LinuxQuestions.org
Download your favorite Linux distribution at LQ ISO.
Go Back   LinuxQuestions.org > Forums > Linux Forums > Linux - Distributions > Slackware
User Name
Password
Slackware This Forum is for the discussion of Slackware Linux.

Notices


Reply
  Search this Thread
Old 06-05-2014, 05:36 PM   #16
dugan
LQ Guru
 
Registered: Nov 2003
Location: Canada
Distribution: distro hopper
Posts: 11,278

Rep: Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346

Two years ago, my 2Ghz Athlon II X3 struggled to play 720p Youtube movies even with CPU frequency scaling turned off. With CPU frequency scaling turned on and the CPU downclocked, the movies looked like slideshows.

Improvements have been made since then to both Google and Adobes' implementations of Flash for Linux, and Youtube is 100% available in HTML5 now, but that experience still made me realize how important it is to have a good CPU.
 
Old 06-05-2014, 07:01 PM   #17
cascade9
Senior Member
 
Registered: Mar 2011
Location: Brisneyland
Distribution: Debian, aptosid
Posts: 3,753

Rep: Reputation: 935Reputation: 935Reputation: 935Reputation: 935Reputation: 935Reputation: 935Reputation: 935Reputation: 935
Quote:
Originally Posted by ReaperX7 View Post
1.5GHz w/ 4 Cores will honestly do more work what a 2.4GHz with 2 Cores.
Only if you are comparing oranges to oranges...AMD and Intel are not directly comprable on core count/MHz.

While I cant find much in the way of real comparisons, the one that I have found does show the 2020M to have a bit more overall CPU power than the A4-5000, and far more per core/thread-

AMD A4-5000-
7-Zip Benchmark (Single Threaded)- 1323
7-Zip Benchmark (Multithreaded)- 4509

Intel Pentium 2020M
7-Zip Benchmark (Single Threaded)- 2856
7-Zip Benchmark (Multithreaded)- 5434

http://www.anandtech.com/show/6981/t...s-latest-apu/2

The AMD does have a bit more graphics grunt, and does use less power. If it was me, I'd probably go for the AMD, but thats mostly because I agree with metaschima in a lot of ways, overheating is an issue. 15w is much easier to dissipate than 35w. Also because if you arent going to run a laptop on batteries, you might as well get a desktop, and if everything else is equal (battery, HDD/SDD, extra addons) the AMD should last longer on battery power.

Last edited by cascade9; 06-06-2014 at 01:30 AM. Reason: typos
 
1 members found this post helpful.
Old 06-05-2014, 07:04 PM   #18
enorbet
Senior Member
 
Registered: Jun 2003
Location: Virginia
Distribution: Slackware = Main OpSys
Posts: 4,830

Rep: Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467Reputation: 4467
These days much multimedia processing is no longer very CPU intensive depending far more on GPU. Many people seem to think graphics is just for gamers but this is not so. We all depend on visuals, application time to be rendered to the screen and content only gets heavier. Much like the old system bugaboo, it used to be said "Nobody needs more than 1MB vram since that will render 1024x768". It wasn't completely correct then and is even less relevant today with larger (and less square) higher resolution screens. It really doesn't pay to scrimp and save on graphics, especially when what was Dream quality just 5-6 years ago can be had for well under $100 US now.

As for CPU, though this is somewhat less an issue these days, it is still true that clock speed is only part of the story. Instructions Per Clock is the other half. Obviously no matter if a CPU had a 10 PetaHz clock, if it did Zero IPS, it could do exactly Zero work.
 
Old 06-05-2014, 08:39 PM   #19
storkus
Member
 
Registered: Jun 2008
Location: Phoenix, Arizona, USA
Distribution: Slackware
Posts: 329

Rep: Reputation: 51
Personally, I would be MUCH more worried whether all the laptops' other subsystems were Linux-compatible: not just graphics, but other peripherals, especially wifi, BIOS, etc.

If you're doing CPU/GPU-heavy work, I'd go to a desktop or even server instead. Unless you're using (near)bleeding edge peripherals, this will almost guarantee Linux compatibility these days.
 
Old 06-05-2014, 09:45 PM   #20
Richard Cranium
Senior Member
 
Registered: Apr 2009
Location: McKinney, Texas
Distribution: Slackware64 15.0
Posts: 3,858

Rep: Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225Reputation: 2225
Quote:
Originally Posted by genss View Post
putting -j over the number of cores will bring no tangible benefit
Unless you are waiting for I/O.

The rule of thumb that I use is to pick a value more than 2x but less than 3x the number of cores for a dedicated build machine. For a machine where I intend to use for something else during a compile/build, I'll use a value less than 2x but more than 1x the number of cores. You should experiment to see what the sweet spot is for your particular system.
 
1 members found this post helpful.
Old 06-06-2014, 01:10 AM   #21
dugan
LQ Guru
 
Registered: Nov 2003
Location: Canada
Distribution: distro hopper
Posts: 11,278

Rep: Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346
Quote:
Originally Posted by Richard Cranium View Post
Unless you are waiting for I/O.

The rule of thumb that I use is to pick a value more than 2x but less than 3x the number of cores for a dedicated build machine. For a machine where I intend to use for something else during a compile/build, I'll use a value less than 2x but more than 1x the number of cores. You should experiment to see what the sweet spot is for your particular system.
I have no idea how these recommendations are determined, but the one I'd heard before was: one more than the number of available cores. That means on an 8-core system, you would set "-j" to 9, not 20.

