Intel has released the final Itanium chips, the generation codenamed Kittson, with up to 8 cores on a 32nm process:
One of Intel's ventures into the historic mainframe space was Itanium: a 64-bit capable processor designed in conjunction with Hewlett Packard. The main reason for Itanium was to run HP-UX and compete against big names, such as Oracle, using a new IA-64 instruction set. The appeal for the original Itanium parts was support for RAS features, ECC, and cores focus on a wide, parallel architecture - the latest cores support 12-wide execution for example. For a short while, there was success: HP's systems based on Itanium are advertised as high-uptime mission critical servers, and a number of customers cling to these systems like a child clings to their favorite blanket due to the way they are integrated at the core of the company. The main purpose was to compete against other mission critical servers and mainframes based on SPARC and IBM Power.
So when the processors were initially delivered to customers, there was potential. However the initial impression was not great - they consumed too much power, were noisy, and needed over the top cooling. Over the years and generations of Itanium, the march into the x86 enterprise space with x86-64 drew potential Itanium customers away, then followed the drop of Microsoft's support for Itanium in 2008, and Oracle's dropped support in 2011. Xeon offerings were becoming popular, with CPUs incorporating the RAS/ECC features required, and Intel decided to slow down Itanium development as a result. In the meantime, due to the way the market was moving, HP transitioned a good part of its product stack to Xeons. Despite this, legal battles between HP and Oracle ensued given predicted support for HP-UX customers. At this point, there were fewer potential Itanium customers each quarter, although existing customers required support.
Today marks the release of the final known variant of Itanium, the 9700 series, beyond assurance testing. Intel spoke to IDG, stating that this generation, code-named Kittson, would be the final member of the Itanium family. These chips are likely to only end up in HP-based Integrity i6 high-uptime servers running HP-UX, and start at $14500. Hewlett Packard Enterprise has stated previously that it will keep support for Itanium-based products until 2025, with the latest OS update (HP-UX 11i v3 2017) coming in June.
(Score: 3, Interesting) by Snotnose on Friday May 12 2017, @11:51AM (15 children)
How does a chip get "noisy"? Do they mean the fans required to keep it cool were noisy?
When the dust settled America realized it was saved by a porn star.
(Score: 3, Funny) by Nuke on Friday May 12 2017, @11:59AM (8 children)
How does a chip get "noisy"? Do they mean the fans required to keep it cool were noisy?
No they can't have meant that because they mention "over the top cooling" as a separate criticism. The chips must have been noisy in themselves like an electronic Super-Brain in a 1950's Hollywood Sci-Fi B-movie sort of way. You know "Whheeaaooww .... Whheeaaooww ..." with the lights dimming up and down at the same time.
(Score: 5, Funny) by theluggage on Friday May 12 2017, @01:15PM (7 children)
You know "Whheeaaooww .... Whheeaaooww ..." with the lights dimming up and down at the same time.
The problem is that the typewriter sound that all computers make when outputting text gets higher in pitch as the speed of the processor increases and, with the itanium, tended to upset local cats and dogs, shatter wineglasses etc. - although it could also cure your kidney stones. AMD-64's introduction of patented VT/ST ("virtual teletype supression technology") was one of the reasons it beat itanium.
Also, intel's failure to deliver a working Jacob's Ladder [wikipedia.org] driver allowed PPC and ARM to get a head-start in serious scientific computing.
However, the main thing that derailed itanium was the Intel's doomed anti-trust legal battle with chip giant Intel, claiming that Intel was abusing it's near-monopoly on the x86 processor architecture to prevent Intel from breaking into the market with it's non-x86 Itanium architecture - the resulting legal singularity eventually forcing them to adopt the rival Amd64 instruction set because their paradox neutralising machine needed working Jacob's Ladder drivers.
Sorry - it must be Friday.
(Score: 1) by khallow on Friday May 12 2017, @10:12PM (5 children)
Also, intel's failure to deliver a working Jacob's Ladder driver allowed PPC and ARM to get a head-start in serious scientific computing.
What is "Jacob's Ladder" here? Sounds pretty big, but Google isn't my friend today and I can't figure out what it's supposed to be.
(Score: 2) by butthurt on Saturday May 13 2017, @04:15AM (4 children)
A Jacob's ladder (more formally, a high voltage traveling arc) is a device for producing a continuous train of large sparks that rise upwards. [...] often seen in films about mad scientists.
-- https://en.wikipedia.org/wiki/Jacob's_ladder_%28electricity%29#Visual_entertainment [wikipedia.org]
(Score: 2) by theluggage on Saturday May 13 2017, @11:52AM (3 children)
...which I did actually link to in the post, using one of those new-fangled hyperlink thingies where you click on the underlined text...
(Score: 1) by khallow on Saturday May 13 2017, @12:02PM (1 child)
(Score: 1) by khallow on Saturday May 13 2017, @12:09PM
However, as the key ingredient in the KSDFT calculations, the exchange and correlation functional to determine the accuracy must be approximated. There are several choices of the so-called Jacob's ladder in the KSDFT calculations, including local density approximation (LDA), generalized gradient approximation (GGA) and hybrid functionals. However, the widely used semi-local LDA and GGA functionals fail to give accurate electronic structures in molecules and semiconductors due to lack of long-range nonlocal Hartree-Fock exchange interaction in the KSDFT calculations.
(Score: 1) by butthurt on Saturday May 13 2017, @12:47PM
Thanks, I see that now. I had looked at khallow's quote, from which your hyperlink had been elided.
(Score: 2) by driven on Wednesday May 17 2017, @07:35PM
Nice post. You should write for Cracked!
(Score: 0) by Anonymous Coward on Friday May 12 2017, @12:43PM
I guess they are speaking about electrical noise.
(Score: 3, Interesting) by RamiK on Friday May 12 2017, @01:13PM (3 children)
Here's another one:
there was success: [...] a number of customers cling to these systems like a child clings to their favorite blanket due to the way they are integrated at the core of the company.
That's not a success. That's a handful of sales. A success is turning a profit or at least making a technical breakthrough and paving the way for others.
So when the processors were initially delivered to customers, there was potential. However the initial impression was not great - they consumed too much power, were noisy, and needed over the top cooling.
No. Itanium failed over the (lack of) compiler optimizations.
compiling...
(Score: 2) by butthurt on Saturday May 13 2017, @04:21AM (2 children)
> A success is turning a profit or at least making a technical breakthrough and paving the way for others.
Maybe one out of three?
Outside embedded processing markets, Intel's Itanium IA-64 explicitly parallel instruction computing (EPIC) and Elbrus 2000 appear as the only examples of a widely used VLIW CPU architectures.
[...]
In December 2015, the first shipment of PCs based on VLIW CPU Elbrus-4s was made in Russia.
-- https://en.wikipedia.org/wiki/Elbrus_2000 [wikipedia.org]
(Score: 2) by RamiK on Saturday May 13 2017, @10:18AM (1 child)
I should have specified "...and paving the way for others to succeed". Maybe add "purposefully" and "causal relations" too... After all, one neither credits the apple tree for the discovery Laws of Motion nor Lamarck for Evolution.
compiling...
(Score: 2) by butthurt on Saturday May 13 2017, @01:15PM
> I should have specified "...and paving the way for others to succeed"
I meant to link to Wikipedia's VLIW article; copy-pasting the one about the Elbrus 2000 was a mistake. I wasn't trying to imply that the Itanium led to the Elbrus.
Isn't Itanium the only exemplar of an EPIC architecture? The fact that it exists and functions could be considered an accomplishment--although I did see the other commenter's remark that the distinction between VLIW and EPIC was a matter of marketing some "tweaks."
(Score: 2, Informative) by Anonymous Coward on Friday May 12 2017, @02:28PM
> How does a chip get "noisy"? Do they mean the fans required to keep it cool were noisy?
What do you expect? Its anandtech. A site by and for 'enthusiasts.'
If it doesn't run call of duty and can't be pimped out with LEDs they don't know how it works.
(Score: 2, Interesting) by KiloByte on Friday May 12 2017, @12:33PM (8 children)
And what systems can actually run on this shiny new chip? Is there even a distribution left?
Windows 2008 Server (dropped in 2012) -- well into LTS.
Debian squeeze (dropped in wheezy) -- EOL.
Red Hat RHEL 5 (dropped in 6) -- LTS.
Suse SLES 11 (dropped in 12) -- well into LTS.
Gentoo -- long since unmaintained.
FreeBSD 10 (dropped in 11).
NetBSD -- long dead dev branch, never released.
Even HP's own HP-UX cancelled their Itanic port before release.
So what's left?
Ceterum censeo systemd esse delendam.
(Score: 2) by epitaxial on Friday May 12 2017, @01:09PM (1 child)
If you're buying Itanium right now then you have existing systems that can't be migrated. You're not running Linux either. I wish OpenVMS had gained more traction and not stayed a niche operating system.
(Score: 1) by technoid_ on Friday May 12 2017, @01:44PM
AT&T had quite a few Superdomes running RH5 last I had access (2013). Some of theses were new systems that existing apps and data were being migrated to.
(Score: 3, Interesting) by AndyTheAbsurd on Friday May 12 2017, @01:32PM (1 child)
Well, there's NonStop Kernel on Itanium-based HPE Integrity Nonstop [hpe.com] systems. But even those are moving to Xeons these days. Apparently they're also now supporting virtualized NonStop, which...ummm... the strength of NonStop was that everything - including the hardware - ran in tandem (in fact, NonStop systems are descended from HP's acquiring of what used to be Tandem Computers), so virtualizing it seems counterproductive unless you're ONLY doing it so that devs have local environments.
Please note my username before responding. You may have been trolled.
(Score: 2) by TheRaven on Friday May 12 2017, @02:20PM
sudo mod me up
(Score: 3, Informative) by butthurt on Friday May 12 2017, @01:44PM (1 child)
Someone took the trouble to list the remaining operating systems for Itanium on Wikipedia:
https://en.wikipedia.org/wiki/Itanium#Software_support [wikipedia.org]
Among the discontinued ones, there's NonStop OS, and as someone else hinted, there's OpenVMS.
https://en.wikipedia.org/wiki/NonStop_OS [wikipedia.org]
https://www.hpe.com/us/en/servers/nonstop.html [hpe.com]
(Score: 2) by butthurt on Saturday May 13 2017, @04:39AM
Clarification: NonStop OS and OpenVMS are the ones that have not been discontinued. Wikipedia's claim of GCOS for Itanium has no citation, and when I checked the GCOS home page there was no mention of Itanium.
(Score: 1, Informative) by Anonymous Coward on Friday May 12 2017, @02:32PM
> Even HP's own HP-UX cancelled their Itanic port before release.
lolwut?
Itanium systems shipped with HP-UX from the very start and for a while it was the only available OS.
(Score: 2) by tempest on Friday May 12 2017, @05:59PM
When NetBSD doesn't support it and still supports the Sega Dreamcast, you know it's a dead architecture.
(Score: 4, Funny) by turgid on Friday May 12 2017, @01:26PM (2 children)
God bless itanic, and all who sail on her.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 0) by Anonymous Coward on Friday May 12 2017, @02:21PM (1 child)
NOT FUNNY
(Score: 2) by bob_super on Friday May 12 2017, @09:40PM
Too soon?
(Score: 0) by Anonymous Coward on Friday May 12 2017, @01:36PM (3 children)
The main feature of this CPU was "wide instruction set computer" (WISC) - it crams multiple instructions into a single 64-bit word and execute them simultaneously. It effectively offloads to the compiler the task of figuring out how to cram multiple op-codes into a single word that can be executed simultaneously. Achieving the parallelism requires some serious compiler voodoo magic.
Don't know if it ever incorporated branch prediction/out-of-order execution logics that other CPUs employed to achieve parallelism.
Anyway, in the real world, FAIL, not just because it's WISC. "WISC" isn't even an entry in wikipedia anymore.
(Score: 0) by Anonymous Coward on Friday May 12 2017, @01:39PM
Actually, wikipedia does have an entry for "VLIW" - very long instruction word.
(Score: 2) by TheRaven on Friday May 12 2017, @02:22PM (1 child)
Anyway, in the real world, FAIL, not just because it's WISC. "WISC" isn't even an entry in wikipedia anymore.
That's because no one apart from you uses WISC. The generic term is VLIW (Very Long Instruction Word), though Intel tried to market their tweaks (which reduced the coupling between architecture and microarchitecture) as Explicitly Parallel Instruction Computing (EPIC). Both of these terms have Wikipedia entries.
sudo mod me up
(Score: 0) by Anonymous Coward on Friday May 12 2017, @04:27PM
I think "WISC" was one of the labels used to refer to this architecture style in its early days (back in the 80s?), but, yeah, VLIW is the common general term.
(Score: 2) by butthurt on Friday May 12 2017, @01:48PM (4 children)
The article says the chips will "start at $14500" but it has a table showing prices from $1300 to $4650.
(Score: 2, Interesting) by Anonymous Coward on Friday May 12 2017, @05:57PM (2 children)
Entry level systems start at 14500 USD.
