V27.9 interation time suddenly doubled

scubabob · 2014-01-04, 06:17

I've been running V27.9 for some time now on this Thinkpad/i5-3320M/2.6 GHz CPU. Suddenly, and for no reason I can determine, the iteration times doubled from a stop caused by coming off the port replicator to battery power. A restart did not clear the problem. I don't think it is a turbo boost issue since the times are pretty much x2. The thread startup line says AVX is being used. The blower fan is on, but by no means is the CPU hot with the fan on full blast.

I am getting these results on the current exponent:

[Main thread Jan 3 21:47] Mersenne number primality test program version 27.9
[Main thread Jan 3 21:47] Optimizing for CPU architecture: Core i3/i5/i7, L2 cache size: 256 KB, L3 cache size: 3 MB
[Main thread Jan 3 21:47] Unable to detect which logical CPUs are hyperthreaded.
[Main thread Jan 3 21:47] Assuming logical CPUs 1 and 2, 3 and 4, etc. are each from one physical CPU core.
[Main thread Jan 3 21:47] To the best of my knowledge this assumption is only valid for Microsoft Windows.
[Main thread Jan 3 21:47] To override this assumption, see AffinityScramble2 in undoc.txt.
[Main thread Jan 3 21:47] Starting worker.
[Main thread Jan 3 21:47] Waiting 2 seconds for boot to complete.

and

[Jan 3 21:47] Worker starting
[Jan 3 21:47] Setting affinity to run worker on any logical CPU.
[Jan 3 21:47] Resuming primality test of M55149103 using AVX FFT length 2880K, Pass1=384, Pass2=7680
[Jan 3 21:47] Iteration: 32587837 / 55149103 [59.09%].
[Jan 3 21:59] Iteration: 32600000 / 55149103 [59.11%]. Per iteration time: 0.062 sec.
[Jan 3 22:47] Iteration: 32650000 / 55149103 [59.20%]. Per iteration time: 0.057 sec.

Task manager shows 25-29% usage with no other CPU bound processes spread among the four logical CPUs. Before the sudden slowdown, I was getting .033-.035 iteration times with one thread and ~.022 with two threads (over weekends). Is anybody else seeing this behavior?

scubabob · 2014-01-24, 16:21

I figured it out by accident. Somehow (I don't know how) the power profile got switched to "Maximum battery life" from "Balanced." The max battery life profile had it's "plugged in" maximum CPU performance setting at 5%, which somehow prevented the CPU from operating at full speed. The iteration time on this PC is now back to being consistent with the others.

TheMawn · 2014-01-24, 16:27

Yes. There are special options associated with power profiles limiting CPU use to whatever % to save power. You can go check that "balanced" doesn't limit the CPU to 25% or 50% or whatever.

Note that if you're running Prime95 on battery, as you probably know, you're going to drain your battery really fast, and probably wear it out if you do this repeatedly.

2014-01-04, 06:17	#1
scubabob Jun 2010 25₁₀ Posts	V27.9 interation time suddenly doubled I've been running V27.9 for some time now on this Thinkpad/i5-3320M/2.6 GHz CPU. Suddenly, and for no reason I can determine, the iteration times doubled from a stop caused by coming off the port replicator to battery power. A restart did not clear the problem. I don't think it is a turbo boost issue since the times are pretty much x2. The thread startup line says AVX is being used. The blower fan is on, but by no means is the CPU hot with the fan on full blast. I am getting these results on the current exponent: [Main thread Jan 3 21:47] Mersenne number primality test program version 27.9 [Main thread Jan 3 21:47] Optimizing for CPU architecture: Core i3/i5/i7, L2 cache size: 256 KB, L3 cache size: 3 MB [Main thread Jan 3 21:47] Unable to detect which logical CPUs are hyperthreaded. [Main thread Jan 3 21:47] Assuming logical CPUs 1 and 2, 3 and 4, etc. are each from one physical CPU core. [Main thread Jan 3 21:47] To the best of my knowledge this assumption is only valid for Microsoft Windows. [Main thread Jan 3 21:47] To override this assumption, see AffinityScramble2 in undoc.txt. [Main thread Jan 3 21:47] Starting worker. [Main thread Jan 3 21:47] Waiting 2 seconds for boot to complete. and [Jan 3 21:47] Worker starting [Jan 3 21:47] Setting affinity to run worker on any logical CPU. [Jan 3 21:47] Resuming primality test of M55149103 using AVX FFT length 2880K, Pass1=384, Pass2=7680 [Jan 3 21:47] Iteration: 32587837 / 55149103 [59.09%]. [Jan 3 21:59] Iteration: 32600000 / 55149103 [59.11%]. Per iteration time: 0.062 sec. [Jan 3 22:47] Iteration: 32650000 / 55149103 [59.20%]. Per iteration time: 0.057 sec. Task manager shows 25-29% usage with no other CPU bound processes spread among the four logical CPUs. Before the sudden slowdown, I was getting .033-.035 iteration times with one thread and ~.022 with two threads (over weekends). Is anybody else seeing this behavior?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Lap Top Suddenly 1/4 speed.	petrw1	Hardware	35	2015-11-07 11:36
Suddenly I'm getting only trivial TF tests	fivemack	Software	34	2015-10-25 16:54
P95 est time to go seems off	bcp19	Software	1	2012-08-03 22:39
Doubled compute power for a day?	Christenson	PrimeNet	19	2011-10-26 08:29
New .dat time?	benjackson	Prime Sierpinski Project	16	2008-07-29 07:26

2014-01-24, 16:21	#2
scubabob Jun 2010 5² Posts	I figured it out by accident. Somehow (I don't know how) the power profile got switched to "Maximum battery life" from "Balanced." The max battery life profile had it's "plugged in" maximum CPU performance setting at 5%, which somehow prevented the CPU from operating at full speed. The iteration time on this PC is now back to being consistent with the others.

2014-01-24, 16:27	#3
TheMawn May 2013 East. Always East. 6BF₁₆ Posts	Yes. There are special options associated with power profiles limiting CPU use to whatever % to save power. You can go check that "balanced" doesn't limit the CPU to 25% or 50% or whatever. Note that if you're running Prime95 on battery, as you probably know, you're going to drain your battery really fast, and probably wear it out if you do this repeatedly.