After this work unit is complete, I will be trying strictly Linux on the server, no VM to run f@h windows clients. Ubuntu, Red Hat and open source server of sorts are the options. Hmmm.. wonder what sort of files I will loose in the process? :roflmao:
Ubuntu Linux folding..... time to see how it goes
- Thread starter Mike
- Start date
Well, got it to work. Time to see if it's stable or will last a WU...
VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV
mike@mike-desktop:~/folding$ ./fah6
Note: Please read the license agreement (fah6 -license). Further
use of this software requires that you have read and accepted this agreement.
8 cores detected
--- Opening Log file [January 18 10:03:28 UTC]
# Linux SMP Console Edition ###################################################
###############################################################################
Folding@Home Client Version 6.29
http://folding.stanford.edu
###############################################################################
###############################################################################
[10:23:12] Project: 6069 (Run 1, Clone 60, Gen 112)
[10:23:12]
[10:23:12] Assembly optimizations on if available.
[10:23:12] Entering M.D.
Starting 8 threads
NNODES=8, MYRANK=7, HOSTNAME=thread #7
NNODES=8, MYRANK=1, HOSTNAME=thread #1
NNODES=8, MYRANK=2, HOSTNAME=thread #2
NNODES=8, MYRANK=4, HOSTNAME=thread #4
NNODES=8, MYRANK=3, HOSTNAME=thread #3
NNODES=8, MYRANK=6, HOSTNAME=thread #6
NNODES=8, MYRANK=0, HOSTNAME=thread #0
NNODES=8, MYRANK=5, HOSTNAME=thread #5
Reading file work/wudata_01.tpr, VERSION 4.0.99_development_20090605 (single precision)
Making 1D domain decomposition 1 x 8 x 1
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
:woott:
VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV
mike@mike-desktop:~/folding$ ./fah6
Note: Please read the license agreement (fah6 -license). Further
use of this software requires that you have read and accepted this agreement.
8 cores detected
--- Opening Log file [January 18 10:03:28 UTC]
# Linux SMP Console Edition ###################################################
###############################################################################
Folding@Home Client Version 6.29
http://folding.stanford.edu
###############################################################################
###############################################################################
[10:23:12] Project: 6069 (Run 1, Clone 60, Gen 112)
[10:23:12]
[10:23:12] Assembly optimizations on if available.
[10:23:12] Entering M.D.
Starting 8 threads
NNODES=8, MYRANK=7, HOSTNAME=thread #7
NNODES=8, MYRANK=1, HOSTNAME=thread #1
NNODES=8, MYRANK=2, HOSTNAME=thread #2
NNODES=8, MYRANK=4, HOSTNAME=thread #4
NNODES=8, MYRANK=3, HOSTNAME=thread #3
NNODES=8, MYRANK=6, HOSTNAME=thread #6
NNODES=8, MYRANK=0, HOSTNAME=thread #0
NNODES=8, MYRANK=5, HOSTNAME=thread #5
Reading file work/wudata_01.tpr, VERSION 4.0.99_development_20090605 (single precision)
Making 1D domain decomposition 1 x 8 x 1
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
:woott:
Straight away noticed a little difference. Don't have exact numbers but at least a half hour decrease and better load balance.
Average load imbalance: 8.4 %
Part of the total run time spent waiting due to load imbalance: 3.4 % Now have imbalance down to 3.1% and ( hour/ns ) is less than 7. Cool
Parallel run - timing based on wallclock.
NODE (s) Real (s) (%)
Time: 618.559 618.559 100.0
10:18
(Mnbf/s) (GFlops) (ns/day) (hour/ns)
Performance: 218.224 14.138 3.097 7.749
Average load imbalance: 8.4 %
Part of the total run time spent waiting due to load imbalance: 3.4 % Now have imbalance down to 3.1% and ( hour/ns ) is less than 7. Cool
Parallel run - timing based on wallclock.
NODE (s) Real (s) (%)
Time: 618.559 618.559 100.0
10:18
(Mnbf/s) (GFlops) (ns/day) (hour/ns)
Performance: 218.224 14.138 3.097 7.749
Last edited:
This is with the Linux Ubuntu 10.4 lts and Wine windows smp client and work unit 6069.
Min. Time / Frame : 00:04:41
Avg. Time / Frame : 00:04:49
This is the same WU # 6069 and Linux Ubuntu 10.4 lts and linux 6.29 smp console client.
Min. Time / Frame : 00:03:59
Avg. Time / Frame : 00:04:12
Will have to wait till benchmark is complete for real accurate times but looks good at a glance. What does this mean?
:roflmao:
Min. Time / Frame : 00:04:41
Avg. Time / Frame : 00:04:49
This is the same WU # 6069 and Linux Ubuntu 10.4 lts and linux 6.29 smp console client.
