Posts filed under ‘CPU’
Many low price laptop these days come with Intel Celeron M. Unlike its competitor, the AMD Mobile Sempron, Celeron M lacks dynamic frequency and voltage management. On processors like Mobile Sempron, Pentium M, Core Duo, and Core 2 Duo, the cpu drops frequency and voltage when there is less work load. So if nothing is going on, your cpu may drop from 1.8 Ghz to just 600 mhz and voltage may drop from 1.2v to 1.0v.
Celeron M in contrast runs at the same frequency and voltage all the time. This is why Celeron M typically have a shorter battery life than their mainstream cousins. Celeron M does have On-Demand Clock Modulation (ODCM). Can this be use to increase battery life?
How does ODCM work?
ODCM conjured up an image of a cpu dynamically changing the clock of the cpu depending on load. In fact, all ODCM does is force a processor to go into idle. Suppose you set the ODCM throttling to 75%. Your processor will work normally 75% of the time, and spend the other 25% slacking off. As a result, your cpu runs at the same clock speed and voltage, but runs slower because the cpu is forced to idle.
Does this actually save you power? Well, that depends on your usage. Suppose your cpu is running always at 100%, you are going to use a lot of power. Suppose you then set ODCM to throttle at 50%. 50% of the cycle will be idle, so you will use less power because the computer is at rest 50% of the time. However, the task will now take twice as long, this is not a good trade off.
If on the other hand, you’re going to do is type something in the computer, then your computer is probably idling most of the time waiting for you to press a key. In that case, power usage with ODCM and with ODCM is probably not all that different in both cases the computer is in idle most of the time. However, what if the forced idle uses less power than the natural idle state. Basically if you were to sit around and your computer goes into idle, does it use more power than if ODCM force the computer to idle? I decided to test this out.
The test platform consists of a Acer Aspire 3680 with a Celeron M 520 1.6 Ghz, 512 Mb RAM, and standard battery. I will run two trials.
- Normal Test – Laptop is turned on and left alone until the computer powers off.
- ODCM Test – Run RMClock and set it to throttle at 12.5%. Leave computer alone until the computer powers off. This causes the computer to go into force idle 87.5% of the time.
In both cases, we set the power manager to not turn off the monitor or hard disk. We also make sure nothing is schedule to run during the test. Wireless is turned off.
In both case, the computer powered off at about 2 hours. This mean that the forced idle uses the same amount of power as the normal idle. There was no power savings at all.
ODCM is not a useful tool for battery conservation. Most mobile processor conserve power by idling at a lower clock speed and voltage. ODCM is a poor substitute that attempts to force a processors to go into idle mode. However, since the force idle mode uses exactly the same power as the normal idle mode, you will not notice any power savings.
Here’s a list of Super PI score (2M) on different computers I have access to:
|Computer||SuperPI 2M Score (min)||Idle Temp (C)||Full Load Temp|
|Acer Aspire 3680. Celeron M 520 (1.6 Ghz) 512 Mb Vista||2:19|
|Acer Aspire 3680. Celeron M 520 (1.6 Ghz) 2 Gb (dual channel) Vista in Vmware||1:43|
|Acer Aspire 3680. Celeron M 520 (1.6 Ghz) 2 Gb (dual channel) Vista||1:39||57||65|
|ASUS eee 901 Intel Celeron M 900Mhz Windows XP||3:22|
|ASUS UL50 Intel Core 2 Duo SU7300 1.3G||1:06|
|Averatec 2370. Turion64 TL-50 (1.66 Ghz) Single RAM Module||2:06||45||70|
|Averatec 2370. Turion64 TL-50 (1.66 Ghz) Two RAM Module||2:06||45||70|
|Compaq 906us. Athlon XPM 1500+ (1.3 Ghz)||3:26|
|Custom Desktop. Athlon XPM 2600 (2.0 Ghz)||2:12|
|Custom Desktop. Athlon II X4 640 (3.00 Ghz)||1:01|
|Dell e1505. Core Duo T2250 (1.73 Ghz). Single RAM Module||1:37||50||65|
|Dell e1505. Core Duo T2250 (1.73 Ghz). Dual RAM Module||1:28||50||65|
|Dell Latitude D630 (2.4 Ghz Core 2 Duo T8300)||0:52|
|Dell Inpiron 4150 (1.7 Ghz Pentium 4 M)||3:51|
|Dell Inspiron 9300 (2.0 Ghz Pentium M)||1:45|
|eMachine T1221. Celeron Tualatin (1.3 Gz)||7:08|
|MSI Wind. Intel Atom N270 1.6 Ghz Windows XP||3:32|
|HP Pavilion dv5-1150us Intel Core2Duo P7350 2.00 Ghz||1:00|
|IBM NetVista. Pentium 4 (1.8 Ghz)||4:00|
|IBM NetVista. Pentium 4 (2.4 Ghz)||2:52|
|Toshiba 1805-S20. Celeron Coppermine (1.1 Ghz)||9:28|
|Toshiba Satellite C655-S5047 Celeron 900 (2.2 Ghz) Windows 7 64-bit||1:04|
Both Turion64 and Core 2 Duo supposedly gain improvement through a matched pair of RAM, but apparently there’s no difference in performance in Turion and only a small gain in the Core 2 Duo. All benchmark and temperature ran under Windows XP unless otherwise specified.
Recently, I gotten my fiancee a Dell e1505 with a T2250 and I got myself a Averatec 2370 with a TL-50. I was curious to see what the difference in performance and temperature the two processors were. I first check the temperature of the two processor at idle and underload.
|Intel Core Duo T2250||50C||60C|
|AMD Turion64x2 TL-50||45C||70C|
This sort of make sense, reviewers indicated that the Turion64 has lower power than the Core Duo on idle, but uses higher power under load. Of course, it could also mean that the e1505 has a better cooling system than the Averatec 2370.
I ran Super PI and here are the results.
|CPU||Max Frequency (Ghz)||Super PI 2M Score (min. Lower is better)|
|Intel Core Duo T2250||1.73||1:37|
|AMD Turion64x2 TL-50||1.6||2:06|
It appears that Core Duo is faster than the Turion at least in Super PI. I’ll do further test when I have a chance.