Just for fun i figured id make a large list of what our phones are running at different markers.
All the apps used to test the phone are free!
If users are interested please put what Rom and/or Kernel you are running and a list of statistics.
Fps2D- (FPS)
Linpack- (MFLOPS)
Quadrant
Neocore
Optional:
Battery Life
Set CPU settings
Nice idea
I'm on cm6 with snap(my first day) and I'm loving it
I get 1555 on quadrant
27.4 on neocore
~46mflops on linpack
And 52-53 fps2d with an stdev of around 3
I have a 002 Nova oc'd at 1.19ghz my t2 is set at 380
Sent from my PC36100 using XDA App
Fps2D- AVG 50-53 FPS
Linpack- 38-45 MFLOPS
Quadrant- 1182
Neocore- 27.4fps
Battery Life- 8hrs
Set CPU settings- Set to performance at 1113mhz
Rom- CM6 Froyo 2.2
So far the rom is working great
cyan mod 6 is giving me nearly 40 on linpack
Too bad, seems like anything not in the development forum gets no reads.
haha, case and point no replies since moved.
Why does the new linpack update run slower.... when i click run benchmark it takes like2 and a half seconds...and benchmarks slower like 32-35 instead of almost 40. Wtf
Instead of doing a single run, it now does multiple and averages them giving a more accurate score.
Mhz vs Quadscore on [beta] desire hd2 #Update
hey here same interesting tests i did
if i locked my phone in power saving mode u get
CPU and Score
Mhz ScoreA CPU Mem IOps Linpack
256Mhz 900 1750 525 1600 11
384Mhz 1050 2580 618 2000 16
460Mhz 1350 3000 990 2600 19
500Mhz 1600 3250 1050 2800 20
768Mhz 2200 4600 1120 3500 29
1000Mhz 2500 5400 1120 4150 35.5
1200Mhz 2700 6200 1120 4380 40
maybe there could be another mode that will use
500mhz most of the time [for normal using or one app using Cpu]
and when playing or using heavy 1000Mhz
[Dev]
i see that memory score is not linear
maybe could fixed by our greatest devers
I hope u do understand that these results are total bullcrap.
The newest dual core chips dont even get 2.7k scores in that benchmark.
Something in the desire HD build farks up the scores simple as that.
If you look at the quadrant advanced score, you'll notice it only gets those scores because of the I/O score. The processor one is about the same or a little more than teh Nexux 2.2, so if you underclock to 500 mhz you may get a total of 1600, but your processor score will suck badly and your phone will lag and really suck.
Keep your max at 998 and your min at 245 and use smartass and you should be fine.
orangekid said:
If you look at the quadrant advanced score, you'll notice it only gets those scores because of the I/O score. The processor one is about the same or a little more than teh Nexux 2.2, so if you underclock to 500 mhz you may get a total of 1600mhz, but your processor score will suck badly and your phone will lag and really suck.
Keep your max at 998 and your min at 245 and use smartass and you should be fine.
Click to expand...
Click to collapse
Agree.
Sent from my HTC HD2
orangekid said:
Keep your max at 998 and your min at 245 and use smartass and you should be fine.
Click to expand...
Click to collapse
+1 for this.
My HD2 has been running smoothly with 245-998MHz smartass for weeks now, battery life has been great too.
i update it with more info check post1
i see that memory score is not linear
maybe could fixed by our greatest devers
Quadrant scores are useless. RUn it, then run it again, you'll get different results.
I don't use SetCPU at all, it's just not needed.
According to my Quad Test im getting a total of 2196 with 245-998mhz with smartass on SetCPU my CPU: 4450 Mem: 1127 I/O: 4716 2D: 273 3D: 412 Work that one out
Spreadsheets for CPU Governors and I/O Schedulers
GOVERNOR RESULTS
I/O SCHEDULERS
Summary of the Results
This is a summary of the six most commonly used governors, listed in order of performance.
Best Performing
#1 - Performance
--- : Use Noop or Deadline
--- : Uses a lot more battery
#2 - SmartassV2
--- : Use Noop or SIO
--- : Uses a little more battery
#3 - LulzactiveV2
--- : Use Deadline or Noop
--- : Uses a lot more battery
#4 - Lazy
--- : Use Deadline or CFQ
--- : Helps save battery as long as SOMF is not enabled
#5 - Ondemand
--- : Use Noop or Deadline
--- : Helps save battery
#6 - Conservative
--- : Use CFQ or Noop
--- : Helps save battery
Check my summary in this thread for more info about how to save battery.
