Develop Bugs

Fresh on Frozen Eclair 537mhz 10 mflops - 1125mhz 21 mflops

Just a general question. Fresh, dc, stock, or baked with even overclock highest mflops i got were 9 mflops. So I added Frozen Eclair JIT to Fresh and was able to get 21 mflops @ 1125mhz. I have balanced my phone now for battery life and performance. I am at 537mhz which is giving me 10mflops and 55 fps on netarchy kernel 3.7.1. My question, Is my phone faster @ 537mhz with jit than stock @ 998mhz? I believe that the higher the mflop the faster the phone right?

mflops, is millions of floating point operations. therefore more is better.

This is a VERY interesting question. I kind of doubt that it will receive a good answer, though. I don't know what the answer is.
I would assume that processor speed is king. But if a lower processor speed gets you higher Mflops than, say, my phone than I guess that Mflops are king. I don't know why that would be that way, though.

SWEET I'll take 10mflops @ 537mhz because everything runs flawlessly internet, streaming video, and phone. When I need to game I can turn up the processor to 998 or 1125 and get 17 to 21 mflops. Thanks. My daily driver for now will be 537mhz. More battery life and more power than 2.1 android.

pongoface said:
SWEET I'll take 10mflops @ 537mhz because everything runs flawlessly internet, streaming video, and phone. When I need to game I can turn up the processor to 998 or 1125 and get 17 to 21 mflops. Thanks. My daily driver for now will be 537mhz. More battery life and more power than 2.1 android.
Click to expand...
Click to collapse
**EDIT**Nevermind............i need to get rid of Sense ui

I had a hard time to getting rid of my raining animations too lol
Sent from my PC36100 using XDA App

Ok, Sense-lovers, what's the best kernel?

I'm back to the dark side for the time being, but I'd like a new kernel that unlocks the FPS without breaking my camera. Also, some overclocking would be nice.

I am actually in need of a kernel that I can flash that has the perf-lock disabled for underclocking but I don't care about the FPS at all. I just want my 8mp camera back.
Any ideas?

Evio 1.1 is great. As for the second post, use it with other kernels and setcpu for great ocing capability
Sent from my PC36100 using XDA App

logie said:
Evio 1.1 is great. As for the second post, use it with other kernels and setcpu for great ocing capability
Sent from my PC36100 using XDA App
Click to expand...
Click to collapse
Uh... EViO looks to be a ROM--not a kernel.

Kingxkernel bfs #3 or #4 best

One of kings kernels..
Sent from my HTC Evo 4G using XDA app.

Read my sig. I've found netarchy's to be a little more stable than king's. Same with Fresh. Snappy with 12-16 hours of moderate use on one charge.
Sent from my PC36100 using XDA App

sombdy said:
Read my sig. I've found netarchy's to be a little more stable than king's. Same with Fresh. Snappy with 12-16 hours of moderate use on one charge.
Sent from my PC36100 using XDA App
Click to expand...
Click to collapse
Ditto, get it here: http://forum.xda-developers.com/showthread.php?t=719763
I've been using it on my stock rooted evo 2.2 and its much snappier now. Unlocked FPS, Overclocking, and a hotspot with Wireless Tether Pre8

netarchy 4.1.8 cfs edition is great

The new htc kernal has been giving me phenomenal battery life, but has 30 fps cap and is not overclockable. I have been pretty happy with kingklick, but with this battery life, I am sticking with the htc. Just a matter of time until this one gets tweaked.

shortbus182 said:
The new htc kernal has been giving me phenomenal battery life, but has 30 fps cap and is not overclockable. I have been pretty happy with kingklick, but with this battery life, I am sticking with the htc. Just a matter of time until this one gets tweaked.
Click to expand...
Click to collapse
yea been hearing great things about the new kernel might jump back to the darkside and give it a test

Here are some numbers I got yesterday, using a brand-new Fresh 3.2.0.0 install:
Fresh Stock
Quadrant: 1173
FPS2D: 30/5.52
Linpack: 32.343
Netarchy 4.1.9.1 (No OC):
Quadrant: 1139
FPS2D: 53/4.73
Linpack: 29.302
Netarchy 4.1.9.1 (OC to 1152, Performance)
Quadrant: 1327
FPS2D: 54/3.18
Linpack: 26.69
King's #8 BFS (No OC):
Quadrant: 1069
FPS2D: 54/3.81
Linpack: 25.59
King's #8 BFS (OC to 1152, Performance):
Quadrant: 1486
FPS2D: 54/1.78
Linpack: 37.61
King's #8 BFS (OC to 1152, OnDemand):
Quadrant: 1553
FPS2D: 54/7.83
Linpack: 39.211
Make of that what you will. I still don't know what the advantage of older Kingx's kernels would be, but I never really looked into it. I'm using #8 BFS now, seems great so far.

