High VRM temp makes OC unstable

[Schmuckley]

My z390 prime a vrm hits 90c during stress test while my cpu stays 70c-75c. Does it cause my oc unstable? I don't see any option to water cool it. Do put a vrm fan help or just buy a new mobo?

You should put a fan onto your vrm sink, at least a Cooler Master-type.

 

[mrthanhnguyen]

Its weird guys. Borrow a 9900ks from a friend and its stable 8hr prime95 5.2ghz at 1.3v on evga dark. Put that chip to my board and it crashes 5s after hit start prime95.

 

[oxrufiioxo]

Its weird guys. Borrow a 9900ks from a friend and its stable 8hr prime95 5.2ghz at 1.3v on evga dark. Put that chip to my board and it crashes 5s after hit start prime95.

The Dark is a substantially better board that probably has a way more stable voltages.

Either way a 400 board isn't worth 100mhz.

 

[JackCarver]

Not weird, the Dark has better vrm. 6 real phases on the Dark vs 4 real phases on your board can make the difference as ripple is higher on your board compared to the Dark.

 

[mrthanhnguyen]

Dark or apex or aorus master? I have money yo throw out.

 

[JackCarver]

Dark or apex or aorus master?

VRM wise the Apex is best, 8 real phases and IR power stages.
VRMs of Dark and Master are very similar, 6 real phases and Doublers used, but as you had good results with the Dark from your friend, I would choose the Dark, if I had to choose between the two.

I have money yo throw out.

 

[mrthanhnguyen]

If vrm on aorus master and the dark are the same, I may try the master because it has 4 ram slots. The dark and apex have only 2 slots. I have 4 sticks of ram.

 

[oxrufiioxo]

If vrm on aorus master and the dark are the same, I may try the master because it has 4 ram slots. The dark and apex have only 2 slots. I have 4 sticks of ram.

The Dark has a better vrm and a way better pcb than the master.

 

[JackCarver]

If vrm on aorus master and the dark are the same, I may try the master because it has 4 ram slots. The dark and apex have only 2 slots. I have 4 sticks of ram.

Not exact the same, they both have 6 real phases and both use doublers to get up to 12 virtual vcore phases. The Master has 40A power stages whereas the Dark uses 60A stages, so the Dark is capable of handling more current than the Master. Question is if the 12x40A=480A power stages of the Master are sufficient and I would say yes cause you also have to cool this huge load. Both have really big heatsinks on top of the vrm, the Dark heatsink seems to be even a little bigger here. But with the Dark you theoretically would have to cool 12x60A=720A max current load.

But as you said, that you had good results with the Dark, I would also choose this one like @oxrufiioxo said.

 

[mrthanhnguyen]

Have a very good deal of apex. Is dark much better than apex?

 

[JackCarver]

Have a very good deal of apex. Is dark much better than apex?

Apex is better, 8 real phases for vcore in twin configuration, so 16 IR power stages, capable of 60A each power stage, so theoretically 16x60A=960A max current load.
Also the ripple effect is less here as it has 8 real phases compared to Dark or Master which have only 6 real phases.

Edit:
I think it's a very good OC mainboard, but also look at the other specs if it's suitable for you.

 

[oxrufiioxo]

Apex is better, 8 real phases for vcore in twin configuration, so 16 IR power stages, capable of 60A each power stage, so theoretically 16x60A=960A max current load.
Also the ripple effect is less here as it has 8 real phases compared to Dark or Master which have only 6 real phases.

The Apex probably has slightly better transient response as well. I still think the actual pcb of the Dark is better though.... Again though whether its 300 or 500 in your local currency that is a high price to pay for an extra 100mhz.

 

[JackCarver]

The Apex probably has slightly better transient response as well....

Yes without Doublers transient response may be also better here. It's a nice OC mainboard, but I didn't look at the other specs of the board like USB/sound and so on.

 

[oxrufiioxo]

I would make sure to watch both these vids before deciding on either. If you haven't already. The output filtering on the Dark is insane.

