[Q] Achieving High Linpack Scores - Nexus One Q&A, Help & Troubleshooting

Ok, so I've experimented a LOT with different ROMs, different kernels, different overclocking etc and I just CANNOT seem to break the 48.333 barrier. I've achieved this several times on CM6 nightly with Wildmonks 1190 kernel.
Tonight, I took a few different ROMs (Rods, CM6 & NAZTech Plain) and modified them removing anything that could drain resources such as Maps, Messenger, Voice Search, etc etc. I re-signed them, flashed them AND flashed Wildmonks experimental 1267 kernel (maxed JUST below because 1267 is an insta-freeze) and I STILL only achieved 48.333 max.
Any tips?
I mean, I'm happy with my phones everyday performance, I'm more or less just trying to clock a linpack score in the high 49's for S&G's.
(The linpack record for N1 is like 50, showing an overclock kernel of 1300 but I highly doubt that as I've never heard of an N1 achieving even 1267 without lockup)

my highest score was 47.6 I think, and for about 2 weeks I kept running linpack days and nights to attempt to get higher score
But after second week I gave up, I could get around 47.6 and no higher.

Anyone?
(10 characters)

uoY_redruM said:
Anyone?
(10 characters)
Click to expand...
Click to collapse
Nobody at all can tell me how to achieve better scores?

1.267 was achieved by 'the jesus phone'

Mine was 41.01 with Stock Nexus One (Froyo)
Now (with the new version of Linpack) I can't get more than 34.9xx

b1337z0r said:
1.267 was achieved by 'the jesus phone'
Click to expand...
Click to collapse
The Jesus phone isn't even in the top ten for Nexus anymore. Check my signature. I hit 50.525 3x using Wildmonk 1267 kernel experimental. OC 1228 mhz stable. The highest for the N1 is 55 legit. Persian.

what is the ideal setup to achieve maximum mlops on the nexus?
1. Rom?
2. Kernel?
3. ?????

Why are benchmark scores so important?!
I find that just using my phone on a daily basis, being able to complete all the tasks I need, and do it with style & grace, is enough for me.
My Nexus still outperforms every single device amongst me and my friends.
I can do way more things, etc. etc. you all know this.
So why bother with trying to get the highest Linpack or Quadrant scores?!
BTW: I read a post the other night in which Cyanogen himself stated that cheating on benchmark scores was easy. so ... hhmm.

Because we can?
Sent from my Nexus One using XDA App

@uoY_redruM
How come I can only score around 36 on my Nexus One? I did full data wipe, cache and dalvik wipe. Then installed CM6 RC3 and wildmonks kernel 2.6.35_SVS-950mV_CFS_AXI_720p_1113MHz_081910. I've also installed SetCPU and overclocked to max of 1113. How can get this above 40?

Jesus' phone was fast, but not as fast as mine!
Love,
GOD
Linpack for Android Top 10
Results of the benchmark speed test for Android cell phones and devices. # Name MFLOPS ROM Notes
1 mtw4991 109.3 Nexus One 1113.6MHz google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys
2 Zeno/ dallas 78.643 Nexus One 998.4MHz google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys
3 Edub Chicago 53.194 PC36100 1190.4MHz sprint/htc_supersonic/supersonic/supersonic:2.2/FRF91/218634:user/release-keys
4 Saiy 52.759 Liberty 600.0MHz google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys
5 MachString 49.212 ADR6300 1190.4MHz google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys
6 br 48.333 ADR6300 google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys
7 br 48.333 ADR6300 google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys
8 Machstring 48.333 ADR6300 1190.4MHz google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys
9 Edub Chicago 47.932 PC36100 1190.4MHz sprint/htc_supersonic/supersonic/supersonic:2.2/FRF91/218634:user/release-keys
10 MachString 47.906 ADR6300 1190.4MHz google/passion/passion/mahimahi:2.2/FRF91/43546:user/release-keys

