Which of these CPUs is most productive? - Page 11

Christenson · 2011-02-03, 07:31

Strange how that works...itś not as pretty as the Windoesn´t version...don´t know why I didn´t think of the -h switch. It looks like this machine is eating NFS@home 16e work units at 40 minutes or so apiece, with three cores working that, one thread doing 0.028 sec/iteration on LL-D on M2552093, and the other two doing P-1ś at M47,74x,xxx taking 520 seconds to complete 10K iterations, or 0.93% of the test. No overclocking. I haven loaded the critical CPU temperature instrumentation software yet.

KingKurly · 2011-02-03, 15:34

Try running 'sensors', it's a command-line utility that will (hopefully) report your CPU temperatures. Should already be installed.

mdettweiler · 2011-02-03, 16:53

Quote:

Originally Posted by KingKurly

Try running 'sensors', it's a command-line utility that will (hopefully) report your CPU temperatures. Should already be installed.

Really? I always had to install it myself on Ubuntu. (FYI, running "sudo apt-get install lm-sensors" should do the trick if it's not already there.)

Christenson · 2011-02-03, 17:21

There isn't a graphical/plotting version of lm-sensors yet, is there? A CPU temperature graph would be a convenient way of indicating a lot....

I have another problem....I manually edited local.txt to indicate my primenet account name and my CPU name. I'm reasonably sure a P-1 test has completed by now, but I don't see the CPU name under "my account". How do I reset it and make sure it is correctly communicated to the V5 server?

KingKurly · 2011-02-03, 19:16

Quote:

Originally Posted by mdettweiler

Really? I always had to install it myself on Ubuntu. (FYI, running "sudo apt-get install lm-sensors" should do the trick if it's not already there.)

Strange, oh well. Thought it was there by default. Either way, your advice would solve that problem quickly enough.

Quote:

Originally Posted by Christenson

There isn't a graphical/plotting version of lm-sensors yet, is there? A CPU temperature graph would be a convenient way of indicating a lot....

I'm a command-line user more than a GUI user, but you might consider 'watch sensors'... watch will run a program every two seconds. This may not be what you want, but perhaps it's a step closer to what you want...?

Quote:

Originally Posted by Christenson

I have another problem....I manually edited local.txt to indicate my primenet account name and my CPU name. I'm reasonably sure a P-1 test has completed by now, but I don't see the CPU name under "my account". How do I reset it and make sure it is correctly communicated to the V5 server?

Use Option 1 (Test/Primenet) inside the mprime menu. It lets you specify user ID, computer name, etc. Also make sure that you set your options under Option 13 (Options/CPU) and optionally Option 14 (Options/Preferences). Hope this helps!

henryzz · 2011-02-03, 20:45

Quote:

Originally Posted by mdettweiler

Really? I always had to install it myself on Ubuntu. (FYI, running "sudo apt-get install lm-sensors" should do the trick if it's not already there.)

Just done exactly that.

Christenson · 2011-02-04, 04:13

Xubuntu 10.10, itś already there, but i´m not sure I am looking at the CPU temperature, but it does fit well with what my finger tells me is the temperature on the heatpipe leading to the CPU. Here´s what I get:
k10temp-pci-00c3
Adapter: PCI adapter
temp1: +33.9°C (high = +70.0°C)
watch sensors fails with ¨Error Opening Terminal: Unknown¨

running ./mprime -m from another terminal window finally got the server to give me status on the new machine, but I have a question -- two of the workers are doing P-1 tests at stage 2, and I told the menus that I had 1500 Meg available for them to use, but the workers insist on using only 250Meg apiece. If I stop and restart the workers, will they be able to use the extra memory?

KingKurly · 2011-02-04, 04:35

Quote:

Originally Posted by Christenson

running ./mprime -m from another terminal window finally got the server to give me status on the new machine, but I have a question -- two of the workers are doing P-1 tests at stage 2, and I told the menus that I had 1500 Meg available for them to use, but the workers insist on using only 250Meg apiece. If I stop and restart the workers, will they be able to use the extra memory?

