User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Raspbian Jessie linux 4.4.9 Severe Performance Degradation

Wed May 11, 2016 3:24 pm

I updated to the new raspberry pi raspbian/linux version (4.4.9), What is the deal with severe performance degradation in Jessie? Why is it so very slow/sluggish compared to the current 4.1 version?

cat /proc/version

Code: Select all

[email protected]:~ $ cat /proc/version
Linux version 4.4.9-v7+ ([email protected]) (gcc version 4.9.3 (crosstool-NG crosstool-ng-1.22.0-88-g8460611) ) #884 SMP Fri May 6 17:28:59 BST 2016
Raspbian Jessie linux kernel 4.4.9
Image

Raspbian Jessie linux kernel 4.1.19
Image


/boot/config.txt

Code: Select all

dtparam=sd_overclock=100
arm_freq=1260
core_freq=500
over_voltage=4
sdram_freq=575
sdram_schmoo=0x02000020
over_voltage_sdram_p=6
over_voltage_sdram_i=4
over_voltage_sdram_c=4
v3d_freq=500
h264_freq=333
gpu_mem=192
Here are some performance benchmarks to illustrate the degradation in 4.4.9

Sdbench 4.4.9
(32GB Samsung EVO+ using "dtparam=sd_overclock=100")
100.000 MHz CORE: 500 MHz, turbo=0 WRITE 15.82 MB/s READ 35.97 MB/s HDPARM 35.99 MB/s

Code: Select all

[email protected]:~ $ sudo ./sdbench.sh
[3;J
CONFIG: 
CLOCK : 100.000 MHz
CORE  : 500 MHz, turbo=0
DATA  : 512 MB, /root/test.dat

HDPARM:
======
 Timing O_DIRECT disk reads: 106 MB in  3.03 seconds =  34.93 MB/sec
 Timing O_DIRECT disk reads: 106 MB in  3.03 seconds =  34.96 MB/sec
 Timing O_DIRECT disk reads: 114 MB in  3.00 seconds =  37.96 MB/sec

WRITE:
=====
536870912 bytes (537 MB) copied, 26.3682 s, 20.4 MB/s
536870912 bytes (537 MB) copied, 30.698 s, 17.5 MB/s
536870912 bytes (537 MB) copied, 45.0785 s, 11.9 MB/s

READ:
====
536870912 bytes (537 MB) copied, 14.2759 s, 37.6 MB/s
536870912 bytes (537 MB) copied, 14.2392 s, 37.7 MB/s
536870912 bytes (537 MB) copied, 14.19 s, 37.8 MB/s

RESULT (AVG):
============
Overlay config                      core_freq   turbo   overclock_50    WRITE        READ        HDPARM
                                       500        0     100.000 MHz   15.82 MB/s   35.97 MB/s   35.99 MB/s
Sdbench 4.1
(32GB Samsung EVO+ using "dtparam=sd_overclock=100")
100.000 MHz CORE: 500 MHz, turbo=1 WRITE 21.51 MB/s READ 40.99 MB/s HDPARM 41.06 MB/s

Code: Select all

[email protected]:~ $ sudo ./sdbench.sh
[3;J
CONFIG: 
CLOCK : 100.000 MHz
CORE  : 500 MHz, turbo=1
DATA  : 512 MB, /root/test.dat

HDPARM:
======
 Timing O_DIRECT disk reads: 126 MB in  3.01 seconds =  41.86 MB/sec
 Timing O_DIRECT disk reads: 124 MB in  3.03 seconds =  40.89 MB/sec
 Timing O_DIRECT disk reads: 122 MB in  3.02 seconds =  40.40 MB/sec

WRITE:
=====
536870912 bytes (537 MB) copied, 25.5738 s, 21.0 MB/s
536870912 bytes (537 MB) copied, 22.3328 s, 24.0 MB/s
536870912 bytes (537 MB) copied, 23.7269 s, 22.6 MB/s

READ:
====
536870912 bytes (537 MB) copied, 12.8048 s, 41.9 MB/s
536870912 bytes (537 MB) copied, 12.3227 s, 43.6 MB/s
536870912 bytes (537 MB) copied, 12.3561 s, 43.4 MB/s

RESULT (AVG):
============
Overlay config                      core_freq   turbo   overclock_50    WRITE        READ        HDPARM
                                       500        1     100.000 MHz   21.51 MB/s   40.99 MB/s   41.06 MB/s
////////////////////////
//////////////////////

Linpack bench 4.4.9
Passed Gflops 5.70

Code: Select all

[email protected]:~ $ ./xhpl
================================================================================
HPLinpack 2.1  --  High-Performance Linpack benchmark  --   October 26, 2012
Written by A. Petitet and R. Clint Whaley,  Innovative Computing Laboratory, UTK
Modified by Piotr Luszczek, Innovative Computing Laboratory, UTK
Modified by Julien Langou, University of Colorado Denver
================================================================================

An explanation of the input/output parameters follows:
T/V    : Wall time / encoded variant.
N      : The order of the coefficient matrix A.
NB     : The partitioning blocking factor.
P      : The number of process rows.
Q      : The number of process columns.
Time   : Time in seconds to solve the linear system.
Gflops : Rate of execution for solving the linear system.

The following parameter values will be used:

N      :    8000 
NB     :     256 
PMAP   : Row-major process mapping
P      :       1 
Q      :       1 
PFACT  :    Left 
NBMIN  :       2 
NDIV   :       2 
RFACT  :   Right 
BCAST  :   2ring 
DEPTH  :       0 
SWAP   : Mix (threshold = 64)
L1     : transposed form
U      : transposed form
EQUIL  : yes
ALIGN  : 8 double precision words

--------------------------------------------------------------------------------

- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
      ||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be               1.110223e-16
- Computational tests pass if scaled residuals are less than                16.0

================================================================================
T/V                N    NB     P     Q               Time                 Gflops
--------------------------------------------------------------------------------
WR02R2L2        8000   256     1     1              59.84              5.706e+00
HPL_pdgesv() start time Wed May 11 11:19:06 2016

HPL_pdgesv() end time   Wed May 11 11:20:06 2016

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)=        0.0025941 ...... PASSED
================================================================================

Finished      1 tests with the following results:
              1 tests completed and passed residual checks,
              0 tests completed and failed residual checks,
              0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests.

Linpack bench 4.1
Passed Gflops 6.60

Code: Select all

[email protected]:~ $ ./xhpl
================================================================================
HPLinpack 2.1  --  High-Performance Linpack benchmark  --   October 26, 2012
Written by A. Petitet and R. Clint Whaley,  Innovative Computing Laboratory, UTK
Modified by Piotr Luszczek, Innovative Computing Laboratory, UTK
Modified by Julien Langou, University of Colorado Denver
================================================================================

An explanation of the input/output parameters follows:
T/V    : Wall time / encoded variant.
N      : The order of the coefficient matrix A.
NB     : The partitioning blocking factor.
P      : The number of process rows.
Q      : The number of process columns.
Time   : Time in seconds to solve the linear system.
Gflops : Rate of execution for solving the linear system.

The following parameter values will be used:

N      :    8000 
NB     :     256 
PMAP   : Row-major process mapping
P      :       1 
Q      :       1 
PFACT  :    Left 
NBMIN  :       2 
NDIV   :       2 
RFACT  :   Right 
BCAST  :   2ring 
DEPTH  :       0 
SWAP   : Mix (threshold = 64)
L1     : transposed form
U      : transposed form
EQUIL  : yes
ALIGN  : 8 double precision words

--------------------------------------------------------------------------------

- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
      ||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be               1.110223e-16
- Computational tests pass if scaled residuals are less than                16.0

================================================================================
T/V                N    NB     P     Q               Time                 Gflops
--------------------------------------------------------------------------------
WR02R2L2        8000   256     1     1              51.69              6.605e+00
HPL_pdgesv() start time Sun Apr 24 09:16:14 2016

HPL_pdgesv() end time   Sun Apr 24 09:17:06 2016

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)=        0.0025941 ...... PASSED
================================================================================

Finished      1 tests with the following results:
              1 tests completed and passed residual checks,
              0 tests completed and failed residual checks,
              0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests.
/////////////////////////////////////////
////////////////////////////////////////

Sysbench 4.4.9
116.0013s

Code: Select all

[email protected]:~ $ sysbench --num-threads=4 --test=cpu --cpu-max-prime=20000 --validate run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 4
Additional request validation enabled.


Doing CPU performance benchmark

Threads started!

Done.

Maximum prime number checked in CPU test: 20000


Test execution summary:
    total time:                          116.0013s
    total number of events:              10000
    total time taken by event execution: 463.8946
    per-request statistics:
         min:                                 45.42ms
         avg:                                 46.39ms
         max:                                138.75ms
         approx.  95 percentile:              50.12ms

Threads fairness:
    events (avg/stddev):           2500.0000/41.10
    execution time (avg/stddev):   115.9736/0.02

Sysbench 4.1
114.4567s

Code: Select all

[email protected]:~ $ sysbench --num-threads=4 --test=cpu --cpu-max-prime=20000 --validate run
[1] 1239

sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 4
Additional request validation enabled.