Last edited by dugan; 06-06-2014 at 01:29 AM.
 
Old 06-06-2014, 07:07 AM   #22
genss
Member
 
Registered: Nov 2013
Posts: 744

Rep: Reputation: Disabled
Quote:
Originally Posted by Richard Cranium View Post
Unless you are waiting for I/O.

The rule of thumb that I use is to pick a value more than 2x but less than 3x the number of cores for a dedicated build machine. For a machine where I intend to use for something else during a compile/build, I'll use a value less than 2x but more than 1x the number of cores. You should experiment to see what the sweet spot is for your particular system.
takes very little time to read a couple kB of the disk, especially if you just untar-ed it so it's still in memory

-jn+1 if you really want that half percent
rationale being that all cores are full when the extra process is starting (reading files from disk, that is)

if you got benchmarks to disprove this statement i'd be happy to know
 
Old 06-06-2014, 07:11 AM   #23
genss
Member
 
Registered: Nov 2013
Posts: 744

Rep: Reputation: Disabled
Quote:
Originally Posted by cascade9 View Post
The AMD does have a bit more graphics grunt, and does use less power. If it was me, I'd probably go for the AMD, but thats mostly because I agree with metaschima in a lot of ways, overheating is an issue. 15w is much easier to dissipate than 35w. Also because if you arent going to run a laptop on batteries, you might as well get a desktop, and if everything else is equal (battery, HDD/SDD, extra addons) the AMD should last longer on battery power.
amd one supports C0 while i found no evidence of the intel one supporting it (could be wrong)
C0 means it completely turns off cores that are not used at the time, and that is most of the time
 
Old 06-06-2014, 07:44 AM   #24
TobiSGD
Moderator
 
Registered: Dec 2009
Location: Germany
Distribution: Whatever fits the task best
Posts: 17,148
Blog Entries: 2

Rep: Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886
Quote:
Originally Posted by genss View Post
putting -j over the number of cores will bring no tangible benefit
On my six-core machine compiling a kernel with -j10 is faster than with -j6.
 
Old 06-06-2014, 07:53 AM   #25
genss
Member
 
Registered: Nov 2013
Posts: 744

Rep: Reputation: Disabled
Quote:
Originally Posted by TobiSGD View Post
On my six-core machine compiling a kernel with -j10 is faster than with -j6.
how about -j7 ?




if someone wants to test, i cba

echo 3 > /proc/sys/vm/drop_caches
time make -jwhatever

should make no difference above ncores+1 (should even be an unnoticeable regression)

hf

PS more then one run per test case , sciencia

Last edited by genss; 06-06-2014 at 07:55 AM.
 
Old 06-06-2014, 10:45 AM   #26
nix84
Member
 
Registered: Apr 2014
Posts: 276

Original Poster
Rep: Reputation: Disabled
Thanks to U all for these comments.
I don't expect to compile the kernel more than once (OK twice for fumble fingers) and doing Encode/Decode of video extensively doesn't seem realistic if I have a desktop team.
I hope I have gotten the gist of everyone's comments correct.
 
Old 06-06-2014, 12:29 PM   #27
dugan
LQ Guru
 
Registered: Nov 2003
Location: Canada
Distribution: distro hopper
Posts: 11,278

Rep: Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346Reputation: 5346
Quote:
Unless you are waiting for I/O.
Er, what? If you're waiting for IO, then you're waiting for IO. Having more CPU parallelization isn't going to make a difference in that case.

Optimizating CPU usage helps you most when IO isn't blocking you.

Last edited by dugan; 06-06-2014 at 12:32 PM.
 
Old 06-08-2014, 04:59 AM   #28
TobiSGD
Moderator
 
Registered: Dec 2009
Location: Germany
Distribution: Whatever fits the task best
Posts: 17,148
Blog Entries: 2

Rep: Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886Reputation: 4886
Quote:
Originally Posted by genss View Post
how about -j7 ?




if someone wants to test, i cba

echo 3 > /proc/sys/vm/drop_caches
time make -jwhatever

should make no difference above ncores+1 (should even be an unnoticeable regression)

hf

PS more then one run per test case , sciencia
-j7 doesn't make much of a difference to -j6. I/O is not an issue here (and therefore dropping the cache not necessary), since I always compile on a tmpfs.
 
  


Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off



Similar Threads
Thread Thread Starter Forum Replies Last Post
Kernel: Processor Family Selection for Generic i686 frozenQueue Linux - Kernel 1 09-23-2010 02:37 AM
Processor selection for a Linux box darasen Linux - General 2 03-03-2010 05:03 PM
Processor Type selection in Kernel kushalkoolwal Linux - Kernel 4 06-16-2009 04:26 AM
processor selection - Slackware 12.1 - Kernel 2.6.5 Condemn Slackware 5 07-17-2008 03:06 PM
Slack 9.0 Selection of kernel dual processor tungsten Slackware 3 06-27-2003 03:50 AM

LinuxQuestions.org > Forums > Linux Forums > Linux - Distributions > Slackware

All times are GMT -5. The time now is 04:41 PM.

Main Menu
Advertisement
My LQ
Write for LQ
LinuxQuestions.org is looking for people interested in writing Editorials, Articles, Reviews, and more. If you'd like to contribute content, let us know.
Main Menu
Syndicate
RSS1  Latest Threads
RSS1  LQ News
Twitter: @linuxquestions
Open Source Consulting | Domain Registration