Announcement

Collapse
No announcement yet.

NVIDIA Tegra K1 Compared To AMD AM1 APUs

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

  • #41
    Originally posted by deppman View Post
    The K1 is looking like a 3.5W-ish chip under "normal" workloads, and 7W-ish chip at full-tilt according to the published pdf. These estimates are from p13 looking at AP+DRAM. NVidia does claim that the actual numbers optimized for mobile will be lower, but I guess we'll see.

    It I am reading this right, all the AM1 chips have a 25W TDP, so they aren't even in the same league. The fact that the K1 meets or beats these procs on many of the benchmarks is really quite impressive.

    On Android, where Intel's code morphing has proven to eat batteries and kill performance for x86, the K1 looks like an easy win over Mullins. Of course, Mullins benefits from the x86 installed base for Linux and Windows platforms.
    I've yet to find any third party power draw benchmarks for the tk1. semiaccurate thought nvidia was being deceptive (http://semiaccurate.com/2014/03/05/n...-number-watts/), and has a history of such behavior (http://semiaccurate.com/2014/01/20/w...eally-perform/). They estimated a power draw of around 35-40W for the board on display.
    I'm really anxious to michael show us the power draw numbers (assuming he has a kill-a-watt, it's easy, but hopefully the firmware is exporting to /sys).

    Comment


    • #42
      Charlie and Power Draw

      Originally posted by liam View Post
      I've yet to find any third party power draw benchmarks for the tk1. semiaccurate thought nvidia was being deceptive (http://semiaccurate.com/2014/03/05/n...-number-watts/), and has a history of such behavior (http://semiaccurate.com/2014/01/20/w...eally-perform/). They estimated a power draw of around 35-40W for the board on display.
      I'm really anxious to michael show us the power draw numbers (assuming he has a kill-a-watt, it's easy, but hopefully the firmware is exporting to /sys).
      Charlie is embarrassingly wrong and needs to hand off writing about nVidia to someone who is more objective. His writings about NV are at best semi-lucid. The 30-40W estimate from his "analysis" is absolutely absurd - especially when you consider a GTX750 runs on the same amount of power. The TK1 has less than 1/3 the CUDA count and is optimized for power effeciency.

      What Charlie was looking at as a T3 powered automotive ECU development board. Since ECU's encounter much harsher conditions than consumer electronics, more robust cooling solutions are required. And justifying his conclusions because NV used an off-the-shelf power adapter instead of spending tens of thousands to build one for demonstrations ... come on, man!

      Nvidia had these chips running all day in a 7" tablet chassis on numerous outings. If these chips were running 30-40w, the chasis would have literally melted (imaging putting a 30-40W lightbulb in a 7" tablet chassis with no active cooling. Can you say "fire hazzard").

      So I am willing to put these claims to the test. I am not an EE, but would this do to measure power draw?

      Comment


      • #43
        Originally posted by deppman View Post
        Charlie is embarrassingly wrong and needs to hand off writing about nVidia to someone who is more objective. His writings about NV are at best semi-lucid. The 30-40W estimate from his "analysis" is absolutely absurd - especially when you consider a GTX750 runs on the same amount of power. The TK1 has less than 1/3 the CUDA count and is optimized for power effeciency.
        GTX750 uses at least 105 watts in crysis3
        Originally posted by deppman View Post
        Nvidia had these chips running all day in a 7" tablet chassis on numerous outings. If these chips were running 30-40w, the chasis would have literally melted (imaging putting a 30-40W lightbulb in a 7" tablet chassis with no active cooling. Can you say "fire hazzard").
        What charlie said was that the benchmarks and power measurements were not done at the same time and maybe not even on the same hardware. So you have a tablet to show that it runs within a tablet. You have a motherboard with big cooler to show that it is really fast in benchmarks. Phoronix has benchmarks of AMD's socs and those computers use 30-40 watts.
        Last edited by Ferdinand; 10 May 2014, 01:24 PM.

        Comment


        • #44
          Originally posted by Ferdinand View Post
          What charlie said was that the benchmarks and power measurements were not done at the same time and maybe not even on the same hardware. So you have a tablet to show that it runs within a tablet. You have a motherboard with big cooler to show that it is really fast in benchmarks. Phoronix has benchmarks of AMD's socs and those computers use 30-40 watts.
          There is no defense for his completely amateur analysis. Judging power draw on an A/C converter by how warm it is? Pure speculation. If my Jetson TK1 is drawing 40 watts, I will eat my hat.

          In any event, we will have all that settled tonight, as I will be purchasing a multimeter and will measure Volts and Amps on other the DC or AC line tonight. Or both.

          Comment


          • #45
            Use a "good brand" power usage monitor - it will tell you how much power the whole PC including it power adapter (it's not 100% efficient) is actually using (and it's safe). My PC build with a 10W J1900 can use up to 37,5W (record during xonotic benchmark) of power (1 SSD, two USB devices, two SODIMMs, the motherboard plus a power brick) and in a fanless mesh case the CPU can get up to ~65C.

            Comment


            • #46
              Originally posted by deppman View Post
              There is no defense for his completely amateur analysis. Judging power draw on an A/C converter by how warm it is? Pure speculation. If my Jetson TK1 is drawing 40 watts, I will eat my hat.

              In any event, we will have all that settled tonight, as I will be purchasing a multimeter and will measure Volts and Amps on other the DC or AC line tonight. Or both.
              Looking forward to that, thanks! The best way to measure is of course ampere clamp and voltage meter on dc end. That 60W powerbrick has max a bit over 90% efficiency anyway(if it's any good).