Doing CPU performance benchmark

Threads started!

Done.

Maximum prime number checked in CPU test: 20000


Test execution summary:
    total time:                          114.4567s
    total number of events:              10000
    total time taken by event execution: 457.7026
    per-request statistics:
         min:                                 45.42ms
         avg:                                 45.77ms
         max:                                 90.88ms
         approx.  95 percentile:              46.32ms

Threads fairness:
    events (avg/stddev):           2500.0000/12.25
    execution time (avg/stddev):   114.4256/0.02
Last edited by Rive on Wed May 11, 2016 7:00 pm, edited 17 times in total.
DNPNWO

Heater
Posts: 13704
Joined: Tue Jul 17, 2012 3:02 pm

Re: Raspbian Jessie 4.4.9 Severe Performance Degradation

Wed May 11, 2016 3:32 pm

No idea, but it's The Linux kernel version that is 4.4.9. Jessie is Debian 8.x where x is whatever it is now a days.
Memory in C++ is a leaky abstraction .

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie 4.4.9 Severe Performance Degradation

Wed May 11, 2016 3:34 pm

Heater wrote:No idea, but it's The Linux kernel version that is 4.4.9. Jessie is Debian 8.x where x is whatever it is now a days.
True, same diff though... Debian 8 on linux kernel 4.4.9 (versus Debian 8 on linux kernel 4.1). You would think the new kernel would be faster....not slower, right?

Code: Select all

sudo rpi-update
DNPNWO

User avatar
MarkHaysHarris777
Posts: 1820
Joined: Mon Mar 23, 2015 7:39 am
Location: Rochester, MN
Contact: Website

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 3:51 pm

Rive wrote:I updated to the new raspberry pi raspbian/linux version (4.4.9), What is the deal with severe performance degradation in Jessie? Why is it so very slow/sluggish compared to the current 4.1 version?
hi Rive, I have no idea either; however, that is the risk run every time the kernel is replaced--- but the question really is where is the correct venue for reporting this. Probably you need to make a bug report on GitHub so that the maintainer(s) know what you've found, and it can then be tracked. I don't think this forum is the right place.

I realize you're having fun 'benching' (nothing wrong with that either) but as a rule I don't upgrade just to be 'current' whatever that means. If there is no reason to upgrade, and things are working well now, well, don't upgrade! I always backup and archive first, and if I run into problems I can back-level easily.
marcus
:ugeek:

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 3:54 pm

MarkHaysHarris777 wrote:
Rive wrote:I updated to the new raspberry pi raspbian/linux version (4.4.9), What is the deal with severe performance degradation in Jessie? Why is it so very slow/sluggish compared to the current 4.1 version?
hi Rive, I have no idea either; however, that is the risk run every time the kernel is replaced--- but the question really is where is the correct venue for reporting this. Probably you need to make a bug report on GitHub so that the maintainer(s) know what you've found, and it can then be tracked. I don't think this forum is the right place.

I realize you're having fun 'benching' (nothing wrong with that either) but as a rule I don't upgrade just to be 'current' whatever that means. If there is no reason to upgrade, and things are working well now, well, don't upgrade! I always backup and archive first, and if I run into problems I can back-level easily.

I always backup, so no worries for me.

As for github, I dont have an account. Perhaps someone who does, can report my findings and link back back to this thread for specifics?
DNPNWO

User avatar
MarkHaysHarris777
Posts: 1820
Joined: Mon Mar 23, 2015 7:39 am
Location: Rochester, MN
Contact: Website

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 4:11 pm

Rive wrote: As for github, I dont have an account. Perhaps someone who does, can report my findings and link back back to this thread for specifics?
Getting an account is easy (and free). I have an account, but I don't know the link|maintainer... thought I did but can't find it... I might even be wrong that its on Github!! (I think it is, though).

Somebody will chime in... it would be nice for you to interface with the developer|maintainer, because they will probably ask you to test or help with benchmarking... you'd probably have fun helping them.
marcus
:ugeek:

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 4:13 pm

MarkHaysHarris777 wrote:
Rive wrote: As for github, I dont have an account. Perhaps someone who does, can report my findings and link back back to this thread for specifics?
Getting an account is easy (and free). I have an account, but I don't know the link|maintainer... thought I did but can't find it... I might even be wrong that its on Github!! (I think it is, though).

Somebody will chime in... it would be nice for you to interface with the developer|maintainer, because they will probably ask you to test or help with benchmarking... you'd probably have fun helping them.
It is done.
DNPNWO

User avatar
jojopi
Posts: 3085
Joined: Tue Oct 11, 2011 8:38 pm

Re: Raspbian Jessie 4.4.9 Severe Performance Degradation

Wed May 11, 2016 4:43 pm

Rive wrote:You would think the new kernel would be faster....not slower, right?
It might not be the kernel version bump ("uname -rv") that is significant, but a GPU firmware bump ("vcgencmd version").

Also, "rpi-update" gets more or less the latest versions that the Foundation engineers are working on. These may be less tested than the recommended stable versions that you get via "sudo apt-get upgrade".

jdb
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 2131
Joined: Thu Jul 11, 2013 2:37 pm

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 5:07 pm

While running the benchmarks that are displaying a large performance difference, can you give the output of "vmstat 2 -w" and post the resulting columns here?
Rockets are loud.
https://astro-pi.org

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 5:28 pm

jdb wrote:While running the benchmarks that are displaying a large performance difference, can you give the output of "vmstat 2 -w" and post the resulting columns here?

All benches run back to back in order shown while vmstat 2- w ran in background for duration.

4.4.9 kernel
Image

Code: Select all

[email protected]:~ $ cat /proc/version
Linux version 4.4.9-v7+ ([email protected]) (gcc version 4.9.3 (crosstool-NG crosstool-ng-1.22.0-88-g8460611) ) #884 SMP Fri May 6 17:28:59 BST 2016

