25. November 2017 at 10:14 #5369
Thread about benchmarking magic dosbox on various android devices. From reports we can build table and people can make decision if their priority on new phone is emulation.
- download Benchmark apk
- best is if you can restart device before benchmark and kill tasks which can affect benchmark
- install and run Benchmark application to the end 3 times
- You will see “timed 2134 gameticks in 1100 realticks”. Please write which device you have and average result of second value(realticks) (which will be always different).
Please note, Benchmark application is not set for absolutly best performance, but for most common settings which usually works in all games. It would be possible set settings for better performance up to 5-10%. Reported performance can be affected by installed background applications like antivirus which decrease overall performance or with current android version.
Lesser score means faster device.
Device Cpu Score Tweak ASUS Zenfone 2 ZE551ML quad core Intel Atom Z3580 @ 2.3 Ghz 1311 Performance mode GPD XD Plus MediaTek MT8176 1144 rooted Honor 5c Kirin 650 1810 Honor 9 Kirin 960 803 After Oreo 8.0 upgrade (before: 920) Honor 10 Lite Kirin 710 1003 Honor Play Kirin 970 843 Oreo 8.1; 831 with Game Acceleration Huawei Mate 9 Kirin 960 729 Huawei Mate 10 Pro Kirin 970 834 Huawei Mate 20 X Kirin 980 792 591 (performance setting) Huawei P8 Lite 2017 Kirin 655 1816 Huawei P9 Lite Kirin 650 1830 Lenovo K8 Note Mediatek Helio X23 1876 Motorola Moto Z Play Snapdragon 625 1867 Android 8.0 w/o root Nokia 8 Snapdragon 835 775 after Pie update (847 on Oreo) Nvidia shield tablet 1130 / 1004 Default settings / OC @2.5GHz Nvidia shield portable Nvidia Tegra 4 1400 Nvidia shield TV (2015) 980 Oneplus 3T Snapdragon 821 2017 Oppo R17 Snapdragon 670 836 Samsung Galaxy A5 (2015) Snapdragon 410 4115 same with performance governor Samsung Galaxy A5 (2017) Exynos 7880 2048 Samsung Galaxy A50 Exynos 9610 817 Samsung Galaxy Alpha Exynos 5430 1203 score 1024 with root + performance governor Samsung Galaxy J5 (2017) Exynos 7870 2220 Samsung Galaxy Note 2 Exynos 4412 1820 Samsung Galaxy Note 4 Snapdragon 805 2180 1380 with root, performance governor, limited to 2 cores (to prevent overheating) Samsung Galaxy Note 8 (N9500) Qualcomm 835 690-870 normal mode (interactive governor) = ~870, performance mode (performance governor) = ~690 Samsung Galaxy S7 Edge Exynos 8890 648 (MM) / 707 (Nougat) root + performance governor Samsung Galaxy S7 Exynos 8890 870 normal mode (not rooted) Samsung Galaxy S9+ Exynos 9810 763 used to get 680 with stock governor, looks like Sammy’s cpu management is completely messed up. Samsung Galaxy S10+ Snapdragon 855 514 Samsung Galaxy S10e Exynos 9820 614 Samsung Galaxy Tab E T560 Spreadtrum SC8830A 3650 Sony Xperia Go NovaThor U8500 3577 Xiaomi Redmi 4X Snapdragon 435 2450 Xiaomi Redmi 5 Plus Snapdragon 625 1834 stock governor, unrooted Xiaomi mi6 Snapdragon 835 836 Xiaomi Pocophone F1 Snapdragon 845 627 miui 10 v8.11.23 beta (android 9) – 568 realtics Xiaomi Redmi Note 3 Snapdragon 650 1114 Rooted, Android Oreo 8.1 (Nitrogen OS), Default interactive governor Xiaomi Redmi Note 4 Snapdragon 825 180325. November 2017 at 15:40 #5383
Samsung Galaxy Alpha (Exynos 5430) – 1203 on average.25. November 2017 at 16:52 #5384
Xiaomi mi6 (snapdragon 835) – 836 Realtics
Proud of this beast :).25. November 2017 at 17:07 #5385
Wtf :d, so its truth, snapdragon is fixed 🙂26. November 2017 at 19:43 #5391
Just a note – Honor 9 has Kirin 960, not 930.26. November 2017 at 21:20 #5392
Average 2017 with Oneplus 3T (snapdragon 821).27. November 2017 at 10:38 #5398
Samsung Galaxy S7 edge (Exynos 8890): 64827. November 2017 at 11:20 #5399
Any overclock and/or performance tweaks?27. November 2017 at 11:33 #5400
Nope, the Exynos is that fast out of the box. With default governor, it can go up to 750-800 realtics occasionally, because it may downclock during the benchmark (not enough load on the cpu ?). For stable, fast performance, I usually set performance governor. The cpu doesn’t overheat either, so it’s really great for Dosbox.27. November 2017 at 11:56 #5401
It will be interesting to see how the 9810 will score when the S9 comes out then. Some leaks suggest a pretty big jump over the 8895.
Edit: Just tried running it with performance governor and got 1024 on my Alpha. I think we should start including it in the tweaks column considering most people probably wouldn’t bother with rooting and switching governors and the difference it makes is not negligible at all.
Result: Samsung Galaxy Alpha – Exynos 5430 – 1024 – Performance governor.28. November 2017 at 0:30 #5409
Somehow the Exynos chips are insanely fast for Dosbox. My old Note 2 (Exynos 4412) scored 1800 realtics, imagine my disappointment when I switched to Snapdragon Note 4, and got 2100.
I managed to get the Note 4 to ~1400 by rooting, using performance governor, and shutting off two of the four cores to prevent overheating. But that was the absolute maximum.28. November 2017 at 0:34 #5410
Here’s an extensive list of timedemo results from real PCs. This will give you a rough idea of what the “real-world” equivalent to Dosbox on your phone is:
https://www.complang.tuwien.ac.at/misc/doombench.html28. November 2017 at 9:37 #5411
Hmm, ok, I am a bit confused, so governor is not default cpu setting? How should I fix table? I like to have in “Score” column value which can be reached without rooting (else I want to have message in “tweak” column what I did)28. November 2017 at 10:50 #5412
I think @acrosstheline1 means that his phone scores 1024 realtics with performance governor, but the score should be added to the tweaks section. Because only a few people actually root and tweak their phones like that.
Maybe you could add another line with the score, and put “root + performance governor” into the tweaks section.28. November 2017 at 10:57 #5413
It (in this case) allows the cpu to keep higher frequency.
Maybe keep both the results if there are more than one with different settings? So in my case – the one that is already in and another result with the tweak mentioned. That way people would be able to see the performance gain too and decide if it is worth it for them.
EDIT: Yeah, what Alex said.
You must be logged in to reply to this topic.