[Q]How accurate is Smartbench 2010? - Nexus S Q&A, Help & Troubleshooting

Hi would like to know if smartbench is accurate. Cause my games index point seems to be higher than other device by alot. 3095 pts
Sent from my Nexus S using XDA Premium App

dplate07 said:
Hi would like to know if smartbench is accurate. Cause my games index point seems to be higher than other device by alot. 3095 pts
Sent from my Nexus S using XDA Premium App
Click to expand...
Click to collapse
i dont trust smartbench. i get games indexes from 6000-1000 while overclocked at 1.544ghz. generally, benchmarks do not mean anything, but are nice to use just to compare devices.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}

Oh that's interesting. simms22, do you get that result every time you run the app or just occasionally? That is a bug that needs to be fixed!
Thanks.

Acei said:
Oh that's interesting. simms22, do you get that result every time you run the app or just occasionally? That is a bug that needs to be fixed!
Thanks.
Click to expand...
Click to collapse
i get a constant 5000-10000. but it ranges between 5000-10000 on every run. its not consistent around a certain score.

Acei said:
Oh that's interesting. simms22, do you get that result every time you run the app or just occasionally? That is a bug that needs to be fixed!
Thanks.
Click to expand...
Click to collapse
i figured theres something wrong, so i havent been publicizing these.
lol, heres a few more

Acei said:
Oh that's interesting. simms22, do you get that result every time you run the app or just occasionally? That is a bug that needs to be fixed!
Thanks.
Click to expand...
Click to collapse
Are you going to throw out all results and go reestablish a new baseline for every device?

Acei said:
Oh that's interesting. simms22, do you get that result every time you run the app or just occasionally? That is a bug that needs to be fixed!
Thanks.
Click to expand...
Click to collapse
morfic said:
Are you going to throw out all results and go reestablish a new baseline for every device?
Click to expand...
Click to collapse
i have to wonder the same the same thing. are ALL results invalid now? if there is a bug, then it cant be trusted with any device until the bug is squashed. who knows which results are valid or not.

Results are relatively easy to hash out, at least so far. These high index scores are caused by one or more sub-tests that somehow fails and ends up with unrealistic scores (well into millions). The server submit script rejects these results so they don't really make it to the DB (although I will check it out manually to make sure this is the case). I do have a script that runs nightly and looks for other anomalies as well.
As you can see in this link, you won't find your scores in the DB:
http://smartphonebenchmarks.com/ind...bench2011:Games&filter_cpu=all&filter_gpu=all
The real question though is why it is reporting these high values - I need to know which test fails and reports false numbers when it is performing the final calculation.
Going back to the same question - is it happening to you on every single test run, or does it only happen occasionally? You are the first person to report this so far.
Thanks again.

morfic said:
Are you going to throw out all results and go reestablish a new baseline for every device?
Click to expand...
Click to collapse
I would do this gladly if I find any evidences that the results contain garbage. So far though, the results either look good or they are close to infinite (very small % of the runs cause this though) so it's pretty easy to figure out which ones are wrong.

Acei said:
I would do this gladly if I find any evidences that the results contain garbage. So far though, the results either look good or they are close to infinite (very small % of the runs cause this though) so it's pretty easy to figure out which ones are wrong.
Click to expand...
Click to collapse
Do you have a way to proof that a 3600 when usual avg is 3400 is not caused by a minor glitch?
Why not change the app behavior to not submit anything when something failed?
How much can the current results be trusted if we rely on "i think nothing else but the 'obvious' ones went through"?

Acei said:
Results are relatively easy to hash out, at least so far. These high index scores are caused by one or more sub-tests that somehow fails and ends up with unrealistic scores (well into millions). The server submit script rejects these results so they don't really make it to the DB (although I will check it out manually to make sure this is the case). I do have a script that runs nightly and looks for other anomalies as well.
As you can see in this link, you won't find your scores in the DB:
http://smartphonebenchmarks.com/ind...bench2011:Games&filter_cpu=all&filter_gpu=all
The real question though is why it is reporting these high values - I need to know which test fails and reports false numbers when it is performing the final calculation.
Going back to the same question - is it happening to you on every single test run, or does it only happen occasionally? You are the first person to report this so far.
Thanks again.
Click to expand...
Click to collapse
i get them every time. but the lower the ghz kernel i use, the lower the score.
as far as i remember(i can be wrong) i had to submit my scores to the site to get added to the DB, but that was months ago. btw, i did get these scores a few weeks ago, but never posted them anywhere because they are just too ridiculous.

morfic said:
Do you have a way to proof that a 3600 when usual avg is 3400 is not caused by a minor glitch?
Click to expand...
Click to collapse
I can't guarantee that every single test run results stored in DB are accurate of course. That's one of the reason why I am continuously capturing more results to determine more accurate average results. But you are right, 3600 vs 3800, there is no way to prove that.
Why not change the app behavior to not submit anything when something failed?
How much can the current results be trusted if we rely on "i think nothing else but the 'obvious' ones went through"?
Click to expand...
Click to collapse
Yes, I have been doing this for a while now. Smartbench is an app written by a human after all. Just as other apps I have written since the past 15 years, this one has bugs too. I have been fixing bugs and I am determined to fix new ones that get discovered as well, that's why I am here right now.
The server-side scripts are there just in case more bugs are discovered on the client-side code. The goal is, as you mentioned here already, fix it at the root which is in the app itself.

simms22 said:
i get them every time. but the lower the ghz kernel i use, the lower the score.
Click to expand...
Click to collapse
Hmm, that is indeed interesting.
Did you get this from day 1 or after some mods?
as far as i remember(i can be wrong) i had to submit my scores to the site to get added to the DB, but that was months ago. btw, i did get these scores a few weeks ago, but never posted them anywhere because they are just too ridiculous.
Click to expand...
Click to collapse
Yes those scores are too ridiculous and I do intend to find out why it is reporting them, on your particular phone and not on others.
Do you let them run without any interruptions?
A while back, some have found a way to "cheat" the app by performing certain actions so I had to add some code to check for those.
Going back to your run, I just had a look in the DB. I can see one of your runs - you got high scores because your JellyFish score was really high. Did your Jelly Fish test run properly from the beginning to the end? All I'm doing here is capturing a time before the test and another one at the end, down to a nano-second accuracy. I could easily reject these results on the client side and report invalid runs but I would prefer to find out why it is causing this in the first place...

Hi guys,
I have just released a v1.2.1 which fixed the problem described here.
If you are interested, you can check out the following article (I wrote it) here:
http://bit.ly/lfxS1a
Thanks!
EDIT: For some strange reason, I don't see the latest version in the Android Market yet. In the past, the update has always been instantaneous. I hope this one gets updated soon.
EDIT2: I see the update now!

Acei said:
Hi guys,
I have just released a v1.2.1 which fixed the problem described here.
If you are interested, you can check out the following article (I wrote it) here:
http://bit.ly/lfxS1a
Thanks!
EDIT: For some strange reason, I don't see the latest version in the Android Market yet. In the past, the update has always been instantaneous. I hope this one gets updated soon.
EDIT2: I see the update now!
Click to expand...
Click to collapse
Coming to think of it, we saw similar in the past, where high scores caused a rejection of the scores even if everything finished through.
Why not add a failed score as a zero, total number of scores / number of scores == avg
Not just succeeded scores, this way devices where tests fail will score low, as would be expected.
BUT, if tests fail, can you be verbose what failed, how far it got before it failed?
I have not seen 2011 complete on my phones in a long time (i mean not gotten a "valid score") and i would like to know where the problem is. And none of those with gfx show signs of early exits, numerical tests i can't get visual clues from.
Thanks,
Daniel

Acei said:
Hi guys,
I have just released a v1.2.1 which fixed the problem described here.
If you are interested, you can check out the following article (I wrote it) here:
http://bit.ly/lfxS1a
Thanks!
EDIT: For some strange reason, I don't see the latest version in the Android Market yet. In the past, the update has always been instantaneous. I hope this one gets updated soon.
EDIT2: I see the update now!
Click to expand...
Click to collapse
i just updated and tested. im still getting high scores.for example, overclocked to 1.1ghz i just scored a 12000+ on games. lol, thats just funny. but i do see the jellyfish test as faster than id expect. but not that much. interesting to note, i had a few failed benchmarks this time. but more high scores after.

morfic said:
Coming to think of it, we saw similar in the past, where high scores caused a rejection of the scores even if everything finished through.
Why not add a failed score as a zero, total number of scores / number of scores == avg
Not just succeeded scores, this way devices where tests fail will score low, as would be expected.
Click to expand...
Click to collapse
I'm pretty sure I'll get arguments from both sides on this. Some units will overclock better than others (even if they are the same devices/models), so for those lucky owners with more headroom will want to see results that don't include failed results. Or at least that is my opinion. At the end of the day, you guys will need to tell me which makes more sense.
In the worst case, I can implement both, and allow you to select which "mode" you want to see.
BUT, if tests fail, can you be verbose what failed, how far it got before it failed?
I have not seen 2011 complete on my phones in a long time (i mean not gotten a "valid score") and i would like to know where the problem is. And none of those with gfx show signs of early exits, numerical tests i can't get visual clues from.
Thanks,
Daniel
Click to expand...
Click to collapse
This is definitely possible. There are currently several checks I do to see if test runs are valid:
1. If an app has lost its focus, I assume the test is invalid. This was done to prevent people from switching out of the app during the test run and end up with some funky scores. Somehow users were getting interesting scores!
2. If any of the test suit throws exceptions (i.e. crashes), then I assume the test is invalid. As you mentioned here already, it is possible to find out at least which test suit failed.
It appears that it is still possible to end the test early without crashing as I have witnessed in this thread. So I will look further into other possibilities as well.

simms22 said:
i just updated and tested. im still getting high scores.for example, overclocked to 1.1ghz i just scored a 12000+ on games. lol, thats just funny. but i do see the jellyfish test as faster than id expect. but not that much. interesting to note, i had a few failed benchmarks this time. but more high scores after.
Click to expand...
Click to collapse
Very interesting! So that means your result is caused by not one but two issues. I may have to go as far as checking to see if it has completed rendering the right # of frames during the test.
simms/morfic, would you guys mind testing for few more scenarios for me? I can produce a special build to do some further experiment.
Thanks much!

Acei said:
Very interesting! So that means your result is caused by not one but two issues. I may have to go as far as checking to see if it has completed rendering the right # of frames during the test.
simms/morfic, would you guys mind testing for few more scenarios for me? I can produce a special build to do some further experiment.
Thanks much!
Click to expand...
Click to collapse
I've always been a fan of your work actually being on the forums and working bugs out. There aren't many devs dedicated like you so anything you want to test, hit me up as well, fortunately/unfortunately mine hasn't shown one of these erroneous scores so I'm not sure if my phone will help, my game index score is always in the 3000 range.
Sent from my Nexus S 4G using Tapatalk

dreamsforgotten said:
I've always been a fan of your work actually being on the forums and working bugs out. There aren't many devs dedicated like you so anything you want to test, hit me up as well, fortunately/unfortunately mine hasn't shown one of these erroneous scores so I'm not sure if my phone will help, my game index score is always in the 3000 range.
Sent from my Nexus S 4G using Tapatalk
Click to expand...
Click to collapse
Thanks for the offer.
Even a positive confirmation that the latest build works is helpful.

