- Joined
- Oct 12, 2001
- Location
- Los Angeles
Guys, I know I seem to take forever to do this, and I do, but it seems I need things to align just so in order to even get one test run in! I have to thank Martin for the awesome graphs! He made 2, one which plots each of the 5 mounts and another which averages all mounts into one final number.
Of particular notation is the AquaComputer Double Impact block (Thanks to RickCa1n) which just astounded me with it's sheer performance for a FLAT block (now bow or step). As you can see, it actually beats out a bowed Apogee GT on certain mounts, note that mount variation plays a BIG role!. Keep in mind this is the MOST restrictive block I have EVER come across! For sure, this would require a loop of it's own! I have not the tools nor the will to do pressure drop testing, I'll leave that to Martin
Also of note is I left out of the graphing blocks that could not compete AND that had less than 5 mounts. I also left out STEPPED blocks of the Apogee, Storm and Fuzion. Why? Well in reality the results are not really relevant since you can't buy these blocks stepped... plus, I didn't do 5 mounts of each either - what I can say is that on a flat block, a step should yeild a 2-3C improvement... on a bowed block the difference is a bit more fuzzy but in my opinion a step is preferable to a bow. I'll not go into details of why here.
I have more to test... I have the EK on the bench which does not fit with the current bracket... dremel time! I have the stepped G5 to pick up, and Cathar said he was sending a stepped G7 my way as well. I would love to test OCZ's new stepped block, the thermalright block, DD's new block and whatever else is out there.
So, now onto the cumulative results... I am going to copy and paste my testing methods below as they have not (and will not be) changed.
Testing procedures. First, the was all done with an Asus P5W-DH Deluxe and a E6600 Intel Core 2 Duo. The CPU was run at STOCK speed (2.4Ghz) but with 1.55v set in bios. Load temps were generated with TAT and temps were attained with TAT via the logging function. TAT takes a reading of each core every 2 seconds. Each cores readings are averaged, then the 2 averages are averaged to get a single average temp. FYI TAT generates MUCH more heat than any other load program period. It has been said that TAT is a 'power virus' and will actually get the chip to output about 20% more than Intel's TDC rating. Pump used was an Iwaki RD-30 at 18v, rad is a Thermochill HE (not PA!) 120.3 with 3 102 CFM sanyo denki fans in pull on a shroud. Fans for testing were always at 12v. A Swiftech micro res was used as well. Ambient air temps were taken and recorded every 10 minutes by the digital sensor in the test room. Water temps were taken by an identical sensor in the water. Only pure distilled water was used, no additives were used whatsoever. Thermal paste used for all tests was Arctic Ceramique and I use the 'dollop' method. D-tek mount hardware was used with ~50lbs of total mount pressure - plastic stops were used so every mount had the same amount of pressure. I tried to test when ambient temps were close, but this is not always possible without an environmental chamber which of course I do not have.
Here is a pic of the test setup:
Graph plotting each of the 5 mounts per block:
Average performance of the 5 mounts:
Of particular notation is the AquaComputer Double Impact block (Thanks to RickCa1n) which just astounded me with it's sheer performance for a FLAT block (now bow or step). As you can see, it actually beats out a bowed Apogee GT on certain mounts, note that mount variation plays a BIG role!. Keep in mind this is the MOST restrictive block I have EVER come across! For sure, this would require a loop of it's own! I have not the tools nor the will to do pressure drop testing, I'll leave that to Martin
Also of note is I left out of the graphing blocks that could not compete AND that had less than 5 mounts. I also left out STEPPED blocks of the Apogee, Storm and Fuzion. Why? Well in reality the results are not really relevant since you can't buy these blocks stepped... plus, I didn't do 5 mounts of each either - what I can say is that on a flat block, a step should yeild a 2-3C improvement... on a bowed block the difference is a bit more fuzzy but in my opinion a step is preferable to a bow. I'll not go into details of why here.
I have more to test... I have the EK on the bench which does not fit with the current bracket... dremel time! I have the stepped G5 to pick up, and Cathar said he was sending a stepped G7 my way as well. I would love to test OCZ's new stepped block, the thermalright block, DD's new block and whatever else is out there.
So, now onto the cumulative results... I am going to copy and paste my testing methods below as they have not (and will not be) changed.
Testing procedures. First, the was all done with an Asus P5W-DH Deluxe and a E6600 Intel Core 2 Duo. The CPU was run at STOCK speed (2.4Ghz) but with 1.55v set in bios. Load temps were generated with TAT and temps were attained with TAT via the logging function. TAT takes a reading of each core every 2 seconds. Each cores readings are averaged, then the 2 averages are averaged to get a single average temp. FYI TAT generates MUCH more heat than any other load program period. It has been said that TAT is a 'power virus' and will actually get the chip to output about 20% more than Intel's TDC rating. Pump used was an Iwaki RD-30 at 18v, rad is a Thermochill HE (not PA!) 120.3 with 3 102 CFM sanyo denki fans in pull on a shroud. Fans for testing were always at 12v. A Swiftech micro res was used as well. Ambient air temps were taken and recorded every 10 minutes by the digital sensor in the test room. Water temps were taken by an identical sensor in the water. Only pure distilled water was used, no additives were used whatsoever. Thermal paste used for all tests was Arctic Ceramique and I use the 'dollop' method. D-tek mount hardware was used with ~50lbs of total mount pressure - plastic stops were used so every mount had the same amount of pressure. I tried to test when ambient temps were close, but this is not always possible without an environmental chamber which of course I do not have.
Here is a pic of the test setup:
Graph plotting each of the 5 mounts per block:
Average performance of the 5 mounts:
