Announcement

Collapse
No announcement yet.

NVIDIA Tegra K1 Compared To AMD AM1 APUs

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

  • #16
    Originally posted by Ferdinand View Post
    Wow. Now it is all down to how much power they both use. If they use the same amount of power AMD only has x86 as the main differential. If Nvidia uses more power it fails.
    I would like to see performance per watt and ipc for an A15 core and a Jaguar core.
    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.
    Last edited by deppman; 05-05-2014, 05:19 PM.

    Comment


    • #17
      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.
      IIRC the highest power Mullins is 4.5W. Maybe you're thinking about Kabini - the AM1 chips that Michael is testing are Kabini, not Beema or Mullins.

      Comment


      • #18
        Originally posted by bridgman View Post
        IIRC the highest power Mullins is 4.5W. Maybe you're thinking about Kabini - the AM1 chips that Michael is testing are Kabini, not Beema or Mullins.
        Yes, I'm sorry. All the tested AMD chips appear to have a 25W TDP. I mentioned Mullins knowing it claims 4.5W, and I didn't mean to misrepresent it1. I was just, well, mulling over how the Mullins might compete with the K1.

        Footnotes:
        1 I wouldn't be surprised if the actual max is a bit higher as AMD of late has been prone "marketing over-zealousness" by doing things like quoting boost frequencies instead of base clocks on graphics cards. Not that they are by any means alone in that kind of behavior.

        Comment


        • #19
          Originally posted by deppman View Post
          Yes, I'm sorry. All the tested AMD chips appear to have a 25W TDP.
          On Windows 7 and prime95 running for all cores Athlon 5350 (on idle it uses around 2.5W) actally never exceed 16 W (mostly it is at 14W and there are a peaks at 15.8... so that is 16W ) . That is for the performance mode , in max power saver mode Athlon 5350 idle at 0.8W and max power it uses with prime95 running for all cores is just 1.8W with peaks at max 2.7W so that lower than 3W for 800MHz .

          I reading those values from the Asus utility , othewise i don't know how to measure wattage only for processor , on linux that sensor does not work .
          Last edited by dungeon; 05-05-2014, 09:34 PM.

          Comment


          • #20
            When Athlon 5350 is overclocked (if someone is interested in that game ) at 2600MHz it uses 4-5W more as that utility says it is around max 21W . So if i reading that right (and assuming that utility is right ) TDP value is just design of maximum 25W and Athlon 5350 alone actually never uses that much power .

            Comment


            • #21
              Michael why it says there 2 cores?

              http://www.phoronix.net/image.php?id...a_tk1_sys_show

              Comment


              • #22
                Originally posted by dungeon View Post
                Just depends if you're running it from an idle state afterwards in regards to its cpufreq driver for report core count.
                Michael Larabel
                http://www.michaellarabel.com/

                Comment


                • #23
                  Aha OK .

                  Too bad Am1 Athlon can't be overclocked more on linux ... Anyway i would like to see per clock comparation: Tegra K1 (lowered by 320MHz) vs J1900 (without burst) vs Athlon 5350 (lowered by 50Mhz)... let say all clocked at the same 2GHz and then do the benchmark with the same amount of the same clocked RAM .
                  Last edited by dungeon; 05-06-2014, 12:34 AM.

                  Comment


                  • #24
                    Originally posted by dungeon View Post
                    Anyway i would like to see per clock comparation: Tegra K1 (lowered by 320MHz) vs J1900 (without burst) vs Athlon 5350 (lowered by 50Mhz)... let say all clocked at the same 2GHz and then do the benchmark with the same amount of the same clocked RAM .
                    Just tried some tests, so when clocked similary then Athlon is better .

                    http://openbenchmarking.org/result/1...PL-ATHLON53550
                    http://openbenchmarking.org/result/1...PL-ATHLON53563

                    So these results @2394MHz can be compared with TK1 clock and J1900 with burst .
                    Last edited by dungeon; 05-06-2014, 04:03 AM.

                    Comment


                    • #25
                      Originally posted by dungeon View Post
                      Just tried some tests, so when clocked similary then Athlon is better .

                      http://openbenchmarking.org/result/1...PL-ATHLON53550
                      http://openbenchmarking.org/result/1...PL-ATHLON53563

                      So these results @2394MHz can be compared with TK1 clock and J1900 with burst .
                      Nah those two are singlethreaded tests Michael , i tought they are multithreaded .

