Slashdot Mirror

← Back to Stories (view on slashdot.org)

Qualcomm Eyes Intel With Centriq 2400 Arm Server Chip (eweek.com)

Posted by msmash on from the tussle-continues dept.

Qualcomm is now challenging rival Intel in the rapidly changing data center market. From a report: The company is now selling its long-awaited Centriq 2400 Arm-based server processor that is aimed at the fast-growing cloud market and that Qualcomm officials say beats Intel in such crucial areas as power efficiency and cost. Officials from Arm and its manufacturing partners have for several years talked about pushing the Arm architecture into the data center as an alternative to Intel, and some manufacturers like Cavium and Applied Micro in recent years have rolled out systems-on-a-chip (SoCs) based on the 64-bit Armv8-A design. However, Qualcomm represents the most significant Arm chip maker in terms of scale and resources to challenge Intel, which holds more than 90 percent of the global server chip market. Qualcomm's Centriq chips offer up to 48 single-threaded cores running up to 2.6GHz and are manufactured on Samsung's 10-nanometer FinFET process. The processors sport a bidirectional segmented ring bus with as much as 250G bps of aggregate bandwidth to avoid performance bottlenecks, 512KB of shared L2 cache for every two cores and 60MB of unified L3 cache. There also are six channels of DDR4 memory and support for up to 768GB of total DRAM with 32 PCIe Gen 3 lanes and six PCIe controllers. They also support Arm's TrustZone security technology and hypervisors for virtualization.

23 comments

  1. First? by Anonymous Coward · · Score: 0

    Qualcomm needs its own Slashdot icon. Same with Broadcom.

    1. Re:First? by msmash+(Top+Editor) · · Score: 1

      Sorry, but the Slashdot logo is already copywritten.

    2. Re: First? by Anonymous Coward · · Score: 0

      For copyright you don't write "copywritten", for precisely the same reason that for human rights you wouldn't write "human written." It makes no sense, and indicates that the author has a poor grasp of basic English.

  2. Long term support ? by romiz · · Score: 5, Insightful

    Is Qualcomm really commited to support the chip on the long term? A good measure of this would be the level of support for the chip in the Linux kernel mainline. All serious server processors have support in the mainline - even Itanium did. With Qualcomm's mobile chips, millions of lines of custom code for a two year old kernel is normal, but for server customers it will not be sufficient.

    1. Re:Long term support ? by Anonymous Coward · · Score: 2, Informative

      I have some experience with the Cavium server platform.

      It's not nearly as 'weird' as embedded ARM, it's pretty normal UEFI style boot, for the most part.

      The challenge has been performance per watt and performance per dollar is crap. Single threaded performance is complete and utter crap, and even though it's a 48 core socket, even 48 of them suck next to a comparable Intel Xeon.

      I have not seen the Qualcomm offering yet, and maybe it closes the gap. However even their marketing statements are suspiciously avoiding the whole performance and avoiding putting the power consumption in the context of delivered performance.

    2. Re:Long term support ? by Hal_Porter · · Score: 3, Interesting

      The challenge has been performance per watt and performance per dollar is crap. Single threaded performance is complete and utter crap, and even though it's a 48 core socket, even 48 of them suck next to a comparable Intel Xeon.

      If you want a whole bunch of cores and you don't care about terrible per core performance Intel can sell you a Xeon Phi. It's got a whole load of Atom cores

      https://en.wikipedia.org/wiki/...

      Knights Landing will be built using up to 72 Airmont (Atom) cores with four threads per core,[66][67] using LGA 3647 socket[68] supporting for up to 384 GB of "far" DDR4 2133 RAM and 8â"16 GB of stacked "near" 3D MCDRAM, a version of the Hybrid Memory Cube. Each core will have two 512-bit vector units and will support AVX-512 SIMD instructions, specifically the Intel AVX-512 Foundational Instructions (AVX-512F) with Intel AVX-512 Conflict Detection Instructions (AVX-512CD), Intel AVX-512 Exponential and Reciprocal Instructions (AVX-512ER), and Intel AVX-512 Prefetch Instructions (AVX-512PF).[69]

      Is it selling well? Probably not. I.e. there's some evidence that people don't want to move from 4-8 Core i7 style large cores to 64-72 Atom cores. I.e. people buying server CPUs care about single thread performance.

      Which makes me think ARM based servers are not going to kill x64 ones.