Code: Select all

[email protected]:~ $ vmstat 2 -w
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 1  0        0   280032    31548   350108    0    0   904    45 2462 2707  9  4 83  3  0
 1  0        0   280032    31548   350136    0    0     0     0 4660 5565  1  1 99  0  0
 0  0        0   276376    31576   350880    0    0   380    75 8888 8654  2  3 91  4  0
 0  0        0   276400    31576   350880    0    0     0     0 14483 14061  0  1 98  0  0
 0  1        0   276392    31576   350880    0    0 36864     0 3273 3061  1  2 75 23  0
 0  0        0   276584    31120   350880    0    0 18432     0 8708 8406  2  2 84 12  0
 0  0        0   276640    31120   350880    0    0     0     0 5801 10523  3  2 96  0  0
 0  1        0   276640    31120   350880    0    0 41984     0 4816 8534  0  1 76 23  0
 0  0        0   276240    31120   350880    0    0 25600     0 16461 15310  1  2 84 14  0
 1  0        0   276360    31120   350880    0    0     0     0 4813 4376  1  1 99  0  0
 0  1        0   276360    31120   350880    0    0 32768     0 6905 6352  1  2 77 20  0
 0  1        0   167940    31140   458036    0    0 22560 24608 11237 10999  1  7 74 18  0
 0  2        0    66172    31152   559480    0    0     4 41472 6385 8997  0  5 60 35  0
 0  2        0    34720    31156   589864    0    0     2 12288 16463 16075  0  2 54 44  0
 0  3        0    28140    31160   596456    0    0     0  7726 4743 6424  1  2 55 43  0
 0  4        0    28236    31160   596456    0    0     0     0 4263 4049  0  1 25 73  0
 1  1        0    23192    31168   601332    0    0     2 12290 8225 7775  1  2 30 68  0
 0  1        0    16588    31184   607852    0    0     4 20498 3488 4663  1  3 47 50  0
 0  1        0    17692    30808   605020    0    0     4 20512 7728 8260  1  8 71 20  0
 0  2        0    16016    30816   606528    0    0     2 15890 8239 15051  0  3 57 40  0
 0  3        0    15984    30816   606536    0    0     0     0 3784 6467  0  1 25 74  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 0  3        0    16100    30824   606356    0    0     2 33284 12363 13959  0  3 26 71  0
 0  3        0    16252    30824   606452    0    0     0  8192 3871 6530  1  1 39 59  0
 0  3        0    16036    30824   606328    0    0     0 10246 2406 3819  1  1 31 67  0
 0  3        0    16284    30828   605724    0    0     0  1548 4504 8046  1  1 60 39  0
 1  4        0    16348    30828   605724    0    0     0     0 2905 4767  0  1 50 49  0
 0  4        0    16348    30828   605724    0    0     0     0 3455 6015  0  2 49 48  0
 1  2        0    14828    30840   607576    0    0     2 30738 14366 14147  1  5 48 46  0
 0  3        0    14828    30840   607576    0    0     0     0 3695 6365  0  0 25 74  0
 0  3        0    14828    30840   607576    0    0     0     0 8452 7989  0  2 24 75  0
 1  3        0    14952    30840   607576    0    0     0     0 6829 6432  1  1 24 75  0
 0  3        0    14952    30840   607576    0    0     0     0 2240 3366  0  1 25 74  0
 0  3        0    14952    30840   607576    0    0     0     0 4500 7986  0  1 25 73  0
 0  1        0    16516    30852   606084    0    0     0 23064 3932 5213  1  3 37 59  0
 0  1        0    16312    30852   605964    0    0     0     0 5890 5554  1  1 74 25  0
 0  1        0    16312    30852   605964    0    0     0     0 6816 7636  1  1 74 24  0
 0  1        0   222244    30872   404128    0    0    46 28788 4488 5825  1 11 73 15  0
 0  1        0   166740    30876   458496    0    0     2 28676 14418 15650  0  1 76 23  0
 0  2        0   111496    30876   512908    0    0     0 35844 11597 13878  0  1 73 26  0
 0  3        0   107024    30880   517312    0    0     0  1046 3706 3546  1  1 57 42  0
 0  4        0   107088    30880   517312    0    0     0     0 8724 8296  1  1 49 49  0
 0  2        0    16832    30892   605892    0    0     2 37406 9205 10291  0  6 67 27  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 0  3        0    17120    30892   605968    0    0     0     0 10305 10088  0  1 50 49  0
 2  3        0    14752    30896   607760    0    0     0 54278 16362 15619  0  4 54 42  0
 0  3        0    16116    30904   606348    0    0     2  5134 7118 6900  0  1 52 46  0
 0  4        0    16204    30904   606348    0    0     0     0 13650 13453  0  2 48 50  0
 1  2        0    14764    30912   607848    0    0     0 36378 8397 9451  1  5 70 24  0
 0  3        0    14796    30912   607832    0    0     0     0 3238 3052  0  2 49 49  0
 0  1        0    18508    30920   603776    0    0     2 34818 15302 15443  1  2 70 27  0
 1  2        0    18508    30924   603776    0    0     0    16 7123 6945  0  1 62 36  0
 0  3        0    18508    30924   603776    0    0     0     0 4968 4520  1  0 49 50  0
 1  0        0   100136    30928   524036    0    0     0     2 8746 8296  1  3 51 46  0
 0  1        0   205280    30940   420364    0    0     0 43084 6448 8767  0  9 75 17  0
 0  2        0   149912    30944   474772    0    0     2 31232 12846 13035  0  2 56 42  0
 0  3        0   126336    30948   498204    0    0     0 15892 6679 7528  0  2 59 39  0
 0  4        0   126368    30948   498204    0    0     0     0 3137 3013  0  1 26 73  0
 0  2        0    53588    30956   569428    0    0     2 41474 9412 9510  1  6 37 57  0
 1  3        0    30960    30960   591852    0    0     0  9742 12021 11920  1  1 52 46  0
 0  4        0    31024    30960   591852    0    0     0     0 8354 8214  0  1 25 75  0
 0  2        0    15352    30964   606876    0    0     0 52226 9558 13668  1  8 44 47  0
 0  2        0    16712    30972   605436    0    0     2  3084 5921 8761  0  1 54 44  0
 0  3        0    16712    30972   605436    0    0     0     0 5828 10520  0  1 48 50  0
 0  3        0    16380    30980   605544    0    0     0 57358 12430 14603  1  8 50 42  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 0  4        0    16420    30980   605508    0    0     0     0 6472 6380  0  0 25 75  0
 0  1        0    16428    30984   605504    0    0     0  8194 8250 9312  1  2 31 66  0
 1  1        0   481144     1476   180476    0    0 21842    82 13237 14305  1  7 73 18  0
 1  1        0   402916     1476   258720    0    0 39104     0 6612 8010  1  4 74 21  0
 1  1        0   331740     1476   330056    0    0 35648     0 8679 9920  1  5 74 20  0
 2  0        0   259920     1484   401324    0    0 35712     8 4461 6477  1  3 76 21  0
 0  1        0   188644     1484   472568    0    0 35648     0 5713 7995  1  4 73 22  0
 0  1        0   116832     1484   544012    0    0 35648     0 8536 9751  1  4 72 23  0
 1  1        0    45816     1492   614916    0    0 35456     8 4086 5273  1  4 73 23  0
 0  1        0   510228     1476   151328    0    0 31546    38 10544 11514  1  8 72 19  0
 0  1        0   424916     1476   236496    0    0 42496     0 8285 15526  1  2 74 24  0
 0  1        0   353740     1476   308288    0    0 35904     0 3444 5864  1  4 74 22  0
 0  1        0   282316     1484   379656    0    0 35712     8 9111 10273  1  5 72 22  0
 1  1        0   210520     1484   451208    0    0 35840     0 5716 6871  1  4 73 22  0
 0  1        0   138352     1484   522908    0    0 35776     0 6726 7950  1  4 74 22  0
 1  1        0    66648     1488   594604    0    0 35842     0 8290 9550  1  4 73 22  0
 1  0        0    34488      252   628368    0    0 34270     8 3428 5537  1  4 73 21  0
 0  1        0   449732     2212   211196    0    0 37552    44 17220 18589  1  5 73 21  0
 0  1        0   372108     2212   289328    0    0 39040     0 5095 9294  0  4 73 23  0
 1  0        0   301056     2212   360400    0    0 35584     0 4866 5914  1  4 74 22  0
 0  2        0   229756     2220   431856    0    0 35712     8 8822 10049  1  5 73 22  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 0  1        0   159200     2220   502628    0    0 35328     0 4916 6555  1  4 74 21  0
 0  1        0    88256     2220   573400    0    0 35456     0 5676 7743  1  4 75 20  0
 0  2        0    16744     2228   644844    0    0 35712     8 6510 9373  1  5 73 21  0
 0  0        0   522600     2728   139072    0    0  9562     0 4397 5578  1  5 87  6  0
 0  0        0   522600     2728   139072    0    0     0     0 9325 15418  0  1 99  0  0
 0  0        0   522600     2736   139072    0    0     0    36 3623 6101  1  1 99  0  0
 0  0        0   522600     2736   139072    0    0     0     0 2293 3481  1  0 99  0  0
 0  0        0   522600     2736   139072    0    0     0     0 6605 7875  1  1 99  0  0
 1  0        0   522600     2744   139072    0    0     0     8 4907 4441  1  1 99  0  0
 0  0        0   522600     2744   139072    0    0     0     0 6224 5734  1  1 98  0  0
 0  0        0   522600     2744   139072    0    0     0     0 7674 7240  1  1 98  0  0
 1  0        0   522600     2752   139072    0    0     0    36 3426 3235  1  0 99  0  0
 4  0        0   521428     3172   139992    0    0   668     0 11185 10155 51  2 47  0  0
 4  0        0   521304     3176   139992    0    0     2     6 13168 12256 99  2  0  0  0
 4  0        0   521304     3184   139992    0    0     0    12 7641 6668 100  1  0  0  0
 4  0        0   521304     3312   140108    0    0   122     0 16578 15661 98  2  0  0  0
 4  0        0   521188     3312   140108    0    0     0    16 10464 9543 99  1  0  0  0
 4  0        0   521196     3320   140108    0    0     0    12 11201 10239 99  1  0  0  0
 4  0        0   520964     3580   140108    0    0   130     0 12872 14874 98  2  0  0  0
 4  0        0   520980     3580   140108    0    0     0     8 6004 7253 99  1  0  0  0
 4  0        0   520980     3588   140108    0    0     0     8 14467 13489 98  2  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 4  0        0   520988     3588   140108    0    0     0     0 14764 13816 99  2  0  0  0
 4  0        0   520492     3588   140108    0    0     0     4 5945 4966 99  1  0  0  0
 5  0        0   520492     3596   140108    0    0     0     8 17183 16218 99  1  0  0  0
 4  0        0   520492     3596   140108    0    0     0     0 9345 11354 99  1  0  0  0
 4  0        0   520368     3596   140108    0    0     0     4 5292 8303 99  1  0  0  0
 4  0        0   520368     3604   140116    0    0     4    10 16291 15550 99  1  0  0  0
 4  0        0   520368     3604   140116    0    0     0     0 6123 8790 99  2  0  0  0
 4  0        0   520368     3604   140116    0    0     0     0 6575 11315 99  1  0  0  0
 5  0        0   520368     3612   140108    0    0     0     8 8537 14493 98  2  0  0  0
 4  0        0   520368     3612   140116    0    0     0     0 4024 6169 99  1  0  0  0
 4  0        0   520368     3612   140116    0    0     0     0 8370 14816 98  2  0  0  0
 4  0        0   520368     3620   140108    0    0     0     8 7475 12990 99  1  0  0  0
 4  0        0   520368     3620   140116    0    0     0     0 3802 5674 99  1  0  0  0
 4  0        0   520368     3620   140116    0    0     0     0 8894 15872 97  3  0  0  0
 4  0        0   520368     3628   140108    0    0     0     8 6254 10501 99  1  0  0  0
 4  0        0   520368     3628   140116    0    0     0     0 5410 8889 99  1  0  0  0
 4  0        0   520368     3628   140116    0    0     0     0 8429 14958 98  2  0  0  0
 4  0        0   520368     3636   140108    0    0     0     8 4937 7989 99  1  0  0  0
 5  0        0   520368     3636   140116    0    0     0     0 7098 12257 98  2  0  0  0
 4  0        0   520368     3636   140116    0    0     0     0 8050 14104 99  2  0  0  0
 4  0        0   520368     3644   140108    0    0     0    12 5396 5671 99  1  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 4  0        0   520368     3644   140116    0    0     0     0 8718 15479 98  2  0  0  0
 4  0        0   520244     3644   140116    0    0     0     4 7159 12358 98  2  0  0  0
 4  0        0   520244     3652   140108    0    0     0     8 4091 6269 99  1  0  0  0
 4  0        0   520244     3652   140116    0    0     0     0 8775 15651 98  2  0  0  0
 4  0        0   520244     3652   140116    0    0     0     0 6084 10157 99  1  0  0  0
 4  0        0   520368     3660   140108    0    0     0     8 5702 9477 99  1  0  0  0
 5  0        0   520368     3660   140116    0    0     0     0 15765 14920 99  2  0  0  0
 4  0        0   520368     3660   140116    0    0     0     0 4626 7323 99  1  0  0  0
 4  0        0   520368     3668   140108    0    0     0     8 7484 13071 99  2  0  0  0
 5  0        0   520368     3668   140116    0    0     0     0 10991 13931 99  1  0  0  0
 4  0        0   520368     3668   140116    0    0     0     0 3415 4932 100  1  0  0  0
 5  0        0   520368     3676   140112    0    0     0     8 17244 16360 99  1  0  0  0
 4  0        0   520368     3676   140120    0    0     0     0 6769 11586 99  1  0  0  0
 4  0        0   520368     3676   140120    0    0     0     0 5047 7834 99  1  0  0  0
 4  0        0   520368     3684   140112    0    0     0    10 8878 15514 97  3  0  0  0
 4  0        0   520368     3684   140120    0    0     0     0 5701 9396 99  1  0  0  0
 4  0        0   520368     3684   140120    0    0     0     6 6222 10523 99  1  0  0  0
 4  0        0   520368     3692   140112    0    0     0    12 8517 14503 98  2  0  0  0
 5  0        0   520368     3692   140120    0    0     0     0 7675 6807 99  1  0  0  0
 4  0        0   520368     3692   140120    0    0     0     6 14883 14179 98  2  0  0  0
 5  0        0   520368     3700   140112    0    0     0     8 14676 13806 99  1  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 5  0        0   520368     3700   140120    0    0     0     0 6113 5200 99  1  0  0  0
 4  0        0   520368     3700   140120    0    0     0     0 9010 16042 98  3  0  0  0
 4  0        0   520368     3708   140112    0    0     0     8 6632 10763 99  1  0  0  0
 4  0        0   520368     3708   140120    0    0     0     0 7100 8490 99  2  0  0  0
 5  0        0   520368     3708   140120    0    0     0     0 16151 15280 99  1  0  0  0
 4  0        0   520244     3716   140112    0    0     0     8 6781 8584 99  1  0  0  0
 4  0        0   520244     3716   140120    0    0     0     0 6842 11690 99  1  0  0  0
 1  0        0   520160     3716   140260    0    0    64     0 15070 14480 50  1 49  0  0
 0  0        0   520176     3724   140252    0    0     0    22 5977 6120  0  0 100  0  0
 0  0        0   520176     3724   140260    0    0     0     0 13278 15244  1  2 97  0  0
 0  0        0   520176     3724   140260    0    0     0     0 12247 13210  3  1 96  0  0
 1  0        0   506084     4004   142652    0    0  1326    18 2558 3486  5  9 85  1  0
 2  0        0   425112     4004   142652    0    0     0     0 14660 15889 22  4 74  0  0
 1  0        0   338436     4004   142652    0    0     0     4 5699 10112 24  3 73  0  0
 1  0        0   251760     4012   142652    0    0     0    18 5759 9355 25  2 74  0  0
 1  0        0   165208     4012   142652    0    0     0    12 9785 14884 25  2 73  0  0
 2  0        0    81516     4012   142652    0    0     0     0 8616 7769 22  3 74  0  0
 1  0        0    15424     4020   142736    0    0    46     8 13508 12953 31 14 56  0  0
 5  0        0    22492     4024   128964    0    0    92     0 13938 13974 92  7  1  0  0
 4  0        0    22492     4024   128964    0    0     0     0 3333 4845 97  3  0  0  0
 5  0        0    22492     4032   128956    0    0     0     8 16767 16324 86 13  1  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 6  0        0    22492     4032   128964    0    0     0     0 7247 11693 97  3  0  0  0
 2  0        0    22368     4032   128964    0    0     0     0 5270 7733 87 13  0  0  0
 5  0        0    22244     4040   129076    0    0    60    22 16111 15642 95  5  0  0  0
 4  0        0    22244     4040   129084    0    0     0     0 5334 8845 97  3  0  0  0
 4  0        0    22244     4040   129084    0    0     0     0 10253 11140 84 15  0  0  0
 5  0        0    22244     4048   129076    0    0     0     8 14268 14463 97  3  0  0  0
 4  0        0    22244     4048   129084    0    0     0     0 4479 6404 86 14  0  0  0
 4  0        0    22244     4048   129084    0    0     0     6 9451 14460 97  4  0  0  0
 5  0        0    22120     4056   129076    0    0     0    12 11265 13157 87 13  0  0  0
 4  0        0    21996     4056   129084    0    0     0     4 3697 5519 98  2  0  0  0
 5  0        0    22120     4056   129084    0    0     0     0 13080 15875 83 17  0  0  0
 5  0        0    22120     4064   129080    0    0     0     8 7627 10444 91  9  0  0  0
 4  0        0    22120     4064   129088    0    0     0     8 5534 8953 92  8  0  0  0
 5  0        0    22120     4064   129088    0    0     0     0 12811 15001 88 12  0  0  0
 5  0        0    22120     4072   129080    0    0     0     8 8673 7983 97  3  0  0  0
 5  0        0    22120     4072   129088    0    0     0     0 9870 12526 88 12  0  0  0
 5  0        0    22120     4072   129088    0    0     0     0 10924 13895 87 14  0  0  0
 5  0        0    22120     4080   129080    0    0     0    10 3946 5453 90 10  0  0  0
 5  0        0    22120     4080   129088    0    0     0     0 16772 16002 91 10  0  0  0
 4  0        0    22120     4080   129088    0    0     0     0 6898 11836 88 12  0  0  0
 4  0        0    21996     4088   129080    0    0     0     8 7596 6884 92  8  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 6  0        0    22120     4088   129088    0    0     0     0 9134 15328 89 12  0  0  0
 5  0        0    22120     4088   129088    0    0     0     0 10370 9673 87 13  0  0  0
 4  0        0    22120     4096   129080    0    0     0     8 11037 10341 89 11  0  0  0
 5  0        0    22120     4096   129088    0    0     0     0 14104 14686 86 14  0  0  0
 5  0        0    22120     4096   129088    0    0     0     0 5601 6928 81 19  0  0  0
 1  0        0    22120     4104   129084    0    0     2     8 8067 13573 36 14 50  0  0
 2  0        0    22120     4104   129092    0    0     0     6 14288 13809 25  1 74  0  0
 1  0        0    22120     4104   129092    0    0     0     0 5799 5223 25  0 75  0  0
 1  0        0    22120     4112   129084    0    0     0    12 14715 16193 25  1 73  0  0
 2  0        0    22120     4112   129092    0    0     0     0 12452 11174 25  1 74  0  0
 2  0        0    22120     4112   129092    0    0     0     0 9186 8684 25  0 74  0  0
 1  0        0   517756     4940   134728    0    0  1498    14 15180 16049 28  8 64  1  0
 1  0        0   527468     4940   131856    0    0   100     0 8401 9005  9  1 90  0  0
100.000 MHz 15.51 MB/s 35.87 MB/s 38.20

