i am not sure how reliable cpu spy is but i doubt it is bullshitting me ;-)
so feel free to correct me if im wrong but the way tegra 3 works on this phone is as follow
1 core = 1.5 ghz
2 cores = 1.4 ghz each core
3 i will pressume 3 cores at 1.3 ghz..
when all cores are running, it is running at 1.2ghz x 4
according to my cpu spy info. the phone is mainly running on everything else but 1.2ghz(=quad core)
http://db.tt/syesXRs2
however i fail to see with us using a lot of dual core and hardly any quad core how our battery drains within less than 3 hours (at least from what i have seen at least)..or is the information i have about how tegra 3 runs is false?
shadehh said:
i am not sure how reliable cpu spy is but i doubt it is bullshitting me ;-)
so feel free to correct me if im wrong but the way tegra 3 works on this phone is as follow
1 core = 1.5 ghz
2 cores = 1.4 ghz each core
3 i will pressume 3 cores at 1.3 ghz..
when all cores are running, it is running at 1.2ghz x 4
according to my cpu spy info. the phone is mainly running on everything else but 1.2ghz(=quad core)
http://db.tt/syesXRs2
however i fail to see with us using a lot of dual core and hardly any quad core how our battery drains within less than 3 hours (at least from what i have seen at least)..or is the information i have about how tegra 3 runs is false?
Click to expand...
Click to collapse
As far as I know, it will only go to 1.4 GHz on multiple cores, even on 4 cores.
Related
Hello,
I read lots of articles about how the Tegra 3 only scores well in benchmarks because of its 4 cores, which are overkill in almost all real world scenarios. So I was interested to find out how the the Tegra 3 would do if you made it a Dual Core, like the S4.
I ran Antutu three times, running stock everything (except root), and the lowest I got was 7114.
I know it is not very reliable to use one benchmark, but in my opinion, neither is using quadrant which is made by Qualcomm.
I find it interesting to see that the Tegra 3 scores considerably more than the S4, even with the same number of cores.
What are your thoughts? What do you think caused this? What does it mean?
Unable to upload screenshot:
Ram - 1202
CPU Integer - 2004
CPU Float - 1550
2D - 295
3D - 1242
Database - 475
SD Read - 150
SD Write - 196
not too sure why it would still win hands down *shrugs*
How does the real world speed with 2 cores disabled though?
And does it seem to save any battery if you've had it going for a while.
I've had it set to two cores only for a few days. It makes NO difference to anything but benchmark scores (even antutu still shows 60 FPS in the graphics tests). Games like Dark Meadow THD run exactly the same as before.
I'm not too sure how it has affected the battery life as I installed a mod that lowers the auto brightness at the same time. All I can say is the combination of the two has dramatically increased the life of the battery
So the GPU is bottleneck (surprise)
Which GPU are you referring it?
maybe the benchmark tests are yet to be fully optimised for 4 cores?
ORStoner said:
Hello,
I read lots of articles about how the Tegra 3 only scores well in benchmarks because of its 4 cores, which are overkill in almost all real world scenarios. So I was interested to find out how the the Tegra 3 would do if you made it a Dual Core, like the S4.
I ran Antutu three times, running stock everything (except root), and the lowest I got was 7114.
I know it is not very reliable to use one benchmark, but in my opinion, neither is using quadrant which is made by Qualcomm.
I find it interesting to see that the Tegra 3 scores considerably more than the S4, even with the same number of cores.
What are your thoughts? What do you think caused this? What does it mean?
Unable to upload screenshot:
Ram - 1202
CPU Integer - 2004
CPU Float - 1550
2D - 295
3D - 1242
Database - 475
SD Read - 150
SD Write - 196
Click to expand...
Click to collapse
My guess is the reason the Tegra 3 with 2 cores running scores lower than the Snapdragon S4 is because the Tegra 3 has 4 A9 cores, whereas the Snapdragon has 2 cores that are closer to the A15 architecture, which is a faster chip. A15 will be quicker than A9 if the same number of cores are being used in each chipset.
