friend'scatdied
[H]ard|Gawd
- Joined
- Feb 20, 2005
- Messages
- 1,283
I want to get my Q9550 to run with as low of a power draw as possible. I'm going to be squeezing the Q9550 and an HD 5850 (which I will also undervolt as aggressively as possible) to a Shuttle 300W 80 Plus PSU (2 12V rails, 14A each for 28A total).
The lowest it will run is 1.025v in the BIOS, which translates to a vDrop of 0.98-0.99v in-OS and vDroop of 0.96v under load. This is with stock clock speeds (333MHzx8.5, 2.83GHz). This setting is OCCT stable for 9 hours (haven't tried more). I am fairly certain these settings will draw less power than the Q9550S that I was originally intent on purchasing (whose claim to fame is a VID of 1.16). The hottest core gets to about 57'C, with the rest at 52-53'C (linpack).
However, it will also run at 400x7 (2.8GHz) with the same settings, also OCCT stable for 9 hours. From what I am told the higher FSB will give marginally better performance even with the similar clock speeds. The temperature on the hottest core is about 3'C higher than with stock, at 60'C (rest of the cores are 54'C).
(Bear in mind my setup is in a Shuttle case with only one 92mm exhaust fan, which also functions as part of the CPU heatsink/fan unit [Shuttle I.C.E. cooler, which has 3 heatpipes and a modest number of fins]. I understand that they will be higher than you would expect with such an aggressive undervolt if it were in a roomy case with good airflow.)
Given that I am prizing low power draw, will the 400x7 setting draw more overall power than the 333x8.5 setting given that all other variables are held constant? I understand that the 400MHz FSB will put more stress on the NB, which could mean greater power draw there. However NB temps were the same under linpack load, and the MCH voltage is still stock.
I'm quite tempted to go back to 333MHzx8.5 because it seems it would be "safer," and also because my system does not reboot properly when I change any FSB/memory settings (even FSB raised by 1MHz or with a reduction in the memory divider to underclock the RAM).
The lowest it will run is 1.025v in the BIOS, which translates to a vDrop of 0.98-0.99v in-OS and vDroop of 0.96v under load. This is with stock clock speeds (333MHzx8.5, 2.83GHz). This setting is OCCT stable for 9 hours (haven't tried more). I am fairly certain these settings will draw less power than the Q9550S that I was originally intent on purchasing (whose claim to fame is a VID of 1.16). The hottest core gets to about 57'C, with the rest at 52-53'C (linpack).
However, it will also run at 400x7 (2.8GHz) with the same settings, also OCCT stable for 9 hours. From what I am told the higher FSB will give marginally better performance even with the similar clock speeds. The temperature on the hottest core is about 3'C higher than with stock, at 60'C (rest of the cores are 54'C).
(Bear in mind my setup is in a Shuttle case with only one 92mm exhaust fan, which also functions as part of the CPU heatsink/fan unit [Shuttle I.C.E. cooler, which has 3 heatpipes and a modest number of fins]. I understand that they will be higher than you would expect with such an aggressive undervolt if it were in a roomy case with good airflow.)
Given that I am prizing low power draw, will the 400x7 setting draw more overall power than the 333x8.5 setting given that all other variables are held constant? I understand that the 400MHz FSB will put more stress on the NB, which could mean greater power draw there. However NB temps were the same under linpack load, and the MCH voltage is still stock.
I'm quite tempted to go back to 333MHzx8.5 because it seems it would be "safer," and also because my system does not reboot properly when I change any FSB/memory settings (even FSB raised by 1MHz or with a reduction in the memory divider to underclock the RAM).