CPU/Memory Overclocking Stable Separately, but Not Together

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530
First, here is my set-up.

Case: Phanteks Evolv mATX
Mobo: Asus ROG STRIX Z-370G (with current BIOS update)
CPU: Intel i7 8700K
GPU: Nvidia RTX 2080 Founders Edition
Memory: Corsair Vengeance RGB Pro 2x8GB 3000Mhz C15
CPU Cooler: Corsair H60 (120mm fan)
PSU: Corsair TX650M 80+ Gold Certified
Boot Drive: Samsung 970 EVO 250GB NVMe M.2

Fans:

Intake: 2x140mm Phanteks (came with the case)
Exhaust: 1x140mm Corsair ML140 Premium Magnetic Levitation

So, here is my issue. When I overclock the CPU at 4.8Ghz with 1.30V and stress test, things are stable. When I overclock the memory using the XMP profile, 15-17-17-35 at 1.35V, things are stable. However, when I apply both of these and run a blend test in Prime95 after the first few passes, like 20 minutes in, the temps get into the mid 80s up to the 90s.

What can I do to get my temps lower? Should I lower the RAM speed? If I lower the RAM speed from 3000Mhz to something lower, how do I know what timings and voltage should be used?

Could it be that my CPU overclock is limiting how fast the RAM can be overclocked?

Any help would be appreciated on this.
 

luckymatt42

Upstanding
May 23, 2018
446
1
360
If this were me, I'd first set the XMP/memory as you have done. Leave cpu totally stock and do stress testing. Once passed, leave xmp settings where they were, THEN start bumping up the cpu gradually, doing stress testing along the way (yes this takes quite some time, be patient).

CPU and mem are not totally independent as far as stability is concerned. You may indeed achieve one clock with mem, one clock with cpu, but they don't work together. Not uncommon at all.
 
Your CPU OC has pretty high voltage. Most seem to be hitting 4.8 on about 1.26 vcore. That's where your extra heat is coming from, possibly causing thermal throttling, and/or unstable OC.

You could try setting the timings and voltage manually for the DIMMS. Maybe the XMP profile is causing issues. Like you said maybe try with a lower speed too. 2933, or 2800, see if it makes any difference.

Did you do a CMOS rest after bios upgrade? If not try it. Sometimes remnants of old microcode can linger so a CMOS rest might help.
 

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530


Thanks for the reply.

I agree the voltage is somewhat higher than what I have seen from others. I originally had it at 1.25 and things were fine. However, after messing with the memory I set everything back to default and overclocked the CPU and for some reason that same voltage didn't seem to be enough, I would get hardware errors in Prime95. So I just kept bumping it up until they went away.

As for the RAM timings this is where I am lost. I have no idea what timings to use for say, 2800Mhz.
 

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530


Thanks for the reply. I'll give this a try, probably just start at 3.8Ghz with the default voltage and just go up from there.
 


Thats cool. We can sort that for you. Install/run CPUZ. Go to the mem tab and take a screenshot, and then the SPD tab. This will show you the Jedec/XMP timings and voltage for particular speed settings on your ram. Note them all down.

Then you will reboot, go into the bios and set XMP off, and enter the same timings and voltage you noted down. 5 timing settings and the voltage. See what happens. You won't kill anything by trying this :) If it doesn't work then use XMP.

Calvin7 has a good point which I overlooked. Along with 1.3 vcore, your cooling is holding you back a little too. I wouldn't say inadequate, but you can't push that hot chip too far without some very decent cooling.

The main objective is a balance of low voltage (as low as you can go - in this case 1.25 stable, or whatever the actual lowest you can get is), temps below 80c, and your cooler isn't hopping off the walls at full pelt to keep the temps low.

You could also try to increase the SOC voltage up a notch. This may help with your mem, and help make it stable when OC the CPU.
 


+1^ Can't disagree with the methodology at all. Get the mem settled at stock. OC the CPU testing along the way. I think a lot of folks just dial in settings and wonder why things aren't working. Just a generalization, Not suggesting the OP is of that ilk. Just being patient though brings best results.
 

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530


How long do you recommend I run Prime95 to verify stability?
 

luckymatt42

Upstanding
May 23, 2018
446
1
360
I'd recommend a MINIMUM of two hours with Prime95 and/or RealBench. Frankly I like RealBench because it is a bit more of a realistic workload (encoding video, multitasking, etc.) Preferably let it run all night...remember that patience thing we discussed ;)

Something else to note during testing...see how long it takes for your temps to reach "steady state", where temps are not going up any more. (If you see it creeping up and up and taking a long time to settle into one "max" temp, it may mean your cooling needs a little help)
 

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530


So, Ive been tweaking the memory today and things have been good so far. Ive been able to see gains from stock memory speeds to the XMP profile. However, after running the benchmark test in RealBench I get the message, unable to properly parse system specs. I believe I received this message when I did a base test with the memory at default.

I tried searching this online but wasnt able to find out what it means. Any idea?
 

luckymatt42

Upstanding
May 23, 2018
446
1
360
I've never seen that particular message, but does it allow you to run the test? If not, I'd say something isn't stable somewhere. Back things down to your last known good value (remember, this is why we only change ONE thing at a time before testing...), then try again.

One change at a time, keep notes, repeatable tests with EXACTLY the same settings from run to run.
 

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530


I get the message after running the test and passing.
 

luckymatt42

Upstanding
May 23, 2018
446
1
360
Oh, well that's odd. Based on the message it sounds like it's having trouble identifying your hardware. I would just enter that exact message into your favorite search engine and see if others are experiencing the same thing. Make sure you're on the latest RealBench version and all drivers are up to date.

But my instinct is to say that message is not overclocking related...I'd just set BIOS back to default and see if I got the same message. (P.S. This is why it's a good idea to utilize BIOS profiles...there should be an option that allows you to save all current settings under a profile, makes it quick and easy to switch between stock settings, overclocked, stable, etc etc. Each would have it's own BIOS profile.)
 

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530


I searched it in google but didn’t really find an answer. I have the profile saved so I’ll revert back to default and see what happens.
 

lopez.mr.22

Commendable
May 1, 2018
26
0
1,530
So I've yet to revert my OC to see if I still get the message mentioned above in RealBench. However, I am now noticing a small issue with my mouse. Every once in a while the mouse cursor lags, kinda like a stutter and it has a trail to it. I noticed it after I did the overclock and after I got a new monitor.

My CPU is at 4.7GHz at 1.26v and the memory at XMP with 1.32v. Could the overclock cause this mouse issue?

I'll be doing some basic troubleshooting on this soon, but wanted to see what you guys thought on this.
 

TRENDING THREADS