My One X scores around 11000 with four cores and, as you can see, 7114 with two cores. In just curious to know why with two cores it scores around 1000 more than the S4 version?
thegregulator said:
My guess is the reason the Tegra 3 with 2 cores running scores lower than the Snapdragon S4 is because the Tegra 3 has 4 A9 cores, whereas the Snapdragon has 2 cores that are closer to the A15 architecture, which is a faster chip. A15 will be quicker than A9 if the same number of cores are being used in each chipset.
Click to expand...
Click to collapse
I think you misunderstood, the Tegra with two A9 scored greater than the Snapdragon. If you were correct, I would not be as surprised as I am and would not have started this thread.
Out of curiosity what are you using to lock the 2 cores?
Wouldn't mind trying it out myself
Open root explorer.
Go to sys/kernel/debug/tegra_hotplug
Open max_cpus in text editor, change 4 to 2 or 3 (Single core does not work).
Open it again to check it has saved properly and it will go back to 4 the next time you reboot the phone.
You know Tegra 3 has 5 cores instead of 4 cores in the A9 architecture right? So you basically did the benchmark with 3 cores instead of 2.
Sent from my Incredible 2 using XDA
5th core is just a low clock speed / power to run idle tasks?
Doubt it would do much to a benchmark test.
Correct me if I'm wrong but the companion core can only be used on its own while the main processor is shut off. Even if it could 'assist' the main processor, it's only around 300mhz and would make very little difference to the score of a benchmark.
david_hume said:
You know Tegra 3 has 5 cores instead of 4 cores in the A9 architecture right? So you basically did the benchmark with 3 cores instead of 2.
Sent from my Incredible 2 using XDA
Click to expand...
Click to collapse
From what i know the 5th companion core is invisible to the system so 2 would be correct in the max_cpu edit.
I refer to my previous question which was why is the dual core tegra 3 doing BETTER than the dual core S4? Is it down to the GPU?
well that's revealing indeed, what's more interesting is how the included governors work
on-demand quad max 1400mhz conservative in jumping to max
interactive three cores only 1400mhz max jumps more often
performance three cores only locks at 1200mhz and jumps to 1400mhz on stress
glowball frame rate suffer badly when running less the 4 cores
you can see at default ondemand tegra3 is always juggling on 4 cores but rarely peaking to max clocks
While it indeed sucks as a useful tool, you should be aware that Quadrant is not a Qualcomm program...perhaps you're confusing it with the antiquated NeoCore benchmark. Vellamo is Qualcomm as well.
Sorry Vellano is Qualcomm not Quadrant. My mistake.
ORStoner said:
I refer to my previous question which was why is the dual core tegra 3 doing BETTER than the dual core S4? Is it down to the GPU?
Click to expand...
Click to collapse
If you are referring to the s4 in the htc one s then id like to know where you got your information indicating that the tegra is still faster in dual core when in quad the s4 still out performs it. The s4 scores over 12000 compared to 11500 that the tegra does and well over 7000 that it does in dual core mode....
The older a9 however found in older phones such as the sensation is another story with that chip scoring in about 6500
anyone got this overclocked
mox123 said:
anyone got this overclocked
Click to expand...
Click to collapse
And instantly overheated? :cyclops:
Yes .
Sent from my HTC One X using Tapatalk 2
Gpu overclock would be more useful than CPU.
Sent from my HTC One X using xda app-developers app
treebill said:
Gpu overclock would be more useful than CPU.
Sent from my HTC One X using xda app-developers app
Click to expand...
Click to collapse
ok gpu overclock then?
I would overclock my HOX...in a block of ice. Or...well, in real life i dont want to overclock it because it would smoke out in my hand
Overheating is a big problem even without overclocking, imagine it running on 1,6ghz...
Sent from my Renovated HTC One X using Tapatalk 2
Can't really see why you would want to overclock the One X, the phone is blazing fast anyway, 4 cores at 1.5 is enough..
