Remember that Xeon E7-Itanium convergence? FUHGEDDABOUDIT If you are an Itanium shop and you were hoping for a big upgrade in performance with the future "Kittson" Itanium processors around two years from now, it looks like you can forget it. It ain't gonna happen – and you can also delete from your memory the idea of a common socket for Xeons and Itaniums. On January 31, on the …
I honestly don't know how anyone could put a positive spin on this, although I am sure we will find out. HP is basically saying Kittson doesn't exist, we are going to goose Poulson a bit in the 32nm process and just call it Kittson. Oracle actually produced some HP internal emails on this topic. HP had a hypothetical plan to continue to produce Kittson under a new name for a few more generations. Basically just tweaking Kittson and calling it some new name to give the appearance of innovation. This means that HP has not seen it worth investment to even produce the first Kittson. It sounds like they are going to tweak Poulson a bit in a year or two and call it Kittson.
".....HP is basically saying Kittson doesn't exist, we are going to goose Poulson a bit in the 32nm process and just call it Kittson....." No, that is what TPM is saying. You obviously missed the bit about "future development", which implies a further development of Itanium. Of course, having to accept that would malke the IBM trolls' heads explode.
Well obviously HP's press release isn't going to be: "Intel just gave up on Itanium. They asked for more cash to develop 22nm. We said it doesn't make financial sense. Intel was more than fine with ending Itanium development at 32nm." I am sure there will be "future development" in the form of boosting the clock speed by a tenth of a Ghz or putting some more cache on the Tukawila/Poulson 32nm process chip and calling it Kittson.
This is the obvious result of Intel moving all of the Itanium engineers to xeon in 2011.
http://www.xbitlabs.com/news/cpu/display/20110417112218_Intel_Relocates_Itanium_Engineers_to_Xeon_Projects_Sources.html
Shocking that intel will not simply move the poulson chip onto 22nm and put a larger cache on it.
How hard could that be, but its all about money. Intel wants more money from HP and HP does not have any money, so that make processors thru 2017 will be just poulsons.
And the end of the story is now.
e99
"....Shocking that intel will not simply move the poulson chip onto 22nm and put a larger cache on it....." It would have been that simple a few years ago with the old process, but now Intel has invested in the new 3D Tri-gate process, and it appears that shrinking Itanium down and introducing the new process in one step has proved too much.
Thanks for the explanation, curious if there will ever be a 22nm chip. Intel said kitson will not be for two years that makes it 2015 and I think the agreement was only till 2017 in making the chips. But I am sure Intel could make enough chips for 10 years in their fab in one week since there is not that much supply.
cheers....e99
"......curious if there will ever be a 22nm chip...." So are we. There were hints coming out of Intel that the Tri-gate process was making deadlines tough to meet, so I suspect Intel simply took the easy route for a quick upgrade they could pin on the roadmap. As it is, we're just getting Poulson systems now, and they do provide a nice performance jump over Tukzilla, so another jump in two or so years will be welcome. The fun is what happens after that, which I suppose will have to wait for another NDA session with hp and/or Intel, probably when they know in a year or so's time.
If that doesn't send warning bells about the future of Itanium, what will?
What happens if Intel later just kills Itanium? Intel and HP have an agreement for chips to be produced until 2017. That doesn't mean Intel will make updates though. So it could be 4 years of the 9500. Maybe a slight update could be done. HP will need to make another payment to Intel. When does HP decide that this arrangement is a failure.
Itanium is a direct evolution of PA-RISC, and there's a strong commonality from PA2 instructions to Itanium instructions - which is why execution of PA2 binaries on Itanium is pretty fast.
The core design, on the other hand, is fairly different - although some design factors like HP's tendency to like processors with huge chunks of fast SRAM and small cores survived. But the PA-8K design was probably nearing the end of its productive lifespan anyway, and at the time a lot of Itanium's design changes (OoO sacrificed in favor of wider issue) seemed like reasonable choices.
Given that Itanium roundly trounced PA-8K in benchmarks from the beginning, I have a hard time qualifying the move from the PA-8K design as a bad idea. That being said, IPF is pretty obviously dead as a doornail at this point.
I always viewed it as Intel royally screwing HP in order to get access to some of the IP that they needed to speed up the Pentium processor.
PA-RISC was a fairly nice processor that HP had developed from discrete logic into multi and then single chip designs. For a preiod of time in 1990-1994, HP, Digital and IBM went head-to-head to try to produce the fastest chip/system. But rapid development of processors is an expensive operation, and HP did not commit to the same level of resource to keep the processor development going.
When Intel wanted to speed up their x86 processors, they needed to use technologies that they did not hold the rights to, mainly for the superscalar and deep pipeline techniques. They approached HP with an offer to take on the development of the EPIC processor which was to succeed the PA-RISC2 64 bit implementation in exchange for the rights to use some of the technologies that HP used in the PA-RISC family. This enabled Intel to produce faster and more competitive x86 cores (I think the first Intel processor to benefit was the P6 Pentium Pro), and in theory removed the cost of developing the next generation 64 bit processor from HP.
When it came to producing the processor that they promised to HP, Intel were a bit tardy. They produced the original Itanium which was something like 2 years late, was not as compatible with PA-RISC2 as it was supposed to have been, and was slower than promised. Compared to other contemporary processors, the Merced implementation of Itanium was considered very disappointing. So Intel benefited greatly with their own processors, and HP suffered.
This left HP with a gap in the late '90s that meant that they had to continue using the PA-8X00 processor family beyond it's natural lifetime. The fact that HP managed to get significant performance increases by iterative evolution was probably a testament to the design of the original processor and the people remaining in HPs processor development team.
Sounds like HP are still being screwed over. I wonder if there is anybody left in HP who regrets the decisions taken back in the mid '90s.
With the greatest possible respect Peter, you seem to have missed the elephant in the room.
The patent dispute that mattered was between Intel and DEC; right now I don't have time to expand on the details but readers can find various accounts, which likely don't paint the HQs at either DEC or Intel in a very positive light. I particularly liked the Intel VP quoted as saying "we ran out of people and companies to copy from" (or words to that effect).
Close but no cigar.
"....They approached HP with an offer to take on the development of the EPIC processor ...." Hp developed the first Merced generation of Itanium. They realised early on in development that the platform was great for porting, and decided that if they could persuade other partners to pick it up then they could maximise their return on investment, so they approached Intel (not the other way round).
".....When it came to producing the processor that they promised to HP, Intel were a bit tardy...." The delays in getting the first Itanium design out, Merced, a development version, was an hp issue, not an Intel one. Intel then took on the major role in developing Itanium2.
"......So Intel benefited greatly with their own processors, and HP suffered....." Hmmmm, hp gained a production and development partner with massive scale, they gained a chip that allowed them to continue making enterprise UNIX servers, plus easily migrate Compaq's old enterprise customers, plus it killed off many of their competitors. Yeah, disaster - NOT!
"Hp developed the first Merced generation of Itanium. They realised early on in development that the platform was great for porting, and decided that if they could persuade other partners to pick it up then they could maximise their return on investment, so they approached Intel (not the other way round)."
If you say so. Personally, I don't know how it went. However, as HP are having to pay Intel to keep the chip going, I guess Intel are happy with the arrangement. Not sure HP are as much though.
"The delays in getting the first Itanium design out, Merced, a development version, was an hp issue, not an Intel one. Intel then took on the major role in developing Itanium2."
Sorry? You're claiming that HP were selling a development version of the chip? Wasn't even a production version? I bet people who purchased servers with Merced in it are happy to hear that.
"Hmmmm, hp gained a production and development partner with massive scale, they gained a chip that allowed them to continue making enterprise UNIX servers, plus easily migrate Compaq's old enterprise customers, plus it killed off many of their competitors. Yeah, disaster - NOT!"
In general, the only chips it killed off were their own!! Is Sparc still there? Yes. Is Power still there? Yes. Is x86 still there? Yes. So, what exactly did Itanium kill off?
You could possibly argue whether Itanium killed off MIPS. Maybe the other vendors had something to do with it as well. However, Itanium only killed off Alpha because they were both owned by HP. Would have been interested to see who would have prefered to stay on a developed Alpha and who would have prefered to jump to Itanium. From the people I know, most would have stayed on Alpha. Almost nobody in Alpha or PA-RISC was "happy" to move.
HP wanted to support only one processor into the future and chose Itanium. Many people would question the wisdom of this.
"....... but the Chinese have taken up the MIPS architecture with their Godson CPUs." The Godson chip is aimed at Xeon, not Itanium. It also seems that Huawei and Inspur both looked at Godson and chose Itanium instead, so if Godson now falters and dies as a commercial enterprise CPU it will be another scalp claimed by Itanium.
".....If you say so. Personally, I don't know how it went......" Why would anyone expect you to know?
"....as HP are having to pay Intel to keep the chip going....." HP are paying a fab partner, just like other companies pay fab companies to make chips for them. An example is Apple who don't make any of their own chips.
".....I guess Intel are happy with the arrangement. Not sure HP are as much though....." Seeing as both are making a profit out of it they're probably both quite happy.
"....You're claiming that HP were selling a development version of the chip?....." Yes. That's exactly how hp sold it, so that early adopters could start on software. Some companies actually bought the first gen servers and used them in production anyway as they had the best floating integer performance going at the time, but otherwise they were slower than the PA-RISC chips of the day. HP sold the Merced boxes cheap to get ISVs on board so they could brag about 1500 Itanium-ready applications come the Itanium2 launch.
"....In general, the only chips it killed off were their own!!....." Apart from Rock, the last and stillborn UltraSPARC variant, did news reach the troll kindergarten of a range of chips called MIPS?
