I have created this thread to focus on overclocking of the gpu, this is still very experimental and it would be great if you all can help test this feature.
There are allot of different settings that can be pushed to the gpu clock, I do not have access to the MSM7200 chip development guide and have no detailed information.
Non camera kernel with keyboard included:
http://tingstenen.dk/data/X1-kernel-modules-2011-03-18-1300457455.zip
How to activate overclocking
Code:
$ su
# cd /sys/module/clock_wince/parameters/
# echo 0xa99 > grp
The number 0xa99 can be replaced with other values to archive different results. See the following gpu overclocking discussion for more details, 0xa99 is the valued I have had positive results with on our x1 of the few options I have tried.
http://forum.xda-developers.com/showthread.php?t=697673
Our MSM7200A chip is default running with CPU 528Mhz and 256Mhz GPU.
I look forward to hear your feedback on this option and hopefully some of you manage to archive even bette results with other settings. It would alse be great if someone found out how to make an widget or app to easily change gpu clock settings.
Will add a non camera kernel with keyboard as modules later.
great job.
X1-kernel-modules-2011-03-18-1300457455 gives me 21.9 fps.
and just tried newest X1-kernel-modules-2011-03-19-1300544600 and its works, GPU overclock is applied automatically, gives me same result.
preston74 said:
great job.
X1-kernel-modules-2011-03-18-1300457455 gives me 21.9 fps.
and just tried newest X1-kernel-modules-2011-03-19-1300544600 and its works, GPU overclock is applied automatically, gives me same result.
Click to expand...
Click to collapse
Hi Preston, thanks for your feedback, I am glad to hear the auto gpu overclock is working. 0xa99 has been hardcoded in the last two kernels. 21.9 is more than I have been able to get, 21.5 with cpu at 614Mhz and 0xa99 applied
im reach 21.9 fps with cpu ovecloched to 672Mhz.
I reached 20.5 fps on Neocore (Sound off) with CPU overclocked to 672MHz and GPU overclocked using 0xa9a - using PureFroyo by Needo, an older build of his. Anything beyond 0xa9a for me, crashes my Xperia once I load apps which required 3D.
n00b3r said:
I reached 20.5 fps on Neocore (Sound off) with CPU overclocked to 672MHz and GPU overclocked using 0xa9a - using PureFroyo by Needo, an older build of his. Anything beyond 0xa9a for me, crashes my Xperia once I load apps which required 3D.
Click to expand...
Click to collapse
Could you try with 0xa99 once more with a slightly lower CPU clock? I have a theory that 0xa99 give us GPU clock half of our CPU clock. It could be the limit of the 3D chip are around 330-340.
duckly said:
Could you try with 0xa99 once more with a slightly lower CPU clock? I have a theory that 0xa99 give us GPU clock half of our CPU clock. It could be the limit of the 3D chip are around 330-340.
Click to expand...
Click to collapse
Sure. Should I try it with 633 MHz? I tried 0xa99 with 672MHz and I get 19.5 instead of 20.5 FPS.
more tests
build: a new F1 Froyo 2.2.2 V3
cpu oc: 692MHz
kernel: X1-kernel-modules-2011-03-19-1300544600
Neocore with sound off: 22.2fps
Can you try with 0xa92 and compare this with your results from 0xa99? It seems to be slightly faster on my 614MHz clocked cpu.
duckly said:
Can you try with 0xa92 and compare this with your results from 0xa99? It seems to be slightly faster on my 614MHz clocked cpu.
Click to expand...
Click to collapse
I just tried 0xa92 with 672 MHz and i got 21.4fps. Got an extra FPS with 0xa92 compared to 0xa9a.
I have done some testing with different values and here is my results so far. The clock calculations is obviously wrong but it is the best I can come up with.
10101 0010 001 0xa91 (614*2)/3 = ~409 neocore not starting
10101 0011 001 0xa99 (614*2)/4 = ~307 21.5 21.2 176(quadrant)
10101 0100 001 0xaa1 (614*2)/5 = ~245 19.5
10101 0001 010 0xa8a (614) /1 = ~614 neocore not starting
10101 0010 010 0xa92 (614) /2 = ~307 21.1 21.0 176(quadrant)
10101 0011 010 0xa9a (614) /3 = ~205 20.1
10101 0100 010 0xaa2 (614) /4 = ~154 17.9
10101 0010 011 0xa93 (614?)/2 = reboot
10101 0011 011 0xa9b (614?)/3 = reboot
10101 0000 100 0xa84 (614/2)/0 = neocore not starting
10101 0001 100 0xa8c (614/2)/1 = ~307 18.1 169(quadrant)
10101 0010 100 0xa94 (614/2)/2 = ~154 14.8 164(quadrant)
10101 0011 100 0xa9c (614/2)/3 =
The best options I have found so far is 0xa99 and 0xa92 being equal. There are still allot of other bits to play around with.
