[Please close thread} Achieve the best performance and batt life without sacrifices

[jsmasterx]

Good day everyone.

After extensive testing I found the best voltages for our nexus 7 cpu.
those tweaks have been made with the kernel m13

ive tested and compared different voltages behavior and found that in almost every kernel and or roms , the voltage is too high and thisis not for the benefits of performance nor stability.

So ive started to underclock and watch the behavior of the N7 and found the perfect settings.

PS : If someone have better settings to share please feel free. Ill not try to push my idea. but yet i found it was the most better settings.

Sent from my Nexus 7 using xda app-developers app

 

[sert00]

for me this is absolutely useless due to the fact cpus are not the same,never and never will be,and voltages optimal for someone can give bootloop to others...a more accurate thing could be do 3 set of voltages optimization related to the 3 variant known of tegra3 t30,variant 1 2 and 3.e known that differences between 3 variants are the power needed to operate withpour rebbot..when variant 3 support and run at high frequencies with predefined voltage,the same freq. on variant 1 need more millivolt to run..with these 3 division,you can list 3 mode of undervolts,i think it's more usefull that now,but you can trace only yours variants in this mode..let's see if the 3ad will grown and others users with other variant will write its best voltages.just my 2 cent mate!
hi!:good:

 

[mlaws90]

Just tested SIMILAR voltages (the frequency steps were not all the same, most 30 Mhz off) on Trinity 4.2 alpha0 and it works great so far, we will see if I get soft resets or bootloops, but I hasn't yet do I don't think it will.

 

[ericdjobs]

With this kind of information, can you PLEASE include the Tegra variant information? Because not all of us are lucky enough to have a variant 3 (as I suspect you have, with that kind of undervolting)...

Take my nexus 32gb for example, it's a standard Variant 1, and 1090mv at 1600MHz would most definitely result in problems if I pushed it hard at that level. I can just barely get away with 1150-1200mv at 1600... I suggest you run the stress testing/burn in application to make sure your voltages are stable as well

APplication is called "Stability test" and it will push your CPU/GPU to the max. If it runs continuously without any issues at that voltage level, you're fine..

See, I can set 1090mv at 1600MHz no problem on my variant 1, but if I run Stability test, I'll get a lock-up after 2-3 passes or so. By the same token, I'll get a lockup running glow ball and other intensive games. Just because it runs stable during a benchmark / every day use, does NOT mean it's stable.

Here's a link ot stability test: https://play.google.com/store/apps/...=W251bGwsMSwxLDMsImNvbS5pbnRvLnN0YWJpbGl0eSJd

Those of us who do a lot of overclocking on the PC etc know how important it is to "burn in" test your overclocks to make sure they are absolutely stable. This usually means running a linpack or other such intensive burn-ins overnight.. and you don't consider it 'stable' until it runs a LONG time, with no errors.

 

[BurnOne77]

how do u know which variant you have?

 

[ericdjobs]

how do u know which variant you have?

You have to be using a kernel that exports this info... I know Faux's does FOR SURE, and I think a few others have actually merged that over into their own kernels.

It would be under /sys/kernel/debug/t3_variant, I believe. So, as root, you can cat /sys/kernel/debug/t3_variant or just browse via ES file explorer or root explorer or whatever root file explorer you choose. The variant is the cpu_process_id

Each variant more than likely is based on where on the wafer or whatever the chip came from, meaning higher quality.... what it boils down to is the higher the variant, the more likely you'll be able to run at a higher frequency with less power.

Here's the post on the One X forums about Tegra 3 variants, http://forum.xda-developers.com/showthread.php?t=1815466

Lots of good information there.

I'd really like to see the OP run multiple passes of Stability test going 1600MHz at 1090mv, though. If you can do that, you have one exceptional chip, and you're very lucky. Remember just because something is stable in everyday use, doesn't mean it's truly stable.. very little that you do will actually stress the GPU/CPU to the max at the same time, which is the only thing that will truly show if you have problems with undervolting. I'd make a habit of running stability test for at least 5-8 passes (I like to do many more than this) to make sure an undervolt is at least somewhat stable.