 

[EarthDog]

The output filtering on the Dark is insane.

While that is true, living in this level of minutia for ambient overclocks is a waste of time. For all intents and purposes, all three boards listed will bring this CPU to the same overclock.... in other words, the seeming hard wall limits these CPUs have.

 

[oxrufiioxo]

While that is true, living in this level of minutia for ambient overclocks is a waste of time. For all intents and purposes, all three boards listed will bring this CPU to the same overclock.... in other words, the seeming hard wall limits these CPUs have.

I agree, but if you're ditching a completely viable board for 5.1ghz you might as well get the best and not worry about it anymore..... I have my doubts that his specific cpu will clock any higher in either of these boards but I would be happy to be wrong.

 

[JackCarver]

While that is true, living in this level of minutia for ambient overclocks

The Apex is designed for LN2 OC. It has sensors to detect condensate

 

[EarthDog]

The Apex is designed for LN2 OC. It has sensors to detect condensate

Well aware.... just saying that this level of minutia isn't terribly helpful.

I agree, but if you're ditching a completely viable board for 5.1ghz you might as well get the best and not worry about it anymore..... I have my doubts that his specific cpu will clock any higher in either of these boards but I would be happy to be wrong.

The choice is his/hers, surely... just saying for ambient overclocking, it won't make a difference.

 

[mrthanhnguyen]

5.2ghz at 1.3v now on the apex. It was 1.36v for 5.2ghz on the prime a. Still fail occt linpack but seem pass realbench 1hr so far. Maybe I need to increase vcore a little bit to be stable in linpack.

 

[JackCarver]

5.2ghz at 1.3v now on the apex. It was 1.36v for 5.2ghz on the prime a. Still fail occt linpack but seem pass realbench 1hr so far. Maybe I need to increase vcore a little bit to be stable in linpack.

Nice one. Also impressive that you get the same results with 0.06 less VCore. That‘s the ripple effect which is way better on an 8 phase board compared to your 4 phase board before. Yes I assume with a little more VCore you get stable in Linpack.

 

[mrthanhnguyen]

1.3v in bios llc6, but why hwinfo and cpuz show only 1.243 during load? This board has llc up to 8 while the old board has only 7 and I chose 6.If I choose 6 in this apex, does this mean 5 in the prime?

 

[JackCarver]

1.3v in bios llc6, but why hwinfo and cpuz show only 1.243 during load? This board has llc up to 8 while the old board has only 7 and I chose 6.If I choose 6 in this apex, does this mean 5 in the prime?


View attachment 147406

That‘s the loadline as intended by Intel. You set a voltage in BIOS and usually you have a slightly drop of this voltage in windows idle which is called vdrop. When you switch from idle to load you have another voltage drop which is called vdroop. This is intended because of the transient response effect. There should always be a drop from idle to load. With llc you can compensate this drop if you get unstable under load. So you can choose various llc level in BIOS and see how vdroop changes. For high oc it‘s often recommended to choose a llc level where there is no vdroop at all, so voltage in idle is the same as voltage under load but that‘s not good for the cpu. Better is choosing an llc level with vdroop, so that voltage under load is less than voltage in idle. But you can try different llc level in BIOS and test in windows.

 

[mrthanhnguyen]

I feel weird because with the prime a mobo, load voltage is always higher than idle. For example, 1,36v in bios and llc6 and it will be 1.376v during load. This apex llc7, 1.25v in bios , 1.243 idle and 1.204v during load.

 

[JackCarver]

I feel weird because with the prime a mobo, load voltage is always higher than idle. For example, 1,36v in bios and llc6 and it will be 1.376v during load. This apex llc7, 1.25v in bios , 1.243 idle and 1.204v during load.

That‘s dependent of the loadline level. I‘m sure there are also loadline levels on the prime where voltage under load is less than voltage in idle.
You can also choose different loadline levels on the apex with different idle/load voltage behaviour from llc1 to llc7.

 

[oxrufiioxo]

You'll have to experiment unless somebody else has a z390 Apex and a 9900k for you to compare with.