back to article Presto crypto: IBM releases gruntier, faster Z14 mainframe

IBM has launched its latest, newest, biggest, baddest mainframe, the z14 system. It features the next generation of IBM's CMOS mainframe chip technology, with 10-core processors using 14nm silicon-on-insulator technology, and running at 5.2GHz, claimed to be the fastest processor in the industry. Each core has hardware …

  1. Paul Woodhouse

    can you play candy crush on it?

    1. Archaon

      How many instances of Doom per core?

    2. stephanh Silver badge

      Candy Crush for z/OS is $200,000/year for a single LPAR with MSU rating 100.

      Sub-Capacity pricing is also available.

  2. 45RPM Silver badge

    But if you wait 10 years you can have all this power, and more besides, in your mobile phone running iOS 21 or Android 'Banoffee Pie'

    1. Anonymous Coward
      Coat

      Darn, my contract renews at 9 years, so I'm going to miss out.

  3. Sil

    No sure why you repeat IBM infotisement as is.

    Clearly as sentence on encryption such as

    "This is 18x faster than compared x86 systems (that today only focus on limited slices of data) and at just five percent of the cost of compared x86-based solutions "

    is absolutely retarded, with zero indication of what is compared to what, nor of methodology: just a paid-for study that isn't even available through hyperlinks.

    The rest of the press release is of the same caliber.

    1. Daniel von Asmuth Bronze badge
      Gimp

      Performance?

      The Z14 is the best IBM mainframe ever just like '10' is supposed to be the best version of Windows, but hard and comparable numbers are hard to come by. Guess old COBOL code is more efficient than .NET.

      170 cores, 32 TB of main memory, those figures resemble a high-end PC.

      1. stephanh Silver badge

        Re: Performance?

        "170 cores, 32 TB of main memory, those figures resemble a high-end PC."

        Had some trouble finding such a high-end PC at MediaMarkt.

        But in any case, IBM mainframe has never been a cost-effective proposal if you are looking for computing muscle. For example, nobody[1] does HPC on a mainframe. The mainframe's USP have traditionally been IO throughput and robustness (doubled CPUs to do error checking, hot-swappable everything.)

        Why IBM marketing then comes up with a story which stresses its (rather unimpressive, for the money) computing prowess is a bit of a mystery but may explain why sales are falling...

        [1] Yeah, there is probably somebody.

        1. returnofthemus

          For example, nobody[1] does HPC on a mainframe.

          That's because it's specifically designed for high volume transactions, but don't worry POWER9 is on the way, if it's general purpose you're after they operate a public cloud with x86 in abundance.

          I think that just about covers it from a hardware perspective

      2. bombastic bob Silver badge
        Trollface

        Re: Performance?

        "Guess old COBOL code is more efficient than .NET."

        IMBO [In My Bombastic Opinion] C64 BASIC is more efficient than '.NET' ('.NOT" much of a comparison, heh)

        Believe it or not, properly written+compiled COBOL can be as good as FORTRAN or maybe even C code, but the thing is, COBOL programmers {like '.NOT' programmers) apparently want to use language features that are NOT efficient, in inherently inefficient ways.

        But yeah, the language syntax isn't all that friendly. I think it was the first lingo to have data structures, though FORTRAN 'EQUIVALENCE' blocks could do it, kinda (like the ASK/MANMAN system had, something I worked on back in the 90's on HP minicomputers).

        As for '.NOT', I have nothing but contempt and snark for it. I'd rather do COBOL.

    2. diodesign (Written by Reg staff) Silver badge

      Re: Sil

      Hi - yup, I've taken the sentence out. On its own, it's meaningless. If you check our other IBM articles you'll see we're not really down with running Big Blue 'press releases' - and our other chip stories go into a lot more detail when comparing features.

      Take this article as-is: a product announcement, not a product endorsement.

      C.

  4. jMcPhee

    Oh goody, another press release from IBM. Someone important must have options getting ready to vest.

  5. zorko

    typo? or my parochialism?

    "The pervasive encryption should encourage security-conscious CIOS to keep mainframe apps on the mainframe and in-house, helping to stem the dykes walling off the X86 server and public cloud seas threatening to breach its proprietary mainframe profit centre.". Dykes?

    1. Drewc (Written by Reg staff) Gold badge

      Re: typo? or my parochialism?

      not a typo.

      1. bombastic bob Silver badge
        Devil

        Re: typo? or my parochialism?

        I "got it" - like the boy that saved Amsterdam by putting his finger in one

        /me left intentionally ambiguous so everyone can have a good laugh

    2. Mike 16 Silver badge

      Re: typo? or my parochialism?

      A co-worker used to have a news clipping tacked to his wall. It was a headline from the Vietnam War era (not that far past, at the time): "Newsmen visit 'bombed' dykes". I did have to wonder if those newsmen were visiting bars in certain parts of San Francisco, rather than farms in North Vietnam (where they would likely be more welcome).

  6. dlc.usa
    Boffin

    Market It Better Than Usual

    Prove to us that there are no backdoors, and that decrypted data cannot be exfiltrated by SEs and HMCs. Please!

  7. Sureo

    Sounds like the perfect computer for the NSA.

    1. FrankAlphaXII

      Probably who they're targeting it towards, as well as banks, credit card companies, and others who need to do a lot of near real time encrypted transactions.

  8. Alistair Silver badge
    Coat

    Mainframe vs x86 vs *risc

    I've covered a lot of ground in the last 30 years of IT, and I honestly don't see why so many chirp about mainframes being obsolete. Or for that matter *risc. X86_64 has done spectacularly well in improving what can be accomplished on the platform over that time, taking the x86 cisc from 'toy' to 'desktop' to 'server' to cluster. There are astonishing things one can accomplish with an x86_64 cluster. I've been building them for some time, and there are great things one can do with a *risc cluster, I"ve built a few over the years. I'm not much on the Mainframe these days, but I did help move some things off x86 to ZOS supported linux at one time and that move, although it seemed quite costly solved a set of issues (at the time) which would have *actually cost far more* to scale out.

