Is there a benchmark test out there that can accurately judge a phone's performance vs another's? I continually see that people disregard quadrant, but i still want to be able to reliably test my phone's capabilities against someone else's.
Sent from my Desire HD using XDA Premium App
rant/
meh....all of it is almost completely useless.
ill tell you this. ive flashes every single rom on here, overclocked all the way up to 1800 and underclocked down to 700. ran every where in between for multiple days, ran every precievable benchmark test i could get my fingers on. getting scores as horrible as 1200, up to 3200 in quadrant. linpack from 20 to 60
and you know what? it has made absolutley ZERO impact on how the phone works with every day use.
when i realized this, was because i THOUGHT i had my phone set to 1800mhz, but didnt actually check the "apply on boot" box, and i was running at 1000mhz for two weeks thinking "wow this overclock makes my phone fast!"
yeah. the only time i see the difference is when performing tests.....oh and underclocking at 700 made the phone laggy, 1800 made it prone to lock up. but i can not tell a differnece in the actually function of the phone in day to day use between 800 and 1600 mhz if its on a good rom.
i get the fun of seeing the scores. but it really is completely useless. i was all about overclocking to see that high linpack and quad score....but now i run steadily underclocked, get crappy scores, and my phone functions flawlesly.
/rant
Look up cf-bench in the market it is supposed to be the most accurate least biased benchmark out there. There is also a thread on it in the general android app section.
Sent through the power of Cthulhu!
Related
So after reading nearly 5 hours and spending my time in the wee hours of morning, I finally did all the "stable" mods for the phone... If you haven't been reading, make sure you guys check out the stuff in the development forum.
After all modifications, I was able to get 2701 points in quadrant benchmark. What mods did I do?
-i9000 eclair flash (JM5)
http://forum.xda-developers.com/showthread.php?t=734871
-Alternative mimocans lag fix
(one click installer http://forum.xda-developers.com/showthread.php?t=749495)
-One click root (googled it for i9000)
-Overclock kernel 1.0Ghz to 1.2Ghz
(http://forum.xda-developers.com/showthread.php?t=746343)
This stuff really does help out your phone folks. Bench it now with quadrant, then take a peak at the other stuff and make magic happen. If anyone needs any additional help setting up their captivate, I'm more than happy to help.
I agree those fixes help speed a lot. But the quadrant score is meaningless. the speed hack creates an io loopback. The loopback just tells quadrant what it wants to hear.
Does your BT work on the european ROM. For me all people hear is a gargeled mess on there end?
Sent from my SAMSUNG-SGH-I897 using Tapatalk
Can you post the results of these tests:
Neocore
Linpack
CPU Benchmark
I keep hearing about this quadrant, does it actually improve real world performance? Or is just for the sake of scores?
jhego said:
I keep hearing about this quadrant, does it actually improve real world performance? Or is just for the sake of scores?
Click to expand...
Click to collapse
Quadrant is just benchmarking software that takes the cpu, gpu and memory read/write speeds into account. It runs a series of tests and spits a score number out at the end, so you can compare your device to others (like comparing boner sizes, but less gay).
It doesn't actually do anything to speed up the device though.
modest_mandroid said:
Quadrant is just benchmarking software that takes the cpu, gpu and memory read/write speeds into account. It runs a series of tests and spits a score number out at the end, so you can compare your device to others (like comparing boner sizes, but less gay).
It doesn't actually do anything to speed up the device though.
Click to expand...
Click to collapse
This is correct and to be honest there is way to much weight in the I/O tests. That is the only reason that the stock Droid X bests the Gal X. It has more weight than cpu and gpu so you can't put to much into those scores. They really don't mean anything more than bragging rights. What I am interested in is real world usage.
Real world use with the hack provides amazing speed gains opening and switching apps. Io heavy apps are very much improved while open too. It's finally as fast as the iphone.
Whats your battery life like after the overclock?
The score ended up getting lower and lower every time i used quadrant. 2701 is the highest I was able to get so far, but that's with a fresh install of the rom and all the stuff before I started loading on apps. Everytime I ran the benchmark, I of course killed the apps beforehand.
The battery life is the same- to be honest. This is me comparing a rooted stock ROM to the somewhat fresh install of the eclair i9000. The phone is very snappy. I came from an iPhone 4 and one of the biggest eye sores to me was the less-fluidness of changing programs, response to buttons (homescreen-back button) and pinch to zoom. After all these changes, it's a whole different story. Browsing is very appealing, especially since pinch to zoom isn't jagged or slow. The smoothness of this functionality is on par to an iPhone. And there is no waiting when I press the home button or back button.
True, maybe these numbers aren't considerably accurate (as far as the lag fix and EXT2) but at least it shows raw computing capability in it's current state... meaning, the usage of a virtual EXT2. Never the less, the phone is still all around faster, even if it isn't exactly the proper way of going about it.
The only problem I've seen so far is that it likes to randomly shut off. Won't respond to anything unless if I soft reset it. I haven't really found what causes it, since the consistency of it happening goes about in a non set pattern.
I didn't see any real world increase .. so I reverted back in about 4 hours.... I'd rather have the memory than a number that don't transfer to real world speeds...
How is it that I see alot of people benchmarking 2400+ on the benchmark? I have done the benchmark on the TnT Lite 2.4.0 and Vegan beta4 and both performed around 600-800. This is just from regular flash. Is it overclocking or killing all apps other than the benchmark test before proceeding with test? Does the livewalpaper matter?
Thanks!
I get consistent 2300+, and what you are getting is the behavior of wakeup bug in which when tablet wakeup from the long sleep like overnight, it gets stuck on 500+ quadrant score with subpar performance.
Sent from my PC36100 using XDA App
TnT results should be around:
1800 for 2.2
2400 for 2.4 though 3.1
I had 570 average scores using TnT 3.1 (out of sleep mode or also perhaps a combo of apps- not sure what consistently caused it) and rolled back to 3.0. Could not figure out why, but rolling back to 3.0 made things 2400 again and got me out of lag city.
Just did mine on TnT Lite 3.1.1 and got a score of CRAP.. never mind.. it just Force Closed on me before I could remember what it was. LOL
Standby... time to re-do it.
Ok result is: 2542
rushless said:
TnT results should be around:
1800 for 2.2
2400 for 2.4 though 3.1
I had 570 average scores using TnT 3.1 (out of sleep mode or also perhaps a combo of apps- not sure what consistently caused it) and rolled back to 3.0. Could not figure out why, but rolling back to 3.0 made things 2400 again and got me out of lag city.
Click to expand...
Click to collapse
I can confirm for stock TNT 3053, about 1800+ consistently.
Jim
I just killed all apps except for 4 essential apps. Got a lower score of 591. The tablet seems responsive to me. Even played some dungeon heros and I did not experience any lag.
I will have to benchmark the other two tablets and see how they do.
Edit: Redid the benchmark after a reboot and got 2356! which is a significant increase-value wise. My eyes and perception of time deceive me though as I do not notice a difference in usability from when I got the 500-600 scores recently to the 2356 after reboot. I will look into the calendar and autostart problems, see if those are what are keeping my scores down.
Reboot, kill any running apps and you should get some high scores (2000+).
jimcpl said:
I can confirm for stock TNT 3053, about 1800+ consistently.
Jim
Click to expand...
Click to collapse
Did 2 more tests, after pressing power button to wake Gtab, and got 1800+ for each test. Didn't do anything special like killing tasks. Again, this is with 3053 stock.
Jim
1800+ avgs seem to be the norm for stock on this device.
I don't know if the 2000+ scores are due to real performance gains or because of the partitioning and swap used in prepping the installation of those ROMs. Since Quadrant is synthetic it doesn't handle certain 'optimizations' added by code well and may misreport the scores. In reality 2000+ may really be 1800 or less, and 600 may also be running higher than reported.
Quadrant Advanced is better in that it gives you details. Much of the score we get on the G Tablet is from the powerful CPU and the abundant memory. TnT lite and VegaN may show 'boosts' to things like the I/O which is normally what is misrepresented as it's not weighted in the app. This is why many Galaxy S phones got 'fake' 2000+ scores due to the Lag Fix and in real testing they were performing at nominally higher than their 900 avging stock, but nowhere close to 2200 which is what most would avg. Testing a Vibrant with lag fix reporting 23xx in fact wasn't even working as fast/quick as a desire HD at 19xx.
I still haven't been able to get my device to slow down after sleep, so I guess that's something I'll have to do when I wake up later. Maybe a longer sleep time will aid me in achieving this. I have also tried everything to get the battery to drain faster, just in case there is some correlation...but I'm hovering at 84% and it's bedtime (according to my wife ).
If anyone else out there has Quadrant Advanced and is on stock and getting the slowdown - if you'd run it and give me some numbers I'd really appreciate it.
Numbers for stock 3899 + Google Apps are as follows:
Avg: 1917 in 3 runs. The avg isn't a calculation, it just means that's the score I got on the 3rd run. Under optimal conditions, the CPU should have maxed out by the 3rd run - but some people keep doing it and use the score that tapers off as max. It should get between 1830 and 1930 with stock.
