Despite HP's healthy rise in Itanium server sales, IBM think its major rival and Intel will have to give up on the processor in the near future due to basic economics. "The end of life for Itanium will occur in the next five years," IBM VP Scott Handy told us, during an interview here in Austin, Texas. "(HP) will have to …
Power6 on everything
The fact that Power ends up all over the place actually is a good reason to bet on it.
Sun may be making some really good chips, but if you recall, SGI had some really good performing systems that just didn't work out economically for SGI because of economies of scale.
And you know what happened with that. They switched to Intel.
Sun might be able to make a really awesome chip; indeed they have. But making it worth the development cost is another matter.
HP deserves to die a slow and painful death for killing the Alpha!
I always believe a company when it trashes a rival. After all, they wouldn't do it just to convince prospective customers of the trashed company that they would be safer elsewhere.
Twasn't architecture but Billy that killed the beast
SGI died because they thought there was a market for hyper-expensive non-x86 Windows NT workstations. Digital did the same with the Alpha.
Certainly not IBM - sowing Fear, Uncertainty and Doubt among competitors' customers..? ;-)
It's A Blue, Blue World
IBM makes it's money off it's mainframes, and only sells anything else to further more sales of their proprietary systems.
The Armonk Monster loves Unix, because it's no threat to their lock on large corporate transaction processing systems of record, which are what everything else depend upon.
If HP had had any wit, they would be working on proprietary hardware and software to support their Guardian/Pathway systems; as those are the only real alternative to the descendents of System 360. But HP have fallen for the commodity fallacy, and their corporate head will be up on IBM's trophy rack in ten years or so, staring with taxidermy eyes at Sun, NCR, DEC and maybe even Microsoft by then.
Ah well, I've got me blue BAL coding sheets around here somewheres...
Regarding SGI and NT
(answering the AC)
The NT from SGI was running on X86 machines, but with SGI chips which required some heavy modifications of NT, they were the Visual Workstation 320 and 540. You couldn't just slap an NT on them (heck, there wasn't even a standard BIOS on them if I recall correctly).
Commercial Reality and Choice
Actually IBM are sadly, probably right, IT is a "Me Too" industry, where customers will, like lemmings hurl themselves at a successful product, often in the face of overwhelming evidence to the contrary. Most of the people who sign the checks don't know one end of a computer from the other, so they figure, I know that name, I won't get fired for buying it. I still get asked on every project I do that isn't Wintel why are you using this, it'll be cheaper to use Windows and SQL.
IBM may be Ok with Power and Cell processors, they have the financial muscle, but don't bank on it. It is not beyond the relms of impossibility the we land up with an Intel world.
Still we can't rule out Intel making a major mistake with x86, there is that hope. As an architect I like choice because I can choose the right solution for my customers, the sensible ones, who have the money, run a heterogeneous estate taking the best each technology can offer.
Intel x86 mistake?
What, like trumpeting Itanium as their strategic 64 bit solution and then having to run the Red Queen's Race when AMD shipped 64 bit x86 hotly pursued by dual-core versions?
There's a reason that Itanium updates have been slow and disappointing. Intel couldn't rush to ship "me too" 64 bit x86 with multicore variants AND come up with set of x86 AMD-squashers to regain their market share longer term AND maintain focus on the Itanic. Something had to give.
The trouble here is that Chipzilla weathered that little cockup and came back as strong as ever and I am sure that should they make another such monumental pig's ear, they'll get away with that too.
HP: guardians (for now) of Tandem and of VMS
"If HP had had any wit, they would be working on proprietary hardware and software to support their Guardian/Pathway systems; as those are the only real alternative to the descendents of System 360."
Allegedly they are. But then again, allegedly they are also still working on OpenVMS.
Both are available on Itanium-based members of the HP Integrity (sic) server family. You'd hardly know it though, because the meejah are barely interested in anything which isn't Windows. Even the HP-sponsored (?) Forrester Research's report on what's special about Integrity forgets to mention any OS except Windows, Linux and HP-UX.
http://h20341.www2.hp.com/integrity/downloads/ (URL wraps here...)
There's a "disaster proof" video from HP on Youtube which is worth a few minutes of anybody's time if high availability and in particular disaster tolerance is of interest. VMS *does* feature in this one, which makes it the exception that proves the "HP never mention VMS" rule. The associated technical backup material is on the HP website too. http://www.youtube.com/watch?v=qMCHpUtJnEI
Big Blue going Norwegian Blue and an ex Parrot
""The customers we call on perceive Sun as increasing its prices," he said. "They're using margins on the high-end to fund lower-margin efforts."
Seems like the Perfect Business Intelligence Strategy for the high-end has money to burn for Systems which work for them rather than for the Systems Vendor. A painful lesson which IBM may be slow/reluctant/unable/disenabled to learn.
And Paris because she's also a bird and ex-bird if now growed up into a Smarter Woman.
More to a system than a processor
Clearly processor architectures matter, but there's more to creating systems which do well in the market place than just betting on the right CPU.
So with each SPARC iteration, Sun have been designing a processor which will fit into their systems, rather than going to a processor parts bin and then trying to put a system around it. Similarly IBM with POWER, HP with PA-RISC and DEC with Alpha.
And if you go to a datacentre which is predominantly IBM, or SUN, you will find that the customer is loyal not so much to the processor architecture itself but to what they can achieve with the systems. So Sun's appeal to customers is not SPARC per se, but SPARC *and* Solaris *and* CMT *and* their overall system design. So their entry into the x64 server space is as much about bringing the Sun twist to this sector as about hedging their bets on SPARC.
Which brings us back to the subject of the article. HP have done some very clever stuff with their C-class blades in the volume space, but converting the PA-RISC and Alpha installed base to Itanium was always going to be a leap of faith, even if Itanium adoption was widespread. Obviously it's in IBM's interests to FUD Itanium as these HP systems come to the end of their natural - even if the defections go half and half to IBM and Sun, there's a huge opportunity.
Whether el Reg should pick up an "IBM FUDs Itanium" press release is another question ...
Any one notice the
PA-RISC Hilton angle?
Itanic Road Map to Xeon
If you look at the Itanic roadmap it's easy to figure out where it will change to Xeon.
Madison - 2003 pretty good chip, but few customers moved from PA-RISC
Montecito - 2006, three years later, finally dual core, huge cache, customers that have to stay on HP-UX finally start moving. Meaningless Linux or Windows sales or sales outside of HP.
Montvale - The Halloween chip. 4% higher GHz and you can't even put the faster FSB chips in HP highend systems. The only reason this chip was not canceled was to try to get the market to think the Itanic has 5 more years. All Itanic systems today are end of life in 2013 according to HP
Tukwila - 2009 in systems. Quad core and finally 65nm. Wish Intel would not be such a laggard in chip fab technology. Expect the cores to have significantly less performance per core than Montvale which will make software costs prohibitively expensive.
Poulson - There is no commitment...just future which allows it to never happen.
Xeon will be available in 64 socket systems by this time, so there is no need for Itanic. Expect HP to provide emulation software on Xeon to support HP-UX. Similar to how Itanium supports x86 now that they took that out of the itanic chip. HP will move their install base to a real commodity processor.
Kittson - it is a cute name on a processor chart but other than saying it could be the 9th itanic chip it is just a name.
Cheers..as it is late.
The real question is the death of HP-UX
Itanic is sure to die, but what about HP-UX? Will the itanic passengers survive or will HP mgmt let they die with the boat? The only thing HP has going for it is it is hard to say when an O/S dies. Expect V4 of HP-UX in 2010, but I am sure it will be on life support along with the comatose TRU-64, VMS and the NonStop Kernel.
Who killed Alpha?
IIRC, Compaq was the hitman, hired by Intel, before HP got involved.
I'm no longer any sort of HP fan (although I'd love to have a VMS/Itanium workstation for much the same reasons I'd love to have a Showman tractor :-), but that particular deed is not, IMHO, their fault.
What, more reports of The Death of Itanium?
Blimey, I've lost count! Reports of Itanium's death haven't just been wildly exaggerated, there's a whole cottage industry behind it! I expect if they don't announce The Death of Itanium at least monthly then somre poor copywriter's kids go hungry.
So, Mr Handy, if Power is just so gosh-darn wonderful why would anyone buy anything else? Surely IBM should have 100% of the enterprise server market. Could it be that customers try find that other systems (HP, Sun, FSC....) just happen to do the job better? And if Power6 is so much better than everything else, why is it only in a very small range of servers? And why is it starting to pinch ideas from Itanium like in-order execution? And if all those recent Itanium chips are so poor, how come they beat Power5 in shoot-outs? I'd bench against Power6 but you just don't make any servers with Power6 to put up against a Superdome.
Oh, and who was blathering on about the IBM mainframe bizz? That would be the IBM mainframe bizz that's so threatened by emulators on Itanium that IBM is suing everyone whom even thinks about it. Get real!
I think Mr Handy is trying to be the next Jonathan Schwartz, no doubt next we'll see him cavorting around in a penguin suit trying to flog Linux on Power6!
Nice article except for the last paragraph
Are you seriously suggesting IBM can't go fast and multicore to handle "multithreading". As if multithreading was really the bottleneck of modern machines.
IBM's got multicore/chip experience in spades not least with the Cell monster. They can squeeze more Power cores on the same die as x86. Intel will continue to pursue x86 optimisation over (Alpha) innovation as long as they can. That in the face of more *efficient* Power/Cell and ARM designs.
As for growth figures: some absolute figures would have helped. Are HP now shifting more than a couple of thousand of boxes a year?
As for the Linux muppet: get yourself a proper unix! ;-)
Bla, Bla, Bla : just look at the darn numbers
There are plenty of accurate public figures concerning server sales, with details concerning Sun, HP's Itanium and IBM's Power sales. IDC, Gartner and the like make a living counting those sales, quarter after quarter, and the IT pros use those sumbers daily to convince their customers...
... but surprisingly not a single precise figure has been cited in the article...
What we see in the numbers is that Sun's server sales are high (again), that HP's Itanium server sales are high, that the blade form factor is a clear winner in the IT industry, that HP is stronger than IBM in blades, and thus that their Itanium-based c-class blades are a threat to IBM.
What will happen in the next 5 years ? ... IBM usually FUDs when they feel they are at risk. I guess they probably are.