Slashdot Mirror


Intel Announces Xeon Scalable Processor Family (hothardware.com)

MojoKid writes: Intel unveiled information regarding a new Xeon processor family today, some of which will be based on the company's Skylake-SP architecture. Intel will have four levels of Xeon processors that scale with respect to feature support and core counts. Intel is calling it the Xeon Scalable Family with Bronze, Silver, Gold and Platinum processors. Today, Xeon model names follow a fairly easy-to-understand format. Take for example the Xeon E5-4640 v4. "E5" in this case means that it is in the middle of Intel's current stack in terms of features and capabilities, where the "4" signifies use in a 4-socket system. Finally, the "v4" represents the architecture. With this change, a model like the one above would become Intel Xeon Gold 4640, as an example. Regardless, the chips will include support for AVX-512 instructions, QuickAssist and Volume Management Device (VMD) technologies that will take advantage of NVMe solid-state drives. The platform will also support complementary processing engines and IO technologies like Intel FPGAs, Xeon Phi accelerators and Silicon Photonics connectivity. Intel notes the processors will be arriving to market this summer.

30 of 52 comments (clear)

  1. Why 4-digits by Luthair · · Score: 2

    When the last one is always zero.

    1. Re:Why 4-digits by OrangeTide · · Score: 2

      The last digit is the number of back doors MOD 10.

      --
      “Common sense is not so common.” — Voltaire
    2. Re:Why 4-digits by ShanghaiBill · · Score: 3, Informative

      When the last one is always zero.

      Because it was designed in America, where three houses on a cul-de-sac will each have a four digit house number. Americans love superfluous digits.

      But I shouldn't ridicule Americans too much, since some other countries have numbering schemes that are even sillier. For instance, in Japan, houses on a block are not numbered in sequence, but in the order in which they were built.

    3. Re:Why 4-digits by crow · · Score: 3, Interesting

      It depends on where you are. In my town, house numbers start in the single digits. This causes a problem if the street is extended in the wrong direction, as happens from time to time. In the town I grew up on, the streets are numbered as if they extended to the center of town, so houses between 1st and 2nd streets are in the 100s, but streets ten miles away have five-digit addresses (it's a large city). That works well, as you can often tell about where something is by the street number even if you've never heard of the street.

    4. Re:Why 4-digits by Carewolf · · Score: 1

      When the last one is always zero.

      Or 7 when they go CRAZY!

    5. Re:Why 4-digits by dargaud · · Score: 1

      This causes a problem if the street is extended in the wrong direction, as happens from time to time.

      I once raged looking for a number on a street that had been extended twice and the blockheads in charged had chosen the following numbering scheme (from the new start of the street): 200..299 1..199 300..1000

      --
      Non-Linux Penguins ?
    6. Re:Why 4-digits by crow · · Score: 1

      When Concord Street was extended in my town, the new portion was named "Concord Street Extension." That makes addresses ugly, but it solves the problem. Of course, Concord Street is one of those crazy streets where you come to a four-way intersection, and you turn left to stay on Concord Street, or you go straight for Concord Street Extension, or you turn right for another street where you then have to turn left at the next intersection to stay on that street. At the other end at another four-way intersection, you go straight or left for Fountain Street and right for Front Street.

      I think it's time to rename and renumber some streets to clean up this mess.

    7. Re:Why 4-digits by radarskiy · · Score: 1

      "For instance, in Japan, houses on a block are not numbered in sequence, but in the order in which they were built."

      And now you know why new housing developments are given large numbers with gaps: so that when you build something between them later the new construction can still get an in-sequence number without renumbering the whole street.

  2. First question by geekymachoman · · Score: 2

    Does it have a brand new NSA backdoor in it ?

    1. Re:First question by thegarbz · · Score: 1

      Does it have a brand new NSA backdoor in it ?

      Why would it? There's nothing wrong with the old one.

  3. Scalable hardware??? by Heebie · · Score: 1

    This sure sounds like a crock to me. Similar to when they created the "486SX" which was a 486DX in which the math coprocessor did not pass quality control, so they disabled it & sold the chip as an SX instead of throwing it in the bin.

