Testing Thermal Performance
Getting to the meat of cooling performance, I’m off to testing thermal performance in my quad radiator, yate loon slow speed fan, and Swiftech MCP-35X pump system. It’s a bit extreme for radiator capacity, but the pump is just a single DDC pump with top which many high-end systems use. For those with smaller radiators you can simply adjust the results up. I saw about a 3C water/air delta using a quad radiator, so double radiators would see 3C more, etc..
To avoid reiterating test methods, I’m using the same exact thermal test method as outlined in my CPU-370 review here except I used Swiftech’s backplate and mounting system because it does fix mounting pressure on all four corners. Even though it was a bit of a pain removing my old back plate system, it was well worth the change to experience the easy mount of this block. It was a pleasure using in comparison to my previous two block runs and here are the results. I did have slightly higher ambients in this series of mounts, but per my earlier 2600K scoping I didn’t find any notable water core differences even when I did vary ambients. I did set my thermostat to 71F, yet the outside temperature was higher resulting in ambients that were approximately 1-2C higher than previous blocks. In the end, I only compared core/water temperature to account for ambient differences.
First off I tried both orientations (Bolts parallel to locking lever which is the recommended and perpendicular), neither of which proved to be measurably different in my limited testing. My best couple of mounts were actually perpendicular, but I suspect that was purely incidental. Per the orientation of the i7 2600K, the core actually lays perpendicular to the socket lever arm, so you would think running the block such that it bows horizontally to match would net better temperatures as recommended in the installation guide. I just couldn’t measure the difference myself with errors present in mounting. I would suggest the manufacturers recommendation, but it might be worthwhile experimenting a bit. Both orientations on Revision 1 seemed to work about the same for me.
Here are the individual mounts and logged runs:
And now for the summary chart which compares the 5 mounts:
My average air in temperature was 24.44C with a tight standard deviation between mounts. Like a broken record here, but the mount itself is again the most elusive variable…giving me a standard deviation of .75C which is not as good as I had hoped. I suspect there could be some improvements to be had by increasing mounting pressure with washers under the springs, but I haven’t tested that yet. I tested the block stock, with the stock back plate, with the stock mounting system..on a stock unlapped IHS as most users would do using a common and easy to use low risk MX-2 thermal compound.
Of coarse, there is always the comparison of other i7 2600K results which is what most readers want. Here I am using the delta between water temperature and core temperature purposely showing the mounts for error context:
Considering my EK Supreme V1 was king of the Q6600 round, any block that excels beyond that is a winner in my book. The XT (rev 1) has been around now for over a year and a very good performer. If you take into consideration the mounting variable and standard deviations present due to testing, mounting, and sample variance errors, it’s a top block for sure. My Koolance CPU-370 sample tested at a slightly lower mean, but the differences here are very small.
Regarding the flow rate sensitivity, I’m skipping it. I’ve been pouring over data of various sources and finding little value in testing at multiple pumping powers and flow rates. Generally, most blocks of today’s micro pin/fin type designs can handle flow rates as low as .5 GPM at the cost of a few degrees. It’s extremely rare for any one block to test differently using different pumping powers of common use. If one block performs better using a D5 at Setting 3, 99% of the time it will also perform better using two pumps in series. Instead, I’m choosing to give you one common pumping power (MCP35X @100%) which is a strong pump with top, but nothing so different from a D5 basic on up to dual pumps that would net a different result. So with that, I’m proceeding with the 5 mount logged and fixed pumping power method. Without a doubt a CPU test needs at least 3 mounts and 5 just gives you a bit more confidence in the result.
So there you have it, very very good thermal performance that is within overlapping standard deviations of error of the top block tested.