Parallella, Raspberry Pi, FPGA & All That Stuff

Anonymous

The rebuilt Pi2 that was

The rebuilt Pi2 that was turning in 70k jobs is now turning in 55k jobs and running 3 consecutive jobs at a temp of ~40C.

I will push the other two Piz (a 2 and a 3) to 4 concurrent jobs and watch the temps.

Anonymous

How about a 314 gig drive

KF7IJZ
KF7IJZ
Joined: 27 Feb 15
Posts: 110
Credit: 6108311
RAC: 0

RE: The rebuilt Pi2 that

Quote:

The rebuilt Pi2 that was turning in 70k jobs is now turning in 55k jobs and running 3 consecutive jobs at a temp of ~40C.

I will push the other two Piz (a 2 and a 3) to 4 concurrent jobs and watch the temps.

I am running my PI2s 7/8 of my Pi 2s overclocked to 1Ghz and usually see 50K or less seconds for the results. My 900Mhz Pi 2 takes nearly 60K seconds. With a large fan and no enclosure of any kind in a 21-22C ambient room, they all keep 44-47C w/ heatsinks.

I am running all of mine off of the Raspbian "Lite" image, as I have no need for X.

My YouTube Channel: https://www.youtube.com/user/KF7IJZ
Follow me on Twitter: https://twitter.com/KF7IJZ

poppageek
poppageek
Joined: 13 Aug 10
Posts: 259
Credit: 2473733872
RAC: 0

RE: How about a 314 gig

I have been toying with getting one when I get a Pi 3 and using it as a network share for the boinc data directory for all the Pis. I have had terrible luck getting a E@H install on a SD card to last more than a few weeks. Start getting MMC errors then no longer boots. Have tried 4-5 different cards and always the same. Distributed.net does very few writes and runs fine for weeks on end. But bought the Pis for E@H.

If I do it, probably will, will post some info on it.

P.S.
Did you see this?

http://wdlabs.wd.com/Support/#downloads

KF7IJZ
KF7IJZ
Joined: 27 Feb 15
Posts: 110
Credit: 6108311
RAC: 0

RE: I have had terrible

Quote:
I have had terrible luck getting a E@H install on a SD card to last more than a few weeks. Start getting MMC errors then no longer boots. Have tried 4-5 different cards and always the same.

I've had no issues with the Samsung Evo 16Gb cards and the EMTEC Class 10 8 gig cards.

My YouTube Channel: https://www.youtube.com/user/KF7IJZ
Follow me on Twitter: https://twitter.com/KF7IJZ

poppageek
poppageek
Joined: 13 Aug 10
Posts: 259
Credit: 2473733872
RAC: 0

Berryboot

Berryboot home

http://www.berryterminal.com/doku.php/berryboot

I tried it and it did install Raspian to a USB stick.

Anonymous

RE: RE: How about a 314

Quote:

I have been toying with getting one when I get a Pi 3 and using it as a network share for the boinc data directory for all the Pis. I have had terrible luck getting a E@H install on a SD card to last more than a few weeks. Start getting MMC errors then no longer boots. Have tried 4-5 different cards and always the same. Distributed.net does very few writes and runs fine for weeks on end. But bought the Pis for E@H.

If I do it, probably will, will post some info on it.

P.S.
Did you see this?

http://wdlabs.wd.com/Support/#downloads

In an earlier post of mine I noted using a USB-to-mSATA drive. In this configuration the /boot partition stays on the microSD card and the /root partition moves to the mSATA drive where the majority of the reads/writes occur. I have a procedure on my website for doing this. I just used it recently for the mSATA effort. It was originally used for a Rasparian distro so the "/dev/paths" might be the same or different. If using some other Pi distro the paths might also be different but "path substitution" in the procedure should be straight forward. Its fairly easy and straight forward and you can probably use it with the 314 gib drive. I had originally used this procedure for a large USB external powered from a hub attached to a Pi_early_early_model and it was fine. I quit using the Pi though because "back then" they really struggled to crunch. Now its a different story.

The /boot partition will be written to during kernel updates so do not "write protect" it. By using this configuration most of the uploads/downloads/processing will take place on this drive and should eliminate any issues with data corruption on an SD card. I have done it both ways and have not experienced the failure rate that you mention.

Anonymous

On a Pi2 when moving from 3

On a Pi2 when moving from 3 concurrent WUs to 4 concurrent Wus my processing time has moved from ~55k to 77k and the 4th job is "waiting for memory". I have reverted to 3 concurrent WUs on this Pi to see if it returns to 55K without the "wait for memory" condition. Two other Piz, a 2 and a 3 are not experiencing this "wait for memory" condition or increase in processing time.

AgentB
AgentB
Joined: 17 Mar 12
Posts: 915
Credit: 513211304
RAC: 0

RE: On a Pi2 when moving

Quote:
On a Pi2 when moving from 3 concurrent WUs to 4 concurrent Wus my processing time has moved from ~55k to 77k and the 4th job is "waiting for memory".

This i think, is boinc deciding, based on what the application says it needs - regards memory.

It shows up in the client_state.xml file as a rsc_memory_bound. I'm assuming you do have free memory to use?

I notice for example an OAS task is about 150M,

Quote:
h1_0080.90_O1C01Cl2In3__O1AS20-100F_81.00Hz_1948
einstein_O1AS20-100F
104
144000000000000.000000
2880000000000000.000000
157286400.000000

but if i monitor it - it does not come close to that typically ~100M. I'm guessing you could override this limit via app_config.xml but the wiki doesn't provide details

edit: i've had a quick look at the source code and no you can't override from app_config.xml, so apologies for the red fish..

hth.

Anonymous

RE: RE: On a Pi2 when

Quote:
Quote:
On a Pi2 when moving from 3 concurrent WUs to 4 concurrent Wus my processing time has moved from ~55k to 77k and the 4th job is "waiting for memory".

This i think, is boinc deciding, based on what the application says it needs - regards memory.

It shows up in the client_state.xml file as a rsc_memory_bound. I'm assuming you do have free memory to use?

I notice for example an OAS task is about 150M,

Quote:
h1_0080.90_O1C01Cl2In3__O1AS20-100F_81.00Hz_1948
einstein_O1AS20-100F
104
144000000000000.000000
2880000000000000.000000
157286400.000000

but if i monitor it - it does not come close to that typically ~100M. I'm guessing you could override this limit via app_config.xml but the wiki doesn't provide details

edit: i've had a quick look at the source code and no you can't override from app_config.xml, so apologies for the red fish..

hth.

looking at my client_state.xml files I see multiple entries like the following:

p2030.20151015.G188.44+00.94.C.b5s0g0.00000_2257
einsteinbinary_BRP4
147
17500000000000.000000
350000000000000.000000
260000000.000000
20000000.000000

When you reference free memory where is this defined? It this set in the parameters: "Use at most XXX% of CPU time"

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.