Announcement

Collapse
No announcement yet.

CompuLab Turns An 8-Core/16-Thread Xeon, 64GB RAM, NVIDIA Quadro RTX 4000 Into Fan-Less Computer

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

  • #11
    Originally posted by bdcomp View Post
    I really miss an updated small factor fanless PC (fitlet3 ?), with AMD APU from CompuLab...
    Ryzen Embedded has integrated graphics. Not Ryzen2 yet though.

    I found this one: https://tranquilpcshop.co.uk/mini-mu...lay-pc-v1605b/

    Comment


    • #12
      The GPU temperature graph, to me, looks like the GPU compute tests weren't long enough for it to ramp up and stabilize. I would definitively expect to reach temperatures near the CPU temperature if let running long enough.

      Edit: Combined CPU and GPU workloads would be interesting to see too. I'm betting it would throttle hard. Of course all workloads don't put heavy combined CPU+GPU pressure.
      Last edited by AsuMagic; 17 July 2019, 04:36 PM.

      Comment


      • #13
        Originally posted by pegasus View Post
        This is so funny ... since on the other side server vendors are whining that they cannot cool 180W+ cpus with fans only and that we should move to water cooling ... Amateurs!
        Correct me if I am wrong, but I believe CompuLab sets up segregated cooling zones, so the heat from the GPU does not mingle with CPU/board heat on its way out the door.

        Modern server vendors rely on fresh air coming up through the bottom of the rack and taken in en masse through the front panel where all system generated heat is pulled through the back and pushed up through the top.

        If the server vendors were to perform some sort of heat channeling or segregation, this would require consuming space and materials, precious in a 2U format, impossible in a 1U.

        Also for maximum effect the Compulab has to be installed vertically, where rack servers are optimized to be installed horizontally.

        Tower server or workstation cases however could learn from CompuLab's research since many of them sit under desks or in closets where air flow may be restricted.

        Some Lenovo and Dell workstations do utilize plastic facets that are hinged over the PCIe slots to try to segregate heat from the rest of the platform, but not to the same level of exclusion that CompuLab has done.

        FWIW: I just checked CompuLab's website and this model appears to be sold out.

        But I could definitely see using their newer Fitlet in remote solar powered relay stations.

        Comment


        • #14
          Originally posted by danmcgrew View Post
          That means that it (the silicon; a metal) has approximately an eight-times (2³) longer life than the silicon in a machine which runs at 80° C under heavy load.
          Simple--and unavoidable--physics, folks.
          You are correct, but meanwhile I haven't seen any cpu that was not overclocked run into any end-of-life scenario because of this. Even the old 286 cpus from the 80's still run, and they didn't have a fan, or even a heatsink sometimes. They would burn the skin off your finger to the touch, so I'd guess they had to be over 70C.

          Intel's reliability study brought forth three types of chemical degradation: Gate oxide dielectric breakdown, via voiding failure and interlayer breakdown. Interlayer breakdown was the most likely, happening after equivalent hours of 4 080 000 hours, or 465 years. Reduce that by 8, and that's 58 years. Obviously those numbers are extrapolated and estimates, but they seem to suggest that it's not a real issue.

          Anecdotely, I run a mini-server 24/7 than has an atom 330. I've had it running for 10 years straight at around 90-92 C core temp average, and it still runs fine, passes stability tests. So if running under better conditions would have resulted in a 80+ year lifespan, I'm sure the capacitors on the board would have long ago died.

          Comment


          • #15
            This offering isn't so special after all - you can get all kinds of passively cooled systems, including ones with Threadripper CPUs and high-end GPUs, also at https://www.deltatronic.de/en/pc-en

            I own two of their tower systems, the first one was bought when the Athlon64 was new on the market, and I am pretty satisfied with how well they work.

            Comment


            • #16
              I wonder how far they are away from an AMD platform. AMD's power/thermal advantage seems pretty big right now, and the IO advantage is huge.
              Plus, even Ryzen can do ECC.

              Comment


              • #17
                Originally posted by danmcgrew View Post
                Looks like a very nice machine. I would probably buy one if it had an active cooling system.
                I'm a firm believer in the immutable laws of physics, and the main one which concerns me is that which states that chemical (and hence, metallurgical) reactions double for every 10° C rise in temperature. My Lenovo T430 perks along at 37° C. Say it gets up to 50° under heavy load. That means that it (the silicon; a metal) has approximately an eight-times (2³) longer life than the silicon in a machine which runs at 80° C under heavy load.
                Simple--and unavoidable--physics, folks.

                It would be an extremely attractive option were CompuLab to offer this machine with active cooling. It would also be an extremely informative experiment were someone to do whatever needed to be done to simulate active cooling with an existing machine, and THEN measure temps.
                Or just buy a slow-rotating floor (or desk) fan and you'll have a perfectly cromulent cooling solution that requires exactly zero redesign effort?

                Comment


                • #18
                  I'd rather have one or two very quiet slow fans included.

                  Comment


                  • #19
                    From #9--

                    "...It would also be an extremely informative experiment were someone to do whatever needed to be done to simulate active cooling with an existing machine, and THEN measure temps."

                    By "...simulating active cooling...", I meant: exposing the innards to ambient (i.e., make it an 'open' system--remove panels, etc.) AND using a fan; I did NOT mean keeping them (the 'innards') locked up in a sealed box and merely blowing a fan ONTO the sealed box (which would do absolutely nothing to cool the sealed-up contents of the box).

                    Wouldn't be at all surprised to find that an 'open system' without a fan would result in a considerable reduction in overall temperature(s).
                    Last edited by danmcgrew; 17 July 2019, 09:49 PM.

                    Comment


                    • #20
                      Originally posted by pegasus View Post
                      This is so funny ... since on the other side server vendors are whining that they cannot cool 180W+ cpus with fans only and that we should move to water cooling ... Amateurs!
                      Well, try shoving 20 or more of these into a rack and see how well it stays cool.

                      Comment

                      Working...
                      X