With Oracle pulling the rug out from underneath Intel's Itanium processor in an effort to undermine rival HP's Unix-server business, two obvious questions arise: What makes Itanium so special that it has roused the disaffections of the software giant? And why hasn't Big Blue's Power Systems platform been tarred with the same …
TPM shows his IBM love again
As usual, TPM never has a crossed word to say about IBM...
> IBM has a tidy business selling Power-derived processors for game-console makers Sony,
> Nintendo, and Microsoft, and also gets a fair amount of development money – hundreds of
> millions of dollars a year – from the US government and supercomputing centers for high-
> end Power-based clusters.
> Itanium does not have these assets backing it up. There's just Intel making them and HP
> selling them at this point, with a few boxes being sold by Bull, NEC, and a few upstarts such
> as Super Micro, Huawei Technologies, and Inspur.
Ah, so you are saying that IBM's "not-so-profitable" Microprocessor division is propped up by making volume processors in the same fabs, and sharing some of the designs.
But Intel's "not-so-profitable" Itanium division is apparently not propped up by making Itanium in the same fabs and sharing some designs with a volume processor? Funny I thought there was this processor out there called Xeon which had a pretty reasonable volume, is made on the same fab lines as Itanium and shares a bunch of chipset technology? If TPM applies this logic to Power, he has to apply it to Itanium as well.
Good luck to IBM in this space, but I suspect for every 100 customers that decide to get off SPARC or Itanium 99 will go to x86 anyway - as the leader in x86 systems, HP will most likely get the majority of these - the systems themselves obviously command much smaller revenues and margins than commercial UNIX systems, but the door into the data center is open for selling the rest of HP's portfolio
Different views of the same picture...
My reading of TPM's articles is that SPARC/Itanium/Power and their associated UNIX versions are all under threat from x86/Linux.
SPARC looks the least healthy - Oracle isn't showing it a lot of love, it's struggled to perfrom at the same levels as it's rivals over the last 5+ years and there is no magic new CPU revision that will change this scheduled for release.
Itanium may be backed by Intel, but it's hard to see why ever since AMD removed the artificial 64-bit barrier for x86. The only reason for Intel not killing Itanium is that they (apparently) use Itanium systems in their fab control systems.
Itanium is backed by HP because it ported it's OS/software to Itanium and can charge a premium for the hardware over the x86 kit where it has to compete with other x86 vendors. If it moves to x86 there will be an associated drop in hardware revenue.
POWER has a roadmap that is likely to happen and has the current performance crown. More importantly, IBM control the destiny of POWER - if they can sell it, it will survive.
As the CPU wars move on from x86 vs high-end RISC to ARM/MIPS vs x86 over the next 5-10 years, SPARC and Itanium will definitely be dead in big, non-HPC servers for anything other than replacing existing systems that haven't been migrated to an alternative platform. The future of POWER isn't as obvious due to the IBM-factor.
Not that so many businesses put there future in the hands of a particular chip, but that they put it in the future of one bit of software.
IIRC Stallmans driver wasn't a cummunist software utopia, as many portray, but a mitigfation against software suppliers yanking the plug on stuff.
Why don't Oracle target IBM?
IBM has a software portfolio that runs on Solaris / OEL and can hit back by dropping Solaris support for MQ, WebSphere, DB2, etc. There's not much HP can do to move people off Solaris.
Yes, that is good insight. Furthermore I think that Oracle is betting on the UNIX marked becoming a two horse race (3 if you call it *NIX like marked). And they want to make sure that they are one of the two left standing.
And right now HP seems to be the weakest of the two that Oracle has to battle. So it makes sense seen in a cool hard business perspective.
Sure us Oracle/IBM/HP customers who wants the UNIX marked to be a 3+ marked, due to competition, are pissed at Oracle. But that is another story.
Oracle HW still sucks
I'm going to be ruthless here: ultimately Itanium will go away, but what will kill it is not Sparc, not IBM: it's x86, where you can not only pay little for the servers, you can get RHEL or CentOS for not much money either.
What Oracle have done here, short-term, hurts HP, but in the long term it says something else "if you keep your data in Oracle DBs, they get to control you".
"...I think that the reason we won't end up in the same boat as Itanium is that the Power market won't end up in the same spot," says Gelardi. "There is no software company in the world that wants to be dependent on one hardware supplier..."
What does he mean? Isn't POWER cpus only from one vendor? Just like Itanium?
SPARC is open sourced, so anyone can manufacture and sell SPARC. In fact, the chinese super computer on TOP500, Tianhe, is a SPARC derivative. And also Fujitsu sells SPARC.
I think what he means is that softies don't want to be left with "just amd64". They were happy to lose MIPS, Alpha, Sparc and Motorola, and they will not mind losing Itanium because there is still Power left. But they won't want to lose Power because it is the last remaining alternative to keep AMD honest.
I'm not sure I find that argument convincing, myself. Microsoft have spent most of the last decade untroubled by their lack of an alternative and when (recently) they finally went looking for one they picked ARM, not Power. So OK, ARM aren't yet playing in the server space, but if they ever decide to then they'll have a huge consumer base to subsidise their initial forays.
"also Fujitsu sells SPARC."
Don't kid yourself that SPARC is more open than POWER due to Fujitsu.
Try and buy one off Fujitsu in the UK and you'll find you are completely stitched up as even though they own the IP in the M Series product line, they're not allowed to sell it as a vendor outside of AP.
Given this is Unix and the land of high list price / big discounts, does it also make you feel happy that Fujitsu "do SPARC" when all bid requests have to be approved by Oracle - how well do you think that works getting you a price that is better than Oracle's?
There were a lot of disappointed Oracle partners out there when they realised they effectively have not choice but to "work" with (or for) Oracle....
Freescale makes POWER chips, although not anywhere near as fast as what IBM makes. But they make them. I believe others could choose to do so too.
Sparc is made by others as well, but again, not with nearly the same performance as the Fujitsu ones. There are some sparc compatible chips made in China, just like there are MIPS64 chips made in China (which are looking very interesting).
The only truly dead ones are Alpha and PA-Risc, and if volumes are anything to go by, Itanium which never lived in the first place.
I think Microsoft went to the ARM route because Windows 8 is aimed at the consumer. But still, Windows is processor independent for some time with the NT Kernel and .NET. The NT kernel has a history of running on many architectures and there are .NET CLR ports for x86, x64, Itanium, ARM (Windows Mobile/Phone) and POWER (Xbox360).
Mind you, I'm not saying it's just a matter of recompiling their software, but it seems they want some independence from Intel after all.
Ok, I did not know this. Thank you for explaining this to me.
A question, it seems only IBM makes POWER cpus? Freescale does PowerPC cpus which is another thing?
So, is it only IBM that does POWER cpus? Is this true?
Well, as far as I know, there are other vendors developing and selling SPARC cpus. Doesnt this make it more open than POWER?
And SPARC has been officially open sourced. You can download the spec, and manufacture a SPARC yourself. Doesnt this make SPARC more open than POWER? Or?
You really need to try to read up on the stuff you are talking about.
It's all called POWER today, and have been for quite some time.
Moving to Linux
Moving to linux would be good option but I can bet any money that next move from Larry would be saying that the only good linux is our linux and dessuporting RHEL and Suse. So by moving from Itanium to linux you'll end up in same basket.
But I must also add that I believe that all oracle users will end up in same basket sooner or later anyway as I expect finally that it will be available only for oracle-owned platforms. With Solaris as a default DB boot loader, Time to start thinking about some serious alternatives to oracle - but where they are when they are needed... DB2 is not, MSSQL neither.
"Intel .. Itanium .. fab systems"
"The only reason for Intel not killing Itanium is that they (apparently) use Itanium systems in their fab control systems."
Yes and no.
When I last checked, what was important to the automated fab systems at Intel wasn't IA64, it was VMS, because that's what the fab applications (commercial and home grown) use.
If VMS was available on something other than IA64 (e.g. AMD64 or Intel copy), it would be just as relevant to Intel's business operations (probably more so ) than it is on IA64.