Thanks to all the developers.
Caveats
The most important thing to remember is that the testing hardware is a nexus s. Although I believe they are essentially the same device, there may be differences in how the galaxy s hardware affects performance. Do some experimenting yourself, and if it feels right, go with it!
Reserved 1
COMING SOON! --- Battery benchmark << Click for details. New, embarrassing VIDEO of the setup, with me mumbling about it! I'm aiming to find out which governor has the best battery life.
Thank you so much for this, I had been struggling to get a game to run smoothly so switched to SmartassV2 and it fixed my problems!
One thing to understand here is that benchmarks hardly represent real life situations.
Governors for one thing cannot be benchmarked as governors are just an algorithm of how one would manage the CPU frequency. The closer the benchmark is to "performance", the worse off your battery life is (unless the governor itself has been finely tuned to provide the appropriate frequency under certain amounts of load).
Also, it is hard to benchmark I/O schedulers so much as background activity significantly affects the the I/O speed. Schedulers such as NOOP performs well in benchmarks here as NOOP doesn't have a starvation detector. NOOP will perform any and all I/O activities as they come, meaning, a long benchmark with largely small amounts of background activity will bring NOOP to the top.
Other schedulers such as SIO/Deadline performs a little worse than NOOP due to their nature of applying starvation counters on I/O so that long I/O activites are pre-empted by other more important I/O activities (such as system based I/O activities). However, NOOP/Deadline/SIO still performs the best as they don't have heavy algorithms to track penalties and spindle location (for rotational disks)
pikachu01 said:
One thing to understand here is that benchmarks hardly represent real life situations.
Governors for one thing cannot be benchmarked as governors are just an algorithm of how one would manage the CPU frequency. The closer the benchmark is to "performance", the worse off your battery life is (unless the governor itself has been finely tuned to provide the appropriate frequency under certain amounts of load).
Click to expand...
Click to collapse
Have you seen the spreadsheet on Governors? All I'm doing is assessing their impact on CPU IO RAM and Graphics. Battery life impact will be published today, have you seen the video? I'm using proper equipment, not an inaccurate app.
pikachu01 said:
Also, it is hard to benchmark I/O schedulers so much as background activity significantly affects the the I/O speed. Schedulers such as NOOP performs well in benchmarks here as NOOP doesn't have a starvation detector. NOOP will perform any and all I/O activities as they come, meaning, a long benchmark with largely small amounts of background activity will bring NOOP to the top.
Click to expand...
Click to collapse
Test conditions are equivalent for all benchmarks, have you looked at the IO Scheduler sheet? I run each test 10 times, calculate the mean and standard deviation, and produce a geometric mean for a result 3 standard deviations below the mean (i.e. 99.7% of all results will be above that point) then look at the median of the data to see how far it is from the mean (should be very close for normally distributed data) - most results are with 1% of each other, then compensate by splitting the difference between the median and mean to create an error margin, for most results of about 0.5%
pikachu01 said:
Other schedulers such as SIO/Deadline performs a little worse than NOOP due to their nature of applying starvation counters on I/O so that long I/O activites are pre-empted by other more important I/O activities (such as system based I/O activities). However, NOOP/Deadline/SIO still performs the best as they don't have heavy algorithms to track penalties and spindle location (for rotational disks)
Click to expand...
Click to collapse
Check the sheet. I've shown the most likely IO winner depends on which governor is in use.
Battery benchmark results are up! See first post for the link!
UPDATE: Steve had a rethink of his Conservative governor parameters, and I've updated the table to include this. Check column L.
Lower scores are natural with conservative. This time, with the new parameters, the scores are even lower. This indicates that the new conservative is throttling the frequencies more aggressively, but it should also be noted that this governor feels much more responsive than the previous incarnation.
Note - the IO score shows a great improvement, but this is due only to a boost in database IO through the recent FSYNC patch. Make sure you use stable voltages to avoid reboots and potential data loss. (No benefit to UV anyway - check the results of my battery study, see link above.)
bedalus said:
UPDATE: Steve had a rethink of his Conservative governor parameters, and I've updated the table to include this. Check column L.
Lower scores are natural with conservative. This time, with the new parameters, the scores are even lower. This indicates that the new conservative is throttling the frequencies more aggressively, but it should also be noted that this governor feels much more responsive than the previous incarnation.