    They're probably doing QC testing on the chips & "downgrading" the ones that don't work... either that or they're charging you a LICENSE FEE for features and cores in the CPU and making their hardware "by subscription" which would be a truly horrific turn of events!

    1. Re: Scalable hardware??? by threephaseboy · · Score: 3, Informative

      Yeah, Intel would never do something like charge people to unlock features present in the CPU they bought

      --
      .
    2. Re: Scalable hardware??? by by+(1706743) · · Score: 2

      This is similarly done with the Raspberry Pi family.

      Of course, I find this significantly more palatable, as a) the whole computer only cost $35, and b) I think it's due to some licensing agreement with a separate entity (Broadcom).

    3. Re:Scalable hardware??? by unixisc · · Score: 1

      Actually, I didn't get the point of this announcement. To me, it looked more like a die shrink, or an extension of an existing CPU platform. Tossing in more cores, changing the packages or upping the frequencies doesn't exactly change things all that much. What would change it would be new instructions, but I doubt that Intel would wanna include new instructions that would force newer software to be recompiled, particularly when its primary competition has been its older CPUs that have really had a robust lifetime. Like this Pentium I bought 3 years ago still serves me well. If Intel were to include newer instructions, there would be little benefit, since software that runs on the older CPUs would run on the newer ones, but the newer ones would have more idle circuitry for the new instructions that would impact #die/wafer, thereby increasing costs on what's already a very expensive process, thanks to the more expensive newer fabs.

      I think what happens in reality is that their QA/QC does the testing, and the chips that have one or more cores fail are binned lower, so that they can be sold at a lower price w/ the malfunctioning cores disabled, maybe by internal mode test vectors. Or they may find that some chips fail the spec'ed frequency, but are better at lower frequencies. Whatever the case, it's a good model to built a high end test platform CPU, sell as much of it as they get as the higher end, and then sell any fallout as lower end CPUs at discounted prices, as opposed to discarding them altogether.

  4. shelf space by LesserWeevil · · Score: 1

    This reminds me of something happening on supermarket shelves everywhere.. Brands expand their offerings from 1 or 2 to half a dozen or more just to take up more shelf space and drive smaller brands out. The world needs a viable server competitor to Intel, be it AMD or someone else.

    1. Re:shelf space by epyT-R · · Score: 2

      Why?

    2. Re:shelf space by unixisc · · Score: 1

      If you buy Mac Pros, it has a server CPU - used in this case as a workstation CPU. People who need workstations have to go w/ Xeon. Personally, I think Intel should include ECC in the i3/5/7 CPUs, and for those who don't need that, offer them the Atoms or the Pentiums

    3. Re:shelf space by Blaskowicz · · Score: 1

      You can use i3, Pentium and even now Celeron with ECC (the real desktop chips, not Celeron-branded Atoms). You need a "workstation" or "server" grade motherboard though.

      Surprise : there are new motherboards that looks like just your standard gamer/office motherboards with a new coat of paint. So you will pay more.
      i5/i7 are branded as "Xeon E3" like before, but this time they don't run in regular motherboards. So there's still an ECC tax but if that was all you wanted it's a bit cheaper now.

  5. Not an "unveiling" -- an accidental leak by djembe2k · · Score: 4, Informative

    Intel didn't "unveil" anything -- they accidentally posted a "product change notification" that listed model numbers for upcoming Skylake Xeon processors, then quickly took it down. What we've learned is that the model numbering system will change, maybe in a way that keeps a similar organization but renames things, and maybe in a way that means substantive changes. We can't tell. In other words, this is semantics. If there's more to be gleaned from the leak, I haven't heard it here.

    1. Re:Not an "unveiling" -- an accidental leak by Anonymous Coward · · Score: 2, Informative

      This summer they are planning on releasing Coffee Lake, which will be 8xxx in desktops (ex: i7-8700K, or whatever). 8 is getting close to 0, so they are probably going to have to rebrand that. The Xeons had a bit more life in them- a coffee lake Xeon would only be "v6", and it isn't out yet. Their enthusiast chips, should they keep that artificial market, would be out at a similar time to the Xeons (some time in the future- the Skylake enthusiasts aren't even out yet), would be something like "i7-9950X" or whatever.