                      Comment


                      • #26
                        are you able to compare power consumption of the boards when doing similar tasks?

                        Comment


                        • #27
                          Looking to those test numbers alone , i would prefer any time a AM1 5350...x86 code compatible, overall better CPU performance, more SATA or other I/O connectors, potential for APU upgrade, etc.

                          Looking forward for Graphics and Power drain tests...

                          Comment


                          • #28
                            Basically very interesting comparison. Still it would need more real life tests, esp. since we compare an x86 to an ARM. Power consumptions would also be highly of interest. But then, well, I don't know about the prices of the NV-ARM board but as a devel board it ought to be somewhat pricy. Also the question is: how free is it? Binary blobs needed to run? Meh. Furthermore: On the x86 side you normally have a lot more periphery at hand. In some cases you don't need it, but in my view you can never have enough interfaces.

                            I hope to have spare time soon so I can finally get my hands on Kabini / AM1 and Mullins. They promise to make so cute little machines.

                            Comment


                            • #29
                              Originally posted by Adarion View Post
                              Basically very interesting comparison. Still it would need more real life tests, esp. since we compare an x86 to an ARM. Power consumptions would also be highly of interest. But then, well, I don't know about the prices of the NV-ARM board but as a devel board it ought to be somewhat pricy. Also the question is: how free is it? Binary blobs needed to run? Meh. Furthermore: On the x86 side you normally have a lot more periphery at hand. In some cases you don't need it, but in my view you can never have enough interfaces.

                              I hope to have spare time soon so I can finally get my hands on Kabini / AM1 and Mullins. They promise to make so cute little machines.
                              I am surprised how well my peripheries worked. I think we truly have reached the "internet of things". Check out the gallery:

                              https://plus.google.com/photos/+Mike...48638600471660

                              1) USB Mouse, Keyboard work immediately
                              2) Fat32, Ext4 formatted SD cards work
                              3) Logitech webcam works flawlessly
                              4) HP 8600 OfficeJet printer works flawlessly with HPLIP.
                              5) 1920x1080p monitor works flawlessly (upscaled to 2560x1440; this is an HDMI limitation).
                              6) Synology NAS works flawlessly
                              7) Multi-card reader ...
                              8) Everyman headphones ...?

                              The only peripheries I have yet to test are 7+8, and I will probably do that tonight. Many of these have moved to the web-services model, and therefore work independent of drivers. ARM NFS works like a champ with the NAS.

                              Almost all productivity software works just great, like Gimp, Inkscape, LibreOffice, etc. There are just a few omissions, and I still have to verify Oracle Java which I need for some other productivity apps (XMLAuthor, Webstorm). Gosh, I wish I had 16GB of dram and a larger boot disk (although I have a 32GB SDCard which I am putting to good use). I am still awaiting my CUDA toolkit from NVIDIA.

                              Don't be surprised if nVidia launches an arm-based steam machine for <$200 or so. You read it here first

                              Comment


                              • #30
                                Originally posted by deppman View Post
                                I am surprised how well my peripheries worked. I think we truly have reached the "internet of things". Check out the gallery:

                                https://plus.google.com/photos/+Mike...48638600471660

                                1) USB Mouse, Keyboard work immediately
                                2) Fat32, Ext4 formatted SD cards work
                                3) Logitech webcam works flawlessly
                                4) HP 8600 OfficeJet printer works flawlessly with HPLIP.
                                5) 1920x1080p monitor works flawlessly (upscaled to 2560x1440; this is an HDMI limitation).
                                6) Synology NAS works flawlessly
                                7) Multi-card reader ...
                                8) Everyman headphones ...?

                                The only peripheries I have yet to test are 7+8, and I will probably do that tonight. Many of these have moved to the web-services model, and therefore work independent of drivers. ARM NFS works like a champ with the NAS.

                                Almost all productivity software works just great, like Gimp, Inkscape, LibreOffice, etc. There are just a few omissions, and I still have to verify Oracle Java which I need for some other productivity apps (XMLAuthor, Webstorm). Gosh, I wish I had 16GB of dram and a larger boot disk (although I have a 32GB SDCard which I am putting to good use). I am still awaiting my CUDA toolkit from NVIDIA.

                                Don't be surprised if nVidia launches an arm-based steam machine for <$200 or so. You read it here first
                                How many games does steam have that have an ARM version?

                                Comment

                                Working...
                                X