"Yes. That's exactly how hp sold it, so that early adopters could start on software. Some companies actually bought the first gen servers and used them in production anyway as they had the best floating integer performance going at the time, but otherwise they were slower than the PA-RISC chips of the day. HP sold the Merced boxes cheap to get ISVs on board so they could brag about 1500 Itanium-ready applications come the Itanium2 launch."
Strangely enough, I spoke with many a HP salesman from their BCS group at the time and investigated buying a Superdome. None of them ever told me it was a 'development' chip just to get ISVs interested. Wouldn't be HP FUD would it? And you go on about IBM FUD!! HP salesmen (and women) never once said to me it was development or anything other than a full offering. Your comment is absolute tripe.
"Apart from Rock, the last and stillborn UltraSPARC variant, did news reach the troll kindergarten of a range of chips called MIPS?"
You're just picking on any chip that didn't make it through that time and claiming Itanium killed it off!! You could just as easily claim Power or Xeon killed it off. In reality, what killed Rock was Sun. It had nothing to do with other chips. As for MIPS. To a degree maybe as I've acknowledged before. However, it has survived in one form of another and again, you could also blame every other chip around at the time for 'killing it'.
".....I spoke with many a HP salesman from their BCS group at the time and investigated buying a Superdome...." Gosh, I'd believe you, only anyone who did actually enquire about Superdome would know they didn't come with the first generation Itanium, Merced, they either came with PA8600 or later PA-RISC CPUs or Itanium 2 CPUs. The "development" systems hp largely gave away were the rx9610 IIRC.
".....Your comment is absolute tripe....." I would suggest it is more a case of you simply not knowing what you're talking about, again. Yet again.
".....You're just picking on any chip that didn't make it through that time and claiming Itanium killed it off!!...." SGI actually bought the MIPS design and formed a compnay called MIPS Technologies, just so they were assured of the longterm longevity of their systems. IIRC they did so for eighteen to twenty years. They expressly ended their MIPS range to build Itanium servers:
http://www.osnews.com/story/15741/SGI_To_Drop_MIPS_Irix_Moves_to_Itanium_Linux/
Evidently the news didn't reach the troll kindergarten. Yet again.
I would suggest that you take time to learn something before your next foray, such as maybe spending a few years actually working in the industry.
This post has been deleted by its author
"Gosh, I'd believe you, only anyone who did actually enquire about Superdome would know they didn't come with the first generation Itanium, Merced, they either came with PA8600 or later PA-RISC CPUs or Itanium 2 CPUs. The "development" systems hp largely gave away were the rx9610 IIRC."
All I can tell you is what I was told. They were referred to as Superdomes by the salesmen. As I don't care what a server is called, but what it does, I took them at their word. If you look up the rx9610 servers you've referred to, you will note they are described as being 'Superdome like' and 'call board based like Superdomes' etc. I guess that's close enough for a HP salesman. Anyway, our interest in them waned significantly when we realised just how poorly they performed, so we didn't get as far as talking about specific servers etc. When HP later came back with Superdomes containing Itanium 2's, we also looked at performance and whilst significantly better than the first offering, it still sucked compared to the competing chips. So, again, no dice.
"SGI actually bought the MIPS design and formed a compnay called MIPS Technologies, just so they were assured of the longterm longevity of their systems. IIRC they did so for eighteen to twenty years. They expressly ended their MIPS range to build Itanium servers:
http://www.osnews.com/story/15741/SGI_To_Drop_MIPS_Irix_Moves_to_Itanium_Linux/
Evidently the news didn't reach the troll kindergarten. Yet again.
I would suggest that you take time to learn something before your next foray, such as maybe spending a few years actually working in the industry."
MIPS was not killed off for one simple reason that may have escaped your attention. It still exists. Different role maybe, but the technology is still there and being developed for other purposes, so it did not die. Yes, SGI moved to Itanium. So what. That didn't kill MIPS. I've added a reference to the MIPS wikipedia page that shows the development continuing till ........ 2012. Good for a dead chip. Also, a supercomputer was built in 2007!!
http://en.wikipedia.org/wiki/MIPS_architecture
So, one company chose Itanium over MIPS, but other carried on with MIPS. It certainly didn't kill it. I also notice you've avoided answering my reply about Rocks.
I really do wish The Register would add an icon not just for 'fail', but 'epic fail' or 'tripe' as it would more accurately represent your reply quality.
I'll give you the challenge again. Name one chip that was actually killed off by Itanium as per your previous comments. Rock wasn't. That was killed by Suns incompetence and the advances of all other chip makers (so maybe an assist might be in order). As I've demonstrated above, MIPS wasn't. I might give you Alpha as stated by another poster. It was killed by Compaq in favour of Itanium. PA-RISC. Well, that was owned by HP, so that's a bit of a cheat really, as they simply didn't want to develop more than one chip. So, name another? Finding it hard? The reality is that Itanium was simply not a great game changer and didn't have anything like the impact predicted/hoped on the industry.