      --
      echo -e 'global _start\n _start:\n mov eax, 2\n int 80h\n jmp _start' > a.asm; nasm a.asm -f elf; ld a.o -o a;
    3. Re:Long term support ? by CajunArson · · Score: 1, Insightful

      "Is it selling well? Probably not. I.e. there's some evidence that people don't want to move from 4-8 Core i7 style large cores to 64-72 Atom cores. I.e. people buying server CPUs care about single thread performance."

      Xeon Phis are selling extremely well and have been widely publicly deployed in the Top 500 supercomputers since earlier this year.

      The fact that you are calling a Xeon Phi a failure because it's not replacing a desktop chip shows your ignorance, not any failure on the part of an extremely innovative product.

      --
      AntiFA: An abbreviation for Anti First Amendment.
    4. Re:Long term support ? by Hal_Porter · · Score: 2

      I hope they do sell well. I always thought Larabee was an interesting idea. The basic idea was that you can get a lot more P54C cores than Cores ones in the same area. P54C was in order and Core was out of order. Core had a much better IPC. Larrabee had the interesting idea of using hyperthreading to keep those simple P54C cores busy by switching threads on a cache miss. Noice.

      https://en.wikipedia.org/wiki/...

      Problem is desktop and server applications are currently too dependent on high IPC cores and single thread performance for this approach to work well there.

      Otherwise Intel would discontinue the brainiac cores in Kabylake chips and move everyone over from a small number of Kabylake cores to a large number of Airmont ones.

      HPC is obviously a special case because people are happy to code to architectures which are a bit out of the ordinary.

      IIRC Google have written a paper saying that, for their applications, fewer cores but more IPC is better than more cores with lower IPC.

      --
      echo -e 'global _start\n _start:\n mov eax, 2\n int 80h\n jmp _start' > a.asm; nasm a.asm -f elf; ld a.o -o a;
    5. Re:Long term support ? by fuzzyfuzzyfungus · · Score: 1

      Qualcomm is committed to 'the chip' as an abstract concept; in the sense that they will be happy to sell you another chip of some sort when you have to throw out your current chip because of lack of software support.

    6. Re:Long term support ? by Anonymous Coward · · Score: 0

      It's been frankly a bit of a mixed bag in the history of Phi.

      First when it was LArabee, it was going to replace the need for an nVidia type adapter with Intel x86 style architecture.

      Then that didn't pan out so they brought it to market as Knights Corner, an accelerator to replace GPGPU in HPC, but with an easier programming situation. Except that it didn't comparre favorably with contempory GPUs, was still a bit of a pain to program for right, and was a really *bizarre* OS on a PCIe card approach.

      They finally hit their stride insofar as they ever will with Phi in releasing Knights Landing as a standalone bootable system. The programming model did get easier, and this did get some nice numbers in the Top500. However, it still can't match P100 (not to mention V100) in performance/watt if you work hard for it, but maybe that's ok with the easier programming model. Except of course with Skylake-X out, they now have a platform with somewhat better aggregate compute horsepower than Knights Landing per watt, with a *whole* lot better single thread performance and much more potential memory capacity for workloads that don't do MPI. So the only thing that Phi has going for it now is that really fast MCDRAM, but that's not quite enough to justify it compared to the more traditional Xeon, particularly since Kinghts Landing is such a different platform and doesn't get to share a lot of costs with other Xeon family members (a frankenstein of the same socket form factor as SkyLake-X with the PCH of a Haswell and no UPI or QPI)..

    7. Re: Long term support ? by Anonymous Coward · · Score: 0

      For many servers, I would think threaded performance would be a huge consideration. Beyond that, in the cloud/virtual world being able to dedicate a few cores to an instance or task is important. More important than the performance of a single thread. Being able to stuff 72 cores into a single socket, as long as each core has enough RAM could allow even more consolidation of into fewer boxes.

      Iâ(TM)m pretty sure itâ(TM)s the gamers that are focused on single threaded performance above all else.

  3. Sooo... by JBMcB · · Score: 4, Interesting

    A few issues...

    1. I'm assuming no VMWare? How well does Xen run on ARM?
    2. Can GCC/CLANG optimize for a server profile? I'm assuming that, until now, all the work for the ARM target has been on code compactness and efficiency over performance.
    3. Looks like the chip has plenty of available bandwidth, does it have the transactional horsepower to fill it?

