Seti android app

Message boards : Number crunching : Seti android app
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 8 · Next

AuthorMessage
Alex2
Volunteer tester

Send message
Joined: 2 Nov 12
Posts: 24
Credit: 11,798,226
RAC: 0
Canada
Message 1351144 - Posted: 27 Mar 2013, 13:10:02 UTC - in response to Message 1350918.  
Last modified: 27 Mar 2013, 13:10:40 UTC

Thanks!
I'll be very happy to do SETI on my Galaxy Note 2 (4x ARMv7 @1.6ghz).

On my Note, I'm used to do 2x 40,000 GFLOPS Milkyway WU for 200 credits per day per core.
Now I expect to do:
0.5x 180,000 GFLOPS SETI WU for 50 credits per day per core.

I don't understand.
Is Milkyway offering more credits per GFLOPS? 4x?
Is it any different on a PC?



I completed my first 2 tasks.

- The first problem I could see is over estimate the completion time:
It estimated at first 60h but it was done in 16h (CPU time?) (30h real time) on ARM7@1.6ghz.

- The second problem is that it trigger more often the thermal protection that half the CPU frequency, 1600mhz to 1000mhz on my custom kernel than other NativeBoinc apps.

Also the dramatic difference between the credit earning from SETI to Milky have been confirmed. See my previous post. I'll appreciate if someone could direct me to an explanation or just more info on that subject.
ID: 1351144 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1351147 - Posted: 27 Mar 2013, 13:23:16 UTC - in response to Message 1351139.  

Well, the first four wu's have been reported and one validated so far.

Due to me misunderstanding (and the wording changing when you check the option) I ran in 'Power Saving Mode' for about half of the time. It was estimating 55 hours and it was looking like it would take that long. Then after about 15 hours I un-checked the power saving thing.

It finished up in just under 30 hours and gave 94.03 cobblestones, credit, score ... what is it called now?

The power saving option got me the first time I ran PrimeGrid or Milkyway with it.
When unchecked it says
Power Saving: Prevent screen lock & CPU underclocking.
So I thought "Yeah I want to do that". However when checked it changes to
Power Saving: Allow screen lock and CPU underclocking.
It does make sense. Power Save checked = on & unchecked = off. Just threw me off by not reading what it was telling me.

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1351147 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1351150 - Posted: 27 Mar 2013, 13:29:47 UTC - in response to Message 1351144.  

Thanks!
I'll be very happy to do SETI on my Galaxy Note 2 (4x ARMv7 @1.6ghz).

On my Note, I'm used to do 2x 40,000 GFLOPS Milkyway WU for 200 credits per day per core.
Now I expect to do:
0.5x 180,000 GFLOPS SETI WU for 50 credits per day per core.

I don't understand.
Is Milkyway offering more credits per GFLOPS? 4x?
Is it any different on a PC?


I completed my first 2 tasks.

- The first problem I could see is over estimate the completion time:
It estimated at first 60h but it was done in 16h (CPU time?) (30h real time) on ARM7@1.6ghz.

- The second problem is that it trigger more often the thermal protection that half the CPU frequency, 1600mhz to 1000mhz on my custom kernel than other NativeBoinc apps.

Also the dramatic difference between the credit earning from SETI to Milky have been confirmed. See my previous post. I'll appreciate if someone could direct me to an explanation or just more info on that subject.

BOINC overestimates completion times by a large margin at first intentionally. I think it is after ~10 completed tasks the estimates will be close to actual or something along those lines.

SETI@Home uses CreditNew. Which is the recent credit system the BOINC developers created. As far as I know SETI@Home is the only project that uses it. Also other projects can more or less "pay" however much credit they would like per flop. Some projects give large amounts of credit as a way to attract users I suppose.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1351150 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1351152 - Posted: 27 Mar 2013, 13:36:36 UTC - in response to Message 1351144.  


- The second problem is that it trigger more often the thermal protection that half the CPU frequency, 1600mhz to 1000mhz on my custom kernel than other NativeBoinc apps.


I would say it's sign of code quality still had to be reached on other projects.
CPU used fully hence higher temps. Only truly optimized stuff can overheat ;)
So I would not label this as problem. It's issue with device cooling perhaps. Also, faster code can be made slower (and reverse much harder to reach). You just need some CPU throttling.


SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1351152 · Report as offensive
Alex2
Volunteer tester

Send message
Joined: 2 Nov 12
Posts: 24
Credit: 11,798,226
RAC: 0
Canada
Message 1351160 - Posted: 27 Mar 2013, 14:22:03 UTC - in response to Message 1351150.  

BOINC overestimates completion times by a large margin at first intentionally. I think it is after ~10 completed tasks the estimates will be close to actual or something along those lines.

SETI@Home uses CreditNew. Which is the recent credit system the BOINC developers created. As far as I know SETI@Home is the only project that uses it. Also other projects can more or less "pay" however much credit they would like per flop. Some projects give large amounts of credit as a way to attract users I suppose.