"All I can tell you is what I was told. They were referred to as Superdomes by the salesmen. As I don't care what a server is called, but what it does, I took them at their word. If you look up the rx9610 servers you've referred to, you will note they are described as being 'Superdome like' and 'call board based like Superdomes' etc....." Wriggle, wriggle, wriggle! The hp term "Superdome" wasn't used until long after the rx9610 had been released, you're just desperately looking at the hp Retired Products list and trying to paper over the immense gaps in your knowledge.
"....When HP later came back with Superdomes containing Itanium 2's, we also looked at performance and whilst significantly better than the first offering, it still sucked compared to the competing chips....." Yeah, was that "look" just as much fantasy as your Superdome rx9610? LOL! Credibility zero!
"......MIPS was not killed off for one simple reason that may have escaped your attention. It still exists. Different role maybe....." More wriggle, wriggle, wriggle. Being an embedded chip in a washing machine is not pass for being an enterprise solution, though it's probably as close as you've ever been to enterprise computing.
".....I really do wish The Register would add an icon not just for 'fail', but 'epic fail' or 'tripe' as it would more accurately represent your reply quality....." Hey, ask nicely and maybe they'll open a junior forum for you and your kindergarten buddies.
"...... Name one chip that was actually killed off by Itanium as per your previous comments....." I already showed how MIPS was replaced by SGI by Itanium, you're just too sulky to admit it. Now go take your afternoon nap, you're just getting a little tetchy.
"......Rock wasn't. That was killed by Suns incompetence...." Their incompetence in dropping the port of Slowaris to Itanium you mean, leaving them no alternative but to go begging to Fujitsu for SPARC64, a chip they spent years slagging off? Truly classic times, still makes me smile thinking about it.
Enjoy!
/SP&L
"I guess we can expect Larry to be in a great humour....." Really? His last round of singing is due to land him with a $500m bill, and then he has to keep funding development of all his software on Itanium for hp's benefit until hp does finally call it a day. If Larry is singing it will be through gritted teeth.
Maybe. On the other hand, as Intel seem to have confirmed Itanium is dead, certainly after 2017, is he really worried. So, it cost him a few hundred million. Chicken feed. He would have to spend that money over the next few years contining to port the code onto Itanium, so little loss there. Presumably, he also hopes to pickup a certain amount of the hardware business as well. In the meantime, whether by telling the truth or not (individuals personal viewpoint), he has pretty effectively killed off the Itanium (whether it was dying or not beforehand).
It'd take a pretty brave soul to implement something from Oracle on Integrity now.
Perhaps future successors of the Xeon E7 could include a small amount of extra logic on the chip to alternatively decode Itanium instructions instead of x86 instructions. That would really give people wanting a choice between Itanium and x86 value for their money, as they could flexibly switch between both in the same box with the same chip.
That would let the Itanium easily take advantage of future x86 improvements without Intel having to devote a lot of effort to supporting it.
Perhaps future successors of the Xeon E7 could include a small amount of extra logic on the chip to alternatively decode Itanium instructions instead of x86 instructions. [...] they could flexibly switch between both in the same box with the same chip.
They did the reverse with the original Itanics: Give it the hardware to decode x86 instructions natively and make the transition from x86 to Itanic easier. Then they found that it sucked in hardware and software emulation worked better.
"......They did the reverse with the original Itanics: Give it the hardware to decode x86 instructions natively and make the transition from x86 to Itanic easier....." That was because the original Itaniums were not intended to use x86 code much, so they put a whole Pentium3 on the die for the expected odd bit of x86 code. This did run x86 code, which was what was originally promised, just not as fast as a Xeon, unsurprisingly. Of course, the IBM and Sun FUD machines went into overdrive insisting that hp and Intel had claimed the Itanium would run x86 code as fast as a Xeon, an outright lie repeated endlessly by the trolls. In later developments it was found the emulators on Itanium could run x86 code better than the on-die Pentium3' so the die space was used for different enhancements.
I suspect that the reasons against doing that are purely commercial, not technical. It'd take a chunk of money, the result would likely be power hungry and run hot, and the market is anyway quite small. So yes it's perfectly feasible from a technica l point of view, but there's probably not much comercial point in doing so.
Some of Itanium's features are making it into x64 anyway. That'll erode Itanium's position further. For example Itanium has had a fused multiply-add instruction for ages, but only now are Intel getting round to putting one into x64. As soon as it's there, there'd be no reason to buy Itanium for its maths performance.
I pick out that one in particular - it's very useful in a lot of scientific computations and has been a glaring omission from x64/x86 for the past 13 years. PowerPCs from the 7400 family onwards (iMac G4 and upwards) had one back in 2000. As a result I was able to get faster runtimes out of 400MHz PowerPCs than 3.8GHz Pentium 4s,,,
Also I think it's clear that Intel don't want to do any more Itanium, not even merely stamping out the same old design on 22nm. What you suggest would be more work, but Intel clearly aren't in the mood for innovating along these lines.
