You can use a diagram of Bland Altman: in this diagram you report the series of the differences versus the means between each values given by the 2 instruments. The mean of the differences tell you if on average the 2 instruments differ (bias or systematic difference). The limits of agreement are given by this mean +/- one or two sigma. First you have to check is the series of difference between the two instruments are normally distributed.
In principle, you can apply all kinds of comparative metrics between the two precipitation products. For example a Q-Q (quantile-quantile) plot will allow you to compare rainfall distribution from each product and identify their differences with respect to rainfall severity as well. Similarly you can apply other metrics/methods (as G.C. suggested above) to help you understand the differences between products.
The problem for not having a ground reference, which is commonly used as the "ground truth", is that you will not be able to identify which of the two products comes closer to the "truth". So if you need to pick one of the two products to use further (e.g. for hydrologic applications etc)...it will be tricky :)
In my opinion, having ground truthing dataset is a must for the two products you mentioned. Saying no ground data for a particular catchment (hourly, daily, or even monthly) is not fully justified. Instead, you could find limited rainfall records to be optimized for your case. Alternatively, you could find a similar climatic zone (e.g. arid) which have enough record and use it for your purpose.
However, plotting TRMM and GSMAP or finding correlations will give you a great idea on the behavior of the two against each other but not representing the conditions on the ground.