# Set Scaling Governor
Recently, I assembled a nice laptop based on barebone system from Eluktronics, NB50TZ.
This system is amazing and I really like an idea of concept to buy separately body, CPU, memory and SSD.
Right now, I am using this laptop to write this post. So, this is I finally built:
- Eluktronics, NB50TZ - $250 from Eluktronics.com
- Intel i7 9700 - $280 from MicroCenter.com
- CORSAIR Vengeance Performance 32GB (2x16GB) 260-Pin DDR4 SO-DIMM DDR4 2666 (PC4 21300) Laptop Memory Model CMSX32GX4M2A2666C18 - $128 from Amazon.com
- WD Blue 3D NAND 1TB Internal PC SSD - SATA III 6 Gb/s, M.2 2280, Up to 560 MB/s - WDS100T2B0B - $88 from Amazon.com
- Ubuntu 19.10 - FREE
Total cost of my assembly is $746 + tax.
In order to test performance on Ubuntu, I installed performance tests suite from phoronix-test-suite (opens new window) with CPU tests.
Second step is the preparation for CPU tests. All programs have to be closed and CPU clock frequency have to be scaled up.
The following picture shows the range of frequencies that CPU is using by default.
As we see, we need to limit range on upper possible frequency to get the best performance tests results. I have 8 independent cores in my i7, therefore I wrote the script to change scaling governor for all of them.
Here is the python script:
#!/bin/python3
import multiprocessing
import os
import sys
def choice_parser(v):
val = int(v)
if val == 1:
return 'powersave'
elif val == 2:
return 'performance'
else:
print('Wrong value: %s' % val)
sys.exit()
cpu_count = multiprocessing.cpu_count()
print('Set scaling governor for #%s cpus' % cpu_count)
print('Select scaling governor:')
print('1 for powersave')
print('2 for performance')
val = input(">>")
mode = choice_parser(val)
print('Your choice is %s' % mode)
for i in range(cpu_count):
print("Set cpu #%s" % i)
cmd = 'echo %s > /sys/devices/system/cpu/cpu%s/cpufreq/scaling_governor' % (mode, i)
print(cmd)
os.system(cmd)
os.system('cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor')
Selecting in this script parameter 2, moves all my settings to performance and changes frequency of cpu cores.
root@XXXXXX:~# ./set_scaling_governor.py
Set scaling governor for #8 cpus
Select scaling governor:
1 for powersave
2 for performance
>>2
Your choice is performance
Set cpu #0
echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
Set cpu #1
echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
Set cpu #2
echo performance > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor
Set cpu #3
echo performance > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor
Set cpu #4
echo performance > /sys/devices/system/cpu/cpu4/cpufreq/scaling_governor
Set cpu #5
echo performance > /sys/devices/system/cpu/cpu5/cpufreq/scaling_governor
Set cpu #6
echo performance > /sys/devices/system/cpu/cpu6/cpufreq/scaling_governor
Set cpu #7
echo performance > /sys/devices/system/cpu/cpu7/cpufreq/scaling_governor
performance
performance
performance
performance
performance
performance
performance
performance
Let's check the frequencies:
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_cur_freq
My output is here:
root@XXXXXX:~# cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_cur_freq
4538651
4542060
4589555
4562472
4596662
4598009
4583924
4579725
As we see here, my CPU now hot is ready for performance tests.
Let's run performance tests:
phoronix-test-suite run pts/cpu
My results are here:
Phoronix Test Suite v9.2.1
System Information
PROCESSOR: Intel Core i7-9700 @ 4.70GHz
Core Count: 8
Extensions: SSE 4.2 + AVX2 + AVX + RDRAND + FSGSBASE
Cache Size: 12288 KB
Microcode: 0xca
Scaling Driver: intel_pstate performance
GRAPHICS: Intel HD 3GB
Frequency: 1200MHz
OpenGL: 4.5 Mesa 19.0.8
Display Driver: modesetting 1.20.4
Monitor: ED273UR
Screen: 4480x1440
MOTHERBOARD: Notebook NB50TZ
BIOS Version: 1.07.03
Chipset: Intel Cannon Lake PCH
Audio: Realtek ALC269VB
Network: Realtek RTL8111/8168/8411 + Intel-AC 9560
MEMORY: 32768MB
DISK: 1000GB Western Digital WDS100T2B0B + 512GB WorkflowD512
File-System: ext4
Mount Options: errors=remount-ro relatime rw
Disk Scheduler: MQ-DEADLINE
OPERATING SYSTEM: Ubuntu 19.04
Kernel: 5.0.0-37-generic (x86_64)
Desktop: GNOME Shell 3.32.2
Display Server: X Server 1.20.4
Compiler: GCC 8.3.0
