• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

RX 470 MSi Armor OC 4G ... Voltage creep over days?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.


New Member
Dec 1, 2016
Hi Overclockers,

I'm new to overclocking. The best I've done in the past is upping the multiplier on locked intel CPUs as far as it will go.

Right now with my new RX 470 I decided I was going to overclock it because it ran so cool. It's going well, except for one confusing issue.

If you've read this far, thanks! Now:

Initial testing with core voltage at "Auto" did not yield ANY stable OC. I've read some folks can get stable at "Auto" up to 1350 MHz. This one simply could not do that.
I then changed the core voltage (using WattMan) at the last DPM state to 1150mV, or about a +50mV change. I had thought this was a limit as it wouldn't let me set it to 1160.
I figured that got be stable up to 1330 MHz, which was nice. All the while in that testing I found that messing with the memory clock did not affect stability.
At 1150mV with stock core clock, 2100 MHz memory clock did not introduce any stability in the same scenarios where 1335 or 1340 MHz would introduce instability on the core clock.
Interestingly most of these crashes by instability were due to browsing the internet. The most demonic game when it came to destroying my core clocks was World of Warships. It was very sensitive to that. But my memory clock? Nothing seemed to care.
So I ran the card at 1330 MHz / 1150 mV core and 2100 MHz / 950 mV memory for a day and a half without issue. But when I woke up and turned the computer on on the third day, the card was no longer any kind of stable at these voltages. It would crash just by opening Google Chrome or, for heaven's sake, WattMan! Thankfully WattMan resets OCs.
At this point I suspected the memory. Increased voltage to 970 mV. No dice. Went to stock memory clock and voltage. No dice. So it was the core.

I went back and found out that my voltage limit in WattMan was actually 1175. So I set my voltage to that. Great! It was stable! Now, how much farther can I push it?
Turns out 1340 MHz was the limit at 1175mV, or +74mV. I also set my memory to 2100 MHz / 970 mV. That was stable for, again, a full two days.
In the meantime I started playing around with the lower DPM states in WattMan by reducing their voltage to try and get some more life out of the card. A bit silly of an idea in hindsight.
In any case, for one full day that was stable if I reduced DPM states 1-6 by 5mV. The next day however, it was unstable again. That was two days ago.
I then eventually came to the conclusion that no matter what I did with the lower DPM states, they were unstable. Clockspeed, voltage, up or down made the system crash. But figuring as I always reset DPM 7 to 1340 / 1175, I now wonder whether or not my max clock was introducing the instability, not the lower clocks.

Now that ran stable again for one more day, and then yesterday I hit instability again when browsing the internet and 3D applications.
Since I had hit the voltage limit in WattMan (or at least I think I did), I installed MSi Afterburner and tuned it to +78 mV. That was stable.
Until this morning, where I just crashed from opening Google Chrome.

Bearing in mind that I did run MSi Kombuster for 5-10 minutes every time it got "stable" to verify two things:
That the OC had actually applied,
and that it was improving the framerate.
It had "passed" every time then.

I just wonder, has anyone else experienced this issue where a card is voltage creeping over days rather than months?
If this is an uncommon issue, I am tempted to just return the card (as it is still in Newegg's return period).
That is kind of sad, as I feel like being stable at 2100 MHz memory is an extreme rarity for the RX series 4GB cards.

It could be an entirely different issue, where the card is unstable until the VRM reaches a certain temperature and instability is simply from cold starts.
For any curious, I just bumped the memory voltage to +50mV (1 volt total) and it seems to be working.

Scores 4k graphics score in 3DMark time spy!