But like everybody else said, the phone would probably burn up..
I wouldnt overclock my device - at least not at the stage we reached now.
Why?
a) As long as there is no way to lower the voltage this might toast your device - its a unibody, keep that in mind!
b) 100 MHZ more would have literally no effect - its a 6 GHZ device, even if you boost it up to 6,4 - you wont notice, it will just drain your battery.
6 GHZ is WAY enough...this is smartphone...I mean...seriously...its got more power than my 4 years old 1K €uro notebook...
Illux said:
I wouldnt overclock my device - at least not at the stage we reached now.
Why?
a) As long as there is no way to lower the voltage this might toast your device - its a unibody, keep that in mind!
b) 100 MHZ more would have literally no effect - its a 6 GHZ device, even if you boost it up to 6,4 - you wont notice, it will just drain your battery.
6 GHZ is WAY enough...this is smartphone...I mean...seriously...its got more power than my 4 years old 1K €uro notebook...
Click to expand...
Click to collapse
well first of all you can't just multiply the frequency by the number of cores. I'd much prefer an actual 6Ghz single core processor over 4x1.5Ghz because it won't have any compatibility and efficiency issues. Assuming they are of the same architecture and power usage of course.
Also the ARM low power SOCs probably don't have comparable number of commands per clock cycle as an x86 high performance CPU, even if it's 4 years old.
jacobgong said:
well first of all you can't just multiply the frequency by the number of cores. I'd much prefer an actual 6Ghz single core processor over 4x1.5Ghz because it won't have any compatibility and efficiency issues. Assuming they are of the same architecture and power usage of course.
Also the ARM low power SOCs probably don't have comparable number of commands per clock cycle as an x86 high performance CPU, even if it's 4 years old.
Click to expand...
Click to collapse
i agree.. when the multi-core CPUs first came out intel said doubling the core number would give as 47% boost in total performance (not x2 like apple says as they do not know it) lets assume that to be %50 to make the math a little bit easier..
so basically we can make the math here as; 4 cores at 1.2Ghz (when the all 4 active the clock is 1.2Ghz) gives us 1.2 x 3/2 x 3/2= 2.7 Ghz single core performance.. this value for SGS3 is; 1.4 x 3/2 x 3/2= 3.15Ghz
and here we can say dual core at (X) Ghz gives us (X) x 3/2=2.7 thus the (X) = 1.8 Ghz.. so, if you overclock any arm9 based Dual CPU to 1.8 Ghz you get the same performance "on paper".. if you want to catch up with SGS3 we need to OC it to 2.1 Ghz which is impossible at the moment i guess..
what makes the difference here is the lower loads or multiple loads on the CPU.. corecontrol users probably would have noticed; sometimes when the all 4 core are active the clock is only 480 or 640 Mhz (even 320 sometimes if i remember correctly) .. the same amount of load could be taken care of by a dual core at about 720 or 960Mhz.. but here the quad core system stays cooler with a little less energy consumed (or wasted) (as long as all the cores are in one uni-body structure, putting 2 or 4 single cores phsically together is not the case for our smartphones) this is how apple made sure about the smoothness of the ipad 2, new ipad and the iphone 4s.. they used lower clocked 2 power vr 543 GPUs.. when the load is little they can clock down to very low speeds and share the load..
and also you can always find an emtpy core waiting for new task when the others are busy..
so, long story for short; if we were dealing with a little amount but hard processes, having a single core at 2.7Ghz would be good since the quad core design would not cut one task into 4 pieces... as long as we were not thinking about the battery life and the heat.. but since we are dealing with lots of tasks which all could be handled by 1.2Ghz power having 4 cores is better for battery saving and having an empty core for a new task to run parallel with the other running tasks in the background..
It is OC out of the box I think Nvidia OC them for us and it's already pushing itself at the very edge of what is possible for it to do based on temperature, I seem to remember Hamdir saying something along those lines once upon a time...
Why bother to OC it's fast enough as it is.
