Thread 'Use performance of computers and smartphones'

Message boards : Questions and problems : Use performance of computers and smartphones
Message board moderation

To post messages, you must log in.

AuthorMessage
Ray

Send message
Joined: 26 Aug 22
Posts: 1
Message 109700 - Posted: 26 Aug 2022, 22:14:59 UTC

Hello.
I tried BOINC on my unused smartphone (Samsung Galaxy S7).
I saw the result like below.

https://universeathome.pl/universe/host_app_versions.php?hostid=631467

In the result page, the average processing rate is just 0.22 GFLOPS.
As I know, the computing power of Samsung Galaxy S7 is around 260 GFlops.

Who can explain why real processing rate is 0.22 GFlops instead of 260 GFlops?
If you explain about this matter, I will grateful you.

Thank you.


[/img]
ID: 109700 · Report as offensive
ProfileJord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15566
Netherlands
Message 109702 - Posted: 26 Aug 2022, 23:01:31 UTC - in response to Message 109700.  

Well, your CPU has 8 cores, 4 of which are the so-called LITTLE cores which are used for more mundane calculations, and which BOINC will use. It has 4 more big cores (from big.LITTLE) which are faster and used only by Android. When you allow BOINC to use 8 cores, it runs the tasks on 4 cores, on the slower LITTLE ones and double loads tasks onto those cores (2 per core).

Running 2 tasks per core slows things down already.

Then I looked up benchmarks of the Galaxy S7 and see for floating point operations that it runs between 1.43 and 4.08 GFlops on a single core (there are multicore numbers there too), depending on which calculation was run. You can compare it to those numbers as each Universe BHspin v2 0.21 arm-android-linux-gnu task should run on a single core only.

So, set preferences correctly to use just 4 cores, and rerun the calculations and see if they differ much from the previous double loaded values. Running single tasks should run much faster.
ID: 109702 · Report as offensive

Message boards : Questions and problems : Use performance of computers and smartphones

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.