Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Tuesday December 20 2016, @07:04AM   Printer-friendly
from the Which-is-larger-a-processing-core-or-a-memory-core? dept.

http://www.anandtech.com/show/10918/qualcomm-demos-48core-centriq-2400-server-soc-in-action-begins-sampling

Qualcomm this month demonstrated its 48-core Centriq 2400 SoC in action and announced that it had started to sample its first server processor with select customers. The live showcase is an important milestone for the SoC because it proves that the part is functional and is on track for commercialization in the second half of next year.

Qualcomm announced plans to enter the server market more than two years ago, in November 2014, but the first rumors about the company's intentions to develop server CPUs emerged long before that. In fact, being one of the largest designers of ARM-based SoCs for mobile devices, Qualcomm was well prepared to move beyond smartphones and tablets. However, while it is not easy to develop a custom ARMv8 processor core and build a server-grade SoC, building an ecosystem around such chip is even more complicated in a world where ARM-based servers are typically used in isolated cases. From the very start, Qualcomm has been rather serious not only about the processors themselves but also about the ecosystem and support by third parties (Facebook was one of the first companies to support Qualcomm's server efforts). In 2015, Qualcomm teamed up with Xilinx and Mellanox to ensure that its server SoCs are compatible with FPGA-based accelerators and data-center connectivity solutions (the fruits of this partnership will likely emerge in 2018 at best). Then it released a development platform featuring its custom 24-core ARMv8 SoC that it made available to customers and various partners among ISVs, IHVs and so on. Earlier this year the company co-founded the CCIX consortium to standardize various special-purpose accelerators for data-centers and make certain that its processors can support them. Taking into account all the evangelization and preparation work that Qualcomm has disclosed so far, it is evident that the company is very serious about its server business.

From the hardware standpoint, Qualcomm's initial server platform will rely on the company's Centriq 2400-series family of microprocessors that will be made using a 10 nm FinFET fabrication process in the second half of next year. Qualcomm does not name the exact manufacturing technology, but the timeframe points to either performance-optimized Samsung's 10LPP or TSMC's CLN10FF (keep in mind that TSMC has a lot of experience fabbing large chips and a 48-core SoC is not going to be small). The key element of the Centriq 2400 will be Qualcomm's custom ARMv8-compliant 64-bit core code-named Falkor. Qualcomm has yet has to disclose more information about Falkor, but the important thing here is that this core was purpose-built for data-center applications, which means that it will likely be faster than the company's cores used inside mobile SoCs when running appropriate workloads.

Here's an older article about Qualcomm's ARM server efforts.


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 0) by Anonymous Coward on Tuesday December 20 2016, @07:21AM

    by Anonymous Coward on Tuesday December 20 2016, @07:21AM (#443597)

    How much more expensive than x86, how much slower than x86. Sure is great for Facebillionaires to collect the world's social data and destroy the prospects of antisocial losers. But I like to think small. How will an ARM cloud benefit me, the little guy with a tiny handful of servers.

  • (Score: 3, Interesting) by FatPhil on Tuesday December 20 2016, @08:11AM

    by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Tuesday December 20 2016, @08:11AM (#443609) Homepage
    The two largest costs for big server farms is (a) powering all the kit, and (b) cooling all the kit that you've pumped all that power into. The only sensible question is "how low is the power consumption/dissipation?". In the days of ARM32 vs Core, ARM had that battle well and truly in the bag, having about 1/10th of the power, but 1/100th of the power dissipation. However, it's really not so clear any more, as Intel finally realised that datacentre users were noticeing their electricity bills.
    --
    Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
    • (Score: 0) by Anonymous Coward on Tuesday December 20 2016, @08:31AM

      by Anonymous Coward on Tuesday December 20 2016, @08:31AM (#443621)

      Jesus Herbert Christ, I knew someone would declare, power savings is all that matters! Nobody cares about performance anymore, power savings is all that matters! Sure your ARM processors don't have lots of registers like x86, and they don't have as wide vector registers like x86, but dude, they're sooooo green, bro! Green servers are the meme of the decade. If you need fast, just buy more servers and run your workload in parallel! Have a serialized workflow that can't run in parallel? Well you're just screwed. It's going to be slow. But you'll save so much power!!!

      Let me clue you in, jerk. The end-user of the cloud does not care about the power consumption. The question, which you ignored, was how does ARM benefit the little guy's bottom line.

      • (Score: 3, Insightful) by theluggage on Tuesday December 20 2016, @12:00PM

        by theluggage (1797) on Tuesday December 20 2016, @12:00PM (#443704)

        Sure your ARM processors don't have lots of registers like x86, and they don't have as wide vector registers like x86

        ...which are likely doing nothing except wasting die space and consuming a bit of power on a server running multiple instances of a LAMP stack or node.JS + mongoDB. The whole basis for the original ARM was the 80/20 rule-of-thumb: a complex processor spends 80% of its time using 20% of its instruction set.

        Have a serialized workflow that can't run in parallel?

        Then you need a workstation, or a cloud-connected supercomputer service, not a "regular" server. Servers tend to cater for large numbers of simultaneous users - i.e. a heavily parallel workflow. They also tend to be bought for a known purpose - so if you're building a cloud computing service for serial workflows, get something with fewer, higher-powered processors*. If you're hosting 1000 WordPress sites (it happens), get something energy efficient.

        *Which might be ARM chips with existing [arm.com] or forthcoming [theregister.co.uk] vector units. An outfit the size of Qualcomm can build pick-n-mix ARM processors to suit their application rather than wait for Intel to produce just the right model.

      • (Score: 0) by Anonymous Coward on Tuesday December 20 2016, @01:52PM

        by Anonymous Coward on Tuesday December 20 2016, @01:52PM (#443765)

        Um... the bottom line is affected by power consumption.

      • (Score: 4, Informative) by TheRaven on Tuesday December 20 2016, @07:11PM

        by TheRaven (270) on Tuesday December 20 2016, @07:11PM (#443982) Journal

        Sure your ARM processors don't have lots of registers like x86

        AArch64 has 32 GPRs, x86-64 has 16 GPRs.

        they don't have as wide vector registers like x86

        SSE (x86) has 128-bit vectors. AVX (x86) has 256-bit registers. AVX-512 (x86, mostly Xeon Phi) has 512-bit registers. NEON (ARM) has 128-bit vectors. SVE (ARM) has up to 2048-bit registers.

        --
        sudo mod me up
  • (Score: 1, Insightful) by Anonymous Coward on Tuesday December 20 2016, @08:15AM

    by Anonymous Coward on Tuesday December 20 2016, @08:15AM (#443610)

    What lower binnings of CPUs have they designed the 48 core SoC to support? Per core binning? 12/24/36/48? Every 4 cores? Some other value based on internal interconnect and disable features?

    Done right this could be a really awesome chip. Personally however, I would have preferred them design it for an AM1 style motherboard platform, and while having a reference design available for sale themselves, provide the reference platform to third party vendors to help drive down costs and innovate on board features for the full range of server platforms.

    Honestly the only reason Intel still has its dominant position is the ineptness of everyone else. There is lots of potential for the x86/datacenter market to be carved up among smaller players, but no smaller players are individually making the right moves to both drive and support consumer demand for their hardware (Example: The AMD AM1 equivalent Opteron boards. Could have ripped Intel a new one on the sub 400 dollar low power server market, but were impossible to find outside big-box server products). The non-x86 ecosystem hasn't banded together to ensure hardware with driver support. For example a BIOS/firmware integrated VM (like java, not like VT/SVM) to allow platform independent peripheral cards, all of which currently support EFI using x86 specific option roms. If for instance companies had pushed OpenFirmware better, and provided a BIOS integratable forth implementation to run it on, Intel wouldn't have been able to ram EFI and its binary based bootloader architecture down our throats and saddled us with EFI on x86_64 AND ARM for the foreseeable future.

    • (Score: 4, Interesting) by DannyB on Tuesday December 20 2016, @03:01PM

      by DannyB (5839) Subscriber Badge on Tuesday December 20 2016, @03:01PM (#443810) Journal

      Intel (and AMD's compatible instruction set chips) have saddled us with "management engines". Those are a computer within the microprocessor. The processor won't run unless that embedded controller is loaded with an undocumented binary blob by the OS or boot loader. That blob can control anything the microprocessor can control. All of the IO, all of the motherboard.

      This is the ultimate compromise. Built right into your hardware.

      --
      People today are educated enough to repeat what they are taught but not to question what they are taught.