Originally Posted by CL3P20
@ Blameless.. Your doing some crazy computing then sir! My hats off to you. Folding while gaming is about the only thing that 'stresses' my daily user.
*As for your RTL issue.. Im curious if your mobo is keeping tRWSR and tRWDRDD too tight for your RAM's liking? Those timings are more IMC bound than RAM.. seems 3-4 is tight for most boards with RAM speeds between 1800-2200mhz .. 5-6 can allow for some higher speeds without increasing voltage for me. You may want to keep VTT a notch higher than SA/IMCv as well..
This system if going to be a jack of all trades, but near absolute stability is important as some of the work I do on it is helping my wife finish her PhD. Whenever she has sequencing problem that won't fit into the 24GiB of memory that her workstation has (the X58/i7 970 build in my sig), it's going to run on this (we haven't run into anything in ARB that can't fit in ~30GiB of memory, yet). Saves us time from having to send it off to be run on a supercomputer somewhere (the supercomputers can finish the tasks far faster, but by the time she's through the queue this system can already have done it). I also Fold or run projects when I'm away on vacation.
Anyway, my board doesn't have a tRWDRDD setting, but all related tertiary timings are quite tight (0-3) and and tRWSR is 3. Loosening them produced less favorable results than simply loosening tRRD and tFAW to 7 and 40 (forcing all IOLs to 5-6, as an auto detect of 4 on some rows was crapping things up, and any manual offset sometimes produced settings of 7, which was just as bad for stability).
Regarding SA vs VTT, this chip with 8 DIMMs seems to prefer equal or greater SA voltage. If I bump VTT past a narrow range (depending on clock speeds) stability rapidly declines. I may be able to tweak a bit further, but I'm hesitant to start testing over after I'm nearly two full days into it (and has taken almost a week of 12 hours a day of tweaking to get to these settings). 1.15/1.15 is well within safe limits, so if it's stable at this point, I don't plan to screw with it too much. Maybe if I'm feeling really bored at some point I'll loosen tertiary timings and shoot for DDR3-2133.
I do think I've got everything stable now. I finished off 24 hours of Large FFTs without error, had to bump vcore a notch to 1.336v load (at 4.4GHz) to stablize 8 hours of Small FFTs, now I'm several hours into a custom Blend test using 28-30GB of memory. All seems well, and I expect it to pass. After that I'll finish off CPU/memory testing with a few hundred loops of LinX and some HCI Memtest.
Currently at 1866, 9-9-9-27-T2, 7-200-15-7-40-7-8 (tRRD-tRFC-tWRT-tRPT-tFAW-tWRD-tWL), 0-2-3-3-3-3-1-1-3-0 (tRRDR-tRRDD-tWWDR-tWWDD-tRWDR-tRWDD-tWRDR-tWRDD-tRWSR-tCCD).
This is with 1.385 vDIMM (idle is about 1.41v, but full load is very near set point), testing in 30C ambients, with FurMark running in the back ground to dump some extra heat in the case.