Snipeuout wrote: My 2 cents on your post. The issue occurs in Bios 2.0 and 2.1 with the ram set to JEDEC speeds of 2133. On a "Just plugged in" start I can stress test and benchmark the system at 2133 or 3200 with no stability issue using 2.0/2.1, however once I shutdown and start the PC (2133 or 3200) it goes into this failed mode where only the GPU is powered.   If 1.8 works fine, how can we not at least suspect that bio 2.0/2.1 are bugged?

So now this is NOT just a wake from Sleep issue? You have it on the second cold boot after restoring power to the board?

I was trying to explain the potential cause for this situation, or why it is happening, not that it is not happening. Plus I'm working from very little information. But I did see information that seemed to explain at least some of the issues. Add to that, my ASRock Z170 board