Power Consumption of 5 C2D mobos with IGP & discrete VGA
Posted: Mon Aug 28, 2006 5:33 am
1. The power consumption for all five motherboards tested when using a Gigabyte GF6200TC (128MB RAM) and it also shows how the two motherboards with integrated graphics compare when using their IGPs. As a comparison it also shows a Socket 479 Core Duo motherboard which is highlighted in red to standout.
The CPU clock speeds for the Core Duo system are very slightly different than for the C2D boards due to the different FSB and CPU multiplier values that the two architectures use - 1.66GHz & 2.16GHz.
2. Comparing the ASRock 775Dual-VSTA with 4 VGA cards.
The FX5200 is an AGP card whereas the rest are PCI Express.
I tested the 3 PCI Express cards with almost all of the boards and under a number of conditions; idle and load. The 6200TC consumed the least power with typically the 7300LE consuming an extra 1.5W and the 7600GS an extra 1W in comparison.
Full details for the VGA cards are listed in the System Information section below.
3. Comparing the ASRock 775Dual-VSTA with PC3200 (DDR-400) & PC2-5300 (DDR2-667) memory.
None of the tests outlined above are particularly memory intensive, so it’s possible that a more memory intensive application would show a different result.
4. The Gigabyte GA-965P-DS3 using DDR2-667 at 3 speed settings.
The FSB was set to the stock value of 266 which gives the possibility of running the RAM at a clock speed of 266, 333, 400 and even higher. The actual RAM speed is twice the clock speed, which is why the values in the table are 533, 667 & 800.
The last test unfortunately wasn’t completed in full, but it gives an indication of how RAM that requires more voltage affects power consumption. The faster the RAM and/or the tighter the timings the more voltage it generally needs.
I didn’t find a way to determine the voltage used by the RAM when its voltage was set to auto which makes the data harder to interpret.
5. The Gigabyte GA-965P-DS3 at multiple FSB settings for the same CPU clock speed.
As you can see from the disparate data in the above test with regard to voltages, this doesn’t give anywhere near a complete picture of how increasing the FSB for a fixed CPU speed affects power consumption. What it does highlight is that when you increase the FSB to the point that it requires extra voltage, the power consumption jumps noticeably.
The second test at 2.4GHz shows more clearly the effect of changing the FSB and RAM speeds whilst keeping the voltages fixed. The board wouldn’t run at 400 * 6 unfortunately, so I couldn’t complete the test as intended.
6. Under-volting the Gigabyte GA-965P-DS3 in the BIOS.
1.168V was the lowest VCore attainable using RMClock. The BIOS allows a VCore range of 0.52 – 2.0V, this is using BIOS version F3.
I think that the C2D used in these tests is at best average and more likely below par when it comes to over-clocking at stock VCore and in under-volting. Therefore, I didn’t spend much time investigating this. The fragility of the Gigabyte board didn’t inspire further experimentation either.
7. Comparing a Core 2 Duo desktop with a Core Duo laptop and desktop.
System Specifications
Core Duo Laptop - T2300E, 945GM, 1GB DDR2-533, GMA950, 60GB HDD, Screen switched off.
Core Duo Desktop - T2600, 945GM, 2GB DDR2-667, GMA950, 250GB HDD.
Core 2 Duo desktop - E6400, VIA PT880 Pro, 1GB DDR2-667, FX5200 128MB, 250GB HDD.
The Core 2 Duo would not drop below 1.6 GHz at idle, so I only show figures at 1GHz for the other two systems.
The difference between 1 and 2 GB of RAM resulted in a 1W difference at idle and load. This had no effect on the test as it was primarily looking at the difference between idle and load power consumption. Since the 1W difference was constant at both states, it cancelled itself out.
Main System Information
Antec P180
Silverstone ST30NF Power Supply (Fanless)
2 Nexus 120mm fans – one in lower P/S chamber, the other as exhaust on the side of the case opposite the CPU.
Intel E6400 2.13 GHz
Scythe Ninja + with Nexus 120mm
Arctic Silver 5 thermal paste
Samsung P120 250GB SATAII
NEC DVD Burner & floppy drive
Corsair XMS2 DDR2 PC2-5400 2 x 1GB
Kingston PC3200 1GB
Windows XP Pro SP2
(I haven’t listed fan speeds as temperatures weren’t reported)
Motherboards
ASRock 775Dual-VSTA
ASRock Conroe945G-DVI
Asus P5VD2-MX
Gigabyte GA-965P-DS3
Asus P5W DH Deluxe
Gigabyte GA-8i945GM
Graphics cards
Gigabyte GV-NX62TC256D8 – Nvidia 6200TC.
XFX PV-T72P-RAMG – Nvidia 7300LE.
Leadtek WinFast PX7600 GS TDH – Nvidia 7600GS.
The FX5200 is probably discontinued but its specification is – 128MB, AGP, DVI, D-SUB & TV-Out.
For more info on the differences between the varying Nvidia chipsets try Wikipedia.
Notes
RMClock 2.1 was used to set the multiplier and VCore from within Windows.
Speedfan 4.28 was used to monitor temperatures & VCore.
CPU-Z 1.33.1 was used to monitor VCore & FSB speed.
In the graphs, Burn-in refers to CPU Burn-in; Burn refers to CPUBurn.
For CPUBurn I ran the version optimised for the Intel P6; this is an old utility and there isn’t a version optimized for more recent CPUs.
CPU Burn-in was version 1.01 and Prime95 was 24.14. Two instances were run in each case, with the affinity being manually set to make sure each instance ran on its own core.