Code: Select all

[email protected]:~ $ sudo ./sdbench.sh
[3;J
CONFIG: 
CLOCK : 100.000 MHz
CORE  : 500 MHz, turbo=0
DATA  : 512 MB, /root/test.dat

HDPARM:
======
 Timing O_DIRECT disk reads: 108 MB in  3.03 seconds =  35.64 MB/sec
 Timing O_DIRECT disk reads: 132 MB in  3.04 seconds =  43.35 MB/sec
 Timing O_DIRECT disk reads: 108 MB in  3.04 seconds =  35.52 MB/sec

WRITE:
=====
536870912 bytes (537 MB) copied, 49.2956 s, 10.9 MB/s
536870912 bytes (537 MB) copied, 30.6966 s, 17.5 MB/s
536870912 bytes (537 MB) copied, 26.3302 s, 20.4 MB/s

READ:
====
536870912 bytes (537 MB) copied, 14.35 s, 37.4 MB/s
536870912 bytes (537 MB) copied, 14.2275 s, 37.7 MB/s
536870912 bytes (537 MB) copied, 14.25 s, 37.7 MB/s

RESULT (AVG):
============
Overlay config                      core_freq   turbo   overclock_50    WRITE        READ        HDPARM
                                       500        0     100.000 MHz   15.51 MB/s   35.87 MB/s   38.20 MB/s
116.0758s

Code: Select all

[email protected]:~ $ sysbench --num-threads=4 --test=cpu --cpu-max-prime=20000 --validate run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 4
Additional request validation enabled.


Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 20000


Test execution summary:
    total time:                          116.0758s
    total number of events:              10000
    total time taken by event execution: 464.1738
    per-request statistics:
         min:                                 45.42ms
         avg:                                 46.42ms
         max:                                113.95ms
         approx.  95 percentile:              51.73ms

Threads fairness:
    events (avg/stddev):           2500.0000/50.42
    execution time (avg/stddev):   116.0435/0.03
5.796e+00

Code: Select all

[email protected]:~ $ ./xhpl
================================================================================
HPLinpack 2.1  --  High-Performance Linpack benchmark  --   October 26, 2012
Written by A. Petitet and R. Clint Whaley,  Innovative Computing Laboratory, UTK
Modified by Piotr Luszczek, Innovative Computing Laboratory, UTK
Modified by Julien Langou, University of Colorado Denver
================================================================================

An explanation of the input/output parameters follows:
T/V    : Wall time / encoded variant.
N      : The order of the coefficient matrix A.
NB     : The partitioning blocking factor.
P      : The number of process rows.
Q      : The number of process columns.
Time   : Time in seconds to solve the linear system.
Gflops : Rate of execution for solving the linear system.

The following parameter values will be used:

N      :    8000 
NB     :     256 
PMAP   : Row-major process mapping
P      :       1 
Q      :       1 
PFACT  :    Left 
NBMIN  :       2 
NDIV   :       2 
RFACT  :   Right 
BCAST  :   2ring 
DEPTH  :       0 
SWAP   : Mix (threshold = 64)
L1     : transposed form
U      : transposed form
EQUIL  : yes
ALIGN  : 8 double precision words

--------------------------------------------------------------------------------

- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
      ||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be               1.110223e-16
- Computational tests pass if scaled residuals are less than                16.0

================================================================================
T/V                N    NB     P     Q               Time                 Gflops
--------------------------------------------------------------------------------
WR02R2L2        8000   256     1     1              58.91              5.796e+00
HPL_pdgesv() start time Wed May 11 13:25:06 2016

HPL_pdgesv() end time   Wed May 11 13:26:05 2016

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)=        0.0025941 ...... PASSED
================================================================================

Finished      1 tests with the following results:
              1 tests completed and passed residual checks,
              0 tests completed and failed residual checks,
              0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests.
///////////////////////////////////////////////////////////////////
/////////////////////////////////////////////////////////////////

4.1.19 kernel
Image

Code: Select all

[email protected]:~ $ cat /proc/version
Linux version 4.1.19-v7+ ([email protected]) (gcc version 4.9.3 (crosstool-NG crosstool-ng-1.22.0-88-g8460611) ) #858 SMP Tue Mar 15 15:56:00 GMT 2016