ms79723 said:
netarchy 4.1.8 cfs edition is great
Click to expand...
Click to collapse
I'm on 4.1.9.1 BFS, Beta. I read this and decided for BFS: http://www.cs.unm.edu/~eschulte/data/bfs-v-cfs_groves-knockel-schulte.pdf

SysFailure0x5a said:
I'm on 4.1.9.1 BFS, Beta. I read this and decided for BFS: http://www.cs.unm.edu/~eschulte/data/bfs-v-cfs_groves-knockel-schulte.pdf
Click to expand...
Click to collapse
Wow, that was an awesome paper! I had absolutely no idea what BFS or CFS were until I read that. If I understand correctly (in a vastly-oversimplified way), CFS assigns each pending task an equal share of the CPU, while BFS decides which tasks will be completed faster and queues them accordingly. CFS improves turnaround time, but BFS lowers latency. With such an interactive system as a smartphone, BFS would seem to be the best choice.
Or maybe I'm completely missing it.

SilverZero said:
With such an interactive system as a smartphone, BFS would seem to be the best choice.
Click to expand...
Click to collapse
For user experience (as measured by interactivity), BFS would appear to be best choice.
However, I would like to see a study that measured power used vs unit of work done. Since CFS's turnaround time appears to be lower, this suggests it may be a more power-efficient kernel?

Netarchy vs Kingx vs New Fresh Stock Kernels
I don't have SetCPU, so my test were done just flashing the kernel and running the benchmarks. Haven't tested the battery yet, but I'm going to keep netarchy on for awhile and see how it does. I'm using the new Fresh 3.2 ROM btw.
Kingx #9 - No OC,
Quadrant - 1327
linpack -35
Netarchy 4.1.9 - No OC
Quadrant - 1405
linpack - 35
New Fresh Stock - based on the new HTC kernel
Quadrant - 1307
linpack - 33.5

With fresh 3.301 with mattedblue theme
No oc
Quadrant 1202
King#11 oc @ 1113
Quadrant 1327

[Q] Linpack scores

Hello, I was just wondering what scores you guys tend to be getting in linpack. I am trying to figure out the fastest set up for my phone. I am currently getting 2.55 with "not so super froyo" with the 10-12-2010 kernel and my phone overclocked to 520 Mhz with the GPU overclocked as well. I also have a lot of widgets open, I am confident I could hit a solid 3 if they were not open.

I've been doing quite a bit of testing with different builds so I usually don't use a lot of widgets. With no OC I usually hit 2.6 in Linpack. OC'd to 500Mhz it is around 3.3.....

I have changed the goals of this thread... I think we should now have a benchmark of everyone's scores that way we know what is slowing them down or making them faster.
Also... for some reason I do not believe I should be almost a full Mflop slower than you when I have more of an overclock so something is definately up with my phone. Maybe with this thread I can find out what is.

I haven't noticed much change between kernel releases. My phone with no OC has consistently got 2.6-2.7 on any of the .32 kernels and fresh froyo releases (2.2 and 2.2.1)

hi
yesterday i did the benchmarking
I have hit 3.95
I'm using the latest kernal .32
overclocked via Rogue Tolls to 561 MHZ

confused...
interesting results for me are (should say yoda)
if i use stock speeds i got 1.7, with 450 got 2.0-1.9 beyond this the score keeps gettin lower and lower till linpack FC when i reach the 550... so i guess i should stick to 450mhz with donut and .25 kernel, wonder if newer kerlens can get me to ur amazing results.

How are you overclocked? CPU through kernel has some pretty funky results. Also it's more unstable that way.

Nope, i use rogue tools only, and yes using atools to oc its not a good idea, i guess its because rogue tools oc the system AFTER the fs has been mounted, instead of ocing it from beggining like atools does, and with kernel .32 having this mounting issues i guess isnt too smart to rush it even more!

Voltage and overclocking on blazer 2.0

Hi i was wondering if anyone had some good voltages to undervolt the blazer 2.0 rom. What apps did you have success with and what voltages at what speeds? Also some ideas for blazer 2.0 rom with desperado kernel. Either or both would be great
Sent from my SPH-D710 using xda premium

Anyone help me with these dang voltages lol. I tried tegrak and setcpu and i tried just tegrak on the stock kernel that comes with the blazer 2.0 ROM. After little success i switched to desperado and tried the same methods and even gave voltage control a try. My only problem is anything over 1.4g i freeze up and reset regardless of now high i overvolt the speeds. I'm looking to run 1.6g smoothly and maybe even get some ideas from you guys on undervolting the lower speeds
Sent from my SPH-D710 using xda premium

Under/Over
I've been looking at the same thing.
For undervolting, I'm starting with -75mV across the board to run for a day or two to evaluate improved battery life.
HondaCop reported success with -100mV across the board with Tegrak:
http://forum.xda-developers.com/showpost.php?p=18259820&postcount=105
Suggestions from that thread include:
a) leave internal voltages alone
b) let the 300 sec "stability watch" finish before testing
c) don't set to apply at boot until you've run long enough to ensure stability
d) 1.6-1.65Mhz is likely the performance ceiling on our devices
e) run 'CF-Bench' as part of stability testing
Some settings I'm testing this AM:
1504Mhz @ 1375mV (then 1275mV)
1600Mhz @ 1375mV (then 1350mV)
1652Mhz @ 1450mV
1700Mhz @ 1500mV (then 1450mV)
I'm keeping a log to share results on the aforementioned thread.
Good luck!