Don't trust in the top devices list.
It's too easy to cheat
I can show you if you want it and put any MFlops you want.
So here is my advice : Use linpack to improve your device, but never compare your results to other, that make no sense.
I wish greenecomputing can solve this, maybe I should contact them.

dmsr66 said:
Don't trust in the top devices list.
It's too easy to cheat
I can show you if you want it and put any MFlops you want.
So here is my advice : Use linpack to improve your device, but never compare your results to other, that make no sense.
I wish greenecomputing can solve this, maybe I should contact them.
Click to expand...
Click to collapse
Yeah, you might be right, however this popped up while driving right after flashing CM6+Wildmonk's 1113 kernel and shocked the heck outta me It's real, not faked....probably inaccurate, but real.

No worries, I didn't mean yours is fake
I just wanted to warn people about the reliability of the list and tell them to use Linpack as a personal device benchmark (ie after a ROM update) and not like a competition because some results may be totally fake.

well i was getting same scores with stock froyo, so just wondering why when overclocked with WM kernel and using CM6, shouldn't it be improved??

I personally usually get about 36 I would love to get 45

Related

How to interpret MFLOPS

http://www.greenecomputing.com/apps/linpack/linpack-by-device/
I'm lost guys. I can't seem to get an answer anywhere. I posted the link above as reference. Look at the Droid section and see how there are 72 MFLOPS as the leader and then a consistent 50-30 range there.
Are the 50-30 MFLOPS devices running at a stable, reliable level or are these one-hit wonders? I have read recent posts of people happy with "11 MFLOPS" on their ROMs. Are these "Top 10" Droids really just expert devs or are these numbers attainable by the "average user?"
Thanks goiys.
Running a rooted 2.2 rom (the stock, leaked one from Moto) I'm seeing a peak of 9.333MFLOPS
Normal is high 8.x's. No clue beyond that.
How did they manage to do eight times that number?
Quite simply, lies and trickery.
If you look at some of the entries, you'll see misplaced scores like Nexus Ones in the Magic chart.
I'm willing to bet several of the outrageous scores are being done via emulators on a PC and then edited to look like phones.
1TallTXn said:
Running a rooted 2.2 rom (the stock, leaked one from Moto) I'm seeing a peak of 9.333MFLOPS
Normal is high 8.x's. No clue beyond that.
Click to expand...
Click to collapse
Weird.
I'm running BB v0.4
smoked glass theme + add-ons
2.2 Froyo and I'm getting 12.737 MFLOPS consistently, the highest being
just over 13 but it only did it once so I didn't upload that to Linpack
'cause I think that'd be cheap.
I couldn't edit my post but it is clocked at 800Mhz
and I'm wanting to try the 1.2 ULV kernel by P3.

Is it weird to have 1774 at Quadrant?

Hello all,
I'm quite shocked when testing quadrant and I got 1774 score. Usually I only got around 1400-1500.
Then I tried that time, I had around 1400 for the first time, and I repeat the test directly and get 1774 score.
I repeat the method the next day and get also a huge difference between the first try and the second try.
anyone can explain this??
I'm running CM6 RC2
1113MHz monks kernel
pretty nice score, possibly some other proceses running in the background, i got a 1200+ score then killed a bunch of stuff running and then got a 1500 score
I usually hit 1700-ish.
disgustip8ted said:
pretty nice score, possibly some other proceses running in the background, i got a 1200+ score then killed a bunch of stuff running and then got a 1500 score
Click to expand...
Click to collapse
try again directly after you got 1500 score. I got 1700 for the second try
samsung galaxy s
I did the benchmark before applying the voodoo lagfix on my SGS, i was the third on the list just after moto droid x, after the lagfix and reboot, i'm now 1469 and on top of all the device, the original SGS score is way below...
the point is, how this thing work ? is this reliable ? i mean what's the point of the app finally ?
I honestly think that the score doesn't matter. Isn't the most important thing most aim for is speed and stability? If your phone feels fast then you shouldn't care for the score.
Sent from my Nexus One using XDA App
1774marks? no it's not strange.
Here is a screenshot i have made about month ago on original-chinese-odexed Miui [email protected] without any hacks, it was fast as lighting (Or diarrhea)
Miui12.31 not so fast

Mflop question

What does this say about my phone
Sent from my HERO200 using XDA App
That you took the screenshot at 11:08 p.m.? Are we playing "I Spy"? ....I give up.
I just mean...
I saw a post and people are getting scores in the 20s and 30s and mine was 3.... is there something I can do to make my phone faster
Sent from my HTC Hero with Froyo...
Vandelay007 said:
I just mean...
I saw a post and people are getting scores in the 20s and 30s and mine was 3.... is there something I can do to make my phone faster
Sent from my HTC Hero with Froyo...
Click to expand...
Click to collapse
I think you are confusing FPS (Frames Per Second that Neocore gives you) with MFLOPS (Million Floating-Point Operations Per Second). Your Hero can get as high as maybe 5.4 MFLOPS, at least that is the fastest I have seen.
All of these are arbitrary measurements that probably mean very little. it is easy to get caught up in it. I run CM6 and consistently get 4.7 MFLOPS at the stock 691 overclocking. If you hero can tolerate faster processor speeds, it will go higher.
Here, read this.....http://bugclub.org/beginners/math/mflops.html
smolck said:
I think you are confusing FPS (Frames Per Second that Neocore gives you) with MFLOPS (Million Floating-Point Operations Per Second). Your Hero can get as high as maybe 5.4 MFLOPS, at least that is the fastest I have seen.
All of these are arbitrary measurements that probably mean very little. it is easy to get caught up in it. I run CM6 and consistently get 4.7 MFLOPS at the stock 691 overclocking. If you hero can tolerate faster processor speeds, it will go higher.
Here, read this.....http://bugclub.org/beginners/math/mflops.html
Click to expand...
Click to collapse
cool.. very educational..... lol
I just got about 22 fps
Sent from my HTC Hero with Froyo...

[Q] Bad performance? 1563 in Quadrant...

Hi everybody!
After installing the DocRom XXJPO v9 with lagfix Ext4 ALL (a=e4), schedule kernel, I/O, etc.. and tweaks, and I get 1563 in Quadrant's benchmark(I've just installed, i never ran the test before)...It's a very low value, isn't it?What am I doing wrong?
Thanks!
Take test scores with a grain of salt. It's not the best way of assuming something is wrong with your phone.
Sent from my GT-I9000M using XDA App
Actually if u look at the other scores unless something has changed I'm pretty sure that is still the highest s ore on that list. No worries.
Sent from my SCH-I500 using XDA App
Uh, yea. I get something between 900 and 1000 depending on the day with official 2.2. Mostly I assume down to the piss poor filesystem speed. It almost always looks like it has hung on the writing test.
Thanks for the responses!
I've read that some people got 2000 and even a little bit more. I guess they overclocked!!
outrun86 said:
Thanks for the responses!
I've read that some people got 2000 and even a little bit more. I guess they overclocked!!
Click to expand...
Click to collapse
People may be using an overclock but to the people with the 2000+ scores are using the EXT2 binds option/OCLF.
When using OCLF quadrants shows huge scores in the filesystem part whereas it may not reflect the actual performance of the phone.
Personally so far, hardcores new K10C kernel is by far the fastest I've used yet. He's updates a few things in K10C and my phone is really, really fast. You notice it more in the app drawer and when the phone is under load. I'm also using the all EXT4 option.
You should give K10C a go and see for yourself.
Paul, I'll have to wipe to install this new kernel?
... Quadrant scores are pretty meaningless. Only be worried if you feel lag when you're using the phone.
my scores breach 2000(by 3) but im not overclocked. look at my signature
I'll try this new kernel, PaulForde, thanks for the tip!
nwsk, you're totally right, but sometimes I notice some little lag in my live wallpaper, and I don't have lots of apps in background. That's what made me think about it, install Quadrant and compare the results with some others.
By the way racerboy3801, I see you have BUJP8 v9 installed...[offtopic]Have you noticed any differences with JPA/JPO?[/offtopic]