Thanks for the info.
OK
I figured that Seti gives 100 credits for 180000GFLOPS task regardless if it is done on CPU, GPU or ARM.

Yea, right, credits deaply vary from one project to another.
But as I prefer to look for ETI than other projects goal, I'll do only SETI on all my platforms and stop watching my BOINC score.


I would say it's sign of code quality still had to be reached on other projects.
CPU used fully hence higher temps. Only truly optimized stuff can overheat ;)
So I would not label this as problem. It's issue with device cooling perhaps. Also, faster code can be made slower (and reverse much harder to reach). You just need some CPU throttling.


Tanks!
I couldn't understand that by myself before reading it.
I tried this morning to run with NativeBoinc lock and a heat sink (!) on my Note 2 and I got no throttling and way better performance.

I use maximum undervoltage possible but it is still too hot at stock 1.6Ghz.
So no need to add GPU crunching on ARM device. It will just throttle more.

ID: 1351160 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1351166 - Posted: 27 Mar 2013, 14:36:24 UTC - in response to Message 1351150.  
Last modified: 27 Mar 2013, 14:48:48 UTC

SETI@Home uses CreditNew. Which is the recent credit system the BOINC developers created. As far as I know SETI@Home is the only project that uses it. Also other projects can more or less "pay" however much credit they would like per flop. Some projects give large amounts of credit as a way to attract users I suppose.

Of the projects I know, Albert@home (Einstein beta project), SIMAP and NumberFields run with CreditNew. MW also runs CreditNew - they certainly tend to run latest server version (with all the problems that entails).

That fact that MW pays so far above parity may be just bad design [and please don't get me started on the CreditNEw mess] or deliberately.
One of the main screws to turn is the estimated fpops. Seti and Albert are very good at estimating that - consequently their apps pay close to what boinc somewhere defines as the baseline.
If MW overestimate their fpops they ultimately hand out more credit. [Edit: and they appear to do so by a magnitude or two at least, juding from their crazily high APRs]
As I said, that may be deliberately or just poor design.
It certainly attracts the credit whores.
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1351166 · Report as offensive
Profile mimo
Volunteer tester
Avatar

Send message
Joined: 7 Feb 03
Posts: 92
Credit: 14,957,404
RAC: 0
Slovakia
Message 1351172 - Posted: 27 Mar 2013, 14:46:10 UTC - in response to Message 1350972.  

not only due gpu. was u trying to use the Linaro toolchain for building ?

ID: 1351172 · Report as offensive
Profile Swordfish
Avatar

Send message
Joined: 5 Aug 06
Posts: 72
Credit: 3,014,493
RAC: 0
United Kingdom
Message 1351215 - Posted: 27 Mar 2013, 17:46:24 UTC

First 2 WU's on Samsung Galaxy Tab2 10.1 have validated successfully.

Still awaiting completion of first WU's on Galaxy Note 10.1 Tab 2.
ID: 1351215 · Report as offensive
Profile skildude
Avatar

Send message
Joined: 4 Oct 00
Posts: 9541
Credit: 50,759,529
RAC: 60
Yemen
Message 1351220 - Posted: 27 Mar 2013, 18:06:43 UTC - in response to Message 1351215.  

First 2 WU's on Samsung Galaxy Tab2 10.1 have validated successfully.

Still awaiting completion of first WU's on Galaxy Note 10.1 Tab 2.

wow and it only took 10X longer than an Intel(R) Core(TM) i5-2500S CPU @ 2.70GHz [x86 Family 6 Model 42 Stepping 7]



I'm seeing this as proof of concept but nothing more. so this proves that it would take 10 phones to produce the same work as single core on an i5.


In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope
ID: 1351220 · Report as offensive
Profile James Sotherden
Avatar

Send message
Joined: 16 May 99
Posts: 10436
Credit: 110,373,059
RAC: 54
United States
Message 1351223 - Posted: 27 Mar 2013, 18:11:00 UTC

I left an app running on my android overnight and the thing was hot as all get out. So how are you guys cooling these things?
[/quote]

Old James
ID: 1351223 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1351231 - Posted: 27 Mar 2013, 18:42:42 UTC - in response to Message 1351223.  

I left an app running on my android overnight and the thing was hot as all get out. So how are you guys cooling these things?

While unplugged & not charging my phone stays quite cool. It is showing 27.4ºC for the battery temp which is quite normal. Running overnight on the battery it only sucked down about 40% of the battery.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1351231 · Report as offensive
Profile Tazz
Volunteer tester
Avatar

Send message
Joined: 5 Oct 99
Posts: 137
Credit: 34,342,390
RAC: 0
Canada
Message 1351261 - Posted: 27 Mar 2013, 19:47:21 UTC - in response to Message 1351223.  