HP are going to end up being an x64 house whether they like it or not. I wonder if they'll sue Intel like they did Oracle?
If Intel had pressed ahead with a common Xeon/Itanium socket HP's ability to differentiate their Itanium hardware would have been greatly diminished anyway. Hypothetically speaking, whatever technical cleverness HP get from their chipset (had they updated it) would then automatically work with Xeon as well (drivers permitting). If HP customers started wanting that cleverness but with x64, would HP have said no?
Reminds me of when Carly was still at the helm of HP and controlled the final years of Alpha processor development. Rather than come out with a true successor to the EV7 that was manufactured on a next-gen process, HP just bumped the EV7 clock speed by a few dozen MHz, slapped a "z" at the end of the product name, and called it a day.
In about 2005 HP flew us all out to their Bristol HQ to try and sell us this new inhouse cloud flexible dynamic computing clustering concept they had invented.
It consisted of being able to copy a drive image (slowly) from a SAN and reboot to gain an extra webserver or database server machine when you needed it. And this from the company that had incorporated the clustering technology of Tandem, Compaq and Dec !
Sure, they lost the overall court case as it seems promises were made to support HP's Itanium boxes, but the other big accusation/denial over the future of Itanium, looks like Larry was right. Bet we don't hear that mentioned much. Guess it's more fun to criticise Oracle...
This post has been deleted by its author
This post has been deleted by its author
This post has been deleted by its author
Why? Alpha was just a RISC design like PA-RISC and was hitting the end of the RISC development capability. It even had less of a market than PA-RISC. Itanium was designed from the ground up as a porting platform and offered far greater development than Alpha, and allowed hp to use one platform for multiple enterprise OSs (OpenVMS, Linux, Windows and hp-ux). Oh, sorry - did I expect you to understand an argument based on facts and logic?
Just for the record, Alpha was also designed from the ground up as a porting platform, and allowed DEC to use one platform for multiple enterprise OSes (VMS, Tru64 UNIX aka OSF/1, Windows and Linux(es) were delivered, as were some others readers probably won't know including for those in the RT market, VxWorks).
Alpha died for reasons of DEC/CPQ/Intel politics; it was bullied (strongarmed?) out. It had plenty of engineering headroom.
In addition, unlike IA64, the Alpha architecture was licenced to a number of other chip designers and manufacturers. One of them, not that well known in the industry at the time, is rather better known these days. Samsung. Samsung even had actual commercial product that shipped, from chips (eg 21164PC?) to board-level stuff (the board that became the DEC AlphaServer DS20????). There was never any real chance of multi vendor support with the IA64 monstrosity; even Compaq's rebadged Unisys systems (sold as Proliant 9000?) were an abysmal failure.
Interested readers who want to know why DEC thought Alpha would technically have remained ahead of IA64 might like to have a read of the 1999 DEC whitepaper comparing architectural (not just implementation) features of Alpha and IA64, still available at
http://www.cs.trinity.edu/~mlewis/CSCI3294-F01/Papers/alpha_ia64.pdf
"Why? Alpha was just a RISC design like PA-RISC and was hitting the end of the RISC development capability."
The Alpha architecture ran out of development budget, it certainly did not die because it was built on RISC principles. It would appear that EPIC is now facing the same end as the Alpha.
That said EPIC does have a fundamental weakness when compared to CISC & RISC. Folks who design cores for a living have pointed out that EPIC requires large (and consequently slow) register files that would ultimately limit the clock rate and burn more power. History has proven them to be correct.
EPIC doesn't fit into the real-world as well as *some* CISC & RISC architectures.
It would be nice if you stuck to the facts instead of making shit up.
Sorry Matt, but I couldn't help but respond. Perhaps you could enlighten us with how many of those operating systems are left with? You know, with Itanium versions in the future?
RedHat linux? No.
Windows? No.
So, the only operating systems that can run on Itanium in the future are HP ones...............
Doesn't matter what was intended, just what's happened.
"Sorry Matt, but I couldn't help but respond....." What, you spotted another thread where you think you can roll with the trolls? What a surprise - not!
"....Perhaps you could enlighten us with how many of those operating systems are left with?....." Apart from hp-ux and OpenVMS? Well, there's the question of what Huawei are going to run on their Itanium servers, but it's probably going to be a Red Flag Asianux Server variant, which already supports Itanium: http://www.redflag-linux.com/en/product_end.php?class1=10&class2=1&productid=1
And whilst you state no RHEL or Windows support, I can still run the supported versions. Should I get bored, I can even load up a fully-supported latest version of Debian: http://cdimage.debian.org/debian-cd/6.0.6/ia64/iso-cd/
Or even Gentoo: http://distfiles.gentoo.org/releases/ia64/autobuilds/current-iso/
Oh dear, it looks like your eagerness to be "one of the gang" has led you into another arena you know SFA about. Now be a good trainee troll and go get your diaper changed, it's almost as full of the brown stuff as teh rest of you.
