Geekbench 3 is now live in the Play Store, measures real world performance

August 17, 2013

Geekbench 3

If you’ve been looking for a new way to benchmark your device, then Primate Labs Inc has just launched an app for you. It’s called Geekbench 3 and its claim to fame is it tests your device based on real world performance as opposed to just graphics performance. So there are still apps like AnTuTu and Quadrant Standard for graphics, but now there is another for actual, real world performance benchmarks.

Geekbench 3 boasts a lot of features that tests your devices in a way that they aren’t usually tested. Here are a few according to their official website.

  • Geekbench 3 includes 15 new benchmark tests that are designed to model real-world processor intensive tasks. The new tests cover a number of different application domains, including encryption, image processing, signal processing, and physics simulation.
  • Geekbench 3 also includes 12 completely rewritten benchmark tests from Geekbench 2. These tests are now more representative of real-world applications and usage scenarios.

Along with that, Geekbench 3 has been specifically designed to test multi-core processors, but it also still tests single-core performance. This is because many apps and processes use only a single core, making it an important metric. So Geekbench 3 helps you not only determine which devices have the best overall performance, but the best performance per core as well.

Primate Labs Inc also noted that since Geekbench 3 uses totally new tests than Geekbench 2 did, the scores from each one are not comparable. Any discrepancies found between the two apps are due to the difference in programming and nothing more. For those that do care about such things, there is also a brand new interface.

Will anyone be giving the new Geekbench 3 a shot? More importantly, should this be a benchmark that devices reviewers use instead of (or alongside) graphic benchmarks like AnTuTu to give a better perspective on how devices run? If you have a moment, let us know in the comments. If you want to give it a shot, you can find it in the Google Play Store here.

Comments

  • noMOREcheating

    someone quickly do a geekbench 3 test on the g4 before sammy can “boost” the result. this way we can compare it to other devices.

    • Lakshmiram Vasireddy

      GS4 has single-core score of 558 and Multi-core score of 1528, and new nexus 7 scored 586 and 1829. In theory GS4 should score better since it has better processor(snapdragon 600)

      • M. Clark

        What do you think the Nexus 7 2013 has in it? (hint: it’s a quadcore snapdragon 600 clocked slower with the exact same GPU)

      • Xuqi99

        Except Samsink don’t/won’t/can’t optimize* anything!
        * – Except to cheat certain benchmark program/s.

      • petusalem

        Samsung S4 Active 680/1888.

        • Roberto Tomás

          see axarm’s comment — these benchmarks haven’t been earmarked as full-speed apps for power management, so scores will be higher after update.

      • axarm

        You are comparing a phone to a tablet. This is a new benchmark so the phone power management is still working and is throttling the phone. I seen a result yesterday for a S800 and it ran at 1.7 not the 2.1 that is advertised. Give Qualcomm a few days and the power management will be bypassed and the phone will run full throttle.

        • Roberto Tomás

          http://browser.primatelabs.com/geekbench3/3091 – just found one that reads as exynos 5410 running at 2.0 Ghz (which is pretty close to your 2.1ghz). This score, 2454, is surprisingly close to the scores for Qualcomm Snapdragon 800′s that they have up (2746 at 2.17Ghz).

      • Roberto Tomás

        which GS4 is tht? the snapdragon 800 or the exynos 5410?
        scores seem to have dropped a lot in version 3 — the 5410 scored ~3500 before.

        • axarm

          In the previous versions, the power management was diabled allowing the phones to “overclock” for benchmarks. As I said before , it time this will be “fixed” and the scores will rise. It is called cheating.

      • chris pinkston

        nexus 7 has same krait 300 cores as gs4. Gs4 has them clocked higher so I don’t know what gives. Maybe software, stock android vs touchwiz

      • Firmino
    • hassan

      haha best comment ever lol

    • Xuqi99

      No chance of that happening. By the time they get their next software update out for the GS4, Geekbench 4 would have been released.

  • MasterMuffin

    Benchmark that costs money? Yea no thanks

    • Luka Mlinar

      I’m gonna read your comment before i start reading the article. Saves time :D

      • MasterMuffin

        The pressure :O

        • Luka Mlinar

          You can do it! :D

    • renz

      is that unusual?

      • MasterMuffin

        All the biggest ones are free or have a free version. I’m not that eager to benchmark!

    • iamdluhansa

      you can contribute to world peace if you buy it

      • Joshua

        That makes no sense??

        • Luka Mlinar

          you can contribute to the developers pocket if you buy it ;)

          • Joshua

            Exactly

  • Suryawan Pranata Kurniawan

    Got score 710 single core 2040 multi core (sgs4 octocore

    • kascollet

      This score doesn’t show in the GeekBench online database. Did you publish it ?

    • Awais

      Same score here. Infact i got 760 in single core and 2028 in multi octocore. Everything stock not rooted.

  • kascollet

    This new version of GeekBench restablishes the iPhone among the best devices. One month before Apple releases another chip for the iPhone 5S… how convenient :-D

    Other interpretation : Apple’s Swift core is actually much more powerfull than Krait and A15. Doesn’t look good for Samsung/NVidia/Qualcomm.

    • Tamadrummer94

      In day to day usage, I really haven’t seen anything beat the speed of the 5. It’s pretty impeccable. They’re doing something right with that homegrown chip and all their optimizations. I wonder what the Moto X scores…

    • Roberto Tomás

      umm, what? the iphone is still scoring very low, half the speed of the top android devices they have on there.