I left an app running on my android overnight and the thing was hot as all get out. So how are you guys cooling these things?


Running on the battery my Nexus 7 was up to 30C and it chewed up 40% of the battery in 2 hours.

Plugged in it's hanging around 27C.
</Tazz>
ID: 1351261 · Report as offensive
Profile Swordfish
Avatar

Send message
Joined: 5 Aug 06
Posts: 72
Credit: 3,014,493
RAC: 0
United Kingdom
Message 1351268 - Posted: 27 Mar 2013, 20:17:05 UTC

Samsung Tablets, don't even get warm, when crunching.

Then again, the tablets are only set to crunch when connected to the power adapters.

ID: 1351268 · Report as offensive
Profile Link
Avatar

Send message
Joined: 18 Sep 03
Posts: 834
Credit: 1,807,369
RAC: 0
Germany
Message 1351275 - Posted: 27 Mar 2013, 20:45:24 UTC - in response to Message 1351166.  

MW also runs CreditNew - they certainly tend to run latest server version (with all the problems that entails).

They only use CreditNew for nBody WUs, separation WUs have fixed Credit depending on how many stars are used. On x86 CPUs they pay only a bit better than SETI stock apps.
ID: 1351275 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1351282 - Posted: 27 Mar 2013, 21:29:16 UTC - in response to Message 1351275.  

MW also runs CreditNew - they certainly tend to run latest server version (with all the problems that entails).

They only use CreditNew for nBody WUs, separation WUs have fixed Credit depending on how many stars are used. On x86 CPUs they pay only a bit better than SETI stock apps.

Ah, thanks. I was looking at NBody data. I assumed you could run either one or the other and not on a per app basis. Something new learned.
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1351282 · Report as offensive
Alex2
Volunteer tester

Send message
Joined: 2 Nov 12
Posts: 24
Credit: 11,798,226
RAC: 0
Canada
Message 1351316 - Posted: 28 Mar 2013, 0:08:38 UTC - in response to Message 1351223.  

I left an app running on my android overnight and the thing was hot as all get out. So how are you guys cooling these things?


--- TO COOL DOWN A CELL PHONE ---
- Close the screen (-10C)
- Do not run other apps like music or GPS. (-5C)
- It is cooler while not charging, unpluged, or fully charged.
- Do not overclock or even downclock. (up to -20C)
- Put a heat sink on it. I took one from a dumped P3. (-10C)
- Under volt the CPU and the GPU. (-10C)

I use CPU Tweak to get CPU temp and STweaks to change voltages.

With all that, I run 4x SETI at CPU 50C to 75C with custom kernel that throttle at 80C and I can do 4x 180GLOPS SETI wu per day.

I think it may be better to buy a used gamming PC for 200$ and do 10x to 50x more.

ID: 1351316 · Report as offensive
Profile SciManStev Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Jun 99
Posts: 6651
Credit: 121,090,076
RAC: 0
United States
Message 1351321 - Posted: 28 Mar 2013, 0:43:11 UTC

My thoughts are that power must be provided for long term crunching. Overall throughput is way less that a low end desktop or laptop, and I don't know how long a phone or tablet will last crunching constantly without extra cooling.

I may be wrong, but it seems like a cheap computer, with a bit of cooling would do a lot more for a much longer time. To me it just doesn't seem worth it, but I am very pleased that it works.

Steve
Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website
ID: 1351321 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34744
Credit: 261,360,520
RAC: 489
Australia
Message 1351330 - Posted: 28 Mar 2013, 1:25:25 UTC - in response to Message 1351321.  

My thoughts are that power must be provided for long term crunching. Overall throughput is way less that a low end desktop or laptop, and I don't know how long a phone or tablet will last crunching constantly without extra cooling.

I may be wrong, but it seems like a cheap computer, with a bit of cooling would do a lot more for a much longer time. To me it just doesn't seem worth it, but I am very pleased that it works.

Steve

Pretty much my own thoughts there, plus the price/performance/time figures alone isn't worth the effort for me even to consider trying to do it.

Cheers.
ID: 1351330 · Report as offensive
Alex2
Volunteer tester

Send message
Joined: 2 Nov 12
Posts: 24
Credit: 11,798,226
RAC: 0
Canada
Message 1351342 - Posted: 28 Mar 2013, 3:11:06 UTC

Hay!
We must listen when guys with SETI RAC over 50000 talk.
ID: 1351342 · Report as offensive
Profile Link
Avatar

Send message
Joined: 18 Sep 03
Posts: 834
Credit: 1,807,369
RAC: 0
Germany
Message 1351424 - Posted: 28 Mar 2013, 8:59:10 UTC - in response to Message 1351316.  

I think it may be better to buy a used gamming PC for 200$ and do 10x to 50x more.

But an old gaming PC will very likely use more than just 10-50x more energy.
ID: 1351424 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 8 · Next

Message boards : Number crunching : Seti android app


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.