Code: Select all

[email protected]:~ $ vmstat 2 -w
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 0  0        0   304128    31748   350728    0    0   717    40  300  342  8  3 87  2  0
 0  0        0   304104    31748   350756    0    0     0     0  708  198  1  0 99  0  0
 1  0        0   304104    31748   350756    0    0     0     0  801  414  2  1 97  0  0
 0  0        0   304104    31748   350756    0    0     0     0  718  170  1  0 99  0  0
 0  0        0   300936    31776   351496    0    0   379    23  909  554  1  2 97  0  0
 0  1        0   300936    31776   351496    0    0 14336     0 1060  777  0  1 91  9  0
 0  1        0   300952    31776   351496    0    0 45056     0  771  487  1  0 75 24  0
 0  0        0   301556    31328   351496    0    0  6144    10  749  248  0  1 95  4  0
 0  1        0   301556    31328   351496    0    0 14336     0  723  226  0  0 92  8  0
 0  1        0   301572    31328   351496    0    0 44032     0  864  591  0  1 75 24  0
 0  0        0   301292    31328   351496    0    0  8192     0  741  243  0  1 94  5  0
 0  1        0   300812    31328   351496    0    0  9216     0  714  169  1  0 92  7  0
 0  1        0   300812    31328   351496    0    0 44032     0  751  465  1  1 75 24  0
 0  2        0   110388    31340   540852    0    0  9244 49454 1011  806  1  8 75 16  0
 1  1        0    70304    31344   579764    0    0     2 20484  870  576  1  1 76 23  0
 0  2        0    33232    31360   616080    0    0     4 22302  850  554  0  2 76 23  0
 0  3        0    33232    31360   616084    0    0     0     0  751  406  0  0 50 49  0
 1  1        0    12292    30988   636012    0    0     4 48646  826  558  0  4 60 36  0
 0  3        0    12524    30992   635660    0    0     0 10770  843  616  0  1 58 41  0
 2  0        0    12296    31000   635924    0    0     2 20506  776  460  0  1 51 48  0
 0  2        0    12644    31028   635404    0    0    12 44060  871  622  0  3 74 22  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 1  3        0    12688    31028   635416    0    0     0     0  806  475  0  0 50 49  0
 0  1        0    14556    31036   633348    0    0     2 46100  818  530  0  6 63 31  0
 0  2        0    14556    31040   633344    0    0     0    16  753  426  0  0 71 28  0
 0  1        0   233028    31056   418536    0    0    42 30828 1095  976  0  9 59 31  0
 0  2        0   134376    31064   516368    0    0     4 39172  902  702  0  3 61 35  0
 0  2        0    91020    31064   558960    0    0     0 24576  871  628  1  1 50 48  0
 0  3        0    80468    31068   569304    0    0     0 12054  794  499  0  0 55 45  0
 0  4        0    80468    31068   569304    0    0     0     0  738  364  0  0 50 50  0
 0  1        0    12444    31076   634872    0    0     0 49670  876  619  0  4 61 35  0
 0  4        0    12564    31080   634844    0    0     0  1550  752  384  0  0 51 48  0
 1  1        0    11460    31088   635840    0    0     0 42242  794  464  0  3 51 46  0
 0  3        0    12848    31092   634540    0    0     0  8974  925  691  0  1 55 45  0
 1  4        0    12848    31092   634540    0    0     0     0  777  389  0  0 50 50  0
 0  1        0    12412    31100   634512    0    0     0 53250  816  484  0  3 57 40  0
 0  2        0    12348    31104   634512    0    0     0    18 1564 1918  0  1 58 41  0
 0  3        0    12412    31104   634512    0    0     0     0 1386 1660  0  0 25 75  0
 0  1        0   172684    31120   478048    0    0     0 61518 1109 1106  0 10 73 16  0
 0  2        0   130948    31124   519240    0    0     0 12548 2502 5592  0  1 64 35  0
 0  1        0   128832    31124   520980    0    0     0     0  686  121  0  0 50 50  0
 0  1        0   128832    31124   520984    0    0     0     0  658   48  0  0 75 25  0
 0  4        0   128800    31128   520984    0    0     0  3860  774  387  0  0 52 48  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 1  3        0    83252    31132   565972    0    0     0 41474  823  476  0  3 46 51  0
 0  3        0    17196    31136   630920    0    0     0 16128  883  613  0  1 25 73  0
 1  1        0    12480    31136   634980    0    0     0 18180  970  823  0  0 49 50  0
 0  3        0    12404    31140   635160    0    0     0 18446  850  574  0  1 58 41  0
 0  4        0    12500    31140   635160    0    0     0     0  735  353  0  0 50 50  0
 0  3        0    11804    31148   635768    0    0     0 45326  968  790  0  4 45 50  0
 2  1        0    12620    31148   634868    0    0     0 32512  894  667  0  1 28 71  0
 0  2        0    12644    31148   634924    0    0     0 12288  889  682  0  1 50 50  0
 0  2        0    12644    31156   634924    0    0     0    16 1286 1427  0  1 50 50  0
 1  0        0   170988      204   511216    0    0     0     0 1418 1659  0  2 52 47  0
 1  0        0   539768     1436   147756    0    0 37246    44 1955 3138  0  4 74 21  0
 0  1        0   454456     1436   233152    0    0 42688     0 1059 2422  0  1 76 23  0
 0  1        0   369516     1436   318020    0    0 42432     0 1092 2432  0  2 75 23  0
 0  1        0   284204     1444   402664    0    0 42368     8 1671 2907  0  3 73 23  0
 0  1        0   199200     1444   487576    0    0 42432     0 1672 2915  0  2 74 23  0
 2  0        0   114040     1444   572716    0    0 42560     0 1963 3312  0  2 74 23  0
 0  1        0   565628     1284   121944    0    0 37646    52 1454 2805  1  4 75 21  0
 0  1        0   480688     1284   206808    0    0 42432     0 1012 2387  1  2 75 22  0
 1  0        0   395872     1284   291672    0    0 42432     0 1011 2388  0  2 75 22  0
 0  1        0   311056     1292   376528    0    0 42432     8 1016 2380  0  4 73 22  0
 0  1        0   226052     1292   461400    0    0 42432     0 1018 2384  0  4 74 22  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 1  1        0   141420     1292   546392    0    0 42624     0 1439 2775  0  3 74 22  0
 0  1        0   591552     1352    95964    0    0 37784    38 1628 2939  0  5 74 21  0
 0  1        0   513820     2100   172944    0    0 38840     0 2262 3592  0  3 72 25  0
 0  1        0   428260     2100   257912    0    0 42496     0 1815 3132  0  2 74 23  0
 2  0        0   343072     2108   342876    0    0 42496    10 1964 3294  0  2 74 23  0
 0  1        0   257612     2108   428428    0    0 42752     0 2020 3333  0  2 74 23  0
 0  1        0   172320     2108   513528    0    0 42560     8 2011 3332  0  3 73 23  0
 0  1        0    86780     2116   599024    0    0 42752     8 2032 3357  0  3 74 23  0
 0  0        0   611716     3012    75600    0    0   802     0  859  472  1  2 96  1  0
 0  0        0   611716     3016    75564    0    0     2    16  737  284  1  0 99  0  0
 4  0        0   610584     3316    76456    0    0   584    44  890  594 12  0 87  1  0
 4  0        0   610048     3316    76440    0    0     0     0 1051   84 100  0  0  0  0
 4  0        0   610048     3316    76588    0    0    58    40 1075  158 100  0  0  0  0
 4  0        0   610048     3324    76556    0    0     0     8 1055   88 100  0  0  0  0
 4  0        0   610048     3324    76556    0    0     0     0 1050   79 100  0  0  0  0
 4  0        0   610048     3324    76556    0    0     0     0 1053   91 100  0  0  0  0
 4  0        0   610048     3332    76556    0    0     0     8 1054   88 100  0  0  0  0
 4  0        0   610048     3332    76556    0    0     0     0 1051   87 100  0  0  0  0
 4  0        0   610048     3332    76556    0    0     0     0 1049   80 100  0  0  0  0
 4  0        0   610048     3340    76556    0    0     0     8 1054   89 100  0  0  0  0
 4  0        0   610048     3340    76556    0    0     0     0 1051   84 100  0  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 5  0        0   610048     3340    76556    0    0     0     0 1051   85 100  0  0  0  0
 4  0        0   610048     3348    76556    0    0     0     8 1053   87 100  0  0  0  0
 4  0        0   610056     3476    76556    0    0    64     0 1052  101 100  0  0  0  0
 4  0        0   609932     3476    76556    0    0     0     0 1050   96 100  0  0  0  0
 4  0        0   609940     3484    76556    0    0     0     8 1053   99 100  0  0  0  0
 4  0        0   609948     3484    76556    0    0     0     0 1050   97 100  0  0  0  0
 4  0        0   609948     3484    76556    0    0     0     0 1045   77 100  0  0  0  0
 4  0        0   609948     3492    76556    0    0     0    12 1055  104 100  0  0  0  0
 4  0        0   609956     3492    76556    0    0     0     0 1049   85 100  0  0  0  0
 4  0        0   609956     3492    76556    0    0     0     0 1048   85 100  0  0  0  0
 4  0        0   609956     3500    76556    0    0     0     8 1053   86 100  0  0  0  0
 4  0        0   609956     3500    76556    0    0     0     0 1420  702 100  0  0  0  0
 4  0        0   609956     3500    76560    0    0     0     0 1049   91 100  0  0  0  0
 4  0        0   609956     3508    76560    0    0     0    10 1053   84 100  0  0  0  0
 4  0        0   609956     3508    76560    0    0     0     0 1081  138 100  0  0  0  0
 4  0        0   609956     3508    76560    0    0     0     0 1082  147 100  0  0  0  0
 4  0        0   609956     3516    76560    0    0     0     8 1059   93 100  0  0  0  0
 4  0        0   609956     3516    76560    0    0     0     0 1047   80 100  0  0  0  0
 4  0        0   609956     3516    76560    0    0     0     0 1060  107 100  0  0  0  0
 4  0        0   609956     3524    76560    0    0     0     8 1060   97 100  0  0  0  0
 4  0        0   609956     3524    76560    0    0     0     0 1051   85 100  0  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 5  0        0   609956     3524    76560    0    0     0     0 1047   82 100  0  0  0  0
 4  0        0   609956     3532    76560    0    0     0     8 1080  127 100  0  0  0  0
 4  0        0   609956     3532    76560    0    0     0     0 1049   89 100  0  0  0  0
 4  0        0   609956     3532    76560    0    0     0     6 1052   99 100  0  0  0  0
 4  0        0   609956     3540    76560    0    0     0    12 1049   82 100  0  0  0  0
 4  0        0   609956     3540    76560    0    0     0     0 1047   79 100  0  0  0  0
 4  0        0   609956     3540    76560    0    0     0     8 1054  105 100  0  0  0  0
 4  0        0   609956     3548    76560    0    0     0     8 1052   87 100  0  0  0  0
 4  0        0   609956     3548    76560    0    0     0     0 1047   81 100  0  0  0  0
 4  0        0   609956     3548    76560    0    0     0     0 1049   86 100  0  0  0  0
 4  0        0   609956     3556    76560    0    0     0     8 1051   83 100  0  0  0  0
 4  0        0   609956     3556    76560    0    0     0     0 1047   82 100  0  0  0  0
 4  0        0   609956     3556    76560    0    0     0     0 1049   84 100  0  0  0  0
 4  0        0   609956     3564    76560    0    0     0     8 1048   82 100  0  0  0  0
 4  0        0   609956     3564    76560    0    0     0     0 1060   98 100  0  0  0  0
 4  0        0   609956     3564    76560    0    0     0     0 1050   82 100  0  0  0  0
 4  0        0   609956     3572    76560    0    0     0     8 1052   86 100  0  0  0  0
 4  0        0   609956     3572    76560    0    0     0     0 1049   87 100  0  0  0  0
 4  0        0   609956     3572    76560    0    0     0     0 1046   82 100  0  0  0  0
 4  0        0   609956     3580    76560    0    0     0     8 1053   92 100  0  0  0  0
 4  0        0   609956     3580    76560    0    0     0     0 1379  635 100  0  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 5  0        0   609956     3580    76560    0    0     0     4 1052   99 100  0  0  0  0
 4  0        0   609956     3588    76560    0    0     0    10 1054   92 100  0  0  0  0
 4  0        0   609956     3588    76560    0    0     0     2 1045   80 100  0  0  0  0
 4  0        0   609956     3588    76560    0    0     0     0 1049   85 100  0  0  0  0
 4  0        0   609956     3596    76560    0    0     0     8 1052   85 100  0  0  0  0
 0  0        0   609624     3596    76700    0    0    64     0  728  233  3  0 97  0  0
 0  0        0   609624     3596    76700    0    0     0     0  713  303  0  0 100  0  0
 1  0        0   593712     3876    79092    0    0  1326    24  875  669  8  7 84  1  0
 1  0        0   509152     3876    79092    0    0     0     0  781  106 24  2 75  0  0
 1  0        0   422972     3876    79092    0    0     0     0  761   81 23  2 75  0  0
 1  0        0   336172     3884    79084    0    0     0    50  785  132 23  3 75  0  0
 1  0        0   250240     3884    79092    0    0     0     0  760   76 23  2 75  0  0
 1  0        0   163192     3884    79096    0    0     0     0  761   82 23  2 75  0  0
 4  0        0    98340     3892    79180    0    0    46     8  859   92 38 11 50  0  0
 4  0        0    97720     3892    79188    0    0     0     0 1042   84 100  0  0  0  0
 4  0        0    97720     3892    79188    0    0     0     0 1044   89 99  2  0  0  0
 4  0        0    97720     3900    79180    0    0     0     8 1048  101 90 10  0  0  0
 4  0        0    97720     3900    79188    0    0     0     0 1042   85 100  0  0  0  0
 4  0        0    97720     3900    79188    0    0     0     0 1044   95 90 10  0  0  0
 4  0        0    97720     3908    79180    0    0     0     8 1042   78 100  0  0  0  0
 4  0        0    97720     3908    79188    0    0     0     0 1043   80 89 11  0  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 5  0        0    97720     3908    79188    0    0     0     0 1047  109 99  1  0  0  0
 4  0        0    97720     3916    79180    0    0     0     8 1043   84 91  9  0  0  0
 4  0        0    97720     3916    79188    0    0     0     0 1045   87 91  9  0  0  0
 4  0        0    97720     3916    79188    0    0     0     0 1042   92 100  0  0  0  0
 4  0        0    97720     3924    79180    0    0     0    32 1057  121 91  9  0  0  0
 4  0        0    97720     3924    79188    0    0     0     0 1045   84 96  4  0  0  0
 4  0        0    97720     3924    79188    0    0     0     0 1045  100 95  5  0  0  0
 4  0        0    97720     3932    79180    0    0     0    18 1050   99 92  9  0  0  0
 4  0        0    97720     3932    79188    0    0     0     0 1262  458 92  8  0  0  0
 4  0        0    97720     3932    79188    0    0     0     0 1043   99 92  8  0  0  0
 4  0        0    97720     3940    79180    0    0     0    10 1045   86 95  5  0  0  0
 4  0        0    97720     3940    79188    0    0     0     0 1043   85 93  7  0  0  0
 4  0        0    97720     3940    79188    0    0     0     0 1048  103 91  9  0  0  0
 4  0        0    97720     3948    79180    0    0     0     8 1045   82 94  6  0  0  0
 4  0        0    97720     3948    79188    0    0     0     0 1042   85 94  6  0  0  0
 4  0        0    97720     3948    79188    0    0     0     0 1047  114 91  9  0  0  0
 4  0        0    97720     3956    79180    0    0     0     8 1044   81 89 11  0  0  0
 4  0        0    97720     3956    79188    0    0     0     0 1045   90 87 14  0  0  0
 1  0        0    97728     3956    79192    0    0     2     0  987  176 61 18 21  0  0
 1  0        0    97728     3964    79184    0    0     0    10  771  102 25  0 75  0  0
 1  0        0    97728     3964    79192    0    0     0     0  782  121 25  0 75  0  0
procs ---------------memory-------------- ---swap-- -----io---- -system-- ------cpu-----
 r  b     swpd     free     buff    cache   si   so    bi    bo   in   cs us sy id wa st
 2  0        0    97728     3964    79192    0    0     0     0  760   81 25  0 75  0  0
 1  0        0    97728     3972    79184    0    0     0    20  766  100 25  0 74  0  0
 1  0        0    97728     3972    79192    0    0     0     0  767   89 25  0 75  0  0
 0  0        0   606656     3972    79304    0    0    56     0  820  166 31  4 64  0  0
 0  0        0   606656     3980    79296    0    0     0    16  667   88  0  0 100  0  0
 0  0        0   606656     3980    79304    0    0     0     0  662   74  0  0 100  0  0
 0  0        0   606656     3980    79304    0    0     0     0  672   84  0  0 100  0  0
 0  0        0   606656     3988    79296    0    0     0     8  676   76  1  0 99  0  0
 0  0        0   606656     3988    79304    0    0     0     0  689  148  1  0 99  0  0
 0  0        0   606532     3988    79304    0    0     0     0  697  248  2  0 98  0  0
 0  0        0   606532     3996    79296    0    0     0     8  691  296  1  1 98  0  0
 0  0        0   606532     3996    79304    0    0     0     0  685  277  1  0 99  0  0
100.000 MHz 20.22 MB/s 41.30 MB/s 41.94 MB/s