"Apart from hp-ux and OpenVMS? Well, there's the question of what Huawei are going to run on their Itanium servers, but it's probably going to be a Red Flag Asianux Server variant, which already supports Itanium: http://www.redflag-linux.com/en/product_end.php?class1=10&class2=1&productid=1"
"And whilst you state no RHEL or Windows support, I can still run the supported versions. Should I get bored, I can even load up a fully-supported latest version of Debian: http://cdimage.debian.org/debian-cd/6.0.6/ia64/iso-cd/"
"Or even Gentoo: http://distfiles.gentoo.org/releases/ia64/autobuilds/current-iso/"
Oh yes. All those so well known enterprise class business operating systems!!!!
You really make me laugh. You stayed away initially, but couldn't help yourself could you!!
"hitting the end of the RISC development capability"
Yeah. After all, there are no RISC based processors around anymore. Oh, hang on, there are. I'm quite sure Alpha could have continued, but like all the processors that were supposedly heading towards a wall, which was then avoided.
The Alpha architecture was intended to have a lifespan of 25 years. Whether it was capable of that is moot, since it was canned after 12, but there were at least two more generations under development when that happened.
Oh and OpenVMS, Linux and Windows already ran on Alpha platforms, as did Tru64 UNIX. Not a bad "porting platform" (whatever that is exactly...).
"allowed hp to use one platform for multiple enterprise OSs (OpenVMS, Linux, Windows and hp-ux). Oh, sorry - did I expect you to understand an argument based on facts and logic?"
Same single-platform story applied to Alpha, as was pointed out to you earlier in the discussion (last time you claimed to be mentioning "facts and logic"?). Alpha also had two implementations of the realtime OS VxWorks, one from WRS and one from DEC, although not that many folk cared. IA64 couldn't reliably handle a real-time system if you tried, owing to the ridiculous context switch/exception handling latencies.
As was also pointed out to you back then, unlike IA64, Alpha also had multiple licenced sources for the chips (designers and fabricators).
Have a read of the 1999 DEC whitepaper comparing architectural (not just implementation) features of Alpha and IA64, still available at
http://www.cs.trinity.edu/~mlewis/CSCI3294-F01/Papers/alpha_ia64.pdf
Why are you even trying to argue that technology killed Alpha when it's perfectly obvious that corporat politics did?
[yes all this has been said before in this very thread. Maybe Matt will actually acknowledge it this time. Or maybe not]
"Same single-platform story applied to Alpha....Alpha also had multiple licenced sources....." Neatly ignoring two simple facts - the size of the Alpha market was already smaller than the Itanium one, and it was far simpler to port all those Alpha OSs to Itanium than port hp-ux (bigger market than OpenVMS and Tru64 combined) to Alpha, because Itanium was designed to be a porting platform. The first fact convinced Compaq to switch from Alpha to Itanium, and the second convinced hp it was the right decision.
"....Why are you even trying to argue that technology killed Alpha when it's perfectly obvious that corporat politics did?...." Sorry, but market facts killed Alpha in favour of Itanium. The decision to kill off Alpha was actually made by Compaq before they were bought by hp because switching to Itanium made more sense. Deny that all you like, you'll just be wrong again.
" The first fact convinced Compaq to switch from Alpha to Itanium"
Compaq switched from Alpha to Itanium because Intel had made it clear that the Compaq/Intel "special relationship" (re x86) was at risk unless CPQ did as they were told. This was around the time when a 64bit x86 was still impossible (according to Intel) and IA64 was still going to be "industry standard 64bit".
The HP-UX market was an HP-UX market. It wasn't specifically a PA-RISC market, and it isn't specifically an IA64 market. People largely continue to buy HP-UX *despite* IA64, not *because* of IA64.
Ditto VMS. It was (and is) a VMS market, not an IA64 market. It was VMS that mattered when it was on VAX (though VAX was relatively unique in the early years), and it was VMS that mattered when it was Alpha, and the same still goes for IA64. People largely continue to buy VMS *despite* IA64, not *because* of IA64.
Customers of HP-UX and VMS have little hardware reason to buy IA64, IA64 does little that can't already be done on an AMD64, and if/when those OSes become available on a subset of AMD64-based Proliant (check out the cores and memory on a high end Proliant), customers will continue to buy those OSes.
VMS and HP-UX both have stuff that is unique to those OSes and provides businesses with some kind of advantage. What does IA64 have that is unique to it which is an *advantage* to the customer, the software developer, the system designer? IA64 has plenty which is a *disadvantage*.
Anyways, readers don't need to take my word for it, they could ask (or look at) the customers, software developers, etc whose investment is in a multi-platform OS such as Linux or Windows. Folks who can choose their hardware platform are not choosing IA64, whatever Intel may once have said about "industry standard 64bit".