      Basically, they'll probably be rebranding all their shit soon. I don't think we even know how to read the new Xeons- is a Gold 2667 comparable to an E5-2667 v4? Basically, how do you tell the difference between a Gold 2667 Skylake, and a Gold 2667 Coffee Lake, without the version number?

  6. So what you're saying is... by Kokuyo · · Score: 2

    The naming scheme wasn't convoluted enough so far?

  7. Obligatory:Intel CPU Backdoor Report (May 1 2017) by Anonymous Coward · · Score: 2, Interesting

    The goal of this report is to make the existence of Intel CPU backdoors a common knowledge and provide information on backdoor removal.

    What we know about Intel CPU backdoors so far:

    TL;DR version

    Your Intel CPU and Chipset is running a backdoor as we speak.

    The backdoor hardware is inside the CPU/Bridge and the backdoor firmware (Intel Management Engine) is in the chipset flash memory.

    30C3 Intel ME live hack:
    @21m43s, keystrokes leaked from Intel ME above the OS, wireshark failed to detect packets.
    [Video Link] 30C3: Persistent, Stealthy, Remote-controlled Dedicated Hardware Malware
    [Quotes] Vortrag:
    "DAGGER exploits Intel's Manageability Engine (ME), that executes firmware code such as Intel's Active Management Technology (iAMT), as well as its OOB network channel."

    "the ME provides a perfect environment for undetectable sensitive data leakage on behalf of the attacker. Our presentation consists of three parts. The first part addresses how to find valuable data in the main memory of the host. The second part exploits the ME's OOB network channel to exfiltrate captured data to an external platform and to inject new attack code to target other interesting data structures available in the host runtime memory. The last part deals with the implementation of a covert network channel based on JitterBug."

    "We have recently improved DAGGER's capabilites to include support for 64-bit operating systems and a stealthy update mechanism to download new attack code."

    "To be more precise, we show how to conduct a DMA attack using Intel's Manageability Engine (ME)."

    "We can permanently monitor the keyboard buffer on both operating system targets."

    Backdoor removal:
    The backdoor firmware can be removed by following this guide using the me_cleaner script.
    Removal requires a Raspberry Pi (with GPIO pins) and a SOIC clip.

    Decoding Intel backdoors:
    The situation is out of control and the Libreboot/Coreboot community is looking for BIOS/Firmware experts to help with the Intel ME decoding effort.

    If you are skilled in these areas, download Intel ME firmwares from this collection and have a go at them, beware Intel is using a lot of counter measures to prevent their backdoors from being decoded (explained below).

    Useful links:
    The Intel ME subsystem can take over your machine, can't be audited
    REcon 2014 - Intel Management Engine Secrets
    Untrusting the CPU (33c3)
    Towards (reasonably) trustworthy x86 laptops
    30C3 To Protect And Infect - The militarization of the Internet
    30c3: To Protect And Infect Part 2 - Mass Surveillance Tools & Software

    1. Introduction, what is Intel ME

    Short version, from Intel staff:

    Re: What Intel CPUs lack Intel ME secondary processor?
    Amy_Intel Feb 8, 2016 9:27 AM

    The Management Engine (ME) is an isolated and protected coprocessor, embedded as a non-optional part in all current Intel chipsets, I even checked wit

  8. Chip names by ledow · · Score: 2

    I gave up trying to understand chip names and model numbers a long time ago.

    GPUs and CPUs are the worst - I have absolutely no clue whether a particular number / model is better / worse than any other.

    I let my suppliers deal with it - I tell them what I want, they send me a spec, I google to ensure the chip numbers do what I want in terms of core-speed, number of cores, etc.

    It's not hard to come up with a sensible numbering / naming system, but if Intel didn't have a page for every chip that just says number of cores, speed, etc. then I'd have absolutely no clue. Changing it to colours isn't helping matters either - that just reeks of marketing where everyone thinks they need to have "Gold" rather than "Silver" when there's probably little gain for most people.

    For the same reason, I stopped buying separate CPUs and motherboards many years ago and nowadays I don't even tend to buy PSUs etc. separately. Because the combination needs to be correct and I'm not going to waste my time and effort only to get it wrong.

    Everything from CPU sockets, to PSU power draw (don't forget to check the 12V rail!), to PCIe speed,to card profile height - after a while it just gets so boring, and I'm paying people to supply me the gear, so I let them do the legwork.