Security: itlb_multihit: KVM: Mitigation of Split huge pages
+ l1tf: Not affected
+ mds: Not affected
+ meltdown: Not affected
+ spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp
+ spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization
+ spectre_v2: Mitigation of Enhanced IBRS IBPB: conditional RSB filling
+ tsx_async_abort: Mitigation of TSX disabled
Tests:
Rodinia 2.4:
pts/rodinia-1.2.2 [Test: OpenMP LavaMD]
Test 1 of 5
Estimated Trial Run Count: 3
Estimated Test Run-Time: 10 Minutes
Estimated Time To Completion: 1 Hour, 2 Minutes [23:41 PST]
Started Run 1 @ 22:40:17
Started Run 2 @ 22:41:12
Started Run 3 @ 22:42:04
Test: OpenMP LavaMD:
50.24
48.828
49.789
Average: 49.619 Seconds
Deviation: 1.45%
Result compared to 2,967 OpenBenchmarking.org samples since 7 November 2013; median result: 126. Box plot of samples:
[ |-------------------*-----------------------------####*#*#*#*]
This Result (71st Percentile): 49.619 ^
AMD Athlon 5150: 1127 ^ Intel Xeon E5-1680 v3: 101 ^
Intel Core i7-7700K: 158 ^
Intel Core i7-5775C: 213 ^
Rodinia 2.4:
pts/rodinia-1.2.2 [Test: OpenMP CFD Solver]
Test 2 of 5
Estimated Trial Run Count: 3
Estimated Test Run-Time: 10 Minutes
Estimated Time To Completion: 52 Minutes [23:34 PST]
Started Run 1 @ 22:43:04
Started Run 2 @ 22:43:39
Started Run 3 @ 22:44:13
Test: OpenMP CFD Solver:
30.535
30.024
30.22
Average: 30.260 Seconds
Deviation: 0.85%
Result compared to 2,419 OpenBenchmarking.org samples since 7 November 2013; median result: 48.08. Box plot of samples:
[ |-------*-------------------------------#*##*#*#*|]
This Result (64th Percentile): 30.260 ^
ARMv7 rev 3: 467 ^ Intel Xeon E3-1280 v5: 54 ^
Intel Core i7-3615QM: 78 ^
Intel Core i7-5600U: 113 ^
NAMD 2.13b1:
pts/namd-1.0.1
Test 3 of 5
Estimated Trial Run Count: 3
Estimated Test Run-Time: 9 Minutes
Estimated Time To Completion: 43 Minutes [23:27 PST]
Started Run 1 @ 22:44:53
Started Run 2 @ 22:47:36
Started Run 3 @ 22:49:40
ATPase Simulation - 327,506 Atoms:
2.6188
2.65091
2.64322
Average: 2.63764 days/ns
Deviation: 0.64%
Result compared to 1,439 OpenBenchmarking.org samples since 19 September 2018; median result: 1.79. Box plot of samples:
[ |----------------*----------------------*-#*###*!##*|]
This Result (40th Percentile): 2.63764 ^
AMD Athlon II X4 630: 13 ^ 2 x Intel Xeon E5-2683 v4: 1 ^
Intel Xeon Silver 4108: 4 ^
Intel Core i7-3770: 5 ^
x265 3.0:
pts/x265-1.1.0
Test 4 of 5
Estimated Trial Run Count: 3
Estimated Test Run-Time: 31 Minutes
Estimated Time To Completion: 35 Minutes [23:25 PST]
Started Run 1 @ 22:51:52
Started Run 2 @ 22:52:14
Started Run 3 @ 22:52:36
H.265 1080p Video Encoding:
33.73
32.94
33.09
Average: 33.25 Frames Per Second
Deviation: 1.26%
Result compared to 1,047 OpenBenchmarking.org samples since 4 February 2019; median result: 24.13. Box plot of samples:
[ |-----------####################!########*##*#####------------*-*------*----|]
This Result (68th Percentile): 33.25 ^
Intel Core i7-8700K: 31 ^ Intel Core i9-7960X: 52 ^
Intel Core i7-9700K: 47 ^
Intel Core i9-7980XE: 46 ^
7-Zip Compression 16.02:
pts/compress-7zip-1.7.0
Test 5 of 5
Estimated Trial Run Count: 3
Estimated Time To Completion: 4 Minutes [22:56 PST]
Started Run 1 @ 22:53:05
Started Run 2 @ 22:53:38
Started Run 3 @ 22:54:11
Compress Speed Test:
32049
31907
32138
Average: 32031 MIPS
Deviation: 0.36%
Result compared to 5,705 OpenBenchmarking.org samples since 3 January 2012; median result: 18883. Box plot of samples:
[-###*!####*####-----------------------*------*------------| * ]
^ This Result (66th Percentile): 32031
^ AMD FX-6300: 15759 AMD Ryzen Threadripper 3970X: 190119 ^
2 x Intel Xeon Gold 6138: 134793 ^
^ POWER9 altivec supported: 114005
Percentile Classification Of Current Benchmark Run
PROCESSOR
Rodinia
OpenMP LavaMD: 71st
OpenMP CFD Solver: 64th
NAMD
ATPase Simulation - 327,506 Atoms: 40th
x265
H.2.1.V.E: 68th
7-Zip Compression
C.S.T: 66th
OpenBenchmarking.org Percentile
My final results are in this pdf file: alex-laptop.pdf
For my opinion i7 9700 is overkill for Eluktronics, NB50TZ. I checked the public benchmark, there are 16 installations and only one is on i7 9th gen (mine).
This laptop I created not for machine learning, but mostly for development purposes. As a compiler machine, it is match my expectations.