Power consumption of PC
Wed, 10/24/2007 - 17:47 — Lubomir Kuzman
This review is oriented to measure of power consumption of PC at various benchmarks, V-core voltages and CPU, GPU and GDDR3 frequencies.
In some cases, measured max power consumptions did frisson to me. In no case I measured the values under magic 100 Watts that means energy
saving computer. I would like to show you the test results.
Introduction
Was you ever interested in power consumption of your PC? To tell the truth, especially me no, till point of time... I said: "It could be around 100W plus power consumption of monitor" However I tried to build by PC that have the best performance and possibility of overclocking considering to the best price and power consumption too. Also my old CRT monitor I exchanged for the power saving LCD. But this question, couldn't to sleep on both ears after I connected my house to internet and I started to the p2p network. "How much kWh will show my electromer this month? Will be better for me to underclock, or overclock my computer? Or will be better to set up the default values? What is the power distinction between underclocked, normal and overclocked mode? I really don't know. But once I'll measure it" After successful download of all what I wanted, I did the measurements of my PC.
Methodology of measurements
The measuring place contained next measuring instruments
- YOKOGAWA WT210 - Digital power meter
- BLOCK BR1000 - Isolating auto-transformer 1kW
- 2*BOHEMIA TRAFO BC250 - Isolating transformer 0,25kW
- extender sockets-cable, cables...
I connected the PC to measuring instrument YOKOGAWA WT210 and at isolating autotransformer I set of 230V~. I was measuring and typing input current, input power, and power factor during all running benchmarks. After this I set input voltage to 115V~ and repeat all measurements also with this input voltage. As I used for my PC two input voltages 230V~ and 115V~ and LCD monitor, whose power consumption I didn't measure, and it could be powered only of 230V~, for powering of LCD monitor and digital power meter I was using Isolating transformers. I used all that isolating transformers because if the input voltage of PC is just 115V~ and the power meter and LCD monitor are powered of 230V~ the all crushing grounding loop through the VGA cable could be closed between the PC and LCD.
PC contains next components:
- CPU: AMD Athlon X2 4000+ Brisbane ( 65W, 2100MHz, 65nm )
- Cooler of CPU: Coolink Silentator + Noctua NF-S12–1200
- MoBo: ASUS m2n sli deluxe, BIOS 1102
- RAM: DDR2 A-DATA 2*1GB / 800MHz
- VGA: MSI GF8800GTS/640
- HDD: 2*WD2500KS @ RAID0, WD3200AAKS
- DVD RW: LG GSA-H44N
- PCI card: DeLock PCI FireWire 800 2+1
- Fans: 2*Noctua NF-S12–1200 push / pull
- PSU: Fortron BS2 500W + Noctua NF-S12–1200
Measured PC contains ordinary components and I suppose that power consumption of majority PCs will be resemble to this measured one. Maybe only the CPU with higher TDP and my "hungry" graphic card or 3*HDD would to make higher differences of standard PC's power consumption. States of PC / benchmarks that ran at PC:
1. idle - WIN-XP-MCE without another running programs
2. ORTHOS
3. 3D Mark 2006 - trial
4. SiSoft Sandra Lite XI - Processor Arithmetic
5. OVERLORD - demo
6. S.T.A.L.K.E.R. Shadow of Chernobyl - demo
To measuring of power consumption I used in a matter of fact 7 modes. Mode zero was turned-off standing-by PC state. Idle mode means running PC with only an operating system WIN-XP-MCE-SP2 without any antivirus, p2p clients and so on. Nothing running only icons at desktop ( CPU=0%, GPU-2D ). As next benchmark was used known program ORTHOS, that loads both CPU's cores and I measured input power of PC ( CPU=100%, GPU-2D ). As next was used the trial version of 3D Mark 2006 and I measured Input power in first "shooting" scene ( CPU=50%, GPU-3D ). In next benchmark named SiSoft Sandra Lite XI in section benchmarks I started Processor Arithmetic ( CPU=50%, GPU 2D ). And as a last two benchmarks I used two demos of famous games OVERLORD and S.T.A.L.K.E.R. Shadow of Chernobyl. For these two I set the details as maximal at resolution 1280*1024 and 4*AA a 16*AF. After a little trying in these games I found the positions where the power consumption was maximal. I was using these positions for reading power consumptions. I suppose, that in these two games were utilization of CPU and GPU at maximal possible levels.
I tried to make tests that the most draw near to usual PC user and not to some kind of freak whom next-to running eMule, ICQ, Skype, DVD-Shrink CATIA and next around 50 necessary programs at background and during defragmentation of both ( 3 ) HDDs runs the newest game, that is HW serious. So :) I measured the power consumption of ordinary PC durring ordinary work conditions and no for sure overloaded and "peging out" PC.