If you simply say "This is the BEST! Why isn't everyone doing this? Why do kernels set the voltage so high?!" and have everyone simply drop down to those levels, you're going to have a TON of people who will start having issues. There's a reason why they are set that way. If everyone could easily get away with undervolting this much, it would be a no brainer to make it the default...

 

[jsmasterx]

Good evening everyone.

ill look for the variant asap and will also runs the stress test as suggested. Even if it did well so far ill post accurates testing results besides the performance quadrant test.

cheers mates

Sent from my Nexus 7 using xda app-developers app

 

[jsmasterx]

stress test so far

at the end i was tired running it but for now no errors or crash

due to some heat generated after 10 minutes the CPU dropped to 1400mhz instead of 1600. I was on performance mode.

edit : ill try to find my cpu/gpu variant. and post it. i am not sure wich application could give me an accurate information but i am actively searching.

Cheers

 

[jsmasterx]

ok so i found the tegra3 variant. i am posting the results

cpu_speedo_id => 7 soc_speedo_id => 1 cpu_process_id => 0 core_process_id => 0

Sent from my Nexus 7 using xda app-developers app

 

[poondog]

for me this is absolutely useless due to the fact cpus are not the same,never and never will be,and voltages optimal for someone can give bootloop to others...a more accurate thing could be do 3 set of voltages optimization related to the 3 variant known of tegra3 t30,variant 1 2 and 3.e known that differences between 3 variants are the power needed to operate withpour rebbot..when variant 3 support and run at high frequencies with predefined voltage,the same freq. on variant 1 need more millivolt to run..with these 3 division,you can list 3 mode of undervolts,i think it's more usefull that now,but you can trace only yours variants in this mode..let's see if the 3ad will grown and others users with other variant will write its best voltages.just my 2 cent mate!
hi!:good:

No offence OP, but this.

All CPUs are created differently, they won't behave the same at the same voltages. What might be stable for someone may be unstable for someone else. Take clemsyn's kernels for example, they are undervolted to the extreme so they might be stable on one device but not even survive a minute on another device.
So what works for you might not work for someone else. So maybe take the values in the OP as a guide, not as the "best performance and batt life without sacrifices". That might be a bit vague for some.
Within a single variant there might be chips that withstand less voltage than others too
Anyway, as the great kernel dev faux123 said:

Not All CHIPS ARE CREATED EQUAL

Peace!
Mark

 

[ericdjobs]

ok so i found the tegra3 variant. i am posting the results

cpu_speedo_id => 7 soc_speedo_id => 1 cpu_process_id => 0 core_process_id => 0

Sent from my Nexus 7 using xda app-developers app

Thank you! Have you switched kernels at all lately, or anyhthing like that? Because I believe you're suffering from the same bug that I am.

When I switched from Faux's kernel to M-Kernel, my variant also reports as "0" when it was very clearly a "1" before.

If you really do have a variant 0 chip that is hitting 1600MHz reliably at 1090mv then you're one of the luckiest people around! haha. But like others have said, no two chips are created equal, it appears you have a very very good one.

 

[BurnOne77]

mine is 7 1 3 0 in that order... I guess I have a lucky 3 This is a clean device running smoothrom and only seen bricked kernel 0.55 so far.

I also cat /proc/cpuinfo
and it says cpu variant is 0x2, my friend has no t3_variant file to see cpu_procces_id but we both get cpu varient 0x2, I guess we both have a 3?

everyone else try cat /proc/cpuinfo
and report back pls.

 

[jsmasterx]

ive cated cpu info and here are the results

as the previous op said hejad revision 0x2 and i have the same so maybe i have a variant 3. But I really need to be 100% sure.

Processor: ARMv7 Processor rev 9 (v7l) processor: 0 BogoMIPS: 1993.93 processor: 1 BogoMIPS: 1993.93 processor: 2 BogoMIPS: 1993.93 processor: 3 BogoMIPS: 1993.93 Features: swp half thumb fastmult vfp edsp neon vfpv3 tls CPU implementer: 0x41 CPU architecture: 7 CPU variant: 0x2 CPU part: 0xc09 CPU revision: 9 Hardware: grouper Revision: 0000 Serial : 0f410a0001440200

Sent from my Nexus 7 using xda app-developers app

 

[jsmasterx]

Hello Mods please close this thread

O'll open a new thread wich will be more clear and more neutral than this one.