Need Help With GSkill XMP

arossetti

Honorable
Feb 22, 2013
401
0
10,960
Hey All,

I'm hoping someone more knowledgeable in overclocking can answer this for me...
I'm getting insanely high CPU core temps if I have an XMP profile set, and I'm not sure why. For the last few weeks I've been trouble shooting and and tweaking and most definitely have identified that the problem on occurs with XMP enabled.

My Specs:
i7 6700 K @ 4.6 Ghz , 1.27v
Asus Maximus Impact VIII
16 GB GSkill Trident Z 3200 Mhz
EVGA GTX 980 Ti OC
Samsung 830 Pro / Samsung 840 Evo
EVGA P2 650 W PSU

Enclosure:
Fractal Design Nano S

Watercooling:
EKWB Monoblock - CPU
EKWB Full Block - GPU
EKWB D5 Pump w/ 40 mm Resevior
Hardwarelabs 240mm Black Ice GTS Radiator
EKWB 120 SE Radiator
AlphaCool compression fittings 3/8 x 5/8
Soft PVC Tubing

Initially I used the automatic OC utility in the ASUS UEFI Bios which gave me a stable 4.6 OC but at 1.45v - not good. I then used an OC guide and was able to drop it down to a stable 1.42v @ 4.6 Ghz - still not good but I had read that some chips require a lot more juice, additionally I enabled the XMP profile for the memory - after all I want it run at its rated speed.

I then stressed tested it with the ROG benchmark. Core temps were spiking into the 90C and sometimes hitting 100C - definitely not good. I tore down my loop, reset the block - made sure I had the proper amount of thermal paste, rebuilt it and tried again. Same results. Next I recalculated my TDP and though the 240 rad should have kept temps within reason, I added the 120 mm to the loop to help - rebuilt and tried again. Same result.

Finally, I used another automated tool in the BIOS and tried the TPU II overclock for water cooling. This is where I had my epiphany. It gave me a stable 4.6Ghz OC at only 1.27v but also shut off the XMP profile so my memory was only running at 2133 Mhz. I reran the stress test and my temps didn't get higher than 75c at full load -very acceptable.

I then went back into the BIOS and re enabled the XMP but then by OC was recalculated to a 4.2Ghz. Again rerunning the stress test, my temps didn't get higher than 75c.

I'm guessing that I'm better off, performance wise, with the higher CPU frequency rather than the higher memory frequency. Can anyone explain exactly why this is happening? I've never - or at least don't remember ever running into this problem before with an XMP profile, and I've been building my own systems for quite some time.

Thanks in advance!

Cheers.
 
Solution
Could try enable XMP, select profile 1 and then set the VCCIO and DRAM voltages up + 0.05 each, i.e. if DRAM voltage is 1.35 set to 1.4 (both are perfectly safe)

arossetti

Honorable
Feb 22, 2013
401
0
10,960
I've definitely concluded that there is something wrong with the XMP profile. I've been able to get my memory to 3200 Mhz by selecting DRAM target frequency. I don't have the option to get it to 3400 Mhz but at this point I'm just glad my temps aren't above 75 c when all cores are at 100% load (handbrake).
 

arossetti

Honorable
Feb 22, 2013
401
0
10,960


Its not about memory stability. Its the fact that for some reason the XMP profile destroys my temps and seems to throttle throughput.

One thing I noticed is that when enabling the XMP profile (there is only one) that it resets my base clock to 102 instead of the 100 I want to use. I'm sure that is part of the problem.