https://bugzilla.kernel.org/show_bug.cgi?id=78221
--- Comment #5 from t3st3r@mail.ru --- Will try that since that bug is nasty enough. Can take some time.
As initial investigation when looking on commit log and matching encounters of bugs, it appears stability issues were fixed at result of commit 0a4ae727d6aa459247b027387edb6ff99f657792 (appears between 3.15-rc8 -> 3.15 release).
So all 3.15 RCs were not stable on R9 270. However, 3.15 release is okay due to these last-minute fixes. Yet 0a4ae727d6aa459247b027387edb6ff99f657792 seems to be composed of few commits, lets chew a bit more on it. Most likely it comes down to 91b0275c0ecd1870c5f8bfb73e2da2d6c29414b3.
I think I would try little experiment first: return CPDMA as it was in 3.15 last minute fix and see if stability returns to 3.16-rc1 with R9 270.