From Fedora Project Wiki

RHEL 5 x Fedora 12

This is the first power consumption comparison of RHEL 5 and Fedora 12. The results was measured by BLTK that gets power consumption data from laptop's battery info. There were used three workloads for this comparison:

  • Idle - simulates no interaction from user side (no screen saver defined)
  • Office - simulates working in Open Office with breaks
  • Reader - simulates reading text in Firefox


Used Hardware and OS

Laptop

LENOVO ThinkPad T61
* CPU Intel Core2 Duo CPU T7500 2.20 GHz
* Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller
* LCD 15" 1680x1050x24
* Memory 2 GB
* HDD 100 GB
* Battery designed and usable capacity 56.16 Wh

RHEL 5

Kernel release 2.6.18-164.el5

Fedora 12

Kernel release 2.6.31-33.fc12.x86_64


Test description

The tests were done on laptop installed with RHEL 5, Fedora 11 and Fedora 12 using shared /home. The workload mentioned above was used with following specification:

Idle test - test was running for 10 minutes

bltk -I -T 600

Office test - Open Office workload was running with three repetitions (work simulation repeated 3x)

bltk -O -n 3

Reader test - Each of tests was running also with three repetitions. Firefox is a reader application by default (bltk.conf).

bltk -R -n 3

Result tables and charts

Idle workload

RHEL 5

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
1000.01 0.07 0.19 0 17.59 0.39 0 0.54 100 0 0 56160 100 53150 94.64 18.06 186
1000 0 0.08 0 17.16 0.36 0 0.55 100 0 0 55600 99 52800 94.01 16.8 200
1000.01 0.07 0.19 0 18.48 0.38 0 0.78 100 0 0 53130 94.6 50050 89.12 18.48 182
1000.01 0 0.18 0 19.09 0.83 0.02 0.62 100 0 0 52760 93.94 49690 88.47 18.42 182
1000.01 0.07 0.24 0 19.33 1 0 0.55 100 0 0 50040 89.1 46900 83.51 18.84 178
1000.01 0 0.3 0 18.44 0.72 0 0.54 100 0 0 48150 85.73 45010 80.14 18.84 178
1000.01 0 0.3 0 16.75 0.33 0 0.55 100 0 0 44920 79.98 41750 74.34 19.02 177
1000.01 0.03 0.21 0 18.12 0.57 0 0.59 100 0 0 18.35 183.29


F12

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
17.19 9.21 0.27 17.41 17.4 0.4 0 0.14 54.55 45.45 0 56160 100 53900 95.97 13.56 248
16.69 9.62 0.26 16.63 19.42 0.34 0 0.14 54.55 45.45 0 52880 94.15 50460 89.85 14.52 232
45.55 15.01 0.67 46.43 22.65 0.57 4.51 1.02 90.91 9.09 0 52240 93.01 49640 88.39 15.6 216
13.84 6.49 0.15 14.96 15.36 0.33 0 0.18 54.55 45.45 0 49600 88.31 47120 83.9 14.88 226
16.6 9.02 0.28 16.19 16.11 0.32 0 0.12 36.36 63.64 0 47810 85.13 45390 80.82 14.52 232
13.79 6.65 0.16 15.19 14.77 0.34 0 0.14 36.36 63.64 0 47100 83.86 44610 79.43 14.94 225
17.76 9.75 0.32 17.67 15.94 0.39 0.02 0.34 54.55 45.45 0 44560 79.34 42080 74.92 14.88 226
20.2 9.39 0.3 20.64 17.38 0.38 0.65 0.3 54.55 45.45 0 14.7 229.29


F12 tuned

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
20.04 9.52 0.33 19.81 13.96 0.37 0 0.18 63.64 36.36 0 54320 96.72 51760 92.16 15.36 219
16.59 6.92 0.26 17.3 10.49 0.35 0 0.34 81.82 18.18 0 51750 92.14 49090 87.41 15.96 211
16.64 6.75 0.27 16.5 12.1 0.31 0 0.16 72.73 27.27 0 49040 87.32 46400 82.62 15.84 212
16.52 7.11 0.26 17.07 11.98 0.39 0 0.18 72.73 27.27 0 46380 82.58 43750 77.9 15.78 213
17.45 7.58 0.28 17.67 12.13 0.36 0 0.22 72.73 27.27 0 15.74 213.75


Office workload

RHEL 5

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
999.81 0.1 2.97 0 43.3 4.28 1.08 4.55 100 0 0 56160 100 45730 81.42 17.37 194
999.79 0.1 3.22 0 43.3 4.52 3.46 5.23 100 0 0 54140 96.4 42220 75.17 19.85 169
1000.2 0.1 3.41 0 47.98 4.28 4.83 5.13 100 0 0 45180 80.44 32230 57.38 21.57 156
1000.21 0.1 3.25 0 41.98 4.23 0 4.49 100 0 0 42210 75.16 29880 53.2 20.54 164
1000.21 0.1 3.25 0 43.02 4.2 0 7.17 100 0 0 30350 54.04 17850 31.78 20.82 161
1000.21 0.1 3.49 0 40.78 3.77 0 4.43 100 0 0 29870 53.18 17240 30.69 21.04 160
1000.07 0.1 3.27 0 43.39 4.21 1.56 5.17 100 0 0 20.2 167.33


