Announcement

Collapse
No announcement yet.

NUVIA Published New Details On Their Phoenix CPU, Talks Up Big Performance/Perf-Per-Watt

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • NUVIA Published New Details On Their Phoenix CPU, Talks Up Big Performance/Perf-Per-Watt

    Phoronix: NUVIA Published New Details On Their Phoenix CPU, Talks Up Big Performance/Perf-Per-Watt

    Since leaving stealth last year and hiring some prominent Linux/open-source veterans to complement their ARM processor design experts, we have been quite eager to hear more about this latest start-up aiming to deliver compelling ARM server products. Today they shared some early details on their initial "Phoenix" processor that is coming within their "Orion" SoC...

    http://www.phoronix.com/scan.php?pag...ix-Orion-Early

  • #2
    I swear I read "NVIDIA" instead of "NUVIA" by accident and almost thought they were gonna make CPUs...

    Although it is true that ARM implementations have been getting better and more on par with x86...
    Like I had this phone from 2014 and despite having a 2.5GHz processor it performed poorly against my dual-core x86 laptop with the same speed...
    ...but in 2019 I bought this other phone, and it surprisingly matches my laptop's performance! (and is faster than the other phone by 400%!)

    Now where is my 4GHz ARM chip...
    Last edited by tildearrow; 08-11-2020, 03:42 PM.

    Comment


    • #3
      Designing yet another CPU based on a reference (N1) is great. but someone has to put up the dough (capex) to actually get them fabbed in quantity. Also you have to have some dough to get an OEM to design and support a planar for your design. Doing a design to tape and getting some test fabs done is one thing, ramping up for a build out is something completely different.

      As I told the Nuvia sales evangelist, no offense, if they work out great, but unless they have some serious coin backing them, the odds are not good. The road of CPU's is littered with the skeletons of previous attempts, mostly because they ran out of money.

      Comment


      • #4
        Aside from the fact that these headlines look like Nvidia is doing something when they aren't (or that Nivea skincare is jumping in the CPU business https://www.niveausa.com/), it's super easy to make bold claims on paper. It's a whole lot harder to make a fully working system stack that's mass deployable on a huge scale, which is what they really need to take on the x86 market in any meaningful way.

        As for ARM, there's nothing magical about ARM vs. x86 vs.... RISC-V [the new cool architecture] and Nuvia is already failing in its own propaganda by targeting the wrong "competitors". If Nuvia really thinks ARM is all that, they shouldn't be targeting AMD or Intel, they should be targeting the in-house proprietary ARM server chips that Amazon and a few others are making for their own cloud deployments. That would actually be a whole lot more interesting than a contrived powerpoint attack on commodity x86 systems.

        Comment


        • #5
          Originally posted by tildearrow View Post
          I swear I read "NVIDIA" instead of "NUVIA" by accident and almost thought they were gonna make CPUs...
          Actually, right now NVIDIA might be in the closing stages of acquiring ARM. 🤫

          Comment


          • #6
            Sure perf/watt is good, but scales well? What about software compilation? how about top ST perf? Many workloads a very ST dependant. First we have to see the chips working and latter compare to the others

            Comment


            • #7
              Originally posted by edwaleni View Post
              Designing yet another CPU based on a reference (N1) is great. but someone has to put up the dough (capex) to actually get them fabbed in quantity. Also you have to have some dough to get an OEM to design and support a planar for your design. Doing a design to tape and getting some test fabs done is one thing, ramping up for a build out is something completely different.
              This is one of the biggest problems I have wit a lot of these start ups, they seem to think it is all about the chip. Worse they are way too focused on the server market and frankly that hasn't gone well at all for many attempts. To get capital to engineer and sell into the server market a company like this really needs a product aimed at the note book and embedded market. In the end they will need to deliver a complete server solution to get buy in from the big users. That means a complete board solution in a box.

              On the other hand the embedded market is far more willing to adopt and build boards for alternative processor solutions. Produce the right chip and somebody will build a board for it. Laptops might be a harder market to crack but that could change real soon.
              As I told the Nuvia sales evangelist, no offense, if they work out great, but unless they have some serious coin backing them, the odds are not good. The road of CPU's is littered with the skeletons of previous attempts, mostly because they ran out of money.
              I hope they listened. I've yet to see a ARM server processor start up that has a plan that has a chance in hell of working. Most of the server world is a commodity market and thus you need to enter into that market and work your way up from there.

              Comment


              • #8
                Originally posted by chuckula View Post
                Aside from the fact that these headlines look like Nvidia is doing something when they aren't (or that Nivea skincare is jumping in the CPU business https://www.niveausa.com/), it's super easy to make bold claims on paper. It's a whole lot harder to make a fully working system stack that's mass deployable on a huge scale, which is what they really need to take on the x86 market in any meaningful way.

