TIM Comparison Results

Let’s now compare the TIMs against each other, first by contact setting then in some other ways. In each of the contact setting groupings, first chart is performance over time, second chart is the average mount temperature at the end of the 10hr runs, third chart is the 10hr temperature from the best mount, and the fourth chart is a zoomed and simplified view of the amount of cure over the last 9.5hrs of the test.

Poor Contact

First comparison is with Poor Contact block + mounting.  This illustrates, roughly, the worst performance you can expect from a TIM (from a poorly designed mounting system, in use with some GPU blocks/sinks, VRM blocks/sinks, or from a seriously botched mount).

 

Moderate Contact

Second comparison is with Moderate Contact block + mounting.  This illustrates, roughly, the typical performance you can expect from a TIM with a well designed mounting system.

 

Great Contact

Second comparison is with Great Contact block + mounting.  This illustrates, roughly, the performance you can expect from a TIM with an extremely well designed mounting system with extremely high mounting pressure and a block with a compressible bow.

 

Overall Performance

This is a new chart, but it’s not new data.  This is a chart of the temperatures used as the input values for the “Relative Performance Score” charts we’ve had on the conclusion pages.  Says the same thing as that chart, but we think it might be easier to relate to–it’s in degrees, not unitless numbers–and represents how good a TIM is overall, without knowledge of how good contact will be.

“Overall Performance Impression” takes into account all three contact settings, “Typical CPU Mount” takes into account Moderate Contact, Great Contact, and Best Mount temperatures.  ”Overall Performance Impression” is an average designed to describe the overall performance of a TIM where “Typical CPU Mount” is an average designed to describe expected performance with CPU blocks and heatsinks, where contact quality tends to be good.

Overall, TX-4 is very good, but is mostly bolstered by its great Poor Contact performance; Typical CPU Mount temperature is little more than .25c ahead of MX-2 and nearly 1.25c behind PK-1. TX-2 trails MX-2 in both metrics, being roughly .2c behind MX-2 overall and more than two-thirds of a degree behind with Typical CPU Mount.

Contact vs. Temperature

  

Contact Dependence takes the best overall mount and the three averages of the three contact settings and finds the average deviation of the four results.  Lower means there’s less difference in performance as contact varies.  TX-4 is one of the least contact dependent pastes we’ve tested to date, being just slightly more dependent than Shin-Etsu X23-7762. It achieves this with a great Poor Contact performance and steady improvement as contact improves.  TX-2 is more contact dependent than TX-4 but overall is still not very contact dependent–it’s the tiniest bit less dependent than PK-1, although the Contact vs. Temperature chart shows very different performance profiles.

Cure Factor looks at the “Curing Performance” charts and takes a weighted average of the absolute values of the lines from the Curing Performance charts–the sooner a TIM flattens to its final value, the lower the Cure Factor.  The less change in performance over time, the lower the Cure Factor.  It’s not a particularly precise metric, Indigo Xtreme should have a 0.00 Cure Factor, but it will show overall trends for which TIMs do have some sort of curing. Neither TX-2 nor TX-4 will have any observable cure to the naked eye. In fact, TX-2 takes the crown for least-curing paste to date, edging out TC Quantum. TX-4 has very little cure, as well.

The Best Mount charts take the best performance from any of the fifteen mounts and is indicative of the absolutely limit of performance for a TIM (i.e., when you do everything right); it’s another way of looking at how good a TIM is.  TX-2 and TX-4 don’t do particularly well in this metric, coming up roughly a quarter and a tenth of a degree behind MX-2, respectively.