Code: Select all

[email protected]:~ $ sudo ./sdbench.sh
[3;J
CONFIG: 
CLOCK : 100.000 MHz
CORE  : 500 MHz, turbo=1
DATA  : 512 MB, /root/test.dat

HDPARM:
======
 Timing O_DIRECT disk reads: 128 MB in  3.04 seconds =  42.15 MB/sec
 Timing O_DIRECT disk reads: 130 MB in  3.01 seconds =  43.21 MB/sec
 Timing O_DIRECT disk reads: 122 MB in  3.01 seconds =  40.51 MB/sec

WRITE:
=====
536870912 bytes (537 MB) copied, 22.7367 s, 23.6 MB/s
536870912 bytes (537 MB) copied, 24.6147 s, 21.8 MB/s
536870912 bytes (537 MB) copied, 29.5037 s, 18.2 MB/s

READ:
====
536870912 bytes (537 MB) copied, 12.3361 s, 43.5 MB/s
536870912 bytes (537 MB) copied, 12.3478 s, 43.5 MB/s
536870912 bytes (537 MB) copied, 12.5074 s, 42.9 MB/s

RESULT (AVG):
============
Overlay config                      core_freq   turbo   overclock_50    WRITE        READ        HDPARM
                                       500        1     100.000 MHz   20.22 MB/s   41.30 MB/s   41.94 MB/s
114.3134s

Code: Select all

[email protected]:~ $ sysbench --num-threads=4 --test=cpu --cpu-max-prime=20000 --validate run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 4
Additional request validation enabled.


Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 20000


Test execution summary:
    total time:                          114.3134s
    total number of events:              10000
    total time taken by event execution: 457.1349
    per-request statistics:
         min:                                 45.42ms
         avg:                                 45.71ms
         max:                                 66.19ms
         approx.  95 percentile:              46.06ms

Threads fairness:
    events (avg/stddev):           2500.0000/11.11
    execution time (avg/stddev):   114.2837/0.02
6.583e+00

Code: Select all

[email protected]:~ $ ./xhpl
================================================================================
HPLinpack 2.1  --  High-Performance Linpack benchmark  --   October 26, 2012
Written by A. Petitet and R. Clint Whaley,  Innovative Computing Laboratory, UTK
Modified by Piotr Luszczek, Innovative Computing Laboratory, UTK
Modified by Julien Langou, University of Colorado Denver
================================================================================

An explanation of the input/output parameters follows:
T/V    : Wall time / encoded variant.
N      : The order of the coefficient matrix A.
NB     : The partitioning blocking factor.
P      : The number of process rows.
Q      : The number of process columns.
Time   : Time in seconds to solve the linear system.
Gflops : Rate of execution for solving the linear system.

The following parameter values will be used:

N      :    8000 
NB     :     256 
PMAP   : Row-major process mapping
P      :       1 
Q      :       1 
PFACT  :    Left 
NBMIN  :       2 
NDIV   :       2 
RFACT  :   Right 
BCAST  :   2ring 
DEPTH  :       0 
SWAP   : Mix (threshold = 64)
L1     : transposed form
U      : transposed form
EQUIL  : yes
ALIGN  : 8 double precision words

--------------------------------------------------------------------------------

- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
      ||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be               1.110223e-16
- Computational tests pass if scaled residuals are less than                16.0

================================================================================
T/V                N    NB     P     Q               Time                 Gflops
--------------------------------------------------------------------------------
WR02R2L2        8000   256     1     1              51.86              6.583e+00
HPL_pdgesv() start time Wed May 11 14:34:17 2016

HPL_pdgesv() end time   Wed May 11 14:35:08 2016

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)=        0.0025941 ...... PASSED
================================================================================

Finished      1 tests with the following results:
              1 tests completed and passed residual checks,
              0 tests completed and failed residual checks,
              0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests.
Last edited by Rive on Wed May 11, 2016 6:55 pm, edited 6 times in total.
DNPNWO

Heater
Posts: 13704
Joined: Tue Jul 17, 2012 3:02 pm

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 5:43 pm

Rive,

What is the bottom line here? Speed down by 10%, 20%, more, less, what?

I don't want to have to wade through all those figures.

I'm still trying to get my Pi 3 to not throttle back to half speed when I run a benchmark :(
Memory in C++ is a leaky abstraction .

leiptrstormr
Posts: 60
Joined: Mon May 18, 2015 12:33 pm

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 5:57 pm

I'm still trying to get my Pi 3 to not throttle back to half speed when I run a benchmark :(
I noticed it was throttling back as well so I installed cpufrequtils and set the governor to performance. It seemed to speed things up a bit.
https://wiki.debian.org/HowTo/CpuFrequencyScaling

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 6:43 pm

The linpack test may well throttle if running without a heatsink - are you seeing the over-temp symbol?
You can also run "vcgencmd get_throttled" after the test (with the updated firmware) to confirm. "0x0" would mean no throttling.