Related

[Q] please post your quadrant score

I saw a store rep demo'ed it and got 2700, and the 3D got whopping 100+fps on "walking up room" part.
It is exceptional.
mingkee said:
I saw a store rep demo'ed it and got 2700, and the 3D got whopping 100+fps on "walking up room" part.
It is exceptional.
Click to expand...
Click to collapse
I ran the first quadrant within thirty minutes of getting the phone, and it scored 2554 or something like that.
Just ran one and scored... 2463
you have to realize this is the same chip as the XOOM and Atrix but its powering a lower resolution so the quadrant scores will be higher BUT you're at a lower resolution. I have a Xoom and it gets 1800 while this gets 2500. I'd rather have a higher resolution than a higher score. Still you shouldn't have any problems playing games on here.
Consecutive runs with no tweaking at all, background sync on and heavy live wallpaper running.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
It's a 3000 phone, not a 2700 phone...
These scores mean nothing except its a very fast phone. People shouldn't get hung up on benchmarks so much.
Sent from my LG-P999 using XDA App
mobilehavoc said:
These scores mean nothing except its a very fast phone. People shouldn't get hung up on benchmarks so much.
Sent from my LG-P999 using XDA App
Click to expand...
Click to collapse
Nobody seems to be getting hung up in this thread?
mobilehavoc said:
These scores mean nothing except its a very fast phone. People shouldn't get hung up on benchmarks so much.
Sent from my LG-P999 using XDA App
Click to expand...
Click to collapse
After scoring 1700 with my Legend with data2ext, I realized that Quadrant scores don't necessarily don't mean much. What I want to see is how much this phone scores in GLBenchmark.
Sent from my HTC Legend using XDA Premium App
mobilehavoc said:
These scores mean nothing except its a very fast phone. People shouldn't get hung up on benchmarks so much.
Sent from my LG-P999 using XDA App
Click to expand...
Click to collapse
I don't get your point. How does it mean nothing but also means it's fast? It's a benchmark test. The point of it is to gauge performance, nothing more. Would you have bought one if it scored 850? After all... its a meaningless number.
djmcnz said:
Consecutive runs with no tweaking at all, background sync on and heavy live wallpaper running.
...
It's a 3000 phone, not a 2700 phone...
Click to expand...
Click to collapse
This is the G2X forum so I think we want scores from that phone only not the O2X.
Also, 3000 + with no tweaking? Really? Is this with stock rom? If not I belive a non stock rom constitutes tweaking.
atkhobby said:
This is the G2X forum so I think we want scores from that phone only not the O2X.
Also, 3000 + with no tweaking? Really? Is this with stock rom? If not I belive a non stock rom constitutes tweaking.
Click to expand...
Click to collapse
It's the same hardware. The benchmarks will be the same except for any software quirks. Not sure about tweaking, I interpret that to mean "optimising for benchmarks". The only real alternative rom available is Paul's and all he's really done is turn the engineering code off, not really a tweak. Nevertheless, I expect the G2x won't be encumbered by these LG oversights (hope?) so should bench around 3000 out of the box.
The point I was making was perfectly relevant for this thread I think?
djmcnz said:
It's the same hardware. The benchmarks will be the same except for any software quirks. Not sure about tweaking, I interpret that to mean "optimising for benchmarks". The only real alternative rom available is Paul's and all he's really done is turn the engineering code off, not really a tweak. Nevertheless, I expect the G2x won't be encumbered by these LG oversights (hope?) so should bench around 3000 out of the box.
The point I was making was perfectly relevant for this thread I think?
Click to expand...
Click to collapse
I don't think there's any problem with you adding the comparative benchmarks from the O2x to the discussion. However, I DO think you should have been very clear about what those benchmarks were from so people aren't wondering why their phone out of the box isn't performing as well as yours.
Also, I'm still not entirely clear from your response. Are you running a custom rom or not?
MWBehr said:
I don't think there's any problem with you adding the comparative benchmarks from the O2x to the discussion. However, I DO think you should have been very clear about what those benchmarks were from so people aren't wondering why their phone out of the box isn't performing as well as yours.
Also, I'm still not entirely clear from your response. Are you running a custom rom or not?
Click to expand...
Click to collapse
I'd hate for anyone to mmisinterpret, I'm running stock LG ROM with no engineering code. What I mean by that is that it's basic Google code but it *does* include the LG framework, so probably slower than stock Android.
Edit: It would be clear to even a casual observer from the screen shots that it was not stock Froyo.
djmcnz said:
I'd hate for anyone to mmisinterpret, I'm running stock LG ROM with no engineering code. What I mean by that is that it's basic Google code but it *does* include the LG framework, so probably slower than stock Android.
Edit: It would be clear to even a casual observer from the screen shots that it was not stock Froyo.
Click to expand...
Click to collapse
I think it was the "no tweaks" part that was confusing. I'm also wondering if the G2x will suffer from the same engineering code issue that the O2x has. Originally I thought we would be getting AOSP, but apparently it's still LG manipulated, just without their launcher and probably a few other things.
MWBehr said:
I think it was the "no tweaks" part that was confusing. I'm also wondering if the G2x will suffer from the same engineering code issue that the O2x has. Originally I thought we would be getting AOSP, but apparently it's still LG manipulated, just without their launcher and probably a few other things.
Click to expand...
Click to collapse
Oh? That's quite interesting news to me, thanks. That's a bit of a bummer actually that its not ASOP, I was looking forward to that. Let's hope the cut is shallow and easy to clean.
I hear CM7 is running on x2x so this all may become moot?
Sorry, I didn't read further.. I see the answers here in the forum. Please ignore this post
djmcnz said:
Consecutive runs with no tweaking at all, background sync on and heavy live wallpaper running.
It's a 3000 phone, not a 2700 phone...
Click to expand...
Click to collapse
Just curious, how did you get the black status bar at the top of your phone with the blue icons... Cause I want to do that with mine asap. Thanks
luckycook82 said:
Just curious, how did you get the black status bar at the top of your phone with the blue icons... Cause I want to do that with mine asap. Thanks
Click to expand...
Click to collapse
In this case you just need to take the png files from the European 2x framework-res.apk and use them to replace the equivalent in the US 2x. Always backup first. /offtopic
djmcnz said:
In this case you just need to take the png files from the European 2x framework-res.apk and use them to replace the equivalent in the US 2x. Always backup first. /offtopic
Click to expand...
Click to collapse
Can you do a bench on Nenamark? Curious to see what the FPS is on that test?
jlevy73 said:
Can you do a bench on Nenamark? Curious to see what the FPS is on that test?
Click to expand...
Click to collapse
Just one run...
quadrant score
My quadrant score with pauls modaco rom is 3200+ which i am extremely happy with. on stock rom i was getting 26-2700.
>_<
O2x owners get the hell out of our forum! Your devices are programmed differently than ours, you've most likely tweaked your devices in some way (who comes to XDA and does not do so?), and you're confusing the situation.
Out of the box, the G2x from T-MOBILE USA scores about 2400-2500 on Quadrant. End of story.

