3DMark Firestrike & TimeSpy
We all know these bits of software. Firestrike has become a word intertwined with 'overclocking' and (for most) the word phrase 'liquid nitrogen.' Well known to overclockers worldwide, the Extreme iterations of these tests are well known to push a system to its breaking point. Today, we ran the regular versions of Firestrike and Timespy, with the former being a 1080p stress test that should offload as much work onto the CPU/RAM as possible.
We saw virtually no difference in Firestrike regular, with the XMP default profile scoring a 13,573, and the manually tweaked memory offering a 37 point increase. Again, here, more speed would be measurably better. With Timespy, again, it was an absolute margin of error. More speed, here, would be useful, but the modules just were not stable beyond 3200Mhz in some cases.
These results, realistically, shouldn't be much of a surprise to anybody. Firestrike tests are, by their nature, GPU bound almost entirely, and the CPU portion of each benchmark is also just that, CPU bound deliberately. I have yet to see a scenario where a lower amount of system memory, slower memory, or tighter memory, will make a significant difference to 3DMark (within reason).