EDIT: would be useful to confirm if it is the firmware or kernel that has the effect.
You can copy start*.elf and fixup*.dat from /boot to somewhere safe, and repeat the tests with old firmware+new kernel and new firmware+old kernel.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 6:45 pm

Post updated for both 4.1.19 and 4.4.9 for vmstat 2 -w running for all benches.
viewtopic.php?p=972291#p972291
Last edited by Rive on Wed May 11, 2016 6:50 pm, edited 1 time in total.
DNPNWO

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 6:49 pm

dom wrote:The linpack test may well throttle if running without a heatsink - are you seeing the over-temp symbol?
You can also run "vcgencmd get_throttled" after the test (with the updated firmware) to confirm. "0x0" would mean no throttling.

EDIT: would be useful to confirm if it is the firmware or kernel that has the effect.
You can copy start*.elf and fixup*.dat from /boot to somewhere safe, and repeat the tests with old firmware+new kernel and new firmware+old kernel.

I cant speak for anyone else, but in my case, I run linpack at 49C
DNPNWO

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 6:51 pm

Heater wrote:Rive,

What is the bottom line here? Speed down by 10%, 20%, more, less, what?

I don't want to have to wade through all those figures.

I'm still trying to get my Pi 3 to not throttle back to half speed when I run a benchmark :(
Approx 20% performance degradation give or take. That is significant I think. :shock:


Let's put it this way, in the above bench for 4.1.19, @ 6.58 Gflops, the Pi3 does almost a BILLION more operations per second, than it was able do in 4.4.9 (@ 5.79 Gflops). A BILLION operations per second...more :o
FLOPS is a standard measure of computing power -- "Floating Point Operations Per Second". A GigaFlop (or Gflop) is a billion FLOPS. A TeraFlps (or TFlop) is a trillion FLOPS.
So a 1 FLOP machine will do one "operation" in a second. A 1 GFlop machine will do a billion operations in a second. A 2 GFlop machine will do two billion operations in a second. ie, by this measure, a 2 GFlop machine will do twice as much computing work in the same time as a 1 GFlop machine.
Last edited by Rive on Wed May 11, 2016 7:44 pm, edited 10 times in total.
DNPNWO

Heater
Posts: 13704
Joined: Tue Jul 17, 2012 3:02 pm

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 6:53 pm

dom,

So far as I can tell the throttling that I see is PSU related. The temps I'm seeing don't look extreme.

All of which is annoying. My power supplies can put out the current without drooping. But closer inspection shows
they have rather rough regulation.

I'm going to nail that issue before I start on the temperature and heat sink thing.

Oh, yeah, can't see any over-temp symbol. Everything runs headless here.
Memory in C++ is a leaky abstraction .

jdb
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 2131
Joined: Thu Jul 11, 2013 2:37 pm

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 7:06 pm

There's nearly an order of magnitude more interrupts per second and context switches per second in the 4.4.9 case compared to 4.1. Interrupts/context switches incur an overhead - but having such a large disparity in numbers between kernel versions is suspicious.
Rockets are loud.
https://astro-pi.org

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 7:09 pm

jdb wrote:There's nearly an order of magnitude more interrupts per second and context switches per second in the 4.4.9 case compared to 4.1. Interrupts/context switches incur an overhead - but having such a large disparity in numbers between kernel versions is suspicious.
Something else i noticed is that the sdbench 4.4.9 shows turbo=0, whereas 4.1.19 shows turbo=1.....and I have no turbo settings enabled.

So it looks like the sd_overclock is doing something different in 4.1, then it is in 4.4.
DNPNWO

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 7:36 pm

Rive wrote: Something else i noticed is that the sdbench 4.4.9 shows turbo=0, whereas 4.1.19 shows turbo=1.....and I have no turbo settings enabled.

So it looks like the sd_overclock is doing something different in 4.1, then it is in 4.4.
Yes, we no longer use io_is_busy, which used to force turbo mode on any sdcard access. We now handle changing core_freq though a different means.
That change is expected.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 7:37 pm

Heater wrote: Oh, yeah, can't see any over-temp symbol. Everything runs headless here.
Report "vcgencmd get_throttled" output after doing some activity. If you get a non-zero result I can decode it.

ejolson
Posts: 3731
Joined: Tue Mar 18, 2014 11:47 am

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 7:47 pm

Rive wrote:
jdb wrote:There's nearly an order of magnitude more interrupts per second and context switches per second in the 4.4.9 case compared to 4.1. Interrupts/context switches incur an overhead - but having such a large disparity in numbers between kernel versions is suspicious.
Something else i noticed is that the sdbench 4.4.9 shows turbo=0, whereas 4.1.19 shows turbo=1.....and I have no turbo settings enabled.

So it looks like the sd_overclock is doing something different in 4.1, then it is in 4.4.
I think you already found the difference yourself as turbo=0 was highlighted in your original post!

There is a possibility that Pi 3 crashes and wrong results are related to the sudden reduction in CPU voltage when the CPU leaves turbo mode while executing NEON instructions. If so, then turning off turbo mode safely at boot would mitigate this fault. Cooling a system well enough that it naturally stays in turbo mode would be an alternative way to mitigate the same problem. Of course all of this is conjecture.

If you type

Code: Select all

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
does the new kernel report powersave, ondemand or performance?
Last edited by ejolson on Wed May 11, 2016 7:59 pm, edited 6 times in total.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 7:49 pm

ejolson wrote: There is a possibility that Pi 3 crashes and wrong results are related to the sudden reduction in CPU voltage when the CPU leaves turbo mode while executing NEON instructions. If so, then turning off turbo mode safely at boot would mitigate this fault. Cooling a system well enough that it naturally stays in turbo mode would be an alternative way to mitigate the same problem. Of course all of this is conjecture.

If you type

Code: Select all

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
does the new kernel report powersave, ondemand or performance?
My Pi3 doesnt crash, and is stable (and thermals never exceed 56C even when executing NEON instructions...e.g., cpuburn-a53 or linpack).
I don't use turbo mode, and it isn't enabled in /boot/config.txt (except apparently by "sd_overclock' in 4.1.19). My governor is 'ondemand' in both kernels (idling at 600 MHz).

cpuburn-a53 (neon)

Code: Select all

[email protected]:~ $ while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 10; done& ./cpuburn-a53
[1] 1243
frequency(45)=1270134000
temp=32.2'C
frequency(45)=1270000000
temp=45.1'C
frequency(45)=1270000000
temp=47.2'C
frequency(45)=1270000000
temp=49.4'C
frequency(45)=1270002000
temp=50.5'C
frequency(45)=1270000000
temp=51.5'C
frequency(45)=1270000000
temp=52.6'C
frequency(45)=1270000000
temp=53.7'C
frequency(45)=1270000000
temp=53.7'C
frequency(45)=1270000000
temp=53.7'C
frequency(45)=1270000000
temp=54.8'C
frequency(45)=1270000000
temp=53.7'C
frequency(45)=1270000000
temp=53.7'C
frequency(45)=1270000000
temp=54.8'C
frequency(45)=1270000000
temp=54.8'C
frequency(45)=1270002000
temp=54.8'C
frequency(45)=1270000000
temp=54.8'C
frequency(45)=1270000000
temp=55.3'C
frequency(45)=1270000000
temp=55.3'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1269998000
temp=55.8'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1269998000
temp=55.8'C
frequency(45)=1270000000
temp=56.4'C
frequency(45)=1270000000
temp=56.4'C
frequency(45)=1270000000
temp=56.4'C
frequency(45)=1270000000
temp=55.8'C
frequency(45)=1270000000
temp=56.4'C
frequency(45)=1270000000
temp=56.4'C
frequency(45)=1270000000
temp=56.4'C
frequency(45)=1270000000
temp=56.4'C
Last edited by Rive on Wed May 11, 2016 8:02 pm, edited 1 time in total.
DNPNWO

ejolson
Posts: 3731
Joined: Tue Mar 18, 2014 11:47 am

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 7:57 pm

Rive wrote:I don't use turbo mode, and it isn't enabled in /boot/config.txt (except apparently by "sd_overclock' in 4.1.19). My governor is 'ondemand' in both kernels (idling at 600 MHz).
What happens if you set the CPU governor to performance in the 4.4.9 kernel?

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Raspbian Jessie linux 4.4.9 Severe Performance Degradati

Wed May 11, 2016 8:03 pm

ejolson wrote:
Rive wrote:I don't use turbo mode, and it isn't enabled in /boot/config.txt (except apparently by "sd_overclock' in 4.1.19). My governor is 'ondemand' in both kernels (idling at 600 MHz).
What happens if you set the CPU governor to performance in the 4.4.9 kernel?
I dunno. I am currently on 4.1.19 atm. But I assume it would always run at 1.26 Ghz, at around 55C ;) if all cores where maxed with Neon instructions (just like it does when maxed out with neon ondemand). The thermals did not significantly vary between kernels.

I have adequate cooling trust me. I wont overheat.
Last edited by Rive on Wed May 11, 2016 8:06 pm, edited 2 times in total.
DNPNWO

Return to “General discussion”