X58 24GB RAM Freezing

xiwong

Reputable
Apr 24, 2014
4
0
4,510
I have a EVGA E758 with a W3520 and 24GB (6x4GB) of Corsair Dominator CL9 RAM. I have ran this fully stable at 3.8GHz for 2 years with:

Bclk: 181
Ram: 9-9-9-24 2T, 1450MHz, 1.5V
VCore: 1.225V
VTT: +75mV
PLL: 1.6V
ICH: 1.15V
Enabled: Speedstep, Virtualization, HT
Everything else stock

Starting last week, I kept getting freezes when starting VMs. There was no error message or blue screen. I raised voltages and even returned to completely stock speeds (with RAM at 1066 and looser timings) but the only way to stop the freezing was to remove 2 sticks of RAM. All sticks have been tested with Memtest on another X58 computer and passed 8 passes. I'm currently running 16GB and have restored the OC settings above and it is fully stable. Are there any suggestions for what voltages to change or is my IMC going out as it can't even handle 24GB of underclocked RAM at stock bclock?
 

Echocookie

Reputable
Mar 12, 2014
225
0
4,760
According to the X58 SLI Handbook/Manual, the motherboard only officially supports up to 12gb of RAM, 1333MHz max. There is probably something I am missing, but are you sure all of that 16gb of ram is being recognized?
 

xiwong

Reputable
Apr 24, 2014
4
0
4,510


I believe it was because there were no 4GB DIMMS back then, but quite a few people have ran 24GB on their E758s before. I had all 24GB recognized in Windows 7 and actually ran out a few times when I forgot to shut down one of my 8GB VMs. My daily workload is 2 6GB VMs and data analysis (6-8GBGB RAM), leaving some left for the host.
 

xiwong

Reputable
Apr 24, 2014
4
0
4,510


Thanks, do you have any baseline on where the voltage would be excessive (i.e. when to call it quits) at stock speeds? From what I remember QPI PLL Vcore and VTT deal most with the IMC, and I sort of stopped testing when I was at VTT +250mV (E-LEET actually read this as 1.45V on the VTT instead of 1.10+250mV=1.35V) and QPI PLL at 1.3V at stock bclock. Then and in the incremental steps in between it was not stable.
 

xiwong

Reputable
Apr 24, 2014
4
0
4,510


tRFC is on Auto which set it to 174, and I have tested RAM up to 1.65V on stock speeds. Testing is very quick as if I run OCCT CPU after boot, it will fail in about 20-30 seconds with the generic error on Core.