I bought 4 2x16GB Corsair Vengeance 3200 LPX kits thinking this would be the same as buying 1 128GB kit. I seemed to have lucked out in that my motherboard (ASRock x399 Fatal1ty) can use all 128GB at close to the rated clock speeds (all this despite the memory not being listed on the manufacturer's compatibility list).
When I was running benchmarks with Passmark, I discovered my RAM read/write speeds and latency got nontrivially worse as I added more RAM. With dual channel 32GB, I was on the high end of AMD machines for that RAM model (2400 passmark score; 20k MB/s uncached read); for quad channel 64GB and 128GB, the low end (sub 2100; 14k MB/s uncached read). I was a little surprised by this.
I'm trying to decide if I should return it and get a 128GB kit (e.g. Vengeance LPX 3000).
- Should I expect benchmarks to fall simply because I added more RAM (i.e. would 32GB from a 128GB kit be expected to benchmark better than all 128GB)?
- Could it be related to the fact that I'm running 4x32GB kits as a 128GB one? (i.e. are larger kits fundamentally engineered differently? I had assumed the differences were primarily packaging and guarantees they were from the same manufacturing batch)
- Are kits and motherboard manufacturer compatibility charts primarily just guarantees the RAM works at all, or are there likely nontrivial performance gains to be had with a "correct"/supported RAM setup?