Min. Time / Frame : 00:03:59
Avg. Time / Frame : 00:04:12
Will have to wait till benchmark is complete for real accurate times but looks good at a glance. What does this mean?
:roflmao:
Last edited:
Average tpf is now 4:06 minutes at just over 40% complete. Looking favorable to try -bigadv again.
:mike:
I've been having a heck of a time figuring out little bugs/<? that has stopped folding progress from time to time. I got a bigger folding job which is taking longer than an hour or two to complete and have noticed issues such as login after screen saver lockup, defunct processes increase frame time and folder location effecting start of f@h as a service. Oh and the occasional seeing 8 cores but actually only calling 4 to 7 cores to do work, frustrating to say the least. Now the truth is: It's been 8 or 9 years since I have really worked with Linux/Unix. With time, remembering scripts has helped with troubleshooting and tracing but, I'm loosing ppd in the process. I really wonder if it's worth it. At first, and if I can get my self created bugs ironed out, time per frame has decreased as much as 15 percent WU to WU, linux to linux wine. I've about decided to install windows 7 64bit, get a benchmark and then decided if this linux thing is really worth the effort it's going to take. Anyway, that's my progress or lack thereof.
I've been having a heck of a time figuring out little bugs/<? that has stopped folding progress from time to time. I got a bigger folding job which is taking longer than an hour or two to complete and have noticed issues such as login after screen saver lockup, defunct processes increase frame time and folder location effecting start of f@h as a service. Oh and the occasional seeing 8 cores but actually only calling 4 to 7 cores to do work, frustrating to say the least
. Now the truth is: It's been 8 or 9 years since I have really worked with Linux/Unix. With time, remembering scripts has helped with troubleshooting and tracing but, I'm loosing ppd in the process. I really wonder if it's worth it. At first, and if I can get my self created bugs ironed out, time per frame has decreased as much as 15 percent WU to WU, linux to linux wine. I've about decided to install windows 7 64bit, get a benchmark and then decided if this linux thing is really worth the effort it's going to take. Anyway, that's my progress or lack thereof.
I'm running a Ubuntu64 machine right now (finally). It is using 10.04 LTS.
Just learning though.
Just learning though.
Cool Pat, are you running Intel or AMD chips?
My experience so far is mixed. 60** WU's are average 17% faster, 6701's wouldn't complete and I got EUE errors so rapidly and often I was below the 80% for extra credit. To remedy the 6701's, I run one smp ( machine 1 ) at smp -6 and one ( machine 2 ) at -smp 2. I had to keep an eye on the machine ( remotely ) and reconfigure depending on the assigned WU. -bigadv was a bigger improvement in time per frame. As you know, I was at about the max tpf for the 6900 even. 6900's improved by almost 20% and I was even able to complete 1 2684 WU. :woott:
Some Linux pointers I'm hoping may help.
If the box is folding dedicated, command line folding can save and help on the % of time "wasted" oops waiting
ctrl-alt-t = opens terminal and disabling gui or xorg will help a bit. :$
sudo update-rc.d -f gdm remove
enabling xorg is easy also by typing:$
sudo update-rc.d -f gdm defaults
Also, typing :
$ top
in a terminal window.
is equal to resource monitor and does not use but 1 percent of one processor and that is only as if refreshes.
More to come, sucks typing on a phone.
My experience so far is mixed. 60** WU's are average 17% faster, 6701's wouldn't complete and I got EUE errors so rapidly and often I was below the 80% for extra credit. To remedy the 6701's, I run one smp ( machine 1 ) at smp -6 and one ( machine 2 ) at -smp 2. I had to keep an eye on the machine ( remotely ) and reconfigure depending on the assigned WU. -bigadv was a bigger improvement in time per frame. As you know, I was at about the max tpf for the 6900 even. 6900's improved by almost 20% and I was even able to complete 1 2684 WU. :woott:
Some Linux pointers I'm hoping may help.
If the box is folding dedicated, command line folding can save and help on the % of time "wasted" oops waiting
ctrl-alt-t = opens terminal and disabling gui or xorg will help a bit. :$
sudo update-rc.d -f gdm remove
enabling xorg is easy also by typing:$
sudo update-rc.d -f gdm defaults
Also, typing :
$ top
in a terminal window.
is equal to resource monitor and does not use but 1 percent of one processor and that is only as if refreshes.
More to come, sucks typing on a phone.
How has your Linux machine done so far Pat?
I think I have enough of a collection in processors to see how my linux machine does by changing processors and memory. Kind of a benchmark machine for me to upgrade from later. Been hitting the " lot " processor sales on ebay lately. Just gotta be home to use them.