    We've seen chip makers trying to push re-purposed, low-powered chips into servers before, and the results have been underwhelming. If the raw CPU throughput is there, and a compiler/OS/server stack can be created that works well with it on existing server workloads, it may have a shot, but that's a lot of if's.

    --
    My Other Computer Is A Data General Nova III.
    1. Re:Sooo... by TheRaven64 · · Score: 1

      1. I'm assuming no VMWare? How well does Xen run on ARM?

      Xen has worked well on AArch64 for a long time. I think KVM also works. Bhyve support is almost there.

      Can GCC/CLANG optimize for a server profile? I'm assuming that, until now, all the work for the ARM target has been on code compactness and efficiency over performance.

      There's very little work done in the compiler on optimising for power. To a first approximation, the best optimisation for power is to finish quickly and let the CPU sleep. This is exactly the same for server and mobile workloads. There are differences in pipeline design, but the LLVM AArch64 back end already supports pipeline-specific tuning for AArch64 targets ranging from dual-issue in-order to wide superscalar out-of-order, so that shouldn't be a problem. I don't know what the ARM status for GCC is now that ARM, Google, and Apple are all investing almost exclusively in LLVM development, but I'd imagine it's not too different there.

      Looks like the chip has plenty of available bandwidth, does it have the transactional horsepower to fill it?

      I have no idea about this without more microarchitectural details being published, but there's nothing at the ISA level for AArch64 that would prevent it from scaling up to the kind of desired performance.

      --
      I am TheRaven on Soylent News
    2. Re:Sooo... by Richard+W.M.+Jones · · Score: 2

      These run KVM and Xen out of the box. I've been using this chip and its predecessor (Hawker) for a while and they are very nice indeed. It is far from being a "re-purposed" mobile chip (although of course Qualcomm do make buckets of those). They've been developing this as a server-specific chip for several years (4 years I believe from reading another press release).

      --
      libguestfs - tools for accessing and modifying virtual machine disk images
  4. Who controls the 'Trust Zone'? by Anonymous Coward · · Score: 0

    Seriously though, unless the stage0/1 bootloader allows the owner/administrator of the Centriq 2400 system to run/attest arbitrary code inside the TrustZone instance, it isn't really their hardware, and as such is just as untrustworthy and insecure as any Intel/AMD processor.

    If by some change they DO allow that, and don't have a 'factory override' certificate that could be used to override the administrator instantiated security provisions these could be excellent chips.

    Having said that, I don't have much trust of our technological oligopoly, which seems to be colluding on cross platform features which enforce the power divide between the 'owners' and the pwn3d.

  5. ARM Servers are Like the Yeti by Anonymous Coward · · Score: 0

    Everyone has heard of them. No one has seen them.

    ARM proponents will pile on shouting "it's really going to happen this time!". Intel is going to eat crow for sure!!

    AMD has about a billion times more credibility, more capability and more capacity to do this than ARM does. Look for AMD EPYC systems to gain some real traction in the server space.

    ARM could do it theoretically, but why this time? Remember SeaMicro, or whatever their name was? Seems more likely that SPARC will come back from the dead, or that Power could make a play for the server space, or that Itanium will rise from the grave like a processor zombie.

    1. Re:ARM Servers are Like the Yeti by Anonymous Coward · · Score: 0

      AMD HAS an ARM server processor too FYI : http://www.amd.com/en-us/products/server/opteron-a-series

    2. Re:ARM Servers are Like the Yeti by Anonymous Coward · · Score: 0

      AMD HAS an ARM server processor too FYI : http://www.amd.com/en-us/products/server/opteron-a-series

      HAD

    3. Re:ARM Servers are Like the Yeti by TheRaven64 · · Score: 1

      Remember how Intel killed off most of the workstation architectures? They produced slower chips, but had a lot more volume and so were able to scale up production, giving them a lot more to invest in R&D. Now look at the ARM ecosystem. There are about half a dozen completely independent implementations of the ISA, all competing, and most with a larger volume than Intel individually.

      A modern mobile phone CPU is more than adequate for most laptop users and we're starting to see things like Chromebooks shipping with them. That's how companies like SGI and Sun lost their market share: a competitor comes in at the low end and takes that part of the market, then is able to slowly move up. This isn't a new thing for ARM. They started off in Acorn desktops and Apple handhelds, but for most of their existence they were very low-end embedded systems. ARM chips that have been plausible laptop CPUs are a relatively new thing - even the StrongARM chips from the early 2000s were nowhere near competitive with even low-end Pentiums.