Used frequencies and voltages of CPU and GPU:
1. CPU @ 1000MHz @ 1,00V; GPU @ 143MHz / 424Mhz
2. CPU @ 1525MHz @ 1,00V; GPU @ 500MHz / 1600Mhz
3. CPU @ 2100MHz @ 1,28V; GPU @ 500MHz / 1600Mhz
4. CPU @ 2625MHz @ 1,33V; GPU @ 575MHz / 1700Mhz
5. CPU @ 3097MHz @ 1,392V; GPU @ 660MHz / 1780Mhz
The frequency of CPU and its V-core I readed-out, during its 100% load by ORTHOS, in program CPU-Z 1.41 and after I fill data in table. The V-core values of loaded and no-loaded CPU during overclock were little shifed ( 0,208V ). Frequencies of the GPU-core and graphic memory I was adjusting in N-Vidia utility in screen properties / GPU clock, and I fill them in the table.
In my test through the use of six benchmarks I measured the power consumption of PC in "deeply underclocked" state of CPU & GPU ( 1 ). This state of PC was mainly orientated to minimal power consumption. That means minimal frequency of CPU by using of minimal usable stable V-core , and also minimal frequencies of core and memory at graphic card. The games in this state ran at approximately 3 FPS ( my guess ) but otherwise PC had no problem with "active" internet using ( torrents, eMule ... ), working in MS-Ofiice and browsing of the photos. Moderate underclocked state ( 2 ) already only of CPU, was state interim the "sick" CPU and GPU in games and low power consumption to normal state. In this state the graphic card stoped abuse to me at last :) and worked at its standard 500/1600MHz. Default state ( 3 ) means default voltages and frequencies for CPU, GPU and GDDR3 too. The next state was "little" overclocked state ( 4 ) where frequencies of CPU, GPU and GDDR3 weren't at such a high frequencies. These are possible in most cases without any special action with "virgins smile on face". The last overclocked state ( 5 ) meant overclocking to such a maximal usable frequencies of CPU etc, to correct running of all benchmarks for minimal half hour.
For stable running of PC in 5th state I did next changes: the V-core-chipset I set to 1.6V, the V-core-CPU I st to 1.55V, but CPU-Z showed 1.6V for unloaded and 1.392 for loaded CPU at 3097MHz. I also set GPU to 660MHz and GDDR3 to 1780MHz and I hope, that overclock at the stock cooler is so nice. The higher frequencies meant not instability, but instant reset of PC with running benchmarks. During this test I made certain of thin deadline between the stability and instability of PC.
I switched off the thermal regulation of fans before all tests started.
Introduction
Was you ever interested in power consumption of your PC? To tell the truth, especially me no, till point of time... I said: "It could be around 100W plus power consumption of monitor" However I tried to build by PC that have the best performance and possibility of overclocking considering to the best price and power consumption too. Also my old CRT monitor I exchanged for the power saving LCD. But this question, couldn't to sleep on both ears after I connected my house to internet and I started to the p2p network. "How much kWh will show my electromer this month? Will be better for me to underclock, or overclock my computer? Or will be better to set up the default values? What is the power distinction between underclocked, normal and overclocked mode? I really don't know. But once I'll measure it" After successful download of all what I wanted, I did the measurements of my PC.
Methodology of measurements
The measuring place contained next measuring instruments
- YOKOGAWA WT210 - Digital power meter
- BLOCK BR1000 - Isolating auto-transformer 1kW
- 2*BOHEMIA TRAFO BC250 - Isolating transformer 0,25kW
- extender sockets-cable, cables...
I connected the PC to measuring instrument YOKOGAWA WT210 and at isolating autotransformer I set of 230V~. I was measuring and typing input current, input power, and power factor during all running benchmarks. After this I set input voltage to 115V~ and repeat all measurements also with this input voltage. As I used for my PC two input voltages 230V~ and 115V~ and LCD monitor, whose power consumption I didn't measure, and it could be powered only of 230V~, for powering of LCD monitor and digital power meter I was using Isolating transformers. I used all that isolating transformers because if the input voltage of PC is just 115V~ and the power meter and LCD monitor are powered of 230V~ the all crushing grounding loop through the VGA cable could be closed between the PC and LCD.