Overclocking the 8.9 Galaxy Tablet

Hey guys,
Posted under general questions and to my surprise not one reply in a week. So, I'm curious to know somethings about overclocking these tablets. What settings are you using? Have you had any issues with overheating and shutting down?What settings are recommended for web browsing, gaming etc? Let's talk linpack and benchmark testing. Where do you rank and how did you get there? Let's talk SetCPU benchmark scores, what kind of results can you squeeze out out of your device? How does someone pull off 41916.7 MFLOPS?! (Nexus) I'll start......
GT-7310 Running 3.2 / Motley Kernal / SetCPU
SetCPU Scores
Long Bench @ 178ms / Native @ 621.5ms
Linpack Scores
Single thread @ 42.4 MFLOPS / Multi thread @ 83.2 MFLOPS
Obtained running 1504MHz max / 216MHz min
Note: I have overheated and shutdown at 1504 web browsing. At 1400MHz I have had no issues. I see some scores are through the roof. I'm not to familiar with Overclocking and just would like to hear what you guys have to say. Thanks in advance. I may of posted this in the wrong area and I apologize. I would like this to be a thread on the topic rather than a Q & A.
Well I personally run my tab @1200Mhz with -25mV on the 1000&1200 steps, but anyhow, due to the interactive governor the most used setps are 216, 456 and 1200. I don't undervolt the lower test for stability reasons as I mostly work on the tablet so I don't want to test yet how low uV can go, don't have the time to test it right now. (Even though I'm pretty sure -25mV even on lower steps is still largely stable)
As for the set CPU test I tried them out on 1504Mhz: 150 and 595ms. I once pushed the tab to 1600Mhz but it ended up in an immediate freeze so I will skip on trying that again.
I don't have linpack but tried running GLBench 2.1 Egypt and the score was pretty embarrasing
Thanks for sharing. Since my post I have flashed aokp-34 galaxiansoup with v-3 of the Motley Kernal. Running 1400 on 216 no undervolting and am seeing a tiny increase in numbers. I'm still stumped on these linpack scores. This thing runs great. No complaints whatsoever. Anyone want to chime in on the best bench testing app or what the stock numbers on these 7310's are?
everyone will tell you that benchmark scores aren't very representative!
run the test 5 times and each time you will get a slightly different score.
i don't really care about benchmarking, as long as i feel that my tab runs smooth (which i think it does) im pretty happy. im normally at 1,4ghz and everything is smooth and stable!
Thanks for sharing. I couldn't agree more. My tab runs great at the same speed settings as yours. I just don't understand the numbers and how certain device's have like 100 times my mflops......I'm a numbers guy by nature, I like to see it on paper....but I am content with my device. Just always looking to make it cleaner and faster.....you know the addiction I'm sure.
Are you happy with this kernel? I found it stuttering and laggy...
zooster said:
Are you happy with this kernel? I found it stuttering and laggy...
Click to expand...
Click to collapse
Yes, I've not ran into any issues. Totally impressed and content with v3 motley. If there is a better kernel out there I'd love to hear about. I have yet to look, no need.
Been running a few custom profiles via Setcpu peaking @1.5GHz on B37 for a few weeks now. Other than a force close or two and an occasional browser freeze, which is normal with these roms, this thing is running pretty snappy. Gotta watch the temp though.
Sent from my GT-P7310 using xda premium
I sure hope that you've set up your profiles so that your tab overclocks to 1.5Ghz only occasionally, it's a thing that you HW can handle it, but will it on the long term?
Also, I regret that you can't overclock the bus frequency to get an overall performance boost (the GPU could really use it..) as on the Galaxy S
Yeah, I'm only pushing 1.5 on occasion, per profile. I normally run 1.4 all day everyday for the last 3 months with heavy use and no issues. Just gotta watch the temp. I've had mine to 113 degrees which kinda scared me.
Sent from my GT-P7310 using xda premium

Categories

Resources