      --
      I am TheRaven on Soylent News
  6. Wait for benchmarks by Hal_Porter · · Score: 3, Interesting

    There has been a lot of talk about Qualcomm ARM chips taking over from Intel. The problem is when you look at the benchmarks they're rather underwhelming. Eg.

    http://weborus.com/snapdragon-...

    The Snapdragon 835 is a great device if you're running Android. If you're running something like Photoshop I predict performance is going to be disappointing. Microsoft's 'Windows on a Snapdragon' video shows Photoshop running. It doesn't mention performance

    https://www.youtube.com/watch?...

    It's the same with server stuff. And of course Intel have threatened people with a patent lawsuit on SIMD

    https://newsroom.intel.com/edi...

    Protecting x86 ISA Innovation

    Intel invests enormous resources to advance its dynamic x86 ISA, and therefore Intel must protect these investments with a strong patent portfolio and other intellectual property rights. The following graph shows that relentless instruction set innovation translates into a deep and dynamic patent portfolio with over 1,600 patents worldwide relating to instruction set implementations.

    https://imgur.com/a/x0K2V

    New x86 Instructions and Related Patents

    Intel carefully protects its x86 innovations, and we do not widely license others to use them. Over the past 30 years, Intel has vigilantly enforced its intellectual property rights against infringement by third-party microprocessors. One of the earliest examples, was Intelâ(TM)s enforcement of its seminal âoeCrawford â(TM)338 Patent.â In the early days of our microprocessor business, Intel needed to enforce its patent rights against various companies including United Microelectronics Corporation, Advanced Micro Devices, Cyrix Corporation, Chips and Technologies, Via Technologies, and, most recently, Transmeta Corporation. Enforcement actions have been unnecessary in recent years because other companies have respected Intelâ(TM)s intellectual property rights.

    However, there have been reports that some companies may try to emulate Intelâ(TM)s proprietary x86 ISA without Intelâ(TM)s authorization. Emulation is not a new technology, and Transmeta was notably the last company to claim to have produced a compatible x86 processor using emulation (âoecode morphingâ) techniques. Intel enforced patents relating to SIMD instruction set enhancements against Transmetaâ(TM)s x86 implementation even though it used emulation. In any event, Transmeta was not commercially successful, and it exited the microprocessor business 10 years ago.

    Only time will tell if new attempts to emulate Intelâ(TM)s x86 ISA will meet a different fate. Intel welcomes lawful competition, and we are confident that Intelâ(TM)s microprocessors, which have been specifically optimized to implement Intelâ(TM)s x86 ISA for almost four decades, will deliver amazing experiences, consistency across applications, and a full breadth of consumer offerings, full manageability and IT integration for the enterprise. However, we do not welcome unlawful infringement of our patents, and we fully expect other companies to continue to respect Intelâ(TM)s intellectual property rights. Strong intellectual property protections make it possible for Intel to continue to invest the enormous resources required to advance Intelâ(TM)s dynamic x86 ISA, and Intel will maintain its vigilance to protect its innovations and investments.

    If Microsoft can't transform SSE instructions into an ARM SIMD instruction set due to patents on SSE, Photoshop will suck if it's run through Microsoft's x86 to ARM64 JIT engine. And the odds are something like Photoshop is using bits of SSE which are still patented and will be for some time.

    Even if you don't emulate and run code nati

    --
    echo -e 'global _start\n _start:\n mov eax, 2\n int 80h\n jmp _start' > a.asm; nasm a.asm -f elf; ld a.o -o a;
  7. What's a xeon 818? by viperidaenz · · Score: 1

    According to Qualcomm officials, the 48-core Centriq 2460, priced at $1,995, offers more than four times the performance per dollar and 45 percent better performance per watt than Intel’s high-end Xeon Platinum 818 “Skylake” chip.

    Current Xeon processors have 4 digit part numbers, not 3...
    Is it the 8158, 8168, 8180, or something completely made up?

    1. Re:What's a xeon 818? by viperidaenz · · Score: 2

      Ok, so they never actually tested a single socket Xeon 8180, they extrapolated results from published benchmarks of dual-socket 8180's with a score of 2710 down to a "single socket" value of 776, to make it similar to their score of 657

      If they're not going to compare apples with apples, why do it at all?

  8. A processor based on arms? by Anonymous Coward · · Score: 0

    Oh, you meant ARM...