      • kascollet

        I’m speaking about single core power for the moment.

        Just had a look into GeekBench 3 online results.

        Single core performance at shipping frequency :
        - Qualcomm Krait 300 @ 1.9GHz (GS4) : 620
        - Qualcomm Krait @ 1.5 GHz (Xperia Z/Nexus 4) : 510
        - ARM A15 @ 2.0GHz (OC-ed GS4 Octa) : 680
        - ARM A9 @ 2.0 GHz (OC-ed GN2) : 500
        - Apple Swift @ 1.3 GHz (iPhone 5) : 710

        Single core performance, normalized at 2 GHz frequency :
        - Qualcomm Krait 300 : 650
        - Qualcomm Krait : 680
        - ARM A15 : 680
        - ARM A9 : 500
        - Apple Swift : 1090

        Swift would be more than 50% more powerful at the same frequency than both Cortex A15 and Krait. How is this even possible ?

        • Roberto Tomás

          yea apple does fun stuff with cache (and ram) that boosts perf but requires a much lower clock rate. that single core score is the same as an amd e-350 I see. :)

          thanks, I see what you mean. it is hard to find that in the browser, thanks for digging it all up and putting it in one place for us!

          • kascollet
          • kascollet

            Now considering full SoCs.

            Multi core performance at shipping frequency :
            - 4XQualcomm Krait 300 @ 1.9GHz (GS4) : 1700
            - 4XQualcomm Krait @ 1.5 GHz (Xperia Z/Nexus 4) : 1500
            - 4XARM A15 @ 1.6GHz (GS4 Octa) : 1800
            - 4XARM A9 @ 1.5 GHz (GS3 Int) : 1500
            - 2XApple Swift @ 1.3 GHz (iPhone 5) : 1250

            Multi core performance, normalized at 2 GHz frequency :
            - 4XQualcomm Krait 300 : 1790
            - 4XQualcomm Krait : 2000
            - 4XARM A15 : 2250
            - 4XARM A9 : 2000
            - 2XApple Swift : 1920 (Dual-Core !!)

            Well, let’s hope Apple doesn’t go quad-core Swift next month ! It would jus be a monster, even at low frequency.

          • Brody Boswell

            Performance doesn’t scale up liniary with clockspeed.I mean there’s no way a9 or krait 200 beats krait 300

          • kascollet

            I too find it strange, but usually performance scales totally linearly with clock speed. What is strange is the huge variability in Android phones results. I tried to pick representative ones to make my maths but the sample is very wide for a given device. On the contrary, iOS devices are very even from one to another. Maybe people often forget to shut down background apps.

        • kascollet

          I forgot the best SoC available ATM : the Snapdragon S800, featuring Qualcomm’s Krait 400 @ 2.2GHz (Z Ultra ?).
          Single core @ shipping fqcy : 920
          Single @ 2 GHz : 840
          Multi : 2800
          Multi @ 2 GHz : 2545 (quad-core)

          • Roberto Tomás

            what a beast —that is comparable to a 1 year old laptop cpu

            Aug 16, 2013 Hewlett-Packard HP Pavilion dv6 Notebook PC AMD A10-4600M 2300 4 Windows 32-bit 920 2975
            Aug 16, 2013 Sony C6802 Qualcomm Qualcomm 2150 4 Android 32-bit 925 2742

            look at what it compares to in the PC world!!!

          • kascollet

            Yes. That’s impressing !
            I think GPU power is still lacking compared to laptop platforms but ARM SoCs are quickly catching up.

      • axarm

        This video shows why the scores are lower. The power management is keeping the S800 phone at about 1.7Ghz not the higher advertised 2+ Ghz. Given time Qualcomm and Samsung will fix this like they did in previous benchmark versions.

        https://www.facebook.com/photo.php?v=10151611410532913&set=vb.555897912&type=3&permPage=1

        • Roberto Tomás

          Im sorry, I cant view anything on facebook. maybe you could repost it somewhere that doesn’t content block, like youtube?

          But the scores I posted say that the CPU actually ran at 2.17Ghz (for the snapdragon 800) and at 2.0Ghz (for the Exynos 5410), not the typical 1.7Ghz or 1.6Ghz for those two devices. — Someone was smart enough to turn off power management I guess. :)

          • axarm

            The video showed a Samsung(unknown) and Sony C6802 both with S800s. He has the paid for Geekbench and it allows him to monitor the test. It shows the phones running nearer 1.7Ghz than the 2+Ghz. This makes sense because a 2.3Ghz will work in tablets but not phones. The Sony scored 889 2403. The Samsung 817 / 2157. He also has a Sony SGP312 S600 with a score of 499 / 1456. Again the S600 was throttled by the power system. GB2 and other benchmarks have been allowing overclocking that would not happen in the “real world”.

  • Roberto Tomás

    I’ve always preferred a suite of tests to all-in-ones like antutu. gfxbench trex (or latest, it is changing soon) + geekbench 3 + octane + disk io? + battery life? would be my choice.

    please note if you are thinking of using geekbench in place of antutu that antutu also includes I/O, you’ll need something to replace that. geekbench tests memory and cpu, but not disk (or, erm, nand here).

  • Joshua

    So I pay money, and in my case, actually see my outdated devices are really outdated. Logic ;)

  • simpleas

    Samsung ownage yet again!

  • Andrew T Roach

    iPhone 5

    Single core: 712
    Multi core: 1288