X99 experienced a loss in a memory channel.
X99 experienced a loss in a memory channel.
Asrock X99 Extreme 3 operated well with a 5820k and 4x4GB RAM (16GB total). Last night I connected my E5 2683 v3 ES, which functioned properly before. All appeared normal until I noticed only 12GB available after booting into Windows. Steps I attempted: reinserted RAM in same slots, moved RAM to different slots, reset CPU, cleaned CPU pads with isopropyl alcohol, checked for bent pins in socket. Everything seemed intact. I adjusted the waterblock mounting tension and cleaned CPU pads again, but nothing resolved the issue. Currently using two modules and getting 8GB as expected. I haven’t tried running only one pair at a time. For now, I might swap the existing pair with the removed one to test if it resolves the problem.
16GB was confirmed detected and the system has been operating continuously on compute for several months. I'm planning to add my Xeon soon as I enter a compute challenge and would benefit from additional processing power. The 5820k was initially considered for replacement but was postponed; instead, I replaced it with another unit that also reports 8GB. The modules appear to be functioning correctly. I'll verify the remaining RAM slots on the motherboard next and anticipate one of these failing soon.
This confirms the previous findings. I've checked that all four modules together use 12GB. Two modules fit in slots 3 and 4 (8GB), so swapping the other two there also works. Moving the remaining two modules to slots 3 and 4 again yields the same result. I'm going to try placing all four modules once more to verify.
After reviewing the setup more closely, I suspect the RAM configuration could be influencing performance. The system runs Ripjaws 4 with a 3333 module, and I was using it at 2133 with the 5820k. Past experiences show instability when using RAM above 3000. It seems very sensitive to the motherboard and CPU pairing. After switching the CPU, it booted fine without any issues. Windows and the motherboard BIOS both indicated the RAM as 2666, even though the Xeon model doesn’t support RAM above 2133. The timing data came from an XMP profile rather than the standard JEDEC settings for 2133. This suggests the timings are actually slower than expected. I don’t see an XMP toggle in the BIOS, so a full reset might be necessary to rule out lingering settings. For the Xeon, using JEDEC 2133 would be more appropriate. I might consider swapping modules from another machine, but I’m worried it could damage the current setup. It could very well be a contributing factor. Otherwise, this hardware feels unusually erratic.