Peak performance is an important benchmark number, as is sustainable performance. A third useful number is performance per watt, how much can be achieved for a unit of power. In this video I look at the performance per watt of the Cortex-A510, the Cortex-A715, and the Cortex-X3 CPU cores in the Google Tensor G3.
—
Twitter: https://twitter.com/garyexplains
Instagram: https://www.instagram.com/garyexplains/
#garyexplains
Big question is are the commands in the chipsets available on less or equal to the different chips. If an add fi takes 3 cycles on the A510 and 4 on the A7… Then you might be better of the near.future
Very useful data … thanks!
What a great video, love the information provided and the work you put into these videos!
Curious as to why x3 speeds up beyond 1.8W or so.
Thank you for the video, Gary! This is, in my opinion, the BEST video when it comes to talking about any CPU! I am currently using Ryzen 5 2400G, and I am very happy with its CPU power! The real question is, however, how much the top ARM mobile chips can compare to it, using just 5-6 Watts for the whole system! My CPU can reach 90 Watt usage after load! Power efficiency is where the real game is!!!
Amazing job i wait for 8gen3
love to see the 8 Gen 3
Thanks Prof Gary!
Please include at least one snapdragon chip with the same cores for comparison next time
Great video thanks Gary
Seconded!
The 8 Gen 3 is astounding with a performance streets ahead of its predecessor.
You should do the same with the 8g3 and current Exynos, because these Tensor numbers are kinda meaningless (but nonetheless really interesting) unless compared to other SoCs
I miss Speedtest G
Soo who can summarize it for me? What does it mean?
Excellent measurements!
But the freq/power plot would have been better with the xy axes swapped
I hope this can be compared to other SOC's like Exynos, SnapDragon or Dimensity.
Gary will you make a video on the 7s Gen 3 and 8s Gen 3?
This is most useful. I have just received my first Pixel phone, a Pixel 8 and this explains a lot. Thank you.