PC contains next components:
- CPU: AMD Athlon X2 4000+ Brisbane ( 65W, 2100MHz, 65nm )
- Cooler of CPU: Coolink Silentator + Noctua NF-S12–1200
- MoBo: ASUS m2n sli deluxe, BIOS 1102
- RAM: DDR2 A-DATA 2*1GB / 800MHz
- VGA: MSI GF8800GTS/640
- HDD: 2*WD2500KS @ RAID0, WD3200AAKS
- DVD RW: LG GSA-H44N
- PCI card: DeLock PCI FireWire 800 2+1
- Fans: 2*Noctua NF-S12–1200 push / pull
- PSU: Fortron BS2 500W + Noctua NF-S12–1200
Measured PC contains ordinary components and I suppose that power consumption of majority PCs will be resemble to this measured one. Maybe only the CPU with higher TDP and my "hungry" graphic card or 3*HDD would to make higher differences of standard PC's power consumption. States of PC / benchmarks that ran at PC:
1. idle - WIN-XP-MCE without another running programs
2. ORTHOS
3. 3D Mark 2006 - trial
4. SiSoft Sandra Lite XI - Processor Arithmetic
5. OVERLORD - demo
6. S.T.A.L.K.E.R. Shadow of Chernobyl - demo
To measuring of power consumption I used in a matter of fact 7 modes. Mode zero was turned-off standing-by PC state. Idle mode means running PC with only an operating system WIN-XP-MCE-SP2 without any antivirus, p2p clients and so on. Nothing running only icons at desktop ( CPU=0%, GPU-2D ). As next benchmark was used known program ORTHOS, that loads both CPU's cores and I measured input power of PC ( CPU=100%, GPU-2D ). As next was used the trial version of 3D Mark 2006 and I measured Input power in first "shooting" scene ( CPU=50%, GPU-3D ). In next benchmark named SiSoft Sandra Lite XI in section benchmarks I started Processor Arithmetic ( CPU=50%, GPU 2D ). And as a last two benchmarks I used two demos of famous games OVERLORD and S.T.A.L.K.E.R. Shadow of Chernobyl. For these two I set the details as maximal at resolution 1280*1024 and 4*AA a 16*AF. After a little trying in these games I found the positions where the power consumption was maximal. I was using these positions for reading power consumptions. I suppose, that in these two games were utilization of CPU and GPU at maximal possible levels.
I tried to make tests that the most draw near to usual PC user and not to some kind of freak whom next-to running eMule, ICQ, Skype, DVD-Shrink CATIA and next around 50 necessary programs at background and during defragmentation of both ( 3 ) HDDs runs the newest game, that is HW serious. So :) I measured the power consumption of ordinary PC durring ordinary work conditions and no for sure overloaded and "peging out" PC.
Used frequencies and voltages of CPU and GPU:
1. CPU @ 1000MHz @ 1,00V; GPU @ 143MHz / 424Mhz
2. CPU @ 1525MHz @ 1,00V; GPU @ 500MHz / 1600Mhz
3. CPU @ 2100MHz @ 1,28V; GPU @ 500MHz / 1600Mhz
4. CPU @ 2625MHz @ 1,33V; GPU @ 575MHz / 1700Mhz
5. CPU @ 3097MHz @ 1,392V; GPU @ 660MHz / 1780Mhz
The frequency of CPU and its V-core I readed-out, during its 100% load by ORTHOS, in program CPU-Z 1.41 and after I fill data in table. The V-core values of loaded and no-loaded CPU during overclock were little shifed ( 0,208V ). Frequencies of the GPU-core and graphic memory I was adjusting in N-Vidia utility in screen properties / GPU clock, and I fill them in the table.
In my test through the use of six benchmarks I measured the power consumption of PC in "deeply underclocked" state of CPU & GPU ( 1 ). This state of PC was mainly orientated to minimal power consumption. That means minimal frequency of CPU by using of minimal usable stable V-core , and also minimal frequencies of core and memory at graphic card. The games in this state ran at approximately 3 FPS ( my guess ) but otherwise PC had no problem with "active" internet using ( torrents, eMule ... ), working in MS-Ofiice and browsing of the photos. Moderate underclocked state ( 2 ) already only of CPU, was state interim the "sick" CPU and GPU in games and low power consumption to normal state. In this state the graphic card stoped abuse to me at last :) and worked at its standard 500/1600MHz. Default state ( 3 ) means default voltages and frequencies for CPU, GPU and GDDR3 too. The next state was "little" overclocked state ( 4 ) where frequencies of CPU, GPU and GDDR3 weren't at such a high frequencies. These are possible in most cases without any special action with "virgins smile on face". The last overclocked state ( 5 ) meant overclocking to such a maximal usable frequencies of CPU etc, to correct running of all benchmarks for minimal half hour.
For stable running of PC in 5th state I did next changes: the V-core-chipset I set to 1.6V, the V-core-CPU I st to 1.55V, but CPU-Z showed 1.6V for unloaded and 1.392 for loaded CPU at 3097MHz. I also set GPU to 660MHz and GDDR3 to 1780MHz and I hope, that overclock at the stock cooler is so nice. The higher frequencies meant not instability, but instant reset of PC with running benchmarks. During this test I made certain of thin deadline between the stability and instability of PC.
I switched off the thermal regulation of fans before all tests started.
