TL/DR: POWER SUPPLY!
Well, curious. I wonder if some of the changes I made over the last year or so might have softened possible transient power spikes? I think I've read the Vega (and maybe 580) have bigger transient spikes than in the past? Anyway despite all the time that has passed since this post and all the things that seemed to help sometimes even for weeks, recently I had another HORRIBLY maddening run of random crashes in game.
PROBLEM SEEMS TO BE SOLVED: power supply. God bless it.
So, I had quality Antec 750 watt power supply that I think I bought in about 2011. Right up until I replaced my R9 280 (first with a 580 then Vega RX56) I had incredible stability so it never occurred to me the PS could be the problem. I'm pretty sure the Vega RX 56 doesn't draw MORE power than an R9 280 but I have read that possibly it does spike quite high sometimes.
Anyway, I bought a brand new Corsair HX850i and have now gamed for at least 12 hours with no crashes. Interestingly, the power drawn through this PS shows that it's not pulling even half of 850 watts so it's total overkill apparently. I mean, the power supply fan hasn't even needed to come on yet.
I still haven't tried re-enabling the TDR feature but that would the big test I suppose.
Hi guys, I had this problem for almost a year (see here my terrible crashing for months fixed - RX 580 and Vega RX56 ) and I was ... losing my mind. In short, I had NO stability issues for almost 3 years with an Intel box and an R9 280. I then upgraded to an RX580 and started to get the amdkmdap error constantly. Then the Vega RX56 came on sale and I hoped that the 580 was defective and the Vega would fix the issue. Nope. These crashes were just constant, no matter what I was doing. Of course I'd tried playing with the Wattman settings, different driver versions, different graphic settings, I fully wiped the hard drive and reinstalled Windows 10, etc etc. Now, I've been PC gaming for 23 years and building my own gaming systems since the late 90's, so I'm no noob (plus I'm a sys admin for a living) but this was driving me nuts. If I had had more time to dedicate to solving this problem instead of just trying to spend my little bit of free time gaming, I would have have been more methodical and figured it out sooner but oh well.
Now, I have a few posts on this forum about this and thought I had it figured out a couple times thinking something else was causing it, but no. I've fixed it now. No more crashing. It took me a very long time to track down the specific problem by starting to take notes to any changes made by me, Windows Updates, or driver updates. I had tried the TDR_Delay fix and that seemed better for a while but wasn't good enough. Also, either this key or another key (I'm not at home to see my notes) kept changing even DURING my gaming sessions. I'd set it, it would be fine as long as it remained then the driver would crash and when I looked at the registry key it had been changed back.
Okay, the fix for me was to set (or add the key) TdrLevel to "0". This means that a graphics driver crash means you'll need to fully reboot the PC, but I had no choice. I had so many nights where I couldn't even play for 3 or 4 minutes without a crash. Here's a URL that describes the key. TDR Registry Keys | Microsoft Docs
I've now gamed for over 20 hours with no more amdkmdap errors, though I have had a crash or two that forced me to hold the power button in for five seconds to power cycle the computer. Not perfect but I'll take it.
One interesting note. On my work PC which is running Windows 10 1803, there are no TDR keys at all. So, I'm wondering if the cause of going from a system that had an AMD card (the R9 280) and was stable for years and then the frequent crashing (after upgrading to the 580) is because AMD added the TDR key themselves with the newer cards.
The only two games I've been playing were NBA 2K17 and Mass Effect Andromeda.
System specs off the top of my head:
Intel i5-4670K (Haswell)
Gigabyte G1.Sniper Z97 motherboard
16 GB 1866 PC3-14900 DDR3 RAM
AMD Vega RX56
Samsung 850 Pro SSD
750 watt power supply