@Chainfire or anyone that knows, Would you mind explaining what the individual test ran during benchmark are testing as far as hardware/software (not for everyone but just if you were to ignore the native and java and then just have stuff like MINE, MDFLOPS etc.)? And an example of what each of these test would be related to in real world use?
That is a low score, something is not right with your phone. I have a HTC One (m7) and here is my score on an optimised, unfinished test version of a kernel.
Sent from my HTC One using Tapatalk
That is a low score, something is not right with your phone. I have a HTC One (m7) and here is my score on an optimised, unfinished test version of a kernel.
Sent from my HTC One using Tapatalk
@Chainfire or anyone that knows, Would you mind explaining what the individual test ran during benchmark are testing as far as hardware/software (not for everyone but just if you were to ignore the native and java and then just have stuff like MINE, MDFLOPS etc.)? And an example of what each of these test would be related to in real world use?
That is a low score, something is not right with your phone. I have a HTC One (m7) and here is my score on an optimised, unfinished test version of a kernel.
Sent from my HTC One using Tapatalk
Me???
Sent from my One X using Tapatalk
My bad, I thought you were on a SGS4
That is a great score for the Desire HD I imagine Sent from my HTC One using Tapatalk
My bad, I thought you were on a SGS4
Sent from my HTC One using Tapatalk
Yes... No. Lol
Sent from my One X using Tapatalk
CF-Bench / Native Disk Write - accuracy
Hei,
browsing the scores through and doing some trialling of my own as well, there's an issue I think with the Native disk write score. It seems to be all over the place and several devices report scores between 4-12 points or so. I do not think those devices are that slow. Could the benchmark be explained for that part, what does it do and how does it measure the performance?
Hei,
browsing the scores through and doing some trialling of my own as well, there's an issue I think with the Native disk write score. It seems to be all over the place and several devices report scores between 4-12 points or so. I do not think those devices are that slow. Could the benchmark be explained for that part, what does it do and how does it measure the performance?
I did some more trials on this and turns out it's depending on some of the tunables, most notably in my case it is highly dependent on the background dirty bytes. I.e. if I set that to be a of a high value I get good scores - so seems this benchmark is not doing a flush() on the file operations and thus it's really not measuring the true file system speed to the actual end memory/disk device. It's testing more of what happens in the filesystem caches.
You can try it out yourself, all you need is adb connection.
cd /proc/sys/vm
echo 5 >dirty_background_ratio
or you can also set the absolute byte value, too.
echo 10485760 >dirty_background_bytes
the original values depend on what your version of Linux/Android has written originally to those. You set either the ratio or bytes, not both at the same time.
Versus my OnePlus One with one of the latest and fastest chip on the market and 3 gigs of RAM...
And that's why you shouldn't trust benchmarks
Verstuurd vanaf mijn One A0001 met Tapatalk
Attachments
The real phone is a Jiayu G4S with a Miui V7 ported ROM.
The images shows an Xiaomi HM Note 1 W , but the results are from the Jiayu
The images shows an Xiaomi HM Note 1 W , but the results are from the Jiayu
Attachments
Understanding CF bench Scores
Hello, I am running the CF benchmark on the Android Devices which I intend to profile for performance debugging.
But I am bit confused on how are these score calculated for the native and Java tests.
I am seeing some ambiguity and inconsistent results with the test case of Java MIPS in specific.
So @Chainfire can help me to understand how do these scores are calculated to end up with those values in the result screen ?
Is it time that it takes to run that test case or what is it based on ?
Hello, I am running the CF benchmark on the Android Devices which I intend to profile for performance debugging.
But I am bit confused on how are these score calculated for the native and Java tests.
I am seeing some ambiguity and inconsistent results with the test case of Java MIPS in specific.
So @Chainfire can help me to understand how do these scores are calculated to end up with those values in the result screen ?
Is it time that it takes to run that test case or what is it based on ?
Similar threads
Top Liked Posts
-
There are no posts matching your filters.
-
73About
CF-Bench is (mainly) CPU and memory benchmark tool specifically designed to be able to handle multi-core devices, produce a fairly stable score, and test both native as well managed code performance.
It tests specific device properties you do not regularly see tested by other benchmarks, and runs in a set timeframe.
It does produce some "final" scores, but as with every benchmark, you should take those with a grain of salt. It is simply not theoretically possible to produce a single number that accurately describes a device's performance.
Why ?
I know, there are lots of other benchmarks. But they don't give me the numbers I'm personally interested in comparing. Few really support multi-core or test both native and Java code. This benchmark is also supposed to give a more "stable" score compared to some others.
Current scores
Here are a number of scores reached by tested devices:
Samsung Galaxy S2 @ 2x1200 - 6498
Samsung Galaxy Tab 10.1v @ 2x1000 - 5188
HTC Sensation @ 2x1188 - 5174
Motorola Atrix @ 2x1000 - 5136
HTC Thunderbolt @ 1920 - 4663
Motorola Droid X @ 1200 - 2792
Samsung Galaxy Tab 7" @ 1000 - 2401
Samsung Galaxy S @ 1000 - 2205
HTC HD2 @ 1536 - 2041
Samsung Apollo @ 667 - 488
HTC Hero @ 748 - 359
HTC Dream (G1) @ 528 - 191
Of course, you should really keep in mind that the three different scores listed at the bottom of the app are pretty meaningless
Download
Market link (mobile): market://details?id=eu.chainfire.cfbench
Market link (desktop): https://market.android.com/details?id=eu.chainfire.cfbench
2Understanding CF bench Scores
Hello, I am running the CF benchmark on the Android Devices which I intend to profile for performance debugging.
But I am bit confused on how are these score calculated for the native and Java tests.
I am seeing some ambiguity and inconsistent results with the test case of Java MIPS in specific.
So @Chainfire can help me to understand how do these scores are calculated to end up with those values in the result screen ?
Is it time that it takes to run that test case or what is it based on ?2
New posts
-
-
[ROM][OFFICIAL][kebab][14] LineageOS 21
- Latest: stickmanChampion