CPU: 4916
Mem: 2730
IO: 1159
2D: 200
3D: 582
I expect the CPU number to be less than 1/2 during a 'slowdown'.
I expect the rest to be roughly the same.
If you're running a ROM like VegaN or TnT...I'm expecting everything to be the same except I/O which will be higher and maybe 2D/3D because of the newer Tegra drivers.
Update! Finally the pad slowed down just now. It's showing 35% on the battery meter. Not that there's any correlation but I've linked it - just because that's the only time I see the slowdown.
Total: 441
CPU: 918
Memory: 547
IO: 293
2D: 47
3D: 398
The slowing of the CPU has taken a toll on everything else. I thought the rest would look alright but that was stupid on my part. When you run a notebook in low power mode (ex: clock it at less than max), it affects performance across the board - so there was no reason that this should be different. Just wasn't thinking :/
Anyway, there we have it. Now to find a way to fix it permanently.
I'm going to set my wifi to turn off when my screen goes off to see if that helps as someone else noted somewhere.
Putting it in airoplane made overnight did not fix the issue. It woke up with bad sleep bug.
Sent from my PC36100 using XDA Appm
Hey guys,
Ive been using SetCPU since it was fixed for the sensation. I am curious about what others have observed as far as battery life and performance goes.
I am running 1.8Ghz with a cpu voltage of 1.265V (1265000uV).
My battery doesnt last all that long despite all the profiles I have in place. I guess thats expected when you overclock by that much, huh? Is it worth it though?
Im sure more of you out there have experience to share. Im not talking Quadrant and benchmark scores, Im talking REAL use!
I am beginning to feel that dropping the frequency down may be in order since I cannot find anything that needs 1.8Ghz to work on this phone. Plus, less frequency means less required voltage. That would mean more battery!
Anyone care to chime in with their overclock frequency + voltage and experience?
Matt
I'm also interested in getting some input from other users, without having to clog up the respective overclocking threads in the Development section.
Yesterday after work, I set up the 1.5Ghz Undervolted option using utking's tool (http://forum.xda-developers.com/showthread.php?t=1168707). I didn't see too much need to push my CPU too hard, and a stable UV always piques my interest.
I created a couple of basic SetCPU profiles (screen off, battery below 20%, temperature < 50C), and scaling on demand up to 1512Mhz, my phone took about another 7-8 hours of moderately heavy use (Navigation, Maps, Yelp, photos, etc.) before it was on its last legs.
Bear in mind that I've been using the Anker 1900mAh battery. But even after the OC-UV that I set up, this was at least as good of battery usage as before without any SetCPU or overclocking, and actually seemed to be a little bit better to be honest.
I just dropped down to 1.6Ghz @ 1.26V...
I couldn't keep my phone from freezing with anything less than 1.26V @ 1.6Ghz. We will see how this goes for a few days and compare to my previous 1.8Ghz @ 1.265V.
So far, speed seems to be FASTER than 1.8Ghz. My Quadrant score (only being used here for comparison reasons) was 2700 right off the bat, whereas 1.8Ghz would peak around 2700-2800 after several tries.
The carousel works much better! I can swipe through quickly and have it scroll with zero lag. I can also fast swipe and watch is spin nicely!
Google Earth and Maps is about the same as 1.8Ghz.
Before, at 1.8Ghz, I could probably get 16 hours out of my phone with VERY LIGHT use. (A few 2min phone calls, check mail throughout the day, check the web a little). I am charging the phone up and will report back later.
Matt
I went down from 1.62 UV to 1.5 UV,because I could not see any difference in general use nad in benchmarks(except Quadrant) and battery consumption is much bigger with 1.62 with exact same profiles on setCpu.I think that this is best compromise between speed and battery life and as far as I know this is native clock speed of 8260 Snapdragon.
mrg02d said:
I just dropped down to 1.6Ghz @ 1.26V...
I couldn't keep my phone from freezing with anything less than 1.26V @ 1.6Ghz. We will see how this goes for a few days and compare to my previous 1.8Ghz @ 1.265V.
So far, speed seems to be FASTER than 1.8Ghz. My Quadrant score (only being used here for comparison reasons) was 2700 right off the bat, whereas 1.8Ghz would peak around 2700-2800 after several tries.
The carousel works much better! I can swipe through quickly and have it scroll with zero lag. I can also fast swipe and watch is spin nicely!
Google Earth and Maps is about the same as 1.8Ghz.
Before, at 1.8Ghz, I could probably get 16 hours out of my phone with VERY LIGHT use. (A few 2min phone calls, check mail throughout the day, check the web a little). I am charging the phone up and will report back later.