For CPU Burn-in the error checking mode was used, which is not the most power hungry. For Prime95 the torture test was used using the in-place large FFT setting, which is reported as the most power hungry.
The CPU clock speeds for the Core Duo system are very slightly different than for the C2D boards due to the different FSB and CPU multiplier values that the two architectures use - 1.66GHz & 2.16GHz.
2. Comparing the ASRock 775Dual-VSTA with 4 VGA cards.
The FX5200 is an AGP card whereas the rest are PCI Express.
I tested the 3 PCI Express cards with almost all of the boards and under a number of conditions; idle and load. The 6200TC consumed the least power with typically the 7300LE consuming an extra 1.5W and the 7600GS an extra 1W in comparison.
Full details for the VGA cards are listed in the System Information section below.
3. Comparing the ASRock 775Dual-VSTA with PC3200 (DDR-400) & PC2-5300 (DDR2-667) memory.
None of the tests outlined above are particularly memory intensive, so it’s possible that a more memory intensive application would show a different result.
4. The Gigabyte GA-965P-DS3 using DDR2-667 at 3 speed settings.
The FSB was set to the stock value of 266 which gives the possibility of running the RAM at a clock speed of 266, 333, 400 and even higher. The actual RAM speed is twice the clock speed, which is why the values in the table are 533, 667 & 800.
The last test unfortunately wasn’t completed in full, but it gives an indication of how RAM that requires more voltage affects power consumption. The faster the RAM and/or the tighter the timings the more voltage it generally needs.
I didn’t find a way to determine the voltage used by the RAM when its voltage was set to auto which makes the data harder to interpret.
5. The Gigabyte GA-965P-DS3 at multiple FSB settings for the same CPU clock speed.
As you can see from the disparate data in the above test with regard to voltages, this doesn’t give anywhere near a complete picture of how increasing the FSB for a fixed CPU speed affects power consumption. What it does highlight is that when you increase the FSB to the point that it requires extra voltage, the power consumption jumps noticeably.
The second test at 2.4GHz shows more clearly the effect of changing the FSB and RAM speeds whilst keeping the voltages fixed. The board wouldn’t run at 400 * 6 unfortunately, so I couldn’t complete the test as intended.
6. Under-volting the Gigabyte GA-965P-DS3 in the BIOS.
1.168V was the lowest VCore attainable using RMClock. The BIOS allows a VCore range of 0.52 – 2.0V, this is using BIOS version F3.
I think that the C2D used in these tests is at best average and more likely below par when it comes to over-clocking at stock VCore and in under-volting. Therefore, I didn’t spend much time investigating this. The fragility of the Gigabyte board didn’t inspire further experimentation either.
7. Comparing a Core 2 Duo desktop with a Core Duo laptop and desktop.
System Specifications
Core Duo Laptop - T2300E, 945GM, 1GB DDR2-533, GMA950, 60GB HDD, Screen switched off.
Core Duo Desktop - T2600, 945GM, 2GB DDR2-667, GMA950, 250GB HDD.
Core 2 Duo desktop - E6400, VIA PT880 Pro, 1GB DDR2-667, FX5200 128MB, 250GB HDD.
The Core 2 Duo would not drop below 1.6 GHz at idle, so I only show figures at 1GHz for the other two systems.
The difference between 1 and 2 GB of RAM resulted in a 1W difference at idle and load. This had no effect on the test as it was primarily looking at the difference between idle and load power consumption. Since the 1W difference was constant at both states, it cancelled itself out.
Main System Information
Antec P180
Silverstone ST30NF Power Supply (Fanless)
2 Nexus 120mm fans – one in lower P/S chamber, the other as exhaust on the side of the case opposite the CPU.
Intel E6400 2.13 GHz
Scythe Ninja + with Nexus 120mm
Arctic Silver 5 thermal paste
Samsung P120 250GB SATAII
NEC DVD Burner & floppy drive
Corsair XMS2 DDR2 PC2-5400 2 x 1GB
Kingston PC3200 1GB
Windows XP Pro SP2
(I haven’t listed fan speeds as temperatures weren’t reported)
Motherboards
ASRock 775Dual-VSTA
ASRock Conroe945G-DVI
Asus P5VD2-MX
Gigabyte GA-965P-DS3
Asus P5W DH Deluxe
Gigabyte GA-8i945GM
Graphics cards
Gigabyte GV-NX62TC256D8 – Nvidia 6200TC.
XFX PV-T72P-RAMG – Nvidia 7300LE.
Leadtek WinFast PX7600 GS TDH – Nvidia 7600GS.
The FX5200 is probably discontinued but its specification is – 128MB, AGP, DVI, D-SUB & TV-Out.
For more info on the differences between the varying Nvidia chipsets try Wikipedia.
Notes
RMClock 2.1 was used to set the multiplier and VCore from within Windows.
Speedfan 4.28 was used to monitor temperatures & VCore.
CPU-Z 1.33.1 was used to monitor VCore & FSB speed.
In the graphs, Burn-in refers to CPU Burn-in; Burn refers to CPUBurn.
For CPUBurn I ran the version optimised for the Intel P6; this is an old utility and there isn’t a version optimized for more recent CPUs.
CPU Burn-in was version 1.01 and Prime95 was 24.14. Two instances were run in each case, with the affinity being manually set to make sure each instance ran on its own core.
For CPU Burn-in the error checking mode was used, which is not the most power hungry. For Prime95 the torture test was used using the in-place large FFT setting, which is reported as the most power hungry.