CF-Bench, StabilityTest (NOT the Dalvik Linpack!) linpackc, Geekbench for CPU. GLBenchmark, BaseMark Taiji for GPU. RgBenchmark and Geekbench again for memory speeds.I like Antutu more but I am against benchmark apps which give overall score.
I think all subsystems should be benched independently.
linpack for cpu, glbenchmark for gpu, sdtools for sd read/write tests...
linpack also gives some fluctuating results but there is a reason for that. it is a good idea to use it with performance governor..
and I think the best benchmark tool is your eyes and feelings...
those benchmark apps are only meaningful when you are comparing different devices. it doesn't make sense to me to bench on the same device over and over again...
these are my ideas as a user who doesn't use benchmark apps unless needed.
Yes they do.
I saw your PM btw, I sure as hell hope you're not using this kernel on that CM port of yours.
I saw your PM btw, I sure as hell hope you're not using this kernel on that CM port of yours.
No it currently wasn't: but here's a kernel with the sharpness fix included. Please test it and if it looks better I will keep the changes.Andre, thanks for this great kernel; what I wonder is, if hardcore's sharpness tweak is included in the kernel or not.
Thanks.
GT-N7100 cihazımdan Tapatalk 2 ile gönderildi
No it currently wasn't: but here's a kernel with the sharpness fix included. Please test it and if it looks better I will keep the changes.
Well what is going on here now? Either the kernel breaks the hardware acceleration or it doesn't. There is no inbetween. If DicePlayer works with the "HW" tag, then it is not the kernel's fault.Led working perfectly fine and HW accel also working PERFECT even tested it using the video pop out function, and scrolling through apps, for now everything perfect in my case, thx! :good:
The hardware acceleration failed to go through my original 3rd party testing. I need a complete last_kmsg with the boot sequence in the log, and if possible also the moment where you try to play something. The Note doesn't currently have the hacks used on the S3 so that part differs, but I see no reason why we couldn't incorporate that in the Note too, that would also mean about 40mB more userspace memory. But please provide a log first.
Well what is going on here now? Either the kernel breaks the hardware acceleration or it doesn't. There is no inbetween. If DicePlayer works with the "HW" tag, then it is not the kernel's fault.
Wow... that really DID do something!
Sent from my GT-N7100 using Tapatalk 2
Thanks for the kernel. Im having some issues connecting to wifi from time to time after flashing this kernel.
Anyone else experiencing this?
CF-Bench, StabilityTest (NOT the Dalvik Linpack!) linpackc, Geekbench for CPU. GLBenchmark, BaseMark Taiji for GPU. RgBenchmark and Geekbench again for memory speeds.
Yes they do.
I saw your PM btw, I sure as hell hope you're not using this kernel on that CM port of yours.
users tend to believe h/w acc. support means games and video, but i got games working fine, but not MP4 and divx playback those gave me the standard "not playable" with stock player. everybody can download a play market soft player and make it work fine.Well what is going on here now? Either the kernel breaks the hardware acceleration or it doesn't. There is no inbetween. If DicePlayer works with the "HW" tag, then it is not the kernel's fault.
Get your head out of your ass, it's not about disagreeing or not, it's about the facts/content you disagree on. You're don't have immunity to criticism and neither do I so don't get upset when you are being called out on something.I'll do that. And I'll be sure you get a pm when I do. Also, you won't have to worry about your device rebooting anymore.
Have a nice day folks. Sorry for disagreeing with the almighty god
The point of undervolting is that it is a disadvantage free method of gaining power efficiency, and it can bring down the dynamic power usage on the SoC down, from, let's pull it out of my ass, 20% by the best estimates from what I have measured over the last year. If you have instabilities then it is your fault in your methodology because you are undervolting too much. We are merely eliminating a manufacturers safety margin by undervolting.Some of these frequencies are set almost -100mv. Processors nowadays are becoming very efficient on power and are absolutely DWARFED by the amount of power your display, radios, and other hw uses. My own personal opinion is to look at other power consumers such as the aforementioned before playing with hw that in the grand scheme of things is not your biggest power consumer culprit, yet is vital to system stability.
No the sampling rate is not when the CPU is polled for load, it's the periods between which the governor does DVFS decisions, load is merely a statistic of the reverse of how much time the CPU is in idle.SAMPLING RATE: 28000 - just to be clear, the sampling rate is the interval of spacing between CPU being polled for load. Use this as you see fit, the small the interval, the more it checks for pending tasks to be executed. I have found great success and performance (and accuracy and decreased latency) with 20000-15000. Battery will not suffer anything noticeable to the end user.
Same thing, again, absolute nonsense as with the CPU voltages above. Internal voltage especially can bring power gains, albeit it is much less than the CPU, but it is a free gain.honestly, I really don't see the need to mess with this either. More potential instability for a minimal gain in battery when analyzed against your main thieves of your voltage - display, radios.
Deadline is alright, but noop just means that you are turning scheduling off. It doesn't bring you any performance gains at all, you are better off using an IOPS and latency designed scheduler like ROW or FIOPS (I'll bring that back when I have time), again, Deadline is ok, and SIO is a compromise of everything.dealine and noop will give you the best performance on these devices in real time, data speeds are also increased significantly with these two IO schedulers.
Maintenance, logging???!!! These are tasks that do not even remotely task the CPU even at 200MHz with any significant load, any single process out there running in the background can be done at 200MHz, it is more power efficient and you are not even aware of it.Also, if i may add this... I would recommend bumping the min frequency to something like 500MHz for the CPU. Reason being is simple - smaller tasks (maintenance, logging, etc) can be handled with ease without causing unnecessary CPU ramping.
So you see how that whole last paragraph of yours is just nonsense and that last phrase is really something one might take out of a PR article. Your settings you suggested were not given more thought.THIS is performance without battery life taking a hit - making your system AGGRESSIVE YET EFFICIENT, not hindering and crippling it to not do what it is programmed to do. Efficiency. "Less is more" as you see thrown around in the Linux world can also be thought of as more is less if you think in terms of more thought, tuning, and logic is less power consumed by your device.
Yes please do deliver your kernel, because I want to see more nonsense delivered with an attitude. We already have a large amount of clueless developers on this site. The problem isn't that they're unskilled or do not know the technical details, everybody was like that including me; It's that they think they are doing some sort of meaningful change without giving the effort of thinking or dwelling into the topic more than 2 minutes. People are suffering from a severe case of the Dunning-Kruger effect (I was so happy when I found out there's a name to that attitude).I'll do that. And I'll be sure you get a pm when I do. Also, you won't have to worry about your device rebooting anymore.