Which approach should be used to compare two different measurement techniques of same samples? I have individually measured failure forces of 8 materials and those recorded with A method and B method in same time: 8 results in each method, A=8 and B=8. The range of data of both measurement techniques vary from 200 to 600 N.
The output of A method is the gold standard method (my reference). As a result, I want to check the accuracy of B using A as a reference. 
I already performed t-test to check the statistical difference and didn't find any difference, which was good. However, since the sample size is small (n=8), the p-value may not mean so much since the p-value is close to .05. 
> diffAB$p.value
[1] 0.068

For instance, if there is one more data, the p-value may be lower than .05. Therefore, I want to check the accuracy of my B method with respect to gold standard (A) in a different way. However, I couldn't find any example case to fit into my question. I did check RMSE of these two outputs, but the value was pretty high:
(rmse=85)

what approach should be used for this case?
 A: It's important to look at the actual data and not just rely on statistical summaries. In addition to simply plotting corresponding B and A results against each other, you should consider a Bland-Altman plot (individual B-A differences plotted against their corresponding means), which was specifically designed for such comparisons even if one of the methods is a "gold standard." With your data (all values positive, and I suspect with inherent errors proportional to the measured values), such plots based on prior log transformations might be more informative.
With only 8 observations, however, you are in a difficult situation. It would take a very large difference between the two methods to be reasonably detectable by any statistical test. If a root-mean-square error of 85 is considered large in this field and A is considered the "gold standard," you should be wary of using method B regardless of the nominal significance of a statistical test value.
A: A different point, but related to the sample size: t-test is not suitable for your purpose.
T-test, by default, tests the null hypothesis that the means in the two groups are identical. Your test failed to reject $H_0$, meaning the evidence is not sufficient to believe that the two methods differ. But, this does not imply (and certainly does not "prove") that they are equivalent. You cannot "accept $H_0$", you can only "fail to reject" it.
If you had more data, it is well possible that the p-value would drop below 0.05, allowing you to reject $H_0$. On the other hand, if you performed fewer measurements, your p-value would likely move further away (up!) from the significance level. You could make your results more "convincing" by simply making fewer measurements!
To drive this idea to extreme: If you performed no measurements at all, you'd have no evidence whatsoever that the two methods differ. But only an insane person would take this as an indication that they are equivalent.
