's far-from-insignificant Ethernet business).

Yup. That's what I'm dealing with right now. My company is still running Alpha's (DS10's and DS20's). We have a lot of proprietary software which needs to function in a specific way, and re-writing that code would cost tens of millions of dollars. (due to nuclear V&V testing and software quality validation) We're currently seeking solutions to run it on Itanium processors with OpenVMS, but finding the correct interface to our systems is proving a challenge.

 

I'm also looking at running VMS in a Virtual Machine with EmuVM (http://emuvm.com/) but I don't think it will have the stability we need.

[Aergan]

As a small note, "same type of approach" is more like "identical copy" which Intel is allowed to do as a result of the giant court battle between Intel and AMD.

 

I was referring to comparable product offerings / strategy more than actual architecture (still relevant though in either view).

[+John Teacake MVC]

The instruction set was different. It was ever aimed at the desktop market.