---EDIT---
hamdir said:
only faux kernel betas allow OC
big warning OC is bad for the HOX given the thermal envelope
you are risking both you battery and processor if you OC
i know you are used to OC from other devices but those had headroom, it is not the case this time, T3 is operating at its max thermal capabilities on the HOX
Click to expand...
Click to collapse
hamdir said:
the snapdragon 2 on the Arc had a lot of headroom
the chipset is rated @ 1.5ghz stable!
not the case with T3 its milking the very maximum of the 40nm process
in other words Nvidia is OCing its T3 out of box because their chips are designed to survive massive amount of heat (sadly it doesnt mean the battery or other components would survive)
it is already Overclocked lol
sometimes you have to listen to the "science" of it and surrender
Click to expand...
Click to collapse
Just want to share some ideas of the Tegra 3 hotplug, The HTC stocks are fast, but consume too much power...
may be kernel developers may have some discussion here.
When power saving core reach 90% usage,
G cluster should be on.
core 1 should be start at 1 GHz, if usage drop, frequency of core1 decrease.
If usage of core1 reach 90%, turn on core2.
then...
a.
If usage of core 2 is lower than 30% and core 1 still high,that's mean single thread program is running, increase frequency and core 2 off.
b.
If usage of core 1 and 2 is about the same, it indicate that multi-thread program running.
if usage of both core is less then 60% lower the frequency.
If both cores usage is high, turn on core 3, if core 3 doesn't help, turn off core 3 and increase frequency.
core 4 turn on only when core 1,2,3 are busy.
There is always this rough idea in my mind, but I have no time to implement it.
Hope this idea help kernel developers create better kernel.
I don't know if it is a good idea to put a post here, If I am wrong, just correct me.
Sent from my HTC One X using xda premium
lets hope showp1984 do similar thing again...he made his own hotplug manager in ics times (mpdescion..its called i think)
when using micro cpu monitor pro, and other cpu monitoring apps only 2 cores are active at most.
Is this a software or kernel error?
I am using stock ROM, kernel, and everything else, besides unlocked bootloader.
Yup its normal. No quad core device is going to use all four cores at once when its not required. Only when gaming/benchmarks will it actually use them. Its just like Samsung's 8 core processor, thats rarely going to use all 8 cores for battery life reasons.
aooga said:
Yup its normal. No quad core device is going to use all four cores at once when its not required. Only when gaming/benchmarks will it actually use them. Its just like Samsung's 8 core processor, thats rarely going to use all 8 cores for battery life reasons.
Click to expand...
Click to collapse
Not quite true. Samsung uses a big.little configuration so either 4 higher clocked or 4 lower clocked but never all 8.
My note's exynos should have a 4 arm cortex-a15 @ 1.90 GHz & a 4 arm cortex-a7 @ 1.30 GHz but my cortex-a7 has somehow been overclocked to 1.90 GHz and the cluster switch between these two has stopped and because of this i think all the cores run at the same time and my tablet over heats and runs out of battery very quick . what should i do ? please help
Mehran M said:
My note's exynos should have a 4 arm cortex-a15 @ 1.90 GHz & a 4 arm cortex-a7 @ 1.30 GHz but my cortex-a7 has somehow been overclocked to 1.90 GHz and the cluster switch between these two has stopped and because of this i think all the cores run at the same time and my tablet over heats and runs out of battery very quick . what should i do ? please help
Click to expand...
Click to collapse
You must be the first man in the 'verse to run HMP (all cores together) on an exynos 5420 tablet
As for your problem you probably run all 4 A15 cores together at all times. This is called the performance governor and it's enough to overheat your tablet ... just change it back to interactive with whatever app the first change was made.
Stevethegreat said:
You must be the first man in the 'verse to run HMP (all cores together) on an exynos 5420 tablet
As for your problem you probably run all 4 A15 cores together at all times. This is called the performance governor and it's enough to overheat your tablet ... just change it back to interactive with whatever app the first change was made.
Click to expand...
Click to collapse
But it is on interactive