Yes (it's option 4 in the menu, both to stop and start) assuming the P-1 bounds are such that using more memory is appropriate. What exponents/bounds are they? What are your other cores (if any) doing? Do you have enough free memory for the P-1 jobs to share the full allotted 1500?

mdettweiler · 2011-02-04, 04:49

Quote:

Originally Posted by Christenson

Xubuntu 10.10, itś already there, but i´m not sure I am looking at the CPU temperature, but it does fit well with what my finger tells me is the temperature on the heatpipe leading to the CPU. Here´s what I get:
k10temp-pci-00c3
Adapter: PCI adapter
temp1: +33.9°C (high = +70.0°C)
watch sensors fails with ¨Error Opening Terminal: Unknown¨

Hmm...I'm assuming this is under full CPU load? 33.9 C is low, but not entirely unrealistic if you have a very good non-stock cooler and aren't overclocking. I'd guess you indeed are looking at the CPU temperature. (Which means that you have a LOT of headroom with which to overclock!

)

Christenson · 2011-02-04, 06:01

Uhh, stock cooler with heatpipes, full load, but with a piece of duct from the case sitting on top of the fan so as not to recirculate the hot air and the mobo out of the case on the table, I didn't have the heart to take the machine down tonight and move it into the case, though I need to as the internal optical drive for the old machine arrived today.

Christenson · 2011-02-05, 17:19

time for more silly questions:
With six CPU cores, when I set up a given worker thread, I can assign it a helper core in addition to its main core. If the goal is to maximize the long-term throughput, is it better to have more threads doing disparate tasks or more helper cores pushing the tasks through fewer threads at once?

2011-02-03, 07:31	#111
Christenson Dec 2010 Monticello 5·359 Posts	Strange how that works...itś not as pretty as the Windoesn´t version...don´t know why I didn´t think of the -h switch. It looks like this machine is eating NFS@home 16e work units at 40 minutes or so apiece, with three cores working that, one thread doing 0.028 sec/iteration on LL-D on M2552093, and the other two doing P-1ś at M47,74x,xxx taking 520 seconds to complete 10K iterations, or 0.93% of the test. No overclocking. I haven loaded the critical CPU temperature instrumentation software yet. Last fiddled with by Christenson on 2011-02-03 at 07:33

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
CPUs assigned	houding	Hardware	10	2015-03-25 00:33
Number of CPUs to use	houding	Software	3	2015-02-26 19:56
cpus running at 10%	wildrabbitt	Information & Answers	9	2015-01-20 13:31
Whither Older CPUs?	Rodrigo	Operation Billion Digits	4	2010-10-20 15:12
A tale of 3 CPUs	chris2be8	Hardware	7	2010-07-20 23:12

2011-02-03, 15:34	#112
KingKurly Sep 2010 Annapolis, MD, USA 3³·7 Posts	Try running 'sensors', it's a command-line utility that will (hopefully) report your CPU temperatures. Should already be installed.

2011-02-03, 17:21	#114
Christenson Dec 2010 Monticello 5·359 Posts	There isn't a graphical/plotting version of lm-sensors yet, is there? A CPU temperature graph would be a convenient way of indicating a lot.... I have another problem....I manually edited local.txt to indicate my primenet account name and my CPU name. I'm reasonably sure a P-1 test has completed by now, but I don't see the CPU name under "my account". How do I reset it and make sure it is correctly communicated to the V5 server?

2011-02-04, 04:13	#117
Christenson Dec 2010 Monticello 5·359 Posts	Xubuntu 10.10, itś already there, but i´m not sure I am looking at the CPU temperature, but it does fit well with what my finger tells me is the temperature on the heatpipe leading to the CPU. Here´s what I get: k10temp-pci-00c3 Adapter: PCI adapter temp1: +33.9°C (high = +70.0°C) watch sensors fails with ¨Error Opening Terminal: Unknown¨ running ./mprime -m from another terminal window finally got the server to give me status on the new machine, but I have a question -- two of the workers are doing P-1 tests at stage 2, and I told the menus that I had 1500 Meg available for them to use, but the workers insist on using only 250Meg apiece. If I stop and restart the workers, will they be able to use the extra memory?

2011-02-04, 06:01	#120
Christenson Dec 2010 Monticello 5×359 Posts	Uhh, stock cooler with heatpipes, full load, but with a piece of duct from the case sitting on top of the fan so as not to recirculate the hot air and the mobo out of the case on the table, I didn't have the heart to take the machine down tonight and move it into the case, though I need to as the internal optical drive for the old machine arrived today.

2011-02-05, 17:19	#121
Christenson Dec 2010 Monticello 5·359 Posts	time for more silly questions: With six CPU cores, when I set up a given worker thread, I can assign it a helper core in addition to its main core. If the goal is to maximize the long-term throughput, is it better to have more threads doing disparate tasks or more helper cores pushing the tasks through fewer threads at once?