Seems expensive compared to an x86_64 box, but if you compare it to a power or sparc instead, it is right in line with the reason 2/3 of those processor architectures are dying off in the next few years (Fujitsu was the last remaining developer of SPARC and is migrating to ARM for their next supercomputer. Oracle already laid off their SPARC hardware staff, and every other company got shafted/bought out by the two back in the 90s with the sparc32 to 64 migration.)
It might be time to look into our own independent chip designs in the near future. I was just reading the Hackaday interview with Chip Gracey and his (Brother? Father?) from Parallax (makers of the propeller and stamp boards) in Rocklin, CA. The most interesting part of the interview was the bit about 250k being enough to get 180nm chip masks/production line spun up. Sure this is only ~ Pentium 3 era technology, but if enough people pulled together/crowd funded an SoC, or (if a million dollars or so could be pulled together plus the hardware engineers.) a complete motherboard chipset + cpu using tech from expired patents (notably PCI/SDRAM/etc up to 1997 if a full 20 year patent period was invoked?) We could work to free ourselves from the Wintel/Armdroid duopoly and begin providing the sort of free market solutions we had back in the late 80s to mid 90s before Intel managed to crush its competition through legalities while the 'big box' manufacturers killed themselves by pricing out of the market.
(Score: 1) by mechanicjay on Friday May 12 2017, @08:15PM
I think this is a pipedream. It's a pipedream that I am now thinking very very hard about.
My VMS box beat up your Windows box.
(Score: 2) by bob_super on Friday May 12 2017, @09:44PM
Good luck becoming a serious competitor to Intel with an architecture that can dodge their patent portfolio...
(Score: 2) by takyon on Friday May 12 2017, @05:59PM
I reckon it's an error or a clumsy attempt at telling us the minimum price of an HP-UX Itanium server.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Insightful) by aim on Friday May 12 2017, @02:07PM (3 children)
I recall that we (organizers of a Linux conference) got ahold of a demo server, back in the early days of Itanium. It came already preloaded with a Linux distribution, can't remember which one (maybe RedHat?). But man, was that box slow. We had experience with all sorts of hardware, including x86, PA/RISC, Alpha, SPARC, SPARC64, which all pretty much blew Itanium out of the water. The architecture was pretty much a non-starter for us, we never went with it, even after the platform had some time getting optimized.
I still keep wishing HP had stayed on the Alpha path instead of betting (and mostly losing) on Itanium. I guess AMD - helped by ex-Alpha people - made the right choice upgrading x86 to x64, which has become the standard (besides ARM in the mobile space).
(Score: 4, Interesting) by TheRaven on Friday May 12 2017, @02:25PM (2 children)
sudo mod me up
(Score: 2, Informative) by Anonymous Coward on Friday May 12 2017, @02:36PM
Hell, it might have even been x86 code running on the 'emulator' / co-processor part of the chip. That was dog slow.
(Score: 2) by turgid on Friday May 12 2017, @05:51PM
I seem to remember reading somewhere rants by Linus implying that any compiler for itanic would have to be very clever indeed, in that it would have to be able predict the future. The decision not to put certain things into the hardware meant that itanic wasn't much use for anything other than tight Fortran loops.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 2) by LoRdTAW on Friday May 12 2017, @10:49PM (1 child)
I came across an Intel/HP Linux Itanium development CD from I got at PC Expo (TechXNY) back in 2001/2002. I think that was the same show where Compaq was demoing the 1GHz Alpha. One thing is certain, Compaq was fucking stupid to sell Alpha to Intel.
(Score: 2) by turgid on Saturday May 13 2017, @12:52PM
I remember there being an itanic box at a Linux Expo back in about 2002/2003 in London. The RedHat people had it. At the time I was working for Sun and UltraSPARC development wasn't going well compared with POWER. Many people were very critical of Sun but what was really funny was when I asked to see the itanic box, I was refused. I asked a couple of times but it transpired that it had been switched off and put in a cupboard. I asked why. Apparently it had overheated.
Sun did an itanic port of Solaris (internal, not released) in about 1999. Sun did not cancel UltraSPARC.
I knew a Debian developer. He had an itanic box in his house for doing the itanic build of Debian. He used it for drying his clothes, I seem to remember.
There was a massive amount of pro-itanic propaganda going about at the time. Daring to criticise or question it on the green site used to result in -1, Flamebait moderations. I also seem to remember various high-profile people who dared present unfavourable reviews being censored, and things mysteriously disappearing from the Internet.
Google Linus's opinions on itanic if you want the truth. It was doomed from the outset. It was the wrong design for the job.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].