    But the situation for consumers is actually completely the flip of that. People ask me about buying laptops. I say "check it has the ports you want". No point paying a fortune for a laptop that doesn't have enough USB, and if you buy enough USB or whatever you want for it, it'll be fast enough to do what you want nowadays. Even the cheapest gaming laptops are stupendous in specification and able to play anything I have on my Steam account (if my 6-year-old laptop can do it, too!).

    Don't care about the numbers, or the exact model of chip, or anything like that. If you want a cheapy thing just for office, buy a cheapy thing just for office. If you want a gamer's thing for high-end gaming, buy something that's pitched at gamers and comes with the Razer gaming mice or whatever. Pretty much, that's a better indicator than faffing with all the statistics - which although I could understand, nobody else does and I just don't have the time to be bothered to do the research.

    Don't even get me started on "does AMD graphics card X perform better than nVidia graphics card Y?"

    1. Re:Chip names by thegarbz · · Score: 1

      It's a CPU. Even if you pick horribly wrong but are still in the ballpark for price what are you going to lose? A few percent of performance?

  9. Re:Intel is making a mistake... by Rockoon · · Score: 1

    Intel does not want to have so many models. Their 14nm yields in particular are so poor without the binning that they have to have so many models.

    --
    "His name was James Damore."
  10. Re:Intel is making a mistake... by PingSpike · · Score: 1

    Not a big deal for Intel. When they see the confused look on your face they'll just hand you an i3 because they've got to get rid of those somehow.

  11. a bad omen by epine · · Score: 1

    I hate this shit. This is going to become such a nightmare in three to five years when these CPUs begin to hit the resale market. Typing "Xeon E5" into eBay proved too damn useful, so now this.

    I'm trying to imagine myself typing "gold + anything" into eBay, and failing miserably.

    Or any other search bar, without requiring "Coinage Act of 1792" peril impervious sunshades.

    Intel is lately taking a sharp turn back to their Edward "RoDRAM" Hyde persona. You see this with Optane and Optame Memory. And now this, too.

    Any margin we want = Intel behaves like "gentleman" Jekyll (those steady royalty cheques from the British East India Company have a salutary, soothing effect).

    Competitive reality = Intel behaves like Hyde on bath salts in a Dutch bordello.

    The thing about precious metals, linguistically, is that it anchors into a one-dimensional pyramid of extreme greed.

    Exclusivity is well-ordered. Exclusivity doesn't trade RPM for towing capacity (well, it does, but that was resolved by the six-car ocean-frontage garage). But in terms of the human metaphorical reflex, the greed anchor flattens the ordering relationship. It's no longer horses for courses (unless, of course, one has so many horses the course no longer matters, and the debate is really blondes vs. redheads over cigars and cognac, because one also has so many blondes and redheads).

    Under this new naming system, Intel will inevitably become less of an engineering company as the marketrons ride again.

    It's a thoroughly bad omen.

    1. Re:a bad omen by epine · · Score: 1

      You know, I would give quite a bit to see the Xeon Pentium D become known in street slang as the "Xeon Redhead".

      Between the Xeon Bronze and the Xeon Silver, I'll take the Xeon Redhead every time.

      Xeon Bronze Age Xeon Silver Fox Xeon Gold Digger Xeon Platinum Bridge Xeon Palladium Anode — Lucille's most likely official name

      ___

      Ricky: You've got some 'splaining to do.

      Lucy: Er, uhhhh, cold fusion?

      Ricky: Really? Cold fusion?

      Lucy: Yeah! Cold fusion!

  12. Software Licencing Implications by Monkey · · Score: 1

    Awesome, but how will this affect my ridiculous Oracle licensing?

  13. Re:Intel is making a mistake... by toddestan · · Score: 1

    It's not even so tough. Pick a handful of levels (number of cores, amount of cache, whatever). If not all the features work for one level, try again at the lower level, and repeat. You could then bin again on clock frequency, but the simple solution would be market them based upon the clock frequency and level only. So as long as you're in the same level, you would have the exact same feature set, so higher clock = faster. Between levels, it may not be so clear, but you could check benchmarks and what is best would likely depend on what you're using them for too.