That is wonderful advice! Thanks so much
Sent from my SPH-D710 using xda premium

Test Results
Glad to be of help!
Here are my test results; as they say YMMV.
Methodology
o used the same governor for each test (ondemand)
o tested with Quandrant, then Linpack, then CF-Bench
o tested with battery cover removed for physical CPU heat check
o gave ~min between benchmark tools
o used phone to update an app, pull up a page, watch a vid, etc
o re-ran benchmarks to confirm consistent results & stability
Config
Baseband/modem EL13 (big up Calkulin)
Kernel 2.6.37-7-SPH-D710.EL13-CL806382 (chris41g via sfhub auto root, danke!)
ROM/Build GB.EK02 (via OFA update)
Tegrak Overclock Ultimate 0.8.4.5 / driver 2.2.0.2
Tested Successfully
1452Mhz @ 1350mV
1452Mhz @ 1300mV : Quadrant 4741, LP 112, CF-B 13636/3910/7800
1500Mhz @ 1375mV
1500Mhz @ 1350mV : Quadrant 4888, Linpack 117, CF-B 13852/4045/7967
1560Mhz @ 1400mV : benchmarks falling off..
1560Mhz @ 1375mV : and going back up?! (nominal heat)
Tested Unsuccessfully
1700Mhz @ 1500mV : immediate freeze
1652Mhz @ 1450mV : froze in Linpack
1608Mhz @ 1425mV : Quadrant FC, froze during Linpack
1608Mhz @ 1400mV : Quadrant looked epic (5180), froze during Linpack
1600Mhz @ 1375mV : Quadrant > 4800, Linpack 120, froze during CF-B
1592Mhz @ 1375mV : Quadrant = 4988, froze during multiple Linpack runs
1564Mhz @ 1350mV
1504Mhz @ 1300mV
1500Mhz @ 1325mV : reboot during CF-B
Observations
There is a V sweet spot for each CPU speed.
Others have reported being able to go up to 1.7Ghz whereas my phone begins to become unstable ~1.6Ghz.
1.5Ghz appears all day long stable on my phone with minimal heat (mean of 4874 from three Quadrant runs, Linpack regularly >110, consistent CF-B results as noted above).
Next to Try
o put into airplane mode
o use the 'performance' governor and set the min = max for benchmark testing (with attention given to CPU temp!)
I set all the slower speeds at -75mV with the max at 1.5Ghz @ 1350mV. I'll report back after a couple days to see what battery life and long-term stability look like.

OK sounds good!
Sent from my SPH-D710 using xda premium

Safest undervolt

Hey was wondering if anyone knows off by hand the lowest we can undervolt our phones @ stock freq or what our stock voltages are? I undervolted mine recently and got amazing results but stability seems to vary, using overclock freqs @ 1.89 running @ 1.087 mv 1.83 @ 1.05mv 1.78 @ 1.0375 1.7 @ 1.025 mv I seem to be fully stable pulling 28905 scores on anutu, my friends s4 isn't even this fast was very impressed phones not even hot at all, but I really want to try to tweak the lower freq to help with battery life/temp

solution to the battery drain problem
Overclocking the CPU frequency will lead to battery drain as u know it yourself....don't download any of the battery savers as they are useless .....try Greenify....bcoz it works....and it just is too awesome.....

Undervolted and got these results with anutu, impressive, looks like undervolting gets good results

Stock Speeds 1.7ghz -> 1.050
Barely over Stock 1.78ghz -> 1.075
Mild overclock 1.836ghz -> 1.0875
Higher overclock 1.89ghz -> 1.100 (highest anutu score)
Pushing it 1.944 ghz -> 1.250 (yielded a score the same as 1.89ghz)

what kernel were you using here and ROM? thanks

hithete
mmikeman said:
Hey was wondering if anyone knows off by hand the lowest we can undervolt our phones @ stock freq or what our stock voltages are? I undervolted mine recently and got amazing results but stability seems to vary, using overclock freqs @ 1.89 running @ 1.087 mv 1.83 @ 1.05mv 1.78 @ 1.0375 1.7 @ 1.025 mv I seem to be fully stable pulling 28905 scores on anutu, my friends s4 isn't even this fast was very impressed phones not even hot at all, but I really want to try to tweak the lower freq to help with battery life/temp
Click to expand...
Click to collapse
Cn u please tell me what kernal u are using