Machine Learning Inference Performance

AIMark 3

AIMark makes use of various vendor SDKs to implement the benchmarks. This means that the end-results really aren’t a proper apples-to-apples comparison, however it represents an approach that actually will be used by some vendors in their in-house applications or even some rare third-party app.

鲁大师 / Master Lu - AIMark 3 - InceptionV3 鲁大师 / Master Lu - AIMark 3 - ResNet34 鲁大师 / Master Lu - AIMark 3 - MobileNet-SSD 鲁大师 / Master Lu - AIMark 3 - DeepLabV3

Unfortunately for the Black Shark 2, the devices lacked the proper drivers to properly run AIMark, and the benchmark repeatedly crashed upon starting the benchmarks. We had the same issue on the OnePlus 7 Pro, pointing out to some software incompatibility.

AIBenchmark 3

AIBenchmark takes a different approach to benchmarking. Here the test uses the hardware agnostic NNAPI in order to accelerate inferencing, meaning it doesn’t use any proprietary aspects of a given hardware except for the drivers that actually enable the abstraction between software and hardware. This approach is more apples-to-apples, but also means that we can’t do cross-platform comparisons, like testing iPhones.

We’re publishing one-shot inference times. The difference here to sustained performance inference times is that these figures have more timing overhead on the part of the software stack from initialising the test to actually executing the computation.

AIBenchmark 3 - NNAPI CPU

We’re segregating the AIBenchmark scores by execution block, starting off with the regular CPU workloads that simply use TensorFlow libraries and do not attempt to run on specialized hardware blocks.

AIBenchmark 3 - 1 - The Life - CPU/FP AIBenchmark 3 - 2 - Zoo - CPU/FP AIBenchmark 3 - 3 - Pioneers - CPU/INT AIBenchmark 3 - 4 - Let's Play - CPU/FP AIBenchmark 3 - 7 - Ms. Universe - CPU/FP AIBenchmark 3 - 7 - Ms. Universe - CPU/INT AIBenchmark 3 - 8 - Blur iT! - CPU/FP

In AI Benchmark’s CPU workloads, the Black Shark 2 ends up with a bit of a odd spread of scores. In the shorter running benchmarks the phone is getting relatively average inference times, while on the longer running tests for some reason the BS2 falls behind other S855 devices. In fact it looks like the BS2 is landing as amongst the worse off S855 devices in the latter listed tests.

AIBenchmark 3 - NNAPI INT8

AIBenchmark 3 - 1 - The Life - INT8 AIBenchmark 3 - 2 - Zoo - Int8 AIBenchmark 3 - 3 - Pioneers - INT8 AIBenchmark 3 - 5 - Masterpiece - INT8 AIBenchmark 3 - 6 - Cartoons - INT8

AIBenchmark 3 - NNAPI FP16

AIBenchmark 3 - 1 - The Life - FP16 AIBenchmark 3 - 2 - Zoo - FP16 AIBenchmark 3 - 3 - Pioneers - FP16 AIBenchmark 3 - 5 - Masterpiece - FP16 AIBenchmark 3 - 6 - Cartoons - FP16 AIBenchmark 3 - 9 - Berlin Driving - FP16 AIBenchmark 3 - 10 - WESPE-dn - FP16

AIBenchmark 3 - NNAPI FP32

AIBenchmark 3 - 10 - WESPE-dn - FP32

In the INT8, FP16 and FP32 accelerated tests which make use of acceleration blocks such as the Hexagon DSP and the GPU, we see the Black Shark 2 perform very well and in line with other Snapdragon 855 devices.

Overall, the Black Shark 2 is a good performer in the machine learning inferencing benchmarks, but like other devices, it’s not quite the very best in every regard, pointing out that the vendor could have improved upon its performance by keeping the software stack more up to date with what Qualcomm is offering, a widespread issue that I expect to persist over the next years as the ecosystem quickly evolves.

System Performance GPU Performance - Worst of S855
POST A COMMENT

63 Comments

View All Comments

  • s.yu - Thursday, September 26, 2019 - link

    I'm quite aware of people who never go for the achievements. When my cousin plays Kingdom Rush he DOESN'T EVEN GO FOR ALL 3 STARS.
    The horror! It's like eating an apple with two or three random bites then throwing away all the rest!
    Reply
  • StormyParis - Wednesday, September 25, 2019 - link

    Reviewing stuff as part of a for-pay arrangement must require very high-level mental, ethics, ... jiu-jitsu. I'm not sure how long the partnership w/Qualcomm will last, but, as a reader, thanks for the heads up. Reply
  • IUU - Wednesday, September 25, 2019 - link

    They are trying to imitate Apple. Too much fluff, so people believe they buy a superior device(it is the price too).
    So they hope they can make some profit out of thin air.
    I hope you should be critical of the iphones in this fashion as well.
    They allocate a bigger amount of transistors to the faster cores , so they
    can claim supremacy, no matter the fact that on general cpu performance
    they are about the same.
    No matter that their gpu is anemic and nowhere near snapdragons.
    Reply
  • melgross - Thursday, September 26, 2019 - link

    Exactly what fluff are you talking about?

    You really are deliberately ignorant.
    Reply
  • edsib1 - Wednesday, September 25, 2019 - link

    Another phone review where the benchmarks are all wrong. Put the phone in game mode - it is easily in the top 5 855 based phones in terms of gaming performance. Reply
  • cfenton - Wednesday, September 25, 2019 - link

    But is that because it cheats and runs without any thermal limitations when it's in game mode? Personally, I don't want my phone getting hot enough to be uncomfortable to hold. Reply
  • edsib1 - Wednesday, September 25, 2019 - link

    But it doesnt get hot. They run at full tilt and dont throttle.

    Read the other reviews on the internet about these phones. It is Anandtech that is out of step with the other sites.

    The gaming phones are basically running in saver mode - unless you turn game mode on.

    Exactly the same problem with the review of the Oppo Reno 10x zoom....
    Reply
  • Andrei Frumusanu - Wednesday, September 25, 2019 - link

    > But it doesnt get hot. They run at full tilt and dont throttle.

    I already demonstrated that's not what happens.

    > Exactly the same problem with the review of the Oppo Reno 10x zoom....

    What's the problem? That phone didn't throttle. It had a performance mode but that's essentially just a cheating button and running all frequencies pegged at max.
    Reply
  • edsib1 - Thursday, September 26, 2019 - link

    Game mode is not a cheating mode if the phone can maintain that speed. If you run CPU throttle on the Oppo for over 10mins the phone maintains a very high score with game mode engaged (around 15% higher than without).

    The Exynos 9820 on the other hard throttles hard after a couple of minutes losing about 25% performance.

    So why would the game mode that doesnt throttle be more of a cheat mode than the 9820 which throttles hard?
    Reply
  • edsib1 - Thursday, September 26, 2019 - link

    Reviewing a gaming phone without using gaming mode is like testing an Audi RS6 without putting it into sport mode - pointless. Reply

Log in

Don't have an account? Sign up now