Note - the IO score shows a great improvement, but this is due only to a boost in database IO through the recent FSYNC patch. Make sure you use stable voltages to avoid reboots and potential data loss. (No benefit to UV anyway - check the results of my battery study, see link above.)
Click to expand...
Click to collapse
now it would be nice to know steves parameters to test them on our sgs
Skkip said:
now it would be nice to know steves parameters to test them on our sgs
Click to expand...
Click to collapse
Okay!
Sampling rate: 40000
Up thresh: 80
Down: 20
Ignore nice: 0
Freq step: 5
Thanks for looking!
bedalus said:
Okay!
Sampling rate: 40000
Up thresh: 80
Down: 20
Ignore nice: 0
Freq step: 5
Thanks for looking!
Click to expand...
Click to collapse
thx, its nearly the same as stock conservative with lowered sampling rate. on cm9 stock there's a min sampling rate of 78124 so i can't test it hope someone knows a kernel where i can test it...
UPDATE: First post edited. Now summarises all findings. Also, makes it easier to read if you are using the XDA app
People pls try out this nd tell me abt perfo ...use governor luzactive ...scheduler vr ....in nstoolls edit up sampling rate to 24000 nd down sampling to 10k nd set max screen off freq as 200 nd pump dow step as 2 ....m able to get 2 days of batt out of dis config!
Personally I find Lulzactive to give the best battery life at idle but would love to be corrected, anyone found better?
thx, confirms my thoughts that smartassv2/sio is the best combination
Sent using TCP/IP
NI3K said:
Personally I find Lulzactive to give the best battery life at idle but would love to be corrected, anyone found better?
Click to expand...
Click to collapse
I suppose it might depend on the settings, but when i measured the amps, it was clearly draining a lot more. This was playing music with the screen off.
kernels ; battery ; ROM ; gov/sched
Now with summaries in the first posts. Convenient for XDA app users!
using poweramp dsnt eat our batt dat much !
ddrulze said:
using poweramp dsnt eat our batt dat much !
Click to expand...
Click to collapse
hahaha no... I wasn't using an app, I was measuring power (amps = Volts/Resistance) It is a measure of the rate of power consumption.
while some of this info can be found, I'm going to try to be as thorough as possible....
I'm posting this so you don't have to
speedtest .net
shows I'm running at full speed
45 Mbs down
6 Mbs up
as to be expected, but trying to be thorough
CPU-ID (from CPU-z)
this link shows most of the info I'll type the rest out myself
http://valid.x86.fr/a/e76tnv
missing info that might be helpful to someone
Scaling Govenor - hotplug
Model - KFFOWI (Full-ford)
board - Ford
Hardware - mt8127
RAM - 867 MB
Available RAM 297 (34%) --
"not running anything and have rebooted once, so FIRE OS is eating up the rest of the ram"
battery 4177 mV
thermals
mtksabb 120°F
mtktsbattery 102°F
mtkts_bts 110°F
mtktspmic 118°F
battery 103°F
mtktscpu 120°F
mtktsmt 110°F
virtual_sensor 100°F
well that's all for cpi-id.. i know some of the info is useless ....
NOTES
CPU's
i notice it stops cores 2 and 3
but i never notice it stopping 0 and 1
them 2 are always running at 598 MHz with just the screen on
take these for what you think they are worth
Quadrant benchmark
score 5736
in the video test no skips and a minimum (only once) of 58 a high of 63 FPS almost a constant 59-60 FPS
AnTuTu benchmark
score 23,603
for comparison
nexus 6 over - 50k
Galaxy Note 4 - 48k
3dMark
score 3332
My score, for 50 bucks "Priceless"
Ricci
a1yet said:
...
CPU's
i notice it stops cores 2 and 3
but i never notice it stopping 0 and 1
them 2 are always running at 598 MHz with just the screen on...
Click to expand...
Click to collapse
It's probably a big.LITTLE implementation.
https://en.wikipedia.org/wiki/ARM_big.LITTLE
http://www.mediatek.com/en/products/mobile-communications/tablet/mt8127/
Also, the cpu cannot be throttled below 598 MHz it seems.
---------- Post added at 03:12 AM ---------- Previous post was at 03:04 AM ----------
Some interesting info here
https://developer.amazon.com/appsan...ications/01-device-and-feature-specifications