The error could be relate with high temperature on the processor, try to confirm all the fan and Temperature settings are correct (defaults) and confirm there is not dust, dirt or any that can affect the airflow on the rig Try to troubleshooting with the board manufacturer and get the best configuration and check for the latest BIOS. If you already narrow the problem on the board or processor try to RMA the board first; processors are very solid units and are very dificult to see a faulty one, but is possible.
Well, I wrote to Asrock's tech support, giving them the same info I posted here, and I got an answer a couple days later telling me how to test the RAM. So I wrote back and told them the CPU was causing the crashes, not the RAM, and they sent me another email telling me... how to test the RAM. /facepalm
Anyway, the board and CPU are new so there's no significant amount of dust on them. Idle temp is around 40C, and when I had the crashes it usually wasn't under load anyway - just web browsing and stuff. I found a big thread over on TweakTown about this issue. I'll check back in once I've slogged thru it.
Well, it seems this problem is more common than I thought. I haven't finished digging thru all the threads, but it appears that the motherboard sets the CPU's voltage too low by default. When the Vcore setting is left at "Auto" and the power-saving features kick in, Vcore can drop below 0.9 volts. My (wild) guess is that when the voltage gets that low, the GPU gets starved for power and that's why it stops responding. The suggested fix is to manually set the Vcore at 1.190 volts. It's working so far, but I'm not sure it's the best option. Is there a list of the voltage tolerances for the 2105? I'm having trouble finding one, and it's not in the manual.
The Vcore on the new processor have a range from 0.2500V to 1.5200V, any number on this range should be OK. Please check the next link, open the PDF and check the document for detail information, more specifically page 78, table 7.10.1:
That makes no sense. the voltage range 0.8 - 0.85 is related to the CPU cores not the Graphics domain of the processor
There are many voltages
GFX_VID 0.2500 — 1.5200 V
VID 0.2500 — 1.5200 V
Voltage for the system agent 0.879 0.925 0.971 V
I/O supply voltage for other than DDR3 -2/-3% 1.05 +2/+3%
As it seems GFX voltage is to be set separately and although it shares the same range of 0.25 - 1.52 it should be totally unrelated to the CPU domain, as are the system agent and the Vccio of the L3 cache and ring. Unless the ring and l2 cache (shared by both the IA and GT cores) actually runs on CPU VID instead of Vccio but it's a bit cloudy to me what does what exactly. Anyway, too much power saving is bad news for stability and performance.
probably what asrock did is make VID and GFX VID being connected and controlled at the same time instead of GFX being constant and unrelated. Big mistake asrock.
Well you could just disable the C-states and the Voltage will not drop below 1.00V i find it totally useless anyway, saves only 5 Watts and does this at the expense of switching needlessly caches on and off like we make any difference with 5 watts. Setting manually 1.19V is in a no way a good idea though. That is 10 more Watts.
Well, I don't know much about voltage settings (especially for integrated video) but the BIOS seems to have only relevant 2 settings: "IGPU Voltage" with options Auto, -100mV, -50 mV, +0 mV, +50 mV, and +100 mV; and the other setting is "IGPU VDROOP" with options With and Without.
Most of the people reporting this problem seem to be using i5-2500k's, but since I'm using an i3-2105 I tried setting the Vcore to 1.1V. It's been stable except for one unexplained Stop Error 0x1E.