Aurora R7, odd errors using 64GB Kingston HyperX Fury 2933 RAM running @ XMP1

​I purchased 3 matched sets of 32GB (2 x 16). They are all Kingston HyperX Fury 2933. I installed 2 sets at a time in the R7. I’m running the most current bios (v. 1.0.18), and only tested using the first XMP profile (2936 DDR4 17 19 19 39). CPUZ is showing 64GB, dual channel DDR4, and frequency is running at 1463.5 +/ 2 mhz (so effectively just under 2933).​

​The odd thing is that the RAM works, and passes memtest with 0 errors, in some configurations, but not in others. For example, there are 8 different ways that I can install Set 1 and Set 2 on the motherboard (without breaking up matched sets). They work with 0 errors in 5 out of 8 possible configurations. The remaining 3 configurations trigger errors on a number of different memtest tests. I tried testing it hot and cold on the error configurations, and it failed regardless. Exact same issue using Set 1 and Set 3 together (works no errors in 5 out of 8 configurations, errors in 3 configurations). Using Set 2 and Set 3 together worked in the first configuration, and triggered no boot and BSOD scenarios in the next 2 configurations, so I stopped testing. ​

​As an aside, Set 1 and Set 2 were both manufactured during the same week in 2018, and the serial number sequence for all 4 sticks are within 15 digits of each other. Set 3 was manufactured 6 months earlier and the warranty sticker looks different. Set 3 is a different color. ​

​I was planning to RMA Set 3, and keep Set 1 and 2 to use as 64GB, XMP1, in one of the 5 working configurations. I realize I could purchase a matched set of 64GB, but that option at $418 was $150 more expensive.​

​I’m curious if anyone has thoughts on why this works fine in certain configurations, but fails in other others? Because, for example, while identical on paper, they weren’t guaranteed matched pairs by Kingston presumably due to the issue noted below; they’re marginal; or the dell MB is just finicky like that? I can live with any of the above. ​

​But I would be concerned if the RAM may be defective and fail in the future due to the defect. Or if there may be an issue with the MB. Thoughts? ​

​Below were the working/nonworking configurations. I identified the RAM slots from left to right as 1 2 3 4, consistent with the service manual. On the MB it’s actually labeled as 2 1 4 3.​

​Using Set 1 and Set 2 for 64GB ​

​Bank 1 2 3 4 = A C B D = FAIL tests 6-13, over 10K errors​

​Bank 1 2 3 4 = B C A D = PASS​

​Bank 1 2 3 4 = B D A C = PASS​

​Bank 1 2 3 4 = A D B C = PASS​

​Bank 1 2 3 4 = C B D A = PASS​

​Bank 1 2 3 4 = C A D B = FAIL tests 6, 8, 9, 13​

​Bank 1 2 3 4 = D A C B = FAIL tests 6-13​

​Bank 1 2 3 4 = D B C A = PASS​

​Using Set 1 and Set 3 for 64GB​

​Bank 1 2 3 4 = E A F B = PASS​

​Bank 1 2 3 4 = E B F A = PASS ​

​Bank 1 2 3 4 = A E B F = PASS​

​Bank 1 2 3 4 = F A E B = PASS​

​Bank 1 2 3 4 = F B E A = PASS​

​Bank 1 2 3 4 = A F B E = FAIL all but first test​

​Bank 1 2 3 4 = B F A E = FAIL, cancelled at errors​

​Bank 1 2 3 4 = B E A F= FAIL, cancelled at errors​

​Based on the above, for anyone who purchased Kingston Hyperx Fury RAM for an R7 and experiences no boots, freezing in Windows, or random BSODs, etc., – you may have better luck if you rotate the order of your RAM modules. ​

​Oh yeah, I’m running Aurora R7 z370 mb, cpu 8700, boot drive 1TB Samsung m2 NVMe, psu evga G3 gold 850, gpu nvidia gtx 1080. Not sure if it makes a difference, but, since it's working fine in windows in the good configurations and BSODs or freezes in bad configurations, I note that I did a clean install of windows 10 when I removed the Optane drive, and did NOT install Intel RST, or Dell's auto driver update software. ​

​ ​