    Mainframes, in my experience, handle *volume* far better than scale out on x86 **in specific cases**. Risc sill does certain things better than x86_64 **in specific cases**. I'll admit my standard position is "right tool for the job at hand" rather than "latest toy we can lay our hands on".

    As for IBM sales literature, it is *far* from unique. Sales literature is sales literature. Everyone selling hardware or software uses tailored results. It is after all sales literature.

    All the above said, sadly, I'm helping to set up three things that may well be the final migration off our Z.

  9. flibbertygibbert

    But...

    will it blend?

  10. Anonymous Coward
    Anonymous Coward

    x86 -> mainframe

    The problem with a mainframe is that it's damn expensive and the ecosystem of peripherals, people, operations and connectivity are all burdened with the same proprietary uplift cost for running the machine.

    I've yet to come across a workload worthy of moving from x86 platform to a mainframe.

    Scale up or scale out? Doesn't matter when it's a question of software - virtualization through distribution to/from nodes.

    TCO always wins out and I've yet to see one that doesn't pay off in <3 years.

    Now, as to migrating off a mainframe, that's tricky. I've been involved in dozens of projects and some are clear (well-defined ERP migrations like JD Edwards etc.) and some lend themselves to an easy x86 replacement (modern EDI software, transaction archiving).

    Custom-written code from the iron age is the trickiest of all and sometimes the only reason for moving this way is a company merger or CxO level ideology to get off the boxen.

    The blurb associated with this release is just to give the gray haired old farts who cling to their legacy stuff a reference point, when they deliver the capital request for their next dinosaur box.

    Nice try to attempt the security angle, but with a software-defined data centre (via NSX or similar), it's already covered.

  11. QuiteEvilGraham

    Plenty of kiddie comments here today. Meanwhile, the adult world (ie. the stuff that keeps money in circulation plus some others) runs on mainframes. And probably will for quite some time to come.

    Fun Fact: back in the 80's IBM were considering a new mainframe line (Summit, IIRC), until someone pointed out that their existing customers then probably had some cumulative 3 Trillion dollars worth of investment in their current, run the buisiness on, software and it was really asking a lot of them to rewrite all that just so they might invest in something new and different. Thus the new z14 will happily run stuff written back in the 60's if that's what you want. And you would be surprised just how much of that stuff is still running and doing the job it was originally written for. You idiots probably think that you would sensibly replace a bulk ore carrier with a fleet of wee boats, because they are cheaper or something.

    Meanwhile tonight, the money in your bank account will continue to be reconciled, mainly, by a big box with zSystems embossed on the side. Flame away if you want; you are not going to persuade a good proportion of the fortune 500 to move their core systems to x86 and #NET or similar real soon now (let alone abandon their 5 9's reliablilty).

    1. stephanh Silver badge

      The Register had an interview with an actual banker last year.

      https://m.theregister.co.uk/2016/07/01/ing_mainframe_strategy/

      Quote:

      “You won’t find a bank without a mainframe, unless recently established.”

      But, he continued, “We are extremely aggressively moving away from them.”

    2. lordminty

      I've not worked on mainframes for over 20 years now (I was a VSE, VM and OS/390(MVS) sysprog manager, and I've worked with x86, Sparc and HP-UX based solutions for over 10 years now.

      I suspect many of the negative comments come from people who have never experience mainframes, or seen a server estate that consists of tens of thousand of the things. 10s of thousands of physicals are not easy to manage and when you need a small army to run them, it makes mainframes look cheap. Mainframes just get a bad rep due to the upfront capital costs not the long term operational costs.

      Quite simply I've seen nothing in the x86 etc. space that comes anywhere close to IBM's mainframe offerings, and I cease to be amazed by the so-called industry 'developments' in the x86 etc. space that mainframes were doing 20, 30 or 40 years ago. Yes some of the tech out there is interesting (think Sparc-based Exadatas, but they ain't cheap!)

      Virtualisation is another one. VMWare ESX and NSX etc. are all interesting but they are still miles behind what VM on mainframes could do over 20 years ago.

      Just look at the number of Linux VMs you can host on z/VM compared to a VCE Vblock, and even hidden in that press release, 2million Docker containers. You get big figures on big iron.

      1. Anonymous Coward
        Anonymous Coward

        "VMWare ESX and NSX etc."

        This isn't the cutting edge. AWS, Google Cloud, etc are the cutting edge. Does any enterprise, tradition enterprise, have anything that can run like, for instance, Google search? Index a few trillion records and provide a few hundred million results nearly instantaneously and do that for a billion users at the same time... never down or slow or inconsistent regardless of location, connection, device... if a few hundred million users decide to unexpectedly jump on the application, no big deal.

        1. Julian S

          Google doesn't work like that!

          Sorry but Google doesn't work that way. Indexes are pre-built using batch processes and replicated across multiple zones and as much as possible cached in memory with pre-built criteria chains for rapid processing.

    3. returnofthemus

      Plenty of kiddie comments here today.

      LOL!

      Well, you know how it is with kids they like to keep active, some are even hyperactive and what better way is there than letting them run round a Windows server farm putting out fires, plus it keeps them away from the mainframe where they could do some real damage!

    4. Anonymous Coward
      Anonymous Coward

      "you are not going to persuade a good proportion of the fortune 500 to move their core systems to x86 and #NET or similar real soon"

      Definitely agree with the .Net comment.

      Legacy enterprises are going to have to get a move on if they don't want to be Amazoned.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019