                As for ARM, there's nothing magical about ARM vs. x86 vs.... RISC-V [the new cool architecture] and Nuvia is already failing in its own propaganda by targeting the wrong "competitors". If Nuvia really thinks ARM is all that, they shouldn't be targeting AMD or Intel, they should be targeting the in-house proprietary ARM server chips that Amazon and a few others are making for their own cloud deployments. That would actually be a whole lot more interesting than a contrived powerpoint attack on commodity x86 systems.
                the attacks should be on the cost of ownership. They could do that buy selling complete systems and frankly by going at the market from the bottom up. I just can't see a business trying to get into a few big server installation ever lasting long because there isn't enough volume there to sustain such a business. Even Intel and AMD get this and offer enough hardware choices so that volume sales can happen into lesser installations. Often this is via commodity sales so cost is a huge factor.

                Comment


                • #9
                  Originally posted by wizard69 View Post
                  To get capital to engineer and sell into the server market a company like this really needs a product aimed at the note book and embedded market.
                  Embedded is oversaturated already, not a good idea.

                  Notebook or even (slim) desktop or AIO is worth a shot, now that Windows is trying to jump into ARM and there is exactly jackshit hardware that is actually capable of running it, and they are so desperate they had to go in and partner with Qualcomm to make a special "overclocked" Snapdragon for their Surface Pro X.

                  But this is something that is unveling now and this means they had to start developing with this goal years ago.

                  Comment


                  • #10
                    <chuckle>......And yet ANOTHER data point to show that x-86 should now be seen as a deprecated, dying ISA for most computing tasks except for very high end HPC and Supercomputing. I got SUCH flak for me saying this under Michael's post earlier a few days ago concerning the long running issue of AMD not being able to get their Fusion Sensor Hub in good shape for inclusion in the Linux Kernel. The typical insular, arrogant, Linux types came out to lambast me. You know the ones....they have the masturbatory flame wars over GPL vs BSD licensing or which is the best text editor or you suck if you want a GUI for something and don't want to clickety clack in the command line like the kewl kids.

                    Nuvia is not the ONLY one who is going to utterly kick down the doors and blow up the framework of the x-86 house. Apple is going to MATCH or EXCEED what Nuvia is doing when they come out with their M-14 desktop ARM based CPU for their desktops. The A13 is already MORE performant in single core tasks than a lot of Intel and AMD CPU's and close enough for rock and roll in multi-core. And THAT's a PHONE chip. The A14 for the next gen iPhone and iPad will completely compete with Intel and AMD. HOWEVER....it's the desktop silicon which will be unencumbered by the thermal constraint of the A series phone chips which are going to completely shock the Wintel world. Microsoft ALREADY knows this which is why they have ALREADY released their Qualcomm designed SoC with more to come. They are ALREADY porting Office over to MacArm and this will make it trivial to port Office right over to WinArm.

                    And where will this leave Linux on the Desktop by 2022-2025. Horribly behind desperately trying to optimize the entire stack for ARM including other projects like LibreOffice who as far as I know don't even have an ARM port at all. It's NOT about recompilation and porting. It's about OPTIMIZING for performance, stability AND new hardware features like built in A.I. and Neural Net processors that Intel nor AMD CPU's have built in like ARM or their Apple or Qualcomm derivatives.

                    ARM, Apple, Qualcomm and Mediatek are ALL going to beat Intel AND AMD to 5nm. And that's REAL 5nm not optimized 7nm but will be marketed as "5nm". And by the time even AMD gets to real 5nm ( and forget about Intel as they are hopelessly lost in fabrication now ) ARM, Apple, Qualcomm and Mediatek will be chugging merrily along at 3nm. With even MORE Power per Watt efficiencies AND ALSO just plain more computational OOMPH. Not to mention built in AI Engines, Neural Net Engines, GPS, Vision Engines NONE of which is in ANY x86 processor either from Intel nor AMD.

                    x86 died the day Apple announced their move to ARM. Just as Floppy Drives, CD/DVD-ROMs, Hard Drives, and Flash died once Apple announced them to be old, dead technology. Everybody poo-pooed those announcements as well. Until Microsoft...as they always have done...copied Apple as well and turned the lumbering Wintel Titanic around to follow Apple's lead.

                    I am NOT an Apple fanbois. I have NEVER owned ANY Apple product. I do NOT like Apple's "Walled Garden" approach to consumer computing. But they ARE the FIRST consumer computer company to reach 1 Trillion dollars in valuation and that was just 2 years ago. They are approaching 2 Trillion dollars this year and will probably reach that with the introduction of the iPhone 12 as this will be a "Super Cycle" upgrade season. But I make this prediction now. Apple will become a FOUR Trillion dollar evaluated company after they begin to release their COMPLETELY harmonized, converged ARM ecosystem that goes from iPhone to iPad, to Macbook to iMac to Mac Mini to Mac Pro, to the Apple Watch and Apple TV. EVERY SINGLE DEVICE running ARM and a version of iOS/MacOS all completely optimized and performanced tuned to ARM.

                    And where will we Linux folks be. Back to trying to load our flavor of Linux on some little puny Chromebook with a shitty Intel or highly defanged AMD Zen CPU just to get decent battery life or on an ARM based Chromebook or eventually Windows based laptop with an unoptimized version of Linux for ARM and hoping we get bettter performance than a typical Raspberry Pi 4 or 5.

                    Tick Tock Linux. The ARM apocalypse is here.



                    Comment

                    Working...
                    X