my cpu and gpu are not so lucky
i can daily use cpu at 614Mhz
tryed gpu oc with faryab V3 + 18 march kernel
0xa9b - restart
0xa99 - artifacts, neocore starts but artifacts and it stops
0xa92 - always artifacts and bad graphics
0xa9a - always artifacts but neocore bench finished ok, about 21 fps
without oc neocore is 17fps, quadrant about 270, linpack 2.8
just a curiosity, i remember in older kernels that quadrant much faster (4/500) and linpack was about 4/5, why newer kernels are slower at the same cpu speed?
pirlano said:
my cpu and gpu are not so lucky
i can daily use cpu at 614Mhz
tryed gpu oc with faryab V3 + 18 march kernel
0xa9b - restart
0xa99 - artifacts, neocore starts but artifacts and it stops
0xa92 - always artifacts and bad graphics
0xa9a - always artifacts but neocore bench finished ok, about 21 fps
without oc neocore is 17fps, quadrant about 270, linpack 2.8
just a curiosity, i remember in older kernels that quadrant much faster (4/500) and linpack was about 4/5, why newer kernels are slower at the same cpu speed?
Click to expand...
Click to collapse
The optimal clock for my CPU is also 614Mhz, it can run at 633 but will eventually crash.
Can you try to downclock you cpu one or two steps further and try with 0xa99? I am curious to know if the GPU overclock values are connected to the CPU overclocking, could be that you successfully could use 0xa99 with one CPU step lower and end up with better neocore results.
faryab V3 + 18 march kernel + custom startup.txt
(to change grp frequency i use clock_wince.grp=2713 on startup.txt (2713 = 0xa99 for example, to no OC just left blank, i think it's better manual setting than auto-OC))
tryed one time, and seems to be working with startup setting
neocore - always sound off
cpu: 633600 - reboot
cpu: 614400 - ok
gpu:
-no OC: 18,9
-0xa9a: 20,2
-0xa99: reboot
cpu: 595200
-no OC: 18,7
-0xa9a: 19,8
-0xa99: 21,4
EDIT: i can confirme that startup.txt setting works with gpu freq too with 18 march kernel
i tryed acpuclock.oc_freq_khz=614400 clock_wince.grp=2713 on startup.txt and this time phone doesn't reboot, and i got 21,4 too
One thing to note though..
The last 3 bits of the ns register are the clock source.
{0 => TCXO (19.2 MHz?), 1 => global PLL, 2 => backup pll 0, 3 => backup pll 1, 4 => modem pll, 5 => plltest_rcvr_out (whatever that is), 6 => usb xtal, 7 => plltest_core_in}
This is from msm7200 datasheets. it may be wrong for msm7200A though
afaict, modem pll (4) is pll0 in acpuclock.c (245 MHz), global pll is pll1 (768 MHz), backup pll0 is pll2 (1056 MHz). When the device is overclocked, then
The next 4 bits are the (divider - 1). That is, get those 4 bits, increment by one and you get the divider. During the overclock, pll2 is adjusted. So yes, until we have a proper clock calculation algo for arbitary frequencies, you cannot overclock both cpu and gpu.. But actually imho overclocking cpu gives no performance gain at all..
I'm just a newbie, so I want to know how to open Internal? I try to get it on Market, installed but not working? Could anyone show me the way?
Pardon my bad eng
sp3dev said:
One thing to note though..
The last 3 bits of the ns register are the clock source.
{0 => TCXO (19.2 MHz?), 1 => global PLL, 2 => backup pll 0, 3 => backup pll 1, 4 => modem pll, 5 => plltest_rcvr_out (whatever that is), 6 => usb xtal, 7 => plltest_core_in}
This is from msm7200 datasheets. it may be wrong for msm7200A though
afaict, modem pll (4) is pll0 in acpuclock.c (245 MHz), global pll is pll1 (768 MHz), backup pll0 is pll2 (1056 MHz). When the device is overclocked, then
The next 4 bits are the (divider - 1). That is, get those 4 bits, increment by one and you get the divider. During the overclock, pll2 is adjusted. So yes, until we have a proper clock calculation algo for arbitary frequencies, you cannot overclock both cpu and gpu.. But actually imho overclocking cpu gives no performance gain at all..
Click to expand...
Click to collapse
Thanks for the thorough explanation alex.I was confused at first about this when I saw the topic
luv116 said:
I'm just a newbie, so I want to know how to open Internal? I try to get it on Market, installed but not working? Could anyone show me the way?
Pardon my bad eng
Click to expand...
Click to collapse
Terminal Emulator works very bad with non-US keyboard layout (you have to use virtual keyboard for _ and > char, and it's very annoying).
It's faster (and works much better because of the fresh boot) to change settings from startup.txt, read my post
sp3dev said:
One thing to note though..
The last 3 bits of the ns register are the clock source.
{0 => TCXO (19.2 MHz?), 1 => global PLL, 2 => backup pll 0, 3 => backup pll 1, 4 => modem pll, 5 => plltest_rcvr_out (whatever that is), 6 => usb xtal, 7 => plltest_core_in}
This is from msm7200 datasheets. it may be wrong for msm7200A though
afaict, modem pll (4) is pll0 in acpuclock.c (245 MHz), global pll is pll1 (768 MHz), backup pll0 is pll2 (1056 MHz). When the device is overclocked, then
The next 4 bits are the (divider - 1). That is, get those 4 bits, increment by one and you get the divider. During the overclock, pll2 is adjusted. So yes, until we have a proper clock calculation algo for arbitary frequencies, you cannot overclock both cpu and gpu.. But actually imho overclocking cpu gives no performance gain at all..
Click to expand...
Click to collapse
Hi Alex, thanks allot for your description. It is nice to know what the different options means instead of blindly changing the values
Are the pll clock speeds different when you overclock the cpu?
yes, they are.
btw, i've played with overclocking a bit. changing gpu to 1056 mhz pll doesn't seem to do anything, but overclocking the 768 mhz pll to 960 mhz or more seems to increase performance linearly - my neocore went from 16.5 to 21.5.. but i needed to set mdp (panel) to another pll..
as for the panel, some tweaks allowed to raise score in fps2d from 27 to 34 fps. i hope we can get to somewhere 45 with a bit of overclocking.
all in all, i think we need to rewrite some clock code to allow to change pll speed at runtime and recalculate clocks on the fly
I'm just curious what everyone's performance settings are in stock cm7. And if you're running with dalingrin's OC kernel and using Nook Color Tweaks what are the settings that's been working well for you.
CM7 Performance:
CPU: conservative min300 max1200
Compcache RAM: 10%
Use JIT: checked
Enable surface dithering: checked
Allow purging of assets: unchecked
Lock home in memory: checked
Lock messaging app in memory: uncheck
VM hep size: 48m
Nook Color Tweaks:
OC 6/18 conservative gov 300 600 800 1100 1200
CM7 Nightly 103 (waiting for next Stable since everything works) -
CM7 Performance:
CPU: conservative min300 max1300
Compcache RAM: 0%
Use JIT: checked
Enable surface dithering: unchecked
Allow purging of assets: checked
Lock home in memory: checked
Lock messaging app in memory: unchecked
VM hep size: 32m
Nook Color Tweaks:
OC 6/18 conservative gov 300 600 900 1150 1300
I have ran across some interesting articles regarding compcache, JIT and vm heap. you would thing bigger is always better, but I've been reading other sites/forums and get conflicting thoughts.
Compcache:
http://wiki.cyanogenmod.com/wiki/Swap_and_Compcache
Just-In-Time Compiler:
http://wiki.cyanogenmod.com/wiki/Just-In-Time_Compiler
vm heap:
http://ryanolson.wordpress.com/2010...ory-per-vm-heap-can-effect-your-rom-cyanogen/
I have read somewhere that there is a cap of 30 FPS in android devices. Also, that higher fps will lead to more smooth transitions etc. Is this correct?
Moreover, even if the above is right or wrong.. is there any way to change fps value in android?
You can increase UI smoothness with app called AutoKiller Memory Optimizer. Open it and go to Advanced system tweaks and turn the UI optimization tweak on. Not sure if this increases FPS? But it still gives smoother UI.
V07A4ER said:
You can increase UI smoothness with app called AutoKiller Memory Optimizer. Open it and go to Advanced system tweaks and turn the UI optimization tweak on. Not sure if this increases FPS? But it still gives smoother UI.
Click to expand...
Click to collapse
This'll just help in having a certain amount of RAM free at all times so that phone doesn't become laggy.. it doesn't increase FPS value...
I always have 70-80+ MB RAM free so this won't help much..
AFIK windows phone and Iphone have an FPS of about 60 and android has an FPS of 30.. that's why those phones have much smoother UI ..
Just wanted to see what you guys are using for MinFree values, and SysCtl tweaks (if any) on your S5.
