![]() |
![]() |
#1 |
Apr 2020
616 Posts |
![]()
Have 2 laptops
#1 is older P6100 @2GHz, 2 cores, 4GB RAM. Current memory settings 100MB day & night. Using task manager, it's running at ~97-99% CPU. Just doing double-checking so far. #2 is newer i5-7200U @ 2.5GHz, 2 hyperthreaded cores, 8GB RAM. Memory at 1000 / 4000. It runs at about 50% CPU. Does mix of double-checking and PRP. Is there something I should set differently to get #2's usage higher, or is this just how the hyperthreaded cores work? Just want to work as efficiently as possible. |
![]() |
![]() |
![]() |
#2 | |
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest
32×13×43 Posts |
![]() Quote:
These GIMPS computations are usually memory-access-speed bound, so more processor registers does not help much. Sometimes, depending on processor and task, a little more performance can be obtained by running with hyperthreads. You can benchmark both ways and make an informed choice. Options, benchmark, in prime95. What's maximum performance can vary by number of workers and fft length (exponent size). If you are concerned about power efficiency, that gets a little more involved. https://ark.intel.com/content/www/us...-3-10-ghz.html You could consider also running mfakto on the HD620 IGP. In my experience that yields about 18GhzD/day of trial factoring progress, and reduces the cpu-side throughput by about half on an i7-7500U. The IGP and cpu share a package TDP limit, and also memory access. See http://www.mersenneforum.org/showpos...91&postcount=2 and https://www.mersenneforum.org/showthread.php?t=23394 Last fiddled with by kriesel on 2020-05-20 at 19:49 |
|
![]() |
![]() |
![]() |
#3 | |
Sep 2006
Brussels, Belgium
11×151 Posts |
![]() Quote:
Jacob |
|
![]() |
![]() |
![]() |
#4 |
Aug 2002
Termonfeckin, IE
276310 Posts |
![]()
Turn off hyper-threading. Run benchmarks to see if you will be better off running 1 worker on two cores rather than 2 workers on 2 cores. I have a 7500U and I found I got slightly more throughput running a single worker.
|
![]() |
![]() |
![]() |
#5 | |
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest
13A716 Posts |
![]() Quote:
Tabulated results of thorough benchmarking for a variety of cpu types can be found at https://www.mersenneforum.org/showpo...18&postcount=4 and https://www.mersenneforum.org/showpo...19&postcount=5 Last fiddled with by kriesel on 2020-05-21 at 13:31 |
|
![]() |
![]() |
![]() |
Thread Tools | |
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
mpi Two Machines Take longer than One for LA | EdH | Msieve | 9 | 2016-12-10 20:53 |
auto off/on machines | wildrabbitt | Hardware | 3 | 2016-05-14 16:31 |
Appeal for machines | dave_dm | GMP-ECM | 0 | 2005-06-29 02:23 |
P-1: two stages, two machines? | James Heinrich | Marin's Mersenne-aries | 23 | 2005-04-05 18:31 |
Machines | R.D. Silverman | NFSNET Discussion | 12 | 2004-07-02 12:12 |