Home > Windows > Microsoft backs Itanium with Longhorn Microsoft backs Itanium with Longhorn Eugenia Loli 2005-04-11 Windows 33 Comments After a faltering start, Microsoft is stepping up its commitment to Intel’s Itanium processor. About The Author Eugenia Loli Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker. Follow me on Twitter @EugeniaLoli 33 Comments 2005-04-11 6:14 pm this will insure Itaniums survivle 2005-04-11 6:28 pm couldnt the itanium be killed off and lived withought quite easily now that x86-64 is in use? opterons and xeons both scale well so far. not to mention the lack of shear 3rd party support for the itanium. why is intel so vested in making this redheaded stepchild work?? 2005-04-11 6:30 pm Windows on a HPC niche platform. I wonder how long MS will pursue this…. PPC would make more sense… 2005-04-11 6:46 pm Let’s hope this will be true competition for x86-64 crufty architecture. 2005-04-11 6:55 pm the true competition for x86-64 is Power 4/5 and I find it interesting that MS is supporting a chip arch that is failed. prediction of where they are headed? 2005-04-11 7:01 pm Itanium 2 is a far better architecture than x86. x86-64 is a good transitional processor type leading to “real” 64-bit computing. Itanium proved that you can just wholesale jump ship to 64-bits. This is where the lack of 3rd party software comes in. As people begin developing software for x86-64 it will make the transition to other 64-bit processors much easier. 2005-04-11 7:12 pm I heard programming for Itanium is very difficult. Don’t believe it will catch on in the consumer market. In fact I thought it was dead anyway… and I find it interesting that MS is supporting a chip arch that is failed. Intel has deep pockets. Plus they’ve bet everything on Itanium, (un)fortunently IA64 killed it. 2005-04-11 7:16 pm why do we need ‘real’ 64 bit? big deal if its a lil faster on complete 64 bit os an aps. by the time we are all ready to transition to ‘real’ 64 bit, x86-128 or sumthing will be out. the key to the industry is compatibility, and amount of aps, not speed, or even quality for that matter. 2005-04-11 7:24 pm Yes.. And MS supported NT on the Alpha chip as well. Look how well that did. 2005-04-11 7:47 pm A last hail-mary play for Craig Barrett to salvage his dismal record as CEO. Sorry Craig, Intel blew it on Itanic, its already dead with practically no chance at this point of market acceptance. And really, hitching your last hope to another perenially delayed product is not what I call a winning strategy. 2005-04-11 7:50 pm >> Intel has deep pockets. >> Plus they’ve bet everything on Itanium i agree with the first line: for better or for worse intel will swim away from the itanic. it will be a massive writedown sometime in 2007-2009 listed quietly at the back of an SEC filing long after the market has factored in its death. intel will go on and itanic will get listed next to PC Jr, OS/2, Microsoft Bob, Longhorn etc as computing’s biggest flops. 2005-04-11 7:52 pm Too bad few people buy them for lack of applications, lack of vendor support, and total redundancy in a field with PowerPC, SPARC, and AMD64 filling most niches already. Also, the rest of the industry is going to massively-multi-core CPUs, such as Niagara. Tons of simple RISC cores on a chip can scale much like Itanium’s VLIW can, but with the added bonus of the opterating system handling run-time scheduling. There’s only so much the compiler can do, which is evidenced by how fast Java can be with JIT enabled (faster than native C and C++, even). 2005-04-11 7:52 pm >> and I find it interesting that MS is supporting a chip arch that is failed. they are using PPC in xbox 2, so clearly someone at MS is thinking about the CPU market. MS is probably just fulfilling some clause in some secret MS/Intel partnership doc (computing’s Warsaw Pact treaty if you will), that requires them to support each other’s dead products upon request. 2005-04-11 7:54 pm >> Also, the rest of the industry is going to massively-multi-core CPUs, such as Niagara wherever the market is headed i can tell you it is not sun hardware. try again. 2005-04-11 8:09 pm Itanium is not necessarily hard to program. I found assembler optimizing for Itanium 1 quite easy. The loop unrolling with parallel loops and shifting register windows is really nice. A for-loop even exists in hardware. But to use that fully is hard to achieve for the compiler, since all 3 units have to be occupied all the time. At least beating GCC back then (2002) was a piece of cake with hand optimizing. 2005-04-11 8:24 pm “I am so glad Microsoft is dying and am equally happy Linux is taking over, nobody uses Windows anymore so it is a true legacy OS”[/i] At least someone still has a sense of humour 2005-04-11 10:30 pm This assumes that the Itanium will still be around whenever Longhorn (finally) ships, and that people will still care. The Alpha had lots of supporters too, I daresay more than the Itanium. Look where it is now. 2005-04-11 11:59 pm “wherever the market is headed i can tell you it is not sun hardware. try again.” Considering that Sun won’t break binary compatibility, even with a such a new chip design, says to me that they have practically infinite better chance of success than Itanium ever will. SPARC/PowerPC very naturally extended to 64-bits, so Sun/IBM didn’t have to gut their architecture during the transition. Intel obviously didn’t have this foresight…and AMD called them on it big time. Intel’s only real asset right now is a bunch of chip factories, because that’s the only thing that will keep them going until they can re-work their entire product line. 2005-04-12 12:08 am Support a failing cpu with an OS that dosn’t quite exist yet. I think this is leftovers from April 1st? 2005-04-12 12:14 am ” “I am so glad Microsoft is dying and am equally happy Linux is taking over, nobody uses Windows anymore so it is a true legacy OS” ” LOL. Yes, the 3ish% of the desktop market that uses Linux and the 25ish% of the server market that uses Linux is so much more than the somewhere in the 90s% that uses Windows on the desktop and 50ish% that uses it on the servers. I am a supporter of Linux….but still…what a dumb comment to make. 2005-04-12 12:19 am “Considering that Sun won’t break binary compatibility, even with a such a new chip design” Sun binary compatibility has nothing to do with the processors as it is done at the Operating System level (Solaris) the same as SGI’s Irix has done for the MIPS processors. If applications were able to rely on the underlying OS API’s and system calls not changing from version to version than this isn’t a problem. This is where Linux fails many times over, and where you will start to see more applications being developed for the I2 processor. 2005-04-12 1:12 am “Sun binary compatibility has nothing to do with the processors as it is done at the Operating System level (Solaris)…” Binary compatibility starts at the processor’s ISA (each version of SPARC is a binary compatible extension of the previous version). Sun then provides ABI and source-level compatibility at the operating system level. Applications developed on Solaris/SPARC will recompile onto Solaris/x86 and vice versa, but SPARC and x86 are not binary-compatible with each other. The Linux failings you mention are likely where the Linux developers break both source-level and binary compatibility. The Itanium 2 can do nothing to help this. 2005-04-12 3:26 am >> Considering that Sun won’t break binary compatibility, even with a such a new chip design, says to me that they have practically infinite better chance of success than Itanium ever will. you speak as if these were mutually exclusive. Itanic is already dead. Sun is nearly there. In five years the Itanic and Sun Microsystems (as we know it now) will be memories. 2005-04-12 3:30 am “In five years the Itanic and Sun Microsystems (as we know it now) will be memories.” hahaha, do u have any idea how much money SUN has in the bank? they are not going anywhere. same thing with Itanium. Intel has put way to much cash into it and intel is damn stuborn, they wont let it die. 2005-04-12 3:37 am “hahaha, do u have any idea how much money SUN has in the bank?” More than $7.5 billion. Sun can lose money for _years_, if they had to, but they’re starting to turn a profit. Sun’s management aren’t as stupid as many people say they are, and they’ve done everything from saving on real estate with Sun Ray to hestitantly doing layoffs to get their expenses in line. They’re also the biggest Opteron reseller on the planet, now, which will only help their bottom line even more. 2005-04-12 11:22 am It has faced similar issues as with IA-32 since the X64 is just a 64bit cleaned up version of X86 ISA(additional prefix to enable 64bit pointers and address). 2005-04-12 11:25 am >they are using PPC in xbox 2, so clearly someone at MS is >thinking about the CPU market. Not much different to Windows NT PPC edition… Microsoft’s main plan is to have their OS running everywhere(within limits of their resources). 2005-04-12 11:54 am “Linux is so much more than the somewhere in the 90s% that uses Windows on the desktop and 50ish% that uses it on the servers. ” I think it was 30% for server use with windows … 2005-04-12 12:16 pm “Intel has deep pockets. Plus they’ve bet everything on Itanium, (un)fortunently IA64 killed it.” You mean AMD64. IA64 is the Itanium. Then again you’re probably right in Itanium killing itself. 2005-04-12 2:26 pm I had spent some time with Intel engineers looking into application development for bioinformatics on the Itanium 2 processor. The Itanium 2 processor was excellent for specific classes of algorithms because of the large number of registers and the ability to process multiple operations (such as iterations of a loop) in parallel — assuming that you used their compiler, and the compiler “did the right thing”. If the compiler didn’t do the right thing, you could tweak the code to force it to. In many cases, you need to restructure code for it to optimize properly (i.e. tweak/rewrite applications). Algorithms that do lots of fetches or have many dependent steps (little room for internal parallelization), performed awful. In short, in practice we found that for specific tasks with extra effort expended, the Itanium 2 was excellent. For general use, it perfomed worse than many of their other processors (and the price was much higher). There’s very little incentive to use Itanium 2 for the general use case. The processor stands a VERY high likelihood of being discontinued before Long debuts (currently slated for 1Q07, IIRC). Even the Intel engineers I was working with admitted that there is a chance of that. 2005-04-12 4:08 pm Itanium is a very good architecture and is the true IA-64. Its hard to say where its going though. If Itanium become the defacto processor standard who would compete with them? It would take AMD years to reverse engineer IA-64 to provide some competetion. The alternative would be for AMD to release a competing chip design but if they could not secure Microsoft Windows Support like Itanium then it would be all for not. Either way hopefully one day we can choose whatever computing platform we want to use and still maintain compatibility with other systems. 2005-04-12 4:49 pm OS/2 has survived for almost 20 years now (and its 32-bit incarnation has been around for 13 years), at one point it sold over a million copies/month via retail (not including contracts or preloads), and it’s still a more viable platform for some of us than any existing flavor of Windows or Linux (it’s far more flexible than Windows and it has slightly better legacy software support than Linux even with Wine, DOSBOX, and DOSEMU installed). It *is* disappointing that it didn’t win more than 15% or 20% marketshare at its height, and it’s admittedly a mixed bad in many ways (the WPS is extensible and flexible but can also be a house of cards, it isn’t a multi-user OS like Linux so there are security issues it doesn’t adequately address, and there are a lot of applications it can run but the majority of them are older programs which are largely unsupported), but I think “flop” is grossly inaccurate by any measure, and putting in the same class as the PC Junior or Bob is either the result of ignorance or a deliberate prod at the folks who still use the platform. 2005-04-12 5:45 pm I don’t know if any of you have seen or read cpu roadmaps but the Itanium is going to be around well into 2007/8. If you think about it and maybe my numbers are off, the itanic arch is very solid and though maybe a little hard to code for over x86, that still doesn’t kill it off. The itanic can process 6 threads at a time, with x86 only doing 2, maybe with dual core now that’s up to 4, I don’t know for sure. but the itanic has great IPC numbers and it’s a FPU monster. The things holding it back right now are it’s size it’s cost, maybe one more thing to like heat but I don’t think so. Intel already has a dual core Itanium in the works also, that’s like 12 instructions per clock going, if you have the right app coded. and switching to 65nm should make it smaller and cheaper, if this holds true and the pricess fall close to or in line with those of the big cache Xeon chips then intel has a winner in it’s hands. Only time will tell though, one last thing the itanic could use is faster 32bit emulation, so far that’s pretty slow on the itanium 2, but that could change. We’ll see.