Long story short, I've read it mentioned somewhere that a bug exists with PCIe slot 2 (which is x16) where it only works at half the speed on cMP 4,1 (maybe 5,1 also). I've done fairly exhaustive searches and can't find anything specific about this and whether there is a solution to the problem.
I experienced this bug myself when I installed my Samsung 960 EVO 500GB M.2 NVMe into PCI slot 2 and only got half the speed for read/writes that I got when it was installed in slots 3 and 4 (read Q2 of this post).
I want to know more info on the problem and if there is a solution so I can get closer to the maximum speeds my Samsung 960 can deliver by using the x16 slot (3200/1900 MB/s). My understanding is that a x4 slot can deliver 4 x 500MB/s = 2000 MB/s. I'm currently getting between 1500 and 1700 in Blackmagic and other benchmark read/write tests, so this sounds right. So would x16 deliver higher read/write speeds that will realise the potential of my M.2 NVMe blade?
I haven't tried it, but can anyone share some advice on what would happen if I put my Sapphire Nitro Radeon R9 380X into slot 2 and I put the NVMe PCI card into slot 1? Would the slot 2 bug reduce/limit my video performance any (and to what degree?), and would I get faster throughout for my NVMe blade from slot 1? Just trying to wring out a little more of the potential of my NVMe before I concede I've done enough (although I still want to add a dual CPU tray (with either X5680 or X5690) eventually one day if I can find one cheap enough).
I experienced this bug myself when I installed my Samsung 960 EVO 500GB M.2 NVMe into PCI slot 2 and only got half the speed for read/writes that I got when it was installed in slots 3 and 4 (read Q2 of this post).
I want to know more info on the problem and if there is a solution so I can get closer to the maximum speeds my Samsung 960 can deliver by using the x16 slot (3200/1900 MB/s). My understanding is that a x4 slot can deliver 4 x 500MB/s = 2000 MB/s. I'm currently getting between 1500 and 1700 in Blackmagic and other benchmark read/write tests, so this sounds right. So would x16 deliver higher read/write speeds that will realise the potential of my M.2 NVMe blade?
I haven't tried it, but can anyone share some advice on what would happen if I put my Sapphire Nitro Radeon R9 380X into slot 2 and I put the NVMe PCI card into slot 1? Would the slot 2 bug reduce/limit my video performance any (and to what degree?), and would I get faster throughout for my NVMe blade from slot 1? Just trying to wring out a little more of the potential of my NVMe before I concede I've done enough (although I still want to add a dual CPU tray (with either X5680 or X5690) eventually one day if I can find one cheap enough).