Fedora 12

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
243.54 32.55 9.29 235.04 60.62 7.41 2.43 1.21 94.59 5.41 0 54950 97.84 44850 79.86 16.82 200
236.49 32.64 8.83 230.32 58.44 7.42 1.62 1.15 92.6 7.4 0 54040 96.22 43890 78.15 16.9 199
222.65 32.83 7.49 228.47 59.41 8.08 0 1.1 93.25 6.75 0 44830 79.82 34400 61.25 17.38 193
226.22 29.28 8.37 217.43 60.51 6.89 0 1.12 89.92 10.08 0 43850 78.08 33210 59.13 17.73 190
223.15 32.39 7.9 224.69 55.19 7.84 0 1.12 92 8 0 34380 61.21 23710 42.21 17.78 189
231.07 28.42 8.45 225.75 57.72 7.46 0 1.05 86.41 13.59 0 33190 59.09 22430 39.93 17.93 187
230.52 31.35 8.39 226.95 58.65 7.52 0.68 1.13 91.46 8.54 0 17.42 193


Fedora 12 tuned

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
237.85 32.17 8.73 233.21 58.43 7.47 1.64 1.12 100 0 0 53250 94.81 43800 77.99 15.74 214
236.82 30.78 8.5 234.68 58.91 7.76 0 1.1 99.72 0.28 0 43780 77.95 34010 60.55 16.28 207
224.49 29.24 8.41 216.69 56 6.62 0 1.03 100 0 0 33990 60.52 23950 42.64 16.73 201
235.12 30.28 8.44 233.31 57.56 7.75 0 1.08 99.91 0.09 0 23930 42.61 13790 24.55 16.89 199
233.57 30.62 8.52 229.47 57.73 7.4 0.41 1.08 99.91 0.09 0 16.41 205.25


Reader workload

RHEL 5

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
997.43 0.24 10.73 0 63.84 11.78 9.47 11.62 100 0 0 56160 100 55540 98.89 19.24 175
998.06 0.16 5.68 0 46.37 9.59 0 9.01 100 0 0 55540 98.89 54950 97.84 21.03 160
998.17 0.19 6.79 0 45.98 7.59 0 9.64 100 0 0 54940 97.82 54360 96.79 20.88 161
1003.51 0.2 7.38 0 61.37 11.12 9.66 10.17 100 0 0 52820 94.05 52200 92.94 20.86 161
1004.98 0.2 8.1 0 48.38 7.02 0 8.29 100 0 0 52200 92.94 51580 91.84 22.32 150
997.03 0.21 7.86 0 48.06 6.76 0 8.28 100 0 0 51580 91.84 50990 90.79 21.24 158
999.86 0.2 7.76 0 52.33 8.98 3.19 9.5 100 0 0 20.93 160.83


Fedora 12

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
301.19 65.69 13.45 303.94 72.75 13.88 0.03 3.16 100 0 0 56080 99.85 55490 98.8 20.23 166
310.11 42.88 14.09 291.02 73.3 10.86 0 2.95 100 0 0 55470 98.77 54910 97.77 19.38 173
325.22 43.5 12.51 338.85 80.57 15.15 0.01 3.74 100 0 0 54890 97.73 54320 96.72 19.92 169
312.94 43.51 14.47 306.45 66.66 13.48 0 3.39 100 0 0 27500 48.96 26830 47.77 22.97 146
280.32 45.93 9.69 310.83 75.95 15.23 0 3.33 100 0 0 26780 47.68 26110 46.49 23.19 145
288.66 41.61 10.36 309.44 69.88 14.23 0 2.81 100 0 0 26080 46.43 25410 45.24 23.19 145
303.07 47.19 12.43 310.09 73.19 13.81 0.01 3.23 100 0 0 21.48 157.33


Fedora 12 tuned

CPU1 CPU2 HDD BAT begin BAT end Drain BAT life
Tmr Irq Load % Tmr Irq Load % Rd Wr Act StnBy Slp [mWh] [%] [mWh] [%] [W] [min]
276.68 43.67 12.8 272.88 63.95 11.73 0 1.97 100 0 0 51940 92.48 51360 91.45 20.47 164
305.3 40.93 12.59 318.8 76.05 15.17 0 2.24 100 0 0 51340 91.41 50760 90.38 20.47 164
273.4 35.17 11.62 282.98 64.34 12.75 0.01 2.21 100 0 0 50730 90.33 50160 89.31 20.12 167
285.13 39.92 12.34 291.55 68.11 13.22 0 2.14 100 0 0 20.35 165


Power consumption and HDD activity in charts

Conclusion

  1. Fedora 12 is more power saving against RHEL 5 but it's not valid in all cases. The results illuminate dependency of power saving on used application. This conclusion will be investigated deeper in the future (e.g. test using another browser).
  2. The next finding shows dependency of power consumption on charging rate. The charts attached to particular work loads display growing power drain during battery discharging. There is currently no clear explanation for this effect. There is probably influence of two process - growing amperage that compensates falling voltage during discharging battery and growing hardware temperature (thermal absorption of hardware components and growing amperage).
  3. The data of CPU displays different ways in using two cores CPUs.
  4. Tuned is able to increase power saving but not in all cases.