Matt
Click to expand...
Click to collapse
How simple is it to temp-root and overclock? I am really only interested in overclocking to speed up Sense, and you said it performs better with a speed boost.
Sent from my HTC Sensation 4G using XDA Premium App
Very,very easy,just read this tread :http://forum.xda-developers.com/showthread.php?t=1168707
I have mine undervolted at VDD_1175000 and 1500MHz. I like it very much...makes a big difference in quadarant score, but not sure in reality how much faster. Battery life is fine (when the phone is on). Have not experiened FC's
I have modified eugenes batch file to push my kernel file and preferred speed after perma-temp-root and now is all in one click...
So here is a little update:
Its been about 8 hours since full charge and I am at 55% battery left.
1.6Ghz @ 1.26V, On Demand.
Ive been checking email, making a few calls, and surfing the net with both wifi and GPRS (was out of the network, away from 4G). I also played Angry birds for a little bit and showed off Google Maps and Earth to my Mom.
Now that Im back home, ive turned back on 4G...
Ive noticed a slight amount of hesitation while opening and closing things, but nothing bad.
I havent had any profiles kick in yet, but they will soon with the battery getting low. I will resist charging the phone and see if I make it through the night, using it as I need it. I will report tomorrow.
Matt
I must be doing something wrong.
I am rooted and running LeeDriods 1.2 Rom. When I launch setcpu the only options i have are 1000 mghz? WTF am i doing wrong?
i think at LeeDroid 1.2 you must not use setcpu but Demon control?
BigBoppa said:
i think at LeeDroid 1.2 you must not use setcpu but Demon control?
Click to expand...
Click to collapse
Hmmm, okay. I saw where set cpu was updated for the sensation in the latest release but it doesn't seem to support overclocking with this rom.
Shame that as i was about to get that rom and flash it onto my phone, glad i came in here now
Running Leedroid 1.0
Kernel @ 1.7ghz max, i have it set to 1.5ghz and 192mhz
Runs hot when i play finger racing or reckless racing or modern combat 2 or something intensive, but besides the heat, it drains just like it normally would to be honest, maybe a little quicker, definite speed increase in overall user experience, without an OC the rom just flops, better than stock but just crap, OC is needed.
I am just wondering why the Sensation is only hitting around 80MFlops at most in Linpack @ 1.5GHz. It seems that this phone should be getting 100+ every time. I have also noticed that the single core may score 55MFlops while the multicore will only raise that to 60MFlops.
Does it have something to do with the Java libraries? I also saw something about NEON libraries?
I think the highest I've ever scored with my Sensation is 91MFlops.
Thanks for any input on this guys. I know it's just a benchmark, but I'm just curious.
EDIT: Just scored a 27. I'm just gonna uninstall this app and use AnTuTu. How does this CPU ACTUALLY perform against other high end CPUs?
snelan said:
I am just wondering why the Sensation is only hitting around 80MFlops at most in Linpack @ 1.5GHz. It seems that this phone should be getting 100+ every time. I have also noticed that the single core may score 55MFlops while the multicore will only raise that to 60MFlops.
Does it have something to do with the Java libraries? I also saw something about NEON libraries?
I think the highest I've ever scored with my Sensation is 91MFlops.
Thanks for any input on this guys. I know it's just a benchmark, but I'm just curious.
EDIT: Just scored a 27. I'm just gonna uninstall this app and use AnTuTu. How does this CPU ACTUALLY perform against other high end CPUs?
Click to expand...
Click to collapse
Qualcomm actually has better floating point calculations than others, it's one of the benefits of the dragons. If you're using Sense, try linpack on Vanilla and it will be higher. Sense & benches don't go together b/c of something HTC screwed up.
http://www.nordichardware.com/news/...lcomm-demonstrates-new-msm8660-processor.html
what exactly is a floating point calculation?
Awesome, thanks!
By Vanilla, do you just mean a stock ROM without Sense? Would CM7 work? I just installed CM7 and set my min max to 1620/1620 and topped out at 101MFlops, but that was only out of 3 runs.
Does this also mean that Sense has slower real-world performance? Or just for benchmarks?
snelan said:
Awesome, thanks!
By Vanilla, do you just mean a stock ROM without Sense? Would CM7 work? I just installed CM7 and set my min max to 1620/1620 and topped out at 101MFlops, but that was only out of 3 runs.
Does this also mean that Sense has slower real-world performance? Or just for benchmarks?
Click to expand...