I know there was a lot of discussion with the S3, and I've had my S5 for a while, and figured I'd post mine to compare
My Current MinFree Values (in MB)
Foreground App: 20
Visible App: 40
Secondary Server: 200
Hidden Application: 240
Content Provider: 280
Empty App: 390
SysCtl Tweaks (*all are set to apply on Boot*)
My Current VM Values -
vm.dirty_background_ratio: 80%
vm.dirty_expire_centisecs: 1000
vm.dirty_ratio: 90%
vm.dirty_writeback_centisecs: 2000
vm.drop_caches: 3
vm.min_free_kbytes: 4096KB
vm.min_free_order_shift: 4
vm.oom_kill_allocating_task: Disabled
vm.page-cluster: 3
vm.swappiness: 30%
vm.vfs_cache_pressure: 10%
My current fs values -
fs.file-max: 229376
fs.lease-break-time: 15
fs.nr_open: 1048576
My current kernel values (using Ktoons kernel)-
kernel.msgmax: 122880 bytes
kernel.msgmni: 2048
kernel.panic: 5 seconds
kernel.panic_on_oops: Enabled
kernel.random.read_wakeup_threshold: 192 bits
kernel.random.write_wakeup_threshold: 384 bits
kernel.sched_child_runs_first: Disabled
kernel.sched_latency_ns: 12000000 nanoseconds
kernel.sched_min_granularity_ns: 1333332 nanoseconds
kernel.sched_wakeup_granularity_ns: 2666666 nanoseconds
kernel.shmall: 2097152 pages
kernel.shmmax: 268435456 bytes
kernel.shmmni: 5120
kernel.threads-max: 229376
I'd like to hear/see yours, as well as any thoughts/inputs/suggestions!
HOLY S*** those min free values are high... Don't know anything about the rest though... Lowering your min free values should improve battery life
You have battery expension?
I do not have an extended battery pack. I am a heavy user and average 1 day of usage and anywhere between 5-7 hours of SOT. I use KT's kernel and LuigiBull23's profiles to go along with it. Also Greenified and Amplified things.
i have tried many kernel...from them...dragon xia and xceed works the best...
the thing is i am a moderate use guy...just anime episodes ...songs...web surfing and lots of calls...
i main aim is to get better battery life....performance doesnt matter to me much...as long as its stable...
from my experience xceed gives better battery life then dragon xia...but since i am new to these kind of stuff....
i would really love to have knowledge from you guys...
i would love to see if someone can provide me with the adiutor settings they have used...
currently i am on resurrection remix rom 5.8.4 (mi5 standard)
i really want a better battery life from my device
Like how much battery life you are getting? SOT as well as overall usage + SOT.. try my setup on dragonxia.. I don't use it now though..
*CPU 3 &4
Min - 307
Max - 1401
Gov - interactive
*CPU 1 & 2
Min - 307
Max - 1036
Gov - interactive
*Input boost freq core 1 - 556
*Input boost freq core 2 - disabled
*Input boost freq core 3 - 729
*Input boost freq core 4 - disabled
*Touch boost - disabled
*Set GPU at max
*Turn on backlight dimmer
*I/O sched - bfq
*Swap - 100
With daily usage on this, I can say that there's no performance decrease..it actually performs well and battery life is pretty awesome..
Tip: you can personally tweak the kernel depending on your usage..rule of thumb, CPU 3 & 4 (cores with lower clock speed) is the main driver over all..means CPU interaction triggers from here most of the time..and CPU 1 & 2 (higher clock speed) will work if you're doing extensive task like gaming..
viking_kong16 said:
Like how much battery life you are getting? SOT as well as overall usage + SOT.. try my setup on dragonxia.. I don't use it now though..
*CPU 3 &4
Min - 307
Max - 1401
Gov - interactive
*CPU 1 & 2
Min - 307
Max - 1036
Gov - interactive
*Input boost freq core 1 - 556
*Input boost freq core 2 - disabled
*Input boost freq core 3 - 729
*Input boost freq core 4 - disabled
*Touch boost - disabled
*Set GPU at max
*Turn on backlight dimmer
*I/O sched - bfq
*Swap - 100
With daily usage on this, I can say that there's no performance decrease..it actually performs well and battery life is pretty awesome..
Tip: you can personally tweak the kernel depending on your usage..rule of thumb, CPU 3 & 4 (cores with lower clock speed) is the main driver over all..means CPU interaction triggers from here most of the time..and CPU 1 & 2 (higher clock speed) will work if you're doing extensive task like gaming..
Click to expand...
Click to collapse
What does swappiness do actually?
Sent from my MI 5 using Tapatalk
cheaterfruity said:
What does swappiness do actually?
Click to expand...
Click to collapse
Speed up read and write speed of the zram which is enabled by default in our RAM.. which means snappy switching of apps..
viking_kong16 said:
Speed up read and write speed of the zram which is enabled by default in our RAM.. which means snappy switching of apps..
Click to expand...
Click to collapse
Consumes more battery?
Sent from my MI 5 using Tapatalk
cheaterfruity said:
Consumes more battery?
Click to expand...
Click to collapse
Nope.. our phone is using zram as swap..not internal storage.. zram is actually a compressed RAM that sits within RAM itself.. try to google it..