Samsung sues Nvidia over false benchmark results

12 November, 2014
Nvidia responded to the accusation saying it ran the tests on an out-of-the-box Note 4 and a Shield tablet.

Sort by:

  • s
  • spase
  • YP1
  • 13 Nov 2014

How dare Nvidia use Samsungs bench mark patent... :P

    • F
    • Fefo
    • msu
    • 13 Nov 2014

    if we can count a Surface Pro 3 as a tablet, Sammy's right

      • ?
      • Anonymous
      • kg3
      • 13 Nov 2014

      Samsung cheats in benchmarks and now they sue Nvidia? A new low.

        • ?
        • Anonymous
        • D3%
        • 13 Nov 2014

        AnonD-136279, 12 Nov 2014Can you imagine how far the tech world would have advanced ... moreAdvancement would be a lot slower. There would be no such thing as industry pressure to get product out the door. There would be no competition to try and outdo one another and create the best and cheapest products.

        That's one of the reasons why collaborations by multiple universities hardly ever lead to new breakthroughs, but national competitions in which there is a reward at stake is when breakthrough innovations are created.

          • ?
          • Anonymous
          • pqp
          • 13 Nov 2014

          darkgamer, 12 Nov 2014Note 4 uses cortex a57 not a15. Also it can run all 8 cores... moreWho cares? Multicore is just gimmick.

          The single core performance is the most important thing about CPU. And Nvidia and Apple CPUs are much better in single core performance.

            • A
            • Anonym
            • NCe
            • 13 Nov 2014

            Anonym, 13 Nov 2014Well it should only be multiplied by a thousand... but no o... more*frequency is in the equation (damn, it's too late in my region to do semantically coherent replies :P)

              • A
              • Anonym
              • NCe
              • 13 Nov 2014

              Anonym, 13 Nov 2014Do the math and see that I have indeed multiplied by the fr... moreWell it should only be multiplied by a thousand... but no one can say I didn't take frequency out of the equation. They are skewed in the power of 3, but would be accurate weren't it for the 8 simultaneous cores in Note 4 (where I only accounted the 4 of A57).

                • A
                • Anonym
                • NCe
                • 13 Nov 2014

                Anonymous, 13 Nov 2014You are saying: "The methodology you dispute is the on... moreDo the math and see that I have indeed multiplied by the frequency. It's not by any coincidence that all values appear multiplied by "1 million".

                  • ?
                  • Anonymous
                  • TqN
                  • 13 Nov 2014

                  Anonym, 12 Nov 2014Actually no. First of all these are a majorant (for Shie... moreYou are saying: "The methodology you dispute is the one given by ARM, which is stated exactly like this (per core per MHz)."

                  I'm not disputing the methodology given by ARM (which is per core per MHz). Infact, that is what I'm advocating. When you use per MHz, you are taking frequency out of the equation.

                  Example:
                  Core A reports 2 DMIPS/Mhz
                  Core B reports 4 DMIPS/Mhz

                  Since both are reporting per Mhz, we can conclude that core A is higher performance than core B. If, however, results are reported without normalizing to some base frequency (as below), you wouldn't be able to conclude which core is better, without knowing the operating freqency.

                  Core A reports 4 DMIPS. (operating @2MHz)
                  Core B reports 4 DMIPS. (operating @1MHz)

                  Now Core A and Core B seems to be at par, but, for the same result, Core B is operating at lower freq (thus lower power). Thus core B is better.

                  Therefore for clarity, it is better to report iso-frequency results.

                    • D
                    • AnonD-110477
                    • 8jp
                    • 13 Nov 2014

                    wow the biggest cheater couldnt handle the beating

                      • A
                      • Anonym
                      • NCe
                      • 12 Nov 2014

                      uf, 12 Nov 2014For comparing performance of different CPU core designs, it... moreActually no.

                      First of all these are a majorant (for Shield) and minorant (for Note 4). They aren't meant to be an exact representation of performance, just an exercise to show if the claims could be -- in any way, shape or form -- possible.

                      The methodology you dispute is the one given by ARM, which is stated exactly like this (per core per MHz).

                      The point that Note 4 can run the 8 cores simultaneously is indeed something which I didn't account for, hence it basically nullifies the OP (like already stated in the reply before this one)

                        • ?
                        • Anonymous
                        • vae
                        • 12 Nov 2014

                        AMD must be cheering

                          • A
                          • Anonym
                          • NCe
                          • 12 Nov 2014

                          darkgamer, 12 Nov 2014Note 4 uses cortex a57 not a15. Also it can run all 8 cores... moreThe A15 is the one in nVidia Shield, not in Note 4 (which I did point to being A57).
                          However I did not account for the 8 cores running simultaneously, which would indeed turn things completely around! (as in no longer being in the realm of a possibility that a nVidia Shield could be faster)

                            • D
                            • AnonD-251095
                            • 3Dx
                            • 12 Nov 2014

                            AnonD-136279, 12 Nov 2014Can you imagine how far the tech world would have advanced ... moreCan't imagine that, it is already advancing very fast much to our disadvantage.

                              Anonym, 12 Nov 2014Cortex A15 may perform up to 4.01 DMIPS/MHz per core (depen... moreNote 4 uses cortex a57 not a15. Also it can run all 8 cores simultaneously. 4 a57 + 4 a53 cores.

                              It has much faster cpu.

                                • u
                                • uf
                                • TqN
                                • 12 Nov 2014

                                Anonym, 12 Nov 2014Cortex A15 may perform up to 4.01 DMIPS/MHz per core (depen... moreFor comparing performance of different CPU core designs, it is best to leave frequency out of the picture, and look at normalized results. Since higher the frequency, higher the dynamic power, you then have to look at power consumption of respective CPUS...

                                  • D
                                  • AnonD-136279
                                  • nR3
                                  • 12 Nov 2014

                                  Can you imagine how far the tech world would have advanced if all these companies actually worked together instead of trying to take each other out in silly "mine is bigger then yours" games?!

                                    • A
                                    • Anonym
                                    • NCe
                                    • 12 Nov 2014

                                    Shield actually runs at 2.2GHz (not 2.4GHz like I've accounted before) so the end result is more like 35 than 38.

                                      • A
                                      • Anonym
                                      • NCe
                                      • 12 Nov 2014

                                      uf, 12 Nov 2014It's actually pretty simple to understand which of the two ... moreCortex A15 may perform up to 4.01 DMIPS/MHz per core (depending on actual implementation) while the Cortex A57 starts at 4.1 DMIPS/MHz per core.

                                      Given the Note 4 has a clock disadvantage (1.9GHz vs 2.4GHz) doing the math we get at least 31.16 * 10^6 DMIPS for Note 4 and up to 38.5 * 10^6 DMIPS for nVidia Shield.

                                      As such I don't believe things are as linear as you try to put.

                                        • N
                                        • No Fan Boy...!
                                        • 7t5
                                        • 12 Nov 2014

                                        AnonD-325269, 12 Nov 2014yea oc samsung the one who manufactured that cpu but its ap... moreBecause Exynos 5433 is more powerful than Apple A8X...!
                                        Samsung manufactures Apple A8X or all Apple CPUs...but Samsung don't copy Apple's design of CPU , they create their on design or use ARMs design for CPU...!