Click to collapse
Yeah cm or aosp, sense has always been slow with benches since the HTC hero. It doesn't play nice with any bench really, compare the same phone on sense and vanilla and scores are different. I don't know much about floating points but its probably related to how fast things are calculated. The hummingbird had a lower linpack than the scorpions of that time. And this doesn't mean sense is slower but it rapes the phone's ram excessively. When ics is out benches should be done to see how this phone really performs, most of the 1 core is sleep anyway or not at max.
Sent from my T-Mobile Sensation 4G using XDA App
One important thing:
When you score in multi-thread is just about 10 points above the single-thread score, it's because your second core (CPU1) didn't go online, or simply too late, so multi-thread was just using one core
Try this: with root explorer, go to sys/devices/system/CPU/CPU1/online
in this file replace 0 by one, then save. Immediatly set the permissions to read only (on "write" is ticked by default, untick it)
Now your both cores are always online, ready to scale up, and multi-thread scores will always be far superior to single-thread.
(it's not sticky, everything we go back as it was after a reboot)
Also my very best score with Linpak was 115, at 1.782ghz, CPU1 forced online, I posted a screenshot somewhere in Bricked kernel thread. This is a real battery drain and hot device to get this score
In real life, 90 or 95 at 1.5ghz is pretty good
i900frenchaddict said:
One important thing:
When you score in multi-thread is just about 10 points above the single-thread score, it's because your second core (CPU1) didn't go online, or simply too late, so multi-thread was just using one core
Try this: with root explorer, go to sys/devices/system/CPU/CPU1/online
in this file replace 0 by one, then save. Immediatly set the permissions to read only (on "write" is ticked by default, untick it)
Now your both cores are always online, ready to scale up, and multi-thread scores will always be far superior to single-thread.
(it's not sticky, everything we go back as it was after a reboot)
Also my very best score with Linpak was 115, at 1.782ghz, CPU1 forced online, I posted a screenshot somewhere in Bricked kernel thread. This is a real battery drain and hot device to get this score
In real life, 90 or 95 at 1.5ghz is pretty good
Click to expand...
Click to collapse
Wow! Never knew about that, I just assumed that the second core's clocks were off or something. Thank you very much good sir!
Hi guys, just got my HOS yesterday and am extremely happy with it! However something has been bugging me. Benchmarking scores on quadrant.
I know that we should all take benchmarking scores with a pinch of salt, as it does not determine the real world usage of the phone. But seeing everyone consistently getting high 4000s to 5000s on the net and me only getting low 4000s is worrying me.
When I first got my phone, I quickly set up my phone and once that was done i installed the quadrant to feast my eye. It scored 4600+, and I did several more times with reboots here and there and it was all nearly the same give and take 100+. I then did a factory reset wondering why I can't get the score any higher and ended up now only scoring from 4200 to 4500+ tops.
Wonder if there's a problem with my device?
Anyone?
Quadrant is basically a random number generator
Sent from my HTC One S using xda premium
Most of us have rooted devices with custom roms/kernels with their own little goodies and enhancements. Overclocking and changing the governor will generally increase benchmark scores.
@above, that is mostly true. This is a valid statement for any benchmarker.
I get 5013 on stock. Stop worrying. Quadrant is not accurate
Yeah just got my One S and experiencing the same, it's not lagging but I've read it should be 5000 stock but I've maxed at about 4600
I've gotten 4600 and 5200, then uninstalled.
blastx said:
Hi guys, just got my HOS yesterday and am extremely happy with it! However something has been bugging me. Benchmarking scores on quadrant.
I know that we should all take benchmarking scores with a pinch of salt, as it does not determine the real world usage of the phone. But seeing everyone consistently getting high 4000s to 5000s on the net and me only getting low 4000s is worrying me.
When I first got my phone, I quickly set up my phone and once that was done i installed the quadrant to feast my eye. It scored 4600+, and I did several more times with reboots here and there and it was all nearly the same give and take 100+. I then did a factory reset wondering why I can't get the score any higher and ended up now only scoring from 4200 to 4500+ tops.
Wonder if there's a problem with my device?
Click to expand...
Click to collapse
I got my US T-Mobile One S on release date with 1.53 build and got 5300+ on quadrant. Rooted with custom kernel I have gotten close to 5800.
I heard that the update to 1.84 build on stock phones brought the quadrant scores down to mid 4000s. So there is nothing wrong with your phones that were bought with 1.84 build.
Sent from my HTC One S using Xparent Cyan Tapatalk 2
stemnin said:
I've gotten 4600 and 5200, then uninstalled.
Click to expand...
Click to collapse
Read in several other threads everyone else is experiencing the issue, so the benchmark drop after the HTC system update is completely normal. The update improves network and signal but apparently lowers quadrant scores, the device is not noticeably slower though.