The IA64 story is only going one way;the same way as Alpha (albeit perhaps for slightly different reasons).
".......Compaq switched from Alpha to Itanium because Intel had made it clear that the Compaq/Intel "special relationship" (re x86) was at risk unless CPQ did as they were told......" And of course, you have some verifiable link to illustrate that little conspiracy theory? No, I didn't think so. Compaq had been making the Alpha in direct competition with Xeon ever since M$ did their NT port for Alpha and Intel hadn't batted an eyelid over the "special relationship", but suddenly you contend they went to town for Itanium? Get real.
".....The HP-UX market was an HP-UX market...." Duh! The fact that hp had announced they intended to port hp-ux off PA-RISC onto Itanium had assured Itanium of viability in the first place. That meant Itanium was the heir to the hp-ux market, which was bigger than VMS and Tru64 combined.
".....It wasn't specifically a PA-RISC market....." Er, yes it was! The only supported platform for hp-ux was PA-RISC, the hp announcement that it was all to go Itanium in the long term (as it has) means the hp-ux market became an Itanium market. Honestly, stop talking male bovine manure.
"......People largely continue to buy HP-UX *despite* IA64, not *because* of IA64....." Companies continue to buy hp-ux because they get all the goodness of hp-ux but with the added speed that continued versions of Itanium have delivered.
".....Customers of HP-UX and VMS have little hardware reason to buy IA64, IA64 does little that can't already be done on an AMD64....." Er, except run OpenVMS and hp-ux you mean? DUH! Major fail. FFS, quit your shrieking and whining, we get it that you have an irrational hatred of Itanium because it killed Alpha. Get over it, move on!
@Matt Bryant
The court case is only almost over.
I'm sure that it is no coincidence that Intel waited until the appeal was denied before dropping this little bombshell.
Still, the next phase is where HP makes it's case for damages. I imagine that will be harder to pin any blame on Larry when Intel has clearly screwed HP as he claimed they would.
http://www.theregister.co.uk/2013/02/01/oracle_loses_hp_itanium_appeal/
The last laugh may still be Larry's, I think!
"I'm sure that it is no coincidence that Intel waited until the appeal was denied before dropping this little bombshell."
Interesting! I wonder how long Intel has had their plan settled? Intel's announcement certainly does make the judgement against Oracle look a lot weaker now. No doubt some lawyer somewhere is busily trying to work out if Intel's timing is entirely innocent and, if not, whether any legal action can ensue.
In a way it doesn't really matter whether the verdict against Oracle remains or is reversed. What matters really is how much cash Oracle have to give HP. I think you're right about HP finding it hard to pin any blame on Oracle now. They could walk away with a legal victory and no damages - hardly worth all the fuss. It would also be bad for HP's management team - it'd be regarded as yet another example of poor decision making on their part.
This post has been deleted by its author
Yes, but in HP/Intel's defense, it is difficult to determine if this is what Oracle knew would happen and that is why they dropped Itanium, or if this happened because Oracle dropped Itanium. When Oracle, the most important ISV in Itanium by a mile (or kilometer in Europe), decided that they were dropping support, they clearly panicked the entire install base who stopped buying HP Itanium and made plans to migrate elsewhere (IBM Power or x86-Linux). It is a bit of a self fulfilling prophecy. Oracle said that Itanium was dead and, as a result of saying it, it became true. I don't think it had a rosy future regardless of the Oracle situation, but it is clear that Oracle expedited the decline. It is difficult to determine causation.
".....The court case is only almost over....." No, it's over, all that needs to be settled is how much Larry has to pay for his massive FUD exercise. The final phase of the trial is equivalent to sentencing, Oracle's "guilt" has already been decided.
".....I'm sure that it is no coincidence that Intel waited until the appeal was denied before dropping this little bombshell....." As bombshells go, it's hardly major. Intel have not cancelled Itanium, as Oracle insisted they were going to, and have mentioned further development, which Oracle claimed would never happen. IBM's Power roadmap has slipped and dropped features plenty of times. Sun's old SPARC gag had more feature drops and slips than Power's and Itanium's combined, including the hilariously drawn out death of "Rock". I suppose the old Sunshiners just want any "good news" to save them from their misery, and the IBM fanbois tired of waiting for IBM's assurance that Power is the Only True Faith to come true.
Since there is a bit of discussion left to decide damages, it is premature to pronounce the court case as being over.
It isn't.
The really interesting thing is its effect on the Enterprise UNIX marketplace. That market is surely shrinking, but there are still mucho profits available there. Key application vendors still prescribe Enterprise UNIX for core functions in large deployments. If HP is losing traction and IBM doesn't advance its technology AND Oracle doesn't get more price competitive. then what happens to users in this market?
This post has been deleted by its author
I'm wondering who wants to run Oracle on Itanium with ridiculous license core factor of 1.0? Given this and performance metrics, there is no sense at all to stick with itanium platform unless the application is not available on other cpus (which is not the case with Oracle database).
