BonzaiDuck
Lifer
Sometimes my brain is a little slow to catch on, sometimes I can be quick.
I've been running PRIME95 torture-test for specific over-clock settings known to provide error-free results under MEMTEST86. Most of these settings have been benchmarked once or twice with PCMark04, which only loads the processor for a short time, and not to the level common to PRIME95.
My failures in PRIME95 occurred as the 12V rail readings in my motherboard monitoring software (Asus Probe) would dip to their lowest. Even a momentary drop just outside a -5% deviation from the 12V requirement would be enough to register a rounding error in PRIME95, and especially errors in which PRIME95 would inform me that it was shutting down for 5 minutes.
Ordinarily, this power-supply -- an ALLIED AL-B500E -- is within the +/-2% range of 12V. But under load conditions, it will drop to as low as 11.612V, and generally stay within the range of 11.612 and 11.795.
My primary focus in building this project has been "effective yet quiet air-cooling" and case modification toward that purpose. The system essentials are:
3.0C P4 @ 3.6 (current, stable error-free configuration with VDIMM @ 2.75v, VCORE @ "auto" and nominal 1.536V, and minimum VDDQ/AGP of 1.5V)
ASUS P4P800
Hitachi 7K250 Dual RAID0 (total 360GB)
ASUS V9980 nVidia FX5950 Ultra
Creative SB Audigy 2 ZS
Sony DRU700A
CPU cooling by ThermalRight XP120 and YS-Tech 120x38mm fan
My brother is a former telephone company employee who spend his last few years installing T1 lines for businesses. We had a discussion about voltage drops, and his immediate answer was that "decreases in voltage would not cause increases in heat dissipation; increases would." But Power is Power, Watts is Watts, and I still remember some things from Physics 4B -- 40 years ago when I was young and fulla-beans. You know my argument: if the device required so much power and voltage dropped, amperage would increase to maintain that power.
Enough research about power-supplies and rail-voltages led me to consider a PSU replacement. It was either the PCPwr&Cooling Turbo-Cool 510 Deluxe at (its near-constant price for two-years running or) $220, or the newer entry by OCZ in the PowerStream Series -- and specifically a Power-Stream 520.
Here is an excerpt from a review of the PowerStream from "The Crucible":
So I assume that my over-clocking efforts AND my cooling-efforts will be improved slightly by replacing the ALLIED with the beefier OCZ Powerstream 520.
However, there is one other issue -- Power Supply Efficiency and how those numbers are provided by the manufacturers. PC Power & Cooling rates their PSUs to be 70% efficient at maximum load across the range of their Turbo-Cool models. ApexTechUSA rates the ALLIED at "a minimum of 65%, but typically 75% at full load." OCZ glibly explains in their product data sheet that the two lower-wattage models -- the 420 and 520 -- provide a minimum of 63% efficiency while the 600W model provides 70%.
This would mean that if the Powerstream had lower efficiency than either the ALLIED or the Turbo-Cool, it would generate slightly more heat inside the computer case, even if that heat were quickly exhausted from the PSU chassis.
But AnandTech's own review of the PowerStream notes that the 520 model is "conservatively rated" and should probably be rated at a higher wattage. If OCZ chose to rate the model "conservatively", what does that say about their efficiency specification for the 520 model?
I would be interested in the remarks of anyone else who has hard data to support the notion that a good strong 12V rail makes the CPU heat up less than one supplying 11.7V or 11.8V. Obviously, respondents would need to have measured temperatures under two different "power-supply regimes".
I've been running PRIME95 torture-test for specific over-clock settings known to provide error-free results under MEMTEST86. Most of these settings have been benchmarked once or twice with PCMark04, which only loads the processor for a short time, and not to the level common to PRIME95.
My failures in PRIME95 occurred as the 12V rail readings in my motherboard monitoring software (Asus Probe) would dip to their lowest. Even a momentary drop just outside a -5% deviation from the 12V requirement would be enough to register a rounding error in PRIME95, and especially errors in which PRIME95 would inform me that it was shutting down for 5 minutes.
Ordinarily, this power-supply -- an ALLIED AL-B500E -- is within the +/-2% range of 12V. But under load conditions, it will drop to as low as 11.612V, and generally stay within the range of 11.612 and 11.795.
My primary focus in building this project has been "effective yet quiet air-cooling" and case modification toward that purpose. The system essentials are:
3.0C P4 @ 3.6 (current, stable error-free configuration with VDIMM @ 2.75v, VCORE @ "auto" and nominal 1.536V, and minimum VDDQ/AGP of 1.5V)
ASUS P4P800
Hitachi 7K250 Dual RAID0 (total 360GB)
ASUS V9980 nVidia FX5950 Ultra
Creative SB Audigy 2 ZS
Sony DRU700A
CPU cooling by ThermalRight XP120 and YS-Tech 120x38mm fan
My brother is a former telephone company employee who spend his last few years installing T1 lines for businesses. We had a discussion about voltage drops, and his immediate answer was that "decreases in voltage would not cause increases in heat dissipation; increases would." But Power is Power, Watts is Watts, and I still remember some things from Physics 4B -- 40 years ago when I was young and fulla-beans. You know my argument: if the device required so much power and voltage dropped, amperage would increase to maintain that power.
Enough research about power-supplies and rail-voltages led me to consider a PSU replacement. It was either the PCPwr&Cooling Turbo-Cool 510 Deluxe at (its near-constant price for two-years running or) $220, or the newer entry by OCZ in the PowerStream Series -- and specifically a Power-Stream 520.
Here is an excerpt from a review of the PowerStream from "The Crucible":
So I assume that my over-clocking efforts AND my cooling-efforts will be improved slightly by replacing the ALLIED with the beefier OCZ Powerstream 520.
However, there is one other issue -- Power Supply Efficiency and how those numbers are provided by the manufacturers. PC Power & Cooling rates their PSUs to be 70% efficient at maximum load across the range of their Turbo-Cool models. ApexTechUSA rates the ALLIED at "a minimum of 65%, but typically 75% at full load." OCZ glibly explains in their product data sheet that the two lower-wattage models -- the 420 and 520 -- provide a minimum of 63% efficiency while the 600W model provides 70%.
This would mean that if the Powerstream had lower efficiency than either the ALLIED or the Turbo-Cool, it would generate slightly more heat inside the computer case, even if that heat were quickly exhausted from the PSU chassis.
But AnandTech's own review of the PowerStream notes that the 520 model is "conservatively rated" and should probably be rated at a higher wattage. If OCZ chose to rate the model "conservatively", what does that say about their efficiency specification for the 520 model?
I would be interested in the remarks of anyone else who has hard data to support the notion that a good strong 12V rail makes the CPU heat up less than one supplying 11.7V or 11.8V. Obviously, respondents would need to have measured temperatures under two different "power-supply regimes".