              Originally posted by Ferdinand View Post
              No, you assume that processor does not consume more power on gaming, which isn't the case. Check tpu's power consumption charts instead, they are taken directly from pcie socket(maximum 76W more real life scenario 64W, nvidia power limiter should kick it near to tdp).

              Comment


              • #47
                Originally posted by tuke81 View Post
                No, you assume that processor does not consume more power on gaming, which isn't the case. Check tpu's power consumption charts instead, they are taken directly from pcie socket(maximum 76W more real life scenario 64W, nvidia power limiter should kick it near to tdp).
                You're right. Didn't think of that.

                Comment


                • #48
                  Originally posted by riklaunim View Post
                  Use a "good brand" power usage monitor - it will tell you how much power the whole PC including it power adapter (it's not 100% efficient) is actually using (and it's safe). My PC build with a 10W J1900 can use up to 37,5W (record during xonotic benchmark) of power (1 SSD, two USB devices, two SODIMMs, the motherboard plus a power brick) and in a fanless mesh case the CPU can get up to ~65C.
                  65?C is to much i think . If you measure that temp during xonotic, then what will happen if you run mprime during hot summer day .

                  I can recommends runing prime , just to know what is max power usage and temps for the CPU . I usually run that when i overclock CPU and test the memory... so run it and tell what is power usage during 2nd test , and also use 1st one to produce max heat for the CPU, etc.

                  Last edited by dungeon; 10 May 2014, 03:03 PM.

                  Comment


                  • #49
                    Originally posted by deppman View Post
                    There is no defense for his completely amateur analysis. Judging power draw on an A/C converter by how warm it is? Pure speculation. If my Jetson TK1 is drawing 40 watts, I will eat my hat.

                    In any event, we will have all that settled tonight, as I will be purchasing a multimeter and will measure Volts and Amps on other the DC or AC line tonight. Or both.
                    Assuming you know the design power, you can actually make a surprisingly accurate estimate about its load based on temp. Power supplies are designed to be most efficient at ROUGHLY half their design power, and worst at very high/low loads. IOW, temp is a pretty good measure of draw for switching power supplies. It would be better if he knew the exact temp, of course.

                    Comment


                    • #50
                      Power Draw Analysis: RESULTS!

                      Originally posted by deppman View Post
                      There is no defense for his completely amateur analysis. Judging power draw on an A/C converter by how warm it is? Pure speculation. If my Jetson TK1 is drawing 40 watts, I will eat my hat.

                      In any event, we will have all that settled tonight, as I will be purchasing a multimeter and will measure Volts and Amps on other the DC or AC line tonight. Or both.
                      Tests are in!

                      Charlie needs to write a retraction. His pseudo-analysis looks to be off by around 250%, or 20-30W.

                      While these results are far from exhaustive, here are some highlights and observations:
                      • I have yet to observe > 12W total system draw under any workload.
                      • I confirmed the fan's power draw by unplugging it and noting the difference in amps.
                      • Nvidia's numbers found in their technical brief (page 13) appear accurate to conservative.
                      • Nvidia's point about drawing comparisons to mobile are valid. Notice we are using a bevy of desktop-level ports and peripherals (e.g. HDMI out, GbE, RAM, etc) instead of low power mobile alternatives.
                      • The base system does indeed appear to draw about 2W with the processor running at 0.66W at idle.
                      • As load ramps, the increased power draw comes from many components, not just the SOC.
                      • In conclusion, this looks like a fine tablet or steambox chip.


                      Code:
                      FULL SYSTEM JETSON TK1 DC POWER ANALYSIS
                        Includes
                          * Audio in/out Active
                          * Gigabit Ethernet
                          * USB 3 driving Keyboard, Mouse, Logitech C210 Webcam
                          * HDMI out at 1920x1080
                          * Cooling Fan (no heat sink)
                          * Installed 64GB SD card
                      
                      BASE MEASUREMENTS
                        Power-in Voltage : 12.15V
                        Amperage and wattage
                          Idle KDE Desktop : 0.22A ( 2.67W)
                          Fan Amp          : 0.07A ( 0.85W)
                          System Amp*      : 0.17A ( 2.06W) *NV measurement
                      
                          Idle Less Fan    : 0.15A ( 1.82W)
                          Idle Less System : 0.05A ( 0.61W)
                      
                      TEST1 : glmark2 -s 1920x1080 --off-screen
                        Score : 282 (Intel Celeron [email protected] = 151)
                        Power Measurements:
                          Base             : 0.22A ( 2.67W)
                          Peak             : 0.62A ( 7.53W)
                          Observed Avg.    : 0.35A ( 4.25W)
                      
                          Avg. Less Fan    : 0.28A ( 3.40W)
                          Avg. Less Sys    : 0.18A ( 2.19W)
                      
                      TEST2 : CUDA Smoke particle demo
                        Power Measurements:
                          Base             : 0.62A ( 7.53W)
                          Peak             : 0.91A (11.06W)
                          Observed Avg.    : 0.88A (10.69W)
                      
                          Avg. Less Fan    : 0.81A ( 9.85W)     
                          Avg. Less Sys    : 0.71A ( 8.26W)
                      
                      TEST3 : VLC streaming 720p video from NAS GbE
                        Power Measurements:
                          Base             : 0.29A ( 3.52W)
                          Peak             : 0.41A ( 4.98W)
                          Observed Avg.    : 0.34A ( 4.13W)
                      
                          Avg. Less Fan    : 0.27A ( 3.28W)     
                          Avg. Less Sys    : 0.17A ( 2.01W)

                      Comment

                      Working...
                      X