Is there any benchmark on the planet where itanium is faster or equal to POWER or x86?
"I'm wondering who wants to run Oracle on Itanium with ridiculous license core factor of 1.0?....." Probably the people that bought IBM Power and didn't realise Turbo mode meant paying for Oracle licences for all those cores switched off, i.e. a core factor of 2!
".....Is there any benchmark on the planet where itanium is faster or equal to POWER or x86?" Obviously the ones where hp managed to sell all those Itanium servers. Duh!
"Is there any benchmark on the planet where itanium is faster or equal to POWER or x86?"
Yes. I suggest to have a look at TPC-C, for single servers running Oracle.
In 2007, Superdome 64p/128c 1.6GHz Montecito broke the 4Million tpmC barrier.
It took 5 years for Oracle to beat this number with a Sun X2-8 8p/80c Xeon E7 at 5M tpmC.
And the best Power result with Oracle is 1.6MtpmC, on a Power5 p5-595 in 2005. Power never ran TPC-C faster with Oracle.
Itanium is soon dead, and the HP FUD is going on. Intel has explicitly said Itanium is dead. Too bad HP is going down the drain. We have switched to IBM POWER because of major instability problems with the HP stuff, but POWER7 gives us everything we could dream of. We have replaced 1.500 HP servers with one single IBM P795 servers. These 64 cpu monsters gives us more performance for a fraction of the price. I have personally assigned 30 cpus to virtualize the HP load, and the other 34 cpus handle our SAP business. Great stuff indeed. Everybody should migrate to IBM, very soon POWER8 is coming. It will be 123% faster than HP Itanium, clock for clock, per core!
Whilst I'm not saying the last post was correct, you should not the following.
"Could be. But I'm sure you are aware, the 795 has a limit of 1000 logical partitions?"
IBM have just lowered the minimum processor entitlement per VP to 0.05. This would theoretically allow 20 partitions per processor (max 20 x 256), although officially they've left the number at 1000. I have to admit though, that anyone attempting to run that many partitions will hit lots of inefficiencies as no virtualisation technology could give that kind of consolidation ratio per core.
"IBM p795 has a maximum of 32 CPUs."
32 CPUs yes. However, that's 256 cores and therefore 256 processors in IBM speak. The 0.05 entitlement per VP is efvectively per processors, hence the 256*20. But yes, the limit is officially still 1000.
"Given that Itanium roundly trounced PA-8K in benchmarks from the beginning, I have a hard time qualifying the move from the PA-8K design as a bad idea. That being said, IPF is pretty obviously dead as a doornail at this point."
It didn't. I spoke with someone locally who was using PA-RISCs. They got a Itanium, and the Superdome they had ***SPANKED*** it. After a good year or so, the compiler improved enough so the Itanium was at least competitive. They basically got one, got all ready to return it, and ended up getting it for almost nothing so whoever at the company could show a successful Itanium sale.
Anyway... I think the Itanium met it's goal for Intel. It seems to me the goal was largely to get more vendors buying from Intel instead of using their own chips. In terms of UNIX systems? MIPS -- gone. PA-RISC -- gone. Alpha -- Gone. Sun didn't abandon SPARC, but got sidetracked enough to delay the SPARC by quite a while. IBM seems to be the only one who was unaffected and still making POWER systems for those who want them. Don't get me wrong, some of these chips were "on the ropes" anyway, but I think Intel may have more than made up for the huge cash investments into Itanium in additional Xeon sales.
"....the demise of Itanium...." See, once again you are confusing reality with your spoonfed fantasies. The announcement is of a slip, not a cancellation. Go get yourself signed up for remedial reading, you're embarrassing the rest of the trolls.
".....Good job he didn't take my bet....." If I wanted a sure bet it would be on you exposing your stupidity with every post you make.
I couldn't let the last post to be from Matt.. it isn't fair.. he appears to be so anti-IBM.. and IBM really tried hard to embrace Itanium:
"As part of Project Monterey, IBM released a beta test version of AIX 5L for the IA-64 (Itanium) architecture in 2001, but this never became an official product due to lack of interest." (http://en.wikipedia.org/wiki/IBM_AIX)
"....and IBM really tried hard to embrace Itanium....." The really fun bit about IBM's Itanium "support" was how hard they tried NOT to sell it! Their salesgrunts never offered the X450 series Itanium servers, they had to be badgered into selling you one. IBM's standard sales patter was to ignore the customer's requirements and suggest a mainframe, if that failed suggest an Power-based AIX server, and if that failed then offer a Xeon server. Then just keep dropping the prices on those three until the customer bit or the opposing vendor pulled out of the deal. But IBM still managed to sell over 10,000 X450 series servers to IBM customers that found the Itanium server was a better solution to their requirements than mainframe, Power-AIX or Xeon, despite IBM trying their best to convince them otherwise.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
