Originally Posted by lordzed83
Its getting better and better
if driver wont help deffo try volts. What memory speed You are on ?? In my case i can boot up with Soc 0.95 but found that anything below 1.125 under heavy load makes me loose performance or crash. Same with VDDP in my case helped with making oc stable where nothing else did.
Btw I'm using LLC2 auto gave me crash here and there not to mention i had to bump cvore so much i would idle at 1.46
1.25 vcore with LLC3 @3.7GHz (currently using the stock cooler and although 3.8 @1.3v seems to also be stable with LLC3 the temps get too high)
I'm using 2666 14-14-14-34 on 1002 BIOS (so 1T) 1.35v dram and vboot dram. 2933 also works although with much looser timings (tried 18-16-16-38) so I'm sticking to 2666 for now waiting for may AGESA update.
With the stock driver and 60000 nvme idle timeout (also tried 1569325055) still got freezing but there's definitely an improvement over the samsung driver. Freezes happen much less frequently and also last a lot shorter (with the samsung driver the freezes sometimes lasted over 30 seconds). I'm gonna have to call it a day for now gotta wake up early tomorrow but hopefully this helps someone and maybe someone more knowledgeable is able to pinpoint the exact cause for this.
EDIT: So one last update. After checking the NVMe SMART information with CrystalDiskInfo I noticed that after all this testing and freezes the Controller Busy Time (0A) and Number of Error Information Log Entries (0F) values had risen (from 22 to 26 and 88 to 90 Decimal). Also I got a bunch of event viewer warnings at one point during a long freeze with the following description and with varying processes: The application \Device\HarddiskVolume4\Windows\System32\svchost.exe with process id 9724 stopped the removal or ejection for the device PCI\VEN_144D&DEV_A804&SUBSYS_A801144D&REV_00\4&2d380d55&0&0009. The ID matches with my Samsung 960 EVO 500gb.Edited by ggdfdgd3 - 4/24/17 at 4:13pm