Just something for you Quadrant lovers to think about

Found and interesting article and i thought i'd throw it out there. Basically when it comes to quadrants you will have two very polarizing point of views. You will have those that LOVE the program and they must test everything. And those that swear never to use it. But i found an interesting command line from the CMD prompt.
Code:
mount -t tmpfs tmpfs /data/data/com.aurorasoftworks.quadrant.ui.standard
Long story short it mounts the Quadrant Standard application on the RAM of the phone so you can get a true reading. Since the I/O is what bottle necks everything
I ran three test with the QS after three test i got a high of 2528, and a low of 1918. 610 point difference. Not to shabby for running CM7N75 O/C'd to 1516 with a performance governor
Then i mounted QS on the RAM of the phone and ran three test. The highest being 2997 and the lowest being 2785. 212 point difference.
just throwing out some food for thought.
Why do people always cry about Quadrant? those lil silly numbers means nothing I love how all the idiots who review the roms and make videos always do it and gives general public this idea that its real.
5th March 2011 said:
As I said before as syntactic benchmark means nothing as it does not translate in to real world performance. If someone wanted to really inflate it lol they wouldn't have to do much other than allocate it on tfs or ram as they will get higher score on I/O and R/W which will inflate the score which is very easy to spot as its never consistent with regular.
Click to expand...
Click to collapse
UsrBkp said:
Why do people always cry about Quadrant? those lil silly numbers means nothing I love how all the idiots who review the roms and make videos always do it and gives general public this idea that its real.
Click to expand...
Click to collapse
first and foremost - personal attacks aren't needed, wanted, or warranted.
I am sorry if you thought I personally attacked you but I assure you that was not the case. As I was speaking more in general, but what I said still stands.
UsrBkp said:
I am sorry if you thought I personally attacked you but I assure you that was not the case. As I was speaking more in general, but what I said still stands.
Click to expand...
Click to collapse
I know you weren't coming after me. But still in general. We are here to help one another and expand horizons. Not belittle each other.
UsrBkp said:
Why do people always cry about Quadrant? those lil silly numbers means nothing I love how all the idiots who review the roms and make videos always do it and gives general public this idea that its real.
Click to expand...
Click to collapse
I personally don't use quadrant, but any time someone talks about performance they can either post quadrant/linpack results, or they can say "it's really fast." You can't call someone who posts quadrant results an idiot unless you call anyone who has ever commented on a phone's performance an idiot. We talk about this stuff all the time, no need to demean anyone unless you have the one-stop end all of performance gauges.
I love it when people talk and talk and talk but they never listen. As once again the message was lost, which still exist on the original message. I don't know where he got the idea I was attacking him as that maybe some type of paranoia or delusion.
darinmc that will never happen due to the different hardware archstructure. Even when device is using ARMv7 the instruction set is interpreted differently from manufacture to manufacture. Great example is Snapdragon vs Hummingbird where NEON is utilized to improve the IOPs. If its allocated on top of davlik it can be cheated its simple as that.
UsrBkp said:
I love it when people talk and talk and talk but they never listen. As once again the message was lost, which still exist on the original message. I don't know where he got the idea I was attacking him as that maybe some type of paranoia or delusion.
Click to expand...
Click to collapse
Sorry for the typo
i guess that " n't " means alot. Sorry again.
Quadrant has become very popular with people to see where their phones stand performance wise, but at times I find it not to be the most accurate...Unlike Neocore or Linpack which I think are better ways to test GPU/CPU instead. The scoring system may need some work to make it balanced. Personally I think it needs work, earlier I tried Faux's Ginger rom and it scored about the same as the stock MT4G rom...This is clearly odd since the ginger is lighter and actually has a much much smoother experience and higher response times. Quadrant doesnt deal with real world usage. Feel free to disagree if you feel differently.
how about smartbench, is that any better than quadrant?
clarknick27 said:
how about smartbench, is that any better than quadrant?
Click to expand...
Click to collapse
From what I seen Quadrant favors Snapdragon while Smartbench favors Hummingbird. But try GLBenchmark thats what we mostly use as its more comprehensive.
"As I said before as syntactic benchmark means nothing as it does not translate in to real world performance." - HKM
UsrBkp said:
From what I seen Quadrant favors Snapdragon while Smartbench favors Hummingbird. But try GLBenchmark thats what we mostly use as its more comprehensive.
"As I said before as syntactic benchmark means nothing as it does not translate in to real world performance." - HKM
Click to expand...
Click to collapse
thanks Ill give it a shot
UsrBkp said:
From what I seen Quadrant favors Snapdragon while Smartbench favors Hummingbird. But try GLBenchmark thats what we mostly use as its more comprehensive.
"As I said before as syntactic benchmark means nothing as it does not translate in to real world performance." - HKM
Click to expand...
Click to collapse
Hahahaha, where have you been? Thought you retire drop by Sensation section help me out with the SamSux troll over there
Sent from my HTC Vision using XDA Premium App
epsix said:
Hahahaha, where have you been? Thought you retire drop by Sensation section help me out with the SamSux troll over there
Sent from my HTC Vision using XDA Premium App
Click to expand...
Click to collapse
LOL I see people still recognize me. I mean hopefully "they" don't know you know who, as ill try to keep low profile for now. Only reason I came back was due to the recent interest in "fail-pu" which I was trying to shed some lights in here few months back and got myself you know what in CM7NB thread. Hopefully "they" don't go crazy and start issuing you know what as originally ordered by the 2 heads. I am sure MT4G community will suffer if they do it, I mean here I hacked the mmcblk and posted the info and risked my device to help others yet Mr New.Sheriff wanted to show himself as the big man. Oooh well ill help ya much as I can and look in to the Sensation section.
What makes you think that mounting the application in RAM (thus minimizing I/O interactions) gives you a "true reading"? It's not like all your applications are stored in RAM. They access the file system also. So the speed of your phone's filesystem obviously affects the overall speed of the phone and should be included in a good benchmark. Unless you're only interested in comparing CPU/GPU speeds.
sundayhustler said:
What makes you think that mounting the application in RAM (thus minimizing I/O interactions) gives you a "true reading"? It's not like all your applications are stored in RAM. They access the file system also. So the speed of your phone's filesystem obviously affects the overall speed of the phone and should be included in a good benchmark. Unless you're only interested in comparing CPU/GPU speeds.
Click to expand...
Click to collapse
I don't think anyone in this thread said allocating it to tfs gives true reading. All the OP was doing was posting how one can easily hack the score. Its which we known ages ago but the point was when you make an standard everyone must follow it. So obviously if you allocate the whole ROM in RAM instead of NAND or SDCard the IOPs will always be higher.
If its software it can be altered simple as that and someone will always do so and try to pretend they have the legit score but for people like me we can easily tell what is real and what is fake. The legit max score verified by me was 3618 I think without any type of hack all I did was strip the rom and made it cleaner. Which you folks can get around 3200-3400 using AOSP with no problem. Now if you scoring 3800-4000 then well you know whats going on. I am not going to name folks but come on they aint fooling anyone.
neidlinger said:
Found and interesting article and i thought i'd throw it out there. Basically when it comes to quadrants you will have two very polarizing point of views. You will have those that LOVE the program and they must test everything. And those that swear never to use it. But i found an interesting command line from the CMD prompt.
Code:
mount -t tmpfs tmpfs /data/data/com.aurorasoftworks.quadrant.ui.standard
Long story short it mounts the Quadrant Standard application on the RAM of the phone so you can get a true reading. Since the I/O is what bottle necks everything
I ran three test with the QS after three test i got a high of 2528, and a low of 1918. 610 point difference. Not to shabby for running CM7N75 O/C'd to 1516 with a performance governor
Then i mounted QS on the RAM of the phone and ran three test. The highest being 2997 and the lowest being 2785. 212 point difference.
just throwing out some food for thought.
Click to expand...
Click to collapse
Off topic, but what font are you using in those screenies? And a link maybe?
UsrBkp said:
darinmc that will never happen due to the different hardware archstructure. Even when device is using ARMv7 the instruction set is interpreted differently from manufacture to manufacture. Great example is Snapdragon vs Hummingbird where NEON is utilized to improve the IOPs. If its allocated on top of davlik it can be cheated its simple as that.
Click to expand...
Click to collapse
That was pretty much my point, that an end-all doesn't exist. I'm no android pro, just think that if nothing can define performance then everything is game. It's all subjective anyway. My phone, for instance, is really really really fast. Yours is probably just really fast. Mine is 2 really's faster. It's how I roll.
darinmc said:
That was pretty much my point, that an end-all doesn't exist. I'm no android pro, just think that if nothing can define performance then everything is game. It's all subjective anyway. My phone, for instance, is really really really fast. Yours is probably just really fast. Mine is 2 really's faster. It's how I roll.
Click to expand...
Click to collapse
lol and that kind of the reason is why we have Quadrant in first place. Its all about false sense of security as the owner of the device is just simply fooling themselves. Now I am not sure if you know as all chips are different which actually depends on each wafer. But in perfect world with same defect rate as 2 let say you and me both have same phone. You running ROM X1.0 and I am also running ROM X1.0 and with the same settings. Now if you score 100 and I score 105 thats discrepancy. For it to be truly applicable it has to yield the same result over and over. Now thinking it would yield different result is known as insanity unless your fan of quantum mechanics that is yet ironically do to that we have chips today lol.
Think of it as 2 Fords they both running in the same road same model and one of them goes ahead. They are still bound by the same hardware but do to other variables it yields different results. Now those variables are not always predefined and it can be altered without any hardware modifications. But for it to be standard it has to be same. Now if you also have a Ford but have 300mph v8 under the hood you just cheated and inflated the score.
I am not sure if what I am saying is making any sense to anyone but to think your MT4G is better than someone else's is just crazy lol good luck putting that on ebay/cl saying you scored +300points extra on quadrant than other owners see how that goes for ya IRL.
my 3800 quadrant phone sold for eleventy million dollars thank you very much (it came with a case).

CF Bench (More accurate benchmark)

http://www.xda-developers.com/android/cf-bench-for-android-devices/
Anyone check it out yet?
You gotta scroll all the way down in the results page to get your overall score.
i installed this program scored a 2210 not bad i guess
FBis251 said:
http://www.xda-developers.com/android/cf-bench-for-android-devices/
Anyone check it out yet?
You gotta scroll all the way down in the results page to get your overall score.
Click to expand...
Click to collapse
Apparently the dev just wants to get some initial data before tweaking the app. I think it'll eventually get to be more accurate than Quad. I saw a post about someone getting a 10k score with only a 1.5ghz OC using CM7 on a tablet... Yeah, made the dev post up a comment saying that he'd fix the bug... Haha.

[Q] Quadrant scores?

Hello. I've bought a nexus s a few days ago and I've started swapping roms.
My question is simple, how do people manage to get 4-5k quadrant scores?
Currently I'm using Pure Bionix 1.6 rom, but the maximum score I was able to get was around 3.3K.
Describe everything you do, from roms/kernels to apps closed etc.
Sorry if this is a newbie question
Install 1.6.1, then use SetCPU to OC the kernel.
That's how I achieved my 5300 quad.
You can also try other kernels as well.
YMMV
Use matrix rom amd just over clock abd run quadrant a few times back to back and you will hit 4500+
Sent from my Nexus S 4G using XDA Premium App
cyanogenmod 7 + trinity 1.44
4500-5000
Oxygen 2.3.1 @ 1300Mhz ->> 4286, highest score for me with this ROM.
And I didn't do anything special to get this score, I only overclocked CPU.
Thanks for the answers. I've done as you've said krylon. I installed 1.6.1 and oced the cpu to 1540 1540 on demand, however the max score I was able to get was around 3400. Am I missing something?
I'm using adv task killer to kill all the running apps before the test.
My nexus s is an i9023 btw.
spartanpg said:
Thanks for the answers. I've done as you've said krylon. I installed 1.6.1 and oced the cpu to 1540 1540 on demand, however the max score I was able to get was around 3400. Am I missing something?
I'm using adv task killer to kill all the running apps before the test.
My nexus s is an i9023 btw.
Click to expand...
Click to collapse
set setcpu to 1540 max/1540 low. bench four times in a row, throw out the first 2 results, the forth should be the highest.
That's what I'm doing atm, yet the max I get is 3400.
Me and my luck with cell phones
spartanpg said:
That's what I'm doing atm, yet the max I get is 3400.
Me and my luck with cell phones
Click to expand...
Click to collapse
its possible that your phone doesnt like that high of a overclock. try trinity t144 kernel. also, have you updated to the most recent bionix?
I flashed the T144 kernel and the phone got a bit slower yet really stable. The thing is the scores went down a little bit, around 3250 which I already expected. Could it be the phone itself?
I've youtubed a bit to watch how other 4000+ runs went and on the first image teste (2d if I'm not mistaken) I'm obtaining a constant 27-28 fps rate, and that was the only difference.
No need to worry about that. Quadrant scores can be faked and really don't mean much when it comes down to it. It's all about the real world performance.
spartanpg said:
I flashed the T144 kernel and the phone got a bit slower yet really stable. The thing is the scores went down a little bit, around 3250 which I already expected. Could it be the phone itself?
I've youtubed a bit to watch how other 4000+ runs went and on the first image teste (2d if I'm not mistaken) I'm obtaining a constant 27-28 fps rate, and that was the only difference.
Click to expand...
Click to collapse
are you using quadrant advanced or standard? if youre using quadrant advanced, could you provide a screenshot please. id really like to see the score breakdown. the score is really low. also, in the op of the Trinity thread in the development section is a dalvik-wiper.zip, would you flash it please, then try again.
---------- Post added at 01:47 PM ---------- Previous post was at 01:44 PM ----------
qbking77 said:
No need to worry about that. Quadrant scores can be faked and really don't mean much when it comes down to it. It's all about the real world performance.
Click to expand...
Click to collapse
orgasms can be faked too, that doesnt mean anybody is faking there quadrant scores here. with such a low quadrant score for these kernels, i bet his real world performance is reduced also.
simms22 said:
orgasms can be faked too, that doesnt mean anybody is faking there quadrant scores here. with such a low quadrant score for these kernels, i bet his real world performance is reduced also.
Click to expand...
Click to collapse
While not necessarily being faked you can't say that a test where you throw out the first three scores and then use the forth because it is always higher is terribly accurate or being honest. Just because you ran quadrant four times and got a score with 500 more points on the last test doesn't automatically mean your real world performance is that much better. Honestly it's just the Android equivlant of buying a really big truck to overcompensate for something else (not to mention the cause of people always asking these types of questions over and over).
So after searching around I found some1 in xda with a similar problem and I ended using this in the terminal:
su(press enter)
mount -o remount,noauto_da_alloc /data /data(press enter)
And believe it or not, after that I managed to get 4500 right away in standard quadrant 1.1.7. After closing apps and getting advanced quadrant 1.1.1, I managed to get 4900, which is more than fine for me. What I'd like to know is what exacly did that command do.
Regarding the dalvik-wiper.zip, I wasn't able to find it. Will it by any means increase the performance further?
Finally I'd like to ask two things:
1) How safe is it to run a Min 100 Max 1440 OC setting?
2) In case of warranty issues, will this fully revert the phone to the original (or updated from original) state? - http://nexusshacks.com/nexus-s-hacks/how-to-unroot-nexus-s-i9020ti9020ai9023/
Thanks for all the help so far
spartanpg said:
So after searching around I found some1 in xda with a similar problem and I ended using this in the terminal:
su(press enter)
mount -o remount,noauto_da_alloc /data /data(press enter)
And believe it or not, after that I managed to get 4500 right away in standard quadrant 1.1.7. After closing apps and getting advanced quadrant 1.1.1, I managed to get 4900, which is more than fine for me. What I'd like to know is what exacly did that command do.
Regarding the dalvik-wiper.zip, I wasn't able to find it. Will it by any means increase the performance further?
Finally I'd like to ask two things:
1) How safe is it to run a Min 100 Max 1440 OC setting?
2) In case of warranty issues, will this fully revert the phone to the original (or updated from original) state? - http://nexusshacks.com/nexus-s-hacks/how-to-unroot-nexus-s-i9020ti9020ai9023/
Thanks for all the help so far
Click to expand...
Click to collapse
i didnt even consider that because its included in the bionix/trinity combo. but it does have to be included in a stock rom. whats odd is, how did you lose it
---------- Post added at 03:26 PM ---------- Previous post was at 03:21 PM ----------
onigiri1692 said:
While not necessarily being faked you can't say that a test where you throw out the first three scores and then use the forth because it is always higher is terribly accurate or being honest. Just because you ran quadrant four times and got a score with 500 more points on the last test doesn't automatically mean your real world performance is that much better. Honestly it's just the Android equivlant of buying a really big truck to overcompensate for something else (not to mention the cause of people always asking these types of questions over and over).
Click to expand...
Click to collapse
you are in a way right and in a way youre wrong, im not going to argue the fact that all benchmarks are inaccurate. because they are. and ive proved it over and over again. BUT, that said.. its a good way to see where the phones performence is and where it should be(if youre not cheating), especially if youre a developer. for example, when morfic makes some changes in a kernel, we run benchmarks to see how and where that change effected the performance.
simms22 said:
you are in a way right and in a way youre wrong, im not going to argue the fact that all benchmarks are inaccurate. because they are. and ive proved it over and over again. BUT, that said.. its a good way to see where the phones performence is and where it should be(if youre not cheating), especially if youre a developer. for example, when morfic makes some changes in a kernel, we run benchmarks to see how and where that change effected the performance.
Click to expand...
Click to collapse
I was just making the point that most of the super high quadrant scores are inflated and not very accurate, but I agree with you in that benchmarks are a good way to make general comparisons like the example you described, but it isn't an end all, be all. When it comes down to it real world, user experience in my opinion is what really matters.
onigiri1692 said:
I was just making the point that most of the super high quadrant scores are inflated and not very accurate, but I agree with you in that benchmarks are a good way to make general comparisons like the example you described, but it isn't an end all, be all. When it comes down to it real world, user experience in my opinion is what really matters.
Click to expand...
Click to collapse
real world use is all that matters, period. but quadrants are still fun to run, as long as no body is cheating. speaking of, this is from a galaxy s 2 at 1.6ghz, the nexus s has officially been dethroned from the top of quadrant.. http://androidcommunity.com/samsung...-benchmarks-nearly-7800-in-quadrant-20111110/
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Any chance I could get these answered please? Would really apreciate it.
1) How safe is it to run a Min 100 Max 1440 OC setting?
2) In case of warranty issues, will this fully revert the phone to the original (or updated from original) state? - http://nexusshacks.com/nexus-s-hacks...0ti9020ai9023/

[Q] Fast, nominal, or slow cpu?

How do you tell if which type of cpu your phone has? I've looked and can't seem to find out how to check my phone to see which I have. My cpu always seems to run hot no matter what type of kernel I try, and I'm wondering if the cpu may be slow and not capable of tweaking.
Any help someone can give me would be great.
To find out which processor you have, use Terminal Emulator from the Play Store. Reboot then type this command:
su
dmesg | grep ACPU
Here's an explanation of what this means and also recommended voltage settings for each processor type courtesy of a very knowledgeable, helpful, and great kernel dev: Imoseyon. Any "thanks" goes to him and not me for providing all this info.
Post #1341, just read Imoseyon's response to me:
http://rootzwiki.com/topic/31329-ke...11012/page__st__1340__p__1034933#entry1034933
Imoseyon post #1418 for recommended voltage settings per cpu type:
http://rootzwiki.com/topic/31329-ke...-kernel-v16-11012/page__st__1410#entry1042976
This is the output I get from that command. How am I supposed to interpret this as fas, nominal, or slow?
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Sent from my SCH-I535 using xda premium
Try it again after rebooting, mine works each time
Sent from my SCH-I535 using xda app-developers app
crjjr said:
How do you tell if which type of cpu your phone has? I've looked and can't seem to find out how to check my phone to see which I have. My cpu always seems to run hot no matter what type of kernel I try, and I'm wondering if the cpu may be slow and not capable of tweaking.
Any help someone can give me would be great.
Click to expand...
Click to collapse
Found it. For anyone else that may be looking, go to this site and follow the directions. You'll need terminal installed on your phone. It works on any GS3.
http://teamuscellular.com/Forum/top...eankernel-tw-jb-imoseyon-droidroidz-usc-v141/
Finally, it worked.
Is there a performance difference between a slow CPU, a nominal CPU, and a fast CPU?
BlueCross said:
Finally, it worked.
Is there a performance difference between a slow CPU, a nominal CPU, and a fast CPU?
Click to expand...
Click to collapse
Not that I know of, you can run pretty much any kernel stock as well as adjust governors and io scheduler. The primary difference is in how much you can UC/OC or even UV/OV and how well your CPU can hold those tweaks without fear of bootlooping or instability. But as far as just flashing kernels and leaving them there, you're fine. Like I said, I'm basing this all off of what Imoseyon has said so he is WAY more knowledgeable in this area if you need additional input. This slow, nominal, fast cpu exists based on the manufacturer process so its not something you can adjust. It's permanently set with the cpu.
SlimSnoopOS said:
Not that I know of, you can run pretty much any kernel stock as well as adjust governors and io scheduler. The primary difference is in how much you can UC/OC or even UV/OV and how well your CPU can hold those tweaks without fear of bootlooping or instability. But as far as just flashing kernels and leaving them there, you're fine. Like I said, I'm basing this all off of what Imoseyon has said so he is WAY more knowledgeable in this area if you need additional input. This slow, nominal, fast cpu exists based on the manufacturer process so its not something you can adjust. It's permanently set with the cpu.
Click to expand...
Click to collapse
The basis for the difference in CPUs is the same as car Manufacturing. Part's are divided by spec tolerances. The closer to spec is the difference. In car terms GM separates parts according to how close they are to spec. Fast would go in Cadillac, Nominal Buick and slow Chevrolet. They all use the same part just have closer spec tolerance. :laugh:
prdog1 said:
The basis for the difference in CPUs is the same as car Manufacturing. Part's are divided by spec tolerances. The closer to spec is the difference. In car terms GM separates parts according to how close they are to spec. Fast would go in Cadillac, Nominal Buick and slow Chevrolet. They all use the same part just have closer spec tolerance. :laugh:
Click to expand...
Click to collapse
I can always count on this guy to drop some thorough knowledge.
Sent from my SCH-I535 using xda app-developers app
I tried this and got an error message :
sh: grep: not found
write: Broken pipe
I'm on stock rooted, and I granted the su permission.
Sent from my SCH-I535 using Tapatalk
HeadlessPonch said:
I tried this and got an error message :
sh: grep: not found
write: Broken pipe
I'm on stock rooted, and I granted the su permission.
Sent from my SCH-I535 using Tapatalk
Click to expand...
Click to collapse
Have to do immediately after a reboot and some custom boot animations will keep it from working.
prdog1 said:
Have to do immediately after a reboot and some custom boot animations will keep it from working.
Click to expand...
Click to collapse
No custom animation. I got rid of my lock screen to get to my homescreen faster and added a shortcut to the terminal app. I started typing as soon as the app would let me and got the same error.
Your busybox install must be borked. Download busybox free from the playstore and do a "smart install"
prdog1 said:
The basis for the difference in CPUs is the same as car Manufacturing. Part's are divided by spec tolerances. The closer to spec is the difference. In car terms GM separates parts according to how close they are to spec. Fast would go in Cadillac, Nominal Buick and slow Chevrolet. They all use the same part just have closer spec tolerance. :laugh:
Click to expand...
Click to collapse
As a general rule, CPUs that come off the same fab are tested and then placed into different "bins", based on how they perform at different speeds and voltages. The better performing "top bin", chips get rated and sold at the highest clock speed. The slower chips get a lower clock speed assigned, and are sold at a lower price.
This method works well for AMD and Intel, who both sell a wide range of clock speeds for a particular generation CPU at different price points. It eliminates waste by enabling them to sell the lower performing chips instead of rejecting or destroying them, and allows them to offer different price points which is beneficial from a marketing perspective.
With phones, the CPUs are often clocked the same (e.g. [email protected] 1.5 GHz), but the CPUs still range in capability depending on their mfg tolerances, and in the case of our Snapdragon S2s, they are tagged accordingly. Whether you end up with a slow or fast chip in your phone is totally random, but frequently your chances of getting a faster chip increase over time as the fab process is fine tuned, and chip yields go up.
That doesn't mean you can't get a fast CPU right off the bat. For example, I pre-ordered mine on day 1 (June 6th?) and still got a fast CPU.
x714x said:
Your busybox install must be borked. Download busybox free from the playstore and do a "smart install"
Click to expand...
Click to collapse
That worked, thanks!
Now I know I'm nominal. The first line says "ACPU PVS: [1],FMAX[0]". It's different from the other 2 posted results in this thread (all 3 are actually different). What do those numbers mean?
I was lucky enough to have a fast CPU; which values on those charts do I want? I opened Kernel Turner and went to "Voltage" and it is a bunch of weird numbers--and I see something for "-1.5 across board," how many times should I hit that? I'm assuming that's how you undervolt.
Kelton Rivas said:
I was lucky enough to have a fast CPU; which values on those charts do I want? I opened Kernel Turner and went to "Voltage" and it is a bunch of weird numbers--and I see something for "-1.5 across board," how many times should I hit that? I'm assuming that's how you undervolt.
Click to expand...
Click to collapse
http://forum.xda-developers.com/showthread.php?p=36743286
Would this apply for his aosp kernel as well? The 4.2 kernel?
Sent from my SCH-I535 using xda premium
Hi,
First post in this forum in a long time. Or maybe ever.
Not sure if this is the right place to ask this, I'm new to this ROM, and had a stable one in my Thunderbolt, so was away from flashing for almost a year. Just got a vz S3, rooted it as soon as I got home, and after doing some research, decided on Liquid Smooth. Downloaded Liquid-JB-v2.1-Beta3.1-d2vzw this morning, so whatever the latest version as of this morning (Jan 23) is what I flashed. Also got the correct gapps. All went well, and so far, I like what I've seen. (Thanks, developers, this has been the EASIEST flash to a new rom I've ever done.
One problem, and I hope someone can point me in the right direction, as I'm googled out on trying to find a solution.
When I go to settings, performance, current, or, maximum CPU frequency, the highest it goes it 1512. The youtube videos on this show choices up to 1900.
Can anyone point me in the right direction on this?
Thank you.
---------- Post added at 01:54 AM ---------- Previous post was at 01:46 AM ----------
Just installed Setcpu, and it would not allow (nor show) any setting higher than 1512.
Thanks again for any insight.
tprophit said:
Hi,
First post in this forum in a long time. Or maybe ever.
Not sure if this is the right place to ask this, I'm new to this ROM, and had a stable one in my Thunderbolt, so was away from flashing for almost a year. Just got a vz S3, rooted it as soon as I got home, and after doing some research, decided on Liquid Smooth. Downloaded Liquid-JB-v2.1-Beta3.1-d2vzw this morning, so whatever the latest version as of this morning (Jan 23) is what I flashed. Also got the correct gapps. All went well, and so far, I like what I've seen. (Thanks, developers, this has been the EASIEST flash to a new rom I've ever done.
One problem, and I hope someone can point me in the right direction, as I'm googled out on trying to find a solution.
When I go to settings, performance, current, or, maximum CPU frequency, the highest it goes it 1512. The youtube videos on this show choices up to 1900.
Can anyone point me in the right direction on this?
Thank you.
Click to expand...
Click to collapse
tprophit said:
Hi,
First post in this forum in a long time. Or maybe ever.
Not sure if this is the right place to ask this, I'm new to this ROM, and had a stable one in my Thunderbolt, so was away from flashing for almost a year. Just got a vz S3, rooted it as soon as I got home, and after doing some research, decided on Liquid Smooth. Downloaded Liquid-JB-v2.1-Beta3.1-d2vzw this morning, so whatever the latest version as of this morning (Jan 23) is what I flashed. Also got the correct gapps. All went well, and so far, I like what I've seen. (Thanks, developers, this has been the EASIEST flash to a new rom I've ever done.
One problem, and I hope someone can point me in the right direction, as I'm googled out on trying to find a solution.
When I go to settings, performance, current, or, maximum CPU frequency, the highest it goes it 1512. The youtube videos on this show choices up to 1900.
Can anyone point me in the right direction on this?
Thank you.
---------- Post added at 01:54 AM ---------- Previous post was at 01:46 AM ----------
Just installed Setcpu, and it would not allow (nor show) any setting higher than 1512.
Thanks again for any insight.
Click to expand...
Click to collapse
You need to run a custom kernel that is overclockable. The stock kernel on liquid isnt overclockable from what I remember.
---------- Post added at 11:24 AM ---------- Previous post was at 11:23 AM ----------
BlueCross said:
Would this apply for his aosp kernel as well? The 4.2 kernel?
Sent from my SCH-I535 using xda premium
Click to expand...
Click to collapse
Ya, its hardware level so it doesnt make a difference what rom you are running.

Categories

Resources