Anyway, I have enough processors for the 4P quad core board and a variety of processors for that board, the 2p quad core board and a 4 processor dual core board. I have a variety of pc2 and pc1 memory modules of various buss. Since I can't make craZy numbers, I thought I would make the best of what I have.
One question though, will changing processors or memory mid WU create EUE's?
If it will, I've learned how to re-run a previous WU, I know I can't submit that unit for points or credit once submitted, but I was thinking I could complete the unit, submit and then re-run that unit just for benchmark purposes. What do you think? Bad idea? Good Idea even though I will be loosing crunch time for other WU's? Or waste of time all together. should I just put the " listed " fastest components in and see what happens! I think, I may have enough pieces waiting to build a few caseless boards or at the minimum, two.
I think I have enough of a collection in processors to see how my linux machine does by changing processors and memory. Kind of a benchmark machine for me to upgrade from later. Been hitting the " lot " processor sales on ebay lately. Just gotta be home to use them.
Anyway, I have enough processors for the 4P quad core board and a variety of processors for that board, the 2p quad core board and a 4 processor dual core board. I have a variety of pc2 and pc1 memory modules of various buss. Since I can't make craZy numbers, I thought I would make the best of what I have.
One question though, will changing processors or memory mid WU create EUE's?
If it will, I've learned how to re-run a previous WU, I know I can't submit that unit for points or credit once submitted, but I was thinking I could complete the unit, submit and then re-run that unit just for benchmark purposes. What do you think? Bad idea? Good Idea even though I will be loosing crunch time for other WU's? Or waste of time all together. should I just put the " listed " fastest components in and see what happens! I think, I may have enough pieces waiting to build a few caseless boards or at the minimum, two.
Last edited:
No, from my experience, interrupting a job and changing memory or CPU's doesn't crash.
At least not yet. I've done it a few times so far.
There is a new Linux client on the way for SMP/BigAdv. Rumor has it that it is FAST on AMD chips.
I need a Linux app that monitors true CPU temp and true CPU clockspeed. Everything I've tried so far doesn't work. Without those, I can't overclock in Linux.
At least not yet. I've done it a few times so far.
There is a new Linux client on the way for SMP/BigAdv. Rumor has it that it is FAST on AMD chips.
I need a Linux app that monitors true CPU temp and true CPU clockspeed. Everything I've tried so far doesn't work. Without those, I can't overclock in Linux.
Thanks Pat, I kinda remember reading here or on FAH forums the best benchmark is with the same WU and changes made during the course for comparison but wanted to make sure.
Have you tried linux sensor package? Works on two of my Linux laptops but not the server.
http://www.lm-sensors.org/
May be dated.
I use apt-get due to no gui.
http://techthrob.com/2009/03/02/enabling-hardware-sensors-in-linux/
Have you tried linux sensor package? Works on two of my Linux laptops but not the server.
http://www.lm-sensors.org/
May be dated.
I use apt-get due to no gui.
http://techthrob.com/2009/03/02/enabling-hardware-sensors-in-linux/
:mike:
I've been having a heck of a time figuring out little bugs/<? that has stopped folding progress from time to time. I got a bigger folding job which is taking longer than an hour or two to complete and have noticed issues such as login after screen saver lockup, defunct processes increase frame time and folder location effecting start of f@h as a service. Oh and the occasional seeing 8 cores but actually only calling 4 to 7 cores to do work, frustrating to say the least
Windows 7 64bit ultimate didn't recognize all my cpu's on the 4p board, only two sockets.
Interesting. You might have to get Windows Server 200x.
I know there is a cap on the # of CPU's that consumer versions of Win supports.
I know there is a cap on the # of CPU's that consumer versions of Win supports.
I did install Windows Server 2003 64 bit ( detected all processors ), got off playing with Linux and Wine and have not revisited. You know how it is!!!!
I did try Windows server 2000 also, old copy, it was not an enterprise version so I didn't think it would work as 32-bit. Not sure now. May have to visit it also. I know, too much messing around.
Here are a couple more commands that have been helpful to me.
$ cat /proc/cpuinfo /* processor info */
$ cat /proc/meminfo /* memory info */
some of this info is available with ' top ' and cat'd properly
$ top --help /* for ' top ' syntax */
Using these I stumbled on my lack of knowledge towards computer hardware and lack of diligent research.
I did try Windows server 2000 also, old copy, it was not an enterprise version so I didn't think it would work as 32-bit. Not sure now. May have to visit it also. I know, too much messing around.
Here are a couple more commands that have been helpful to me.
$ cat /proc/cpuinfo /* processor info */
$ cat /proc/meminfo /* memory info */
some of this info is available with ' top ' and cat'd properly
$ top --help /* for ' top ' syntax */
Using these I stumbled on my lack of knowledge towards computer hardware and lack of diligent research.
Last edited: