Apple’s A15 Bionic Scores Better than Google’s Tensor Chip in ML Benchmarks
The benchmark scores of theGoogle Tensor chiphave started to show up online as early adopters and the media have received their Pixel 6 and 6 Pro units. Although Google is highlighting the ML capabilities of the Tensor chip, synthetic benchmarks like GeekBench ML are not showing promising results. The Apple A15 Bionic chip seems to be way powerful than Google’s first in-house chip on the machine learning front as well.
Apple A15 Bionic vs Google Tensor Chip: GeekBench ML Scores
As first highlighted by Max Weinbach on Twitter andcorroboratedbyNotebookcheck,Apple’s A15 Bionic shows better results on GeekBench MLacross TensorFlow Lite CPU, GPU, and ML tests. While Tensor scored 313, 1359, and 1722 in CPU, GPU, and NNAPI, A15 Bionic scored 945, 2061, and 2212 in CPU, GPU, and Core ML categories. That’s a clear victory for A15 Bionic across the board, and well, Google Tensor is not even close.Source:Max Weinbach on X (formerly Twitter)
While the scores might be better on A15 Bionic, itdoesn’t mean Tensor’s ML capabilities are not good. In fact, Google’s objective with Tensor was to design a chip that works well with its own ML models. And going by the heterogeneous computing approach Google is taking with Tensor, it’s evident that we shouldn’t entirely rely on benchmarks here.
“We ensured different subsystems inside Tensor work really well together, rather than optimizing individual elements for peak speeds. Peak CPU and GPU speeds look great in benchmarks, but they don’t always reflect real-world user experience,”said Google Product Manager Monika Gupta at the Pixel Fall launch event keynote.
If you are interested to know how Google Tensor compares to chipsets found on other Android flagships, Anshel Sag from Moor Insights & Strategy has compared the Pixel 6 series Geekbench scores against S21 Ultra, Redmagic 6S Pro, ROG Phone 5, Snapdragon Insider Phone, and Surface Duo 2. You can check that out below:As you can see here, not many surprises on the single-core and multi-core CPU@geekbenchbenchmarks, however in the#AIbenchmark I used the NNAPI test to measure performance to make it as fair as possible and I was expecting#Tensor’s AI performance to be higherpic.twitter.com/mkBHE9mWiP— Anshel Sag (@anshelsag)October 25, 2021
As you can see here, not many surprises on the single-core and multi-core CPU@geekbenchbenchmarks, however in the#AIbenchmark I used the NNAPI test to measure performance to make it as fair as possible and I was expecting#Tensor’s AI performance to be higherpic.twitter.com/mkBHE9mWiP— Anshel Sag (@anshelsag)October 25, 2021
Now, apart from the ML benchmark scores, this isn’t a pretty sight either. Google’s first chipset is powerful but it isn’t quite comparable to the Snapdragon 888 chip on most Android flagship phones nowadays. We have also comparedGoogle Tensor against Snapdragon 888 and A15 Bionicin our linked article if you’re interested.
Subin B
Subin writes about consumer tech, software, and security. He secretly misses the headphone jack while pretending he’s better off with the wireless freedom.
Add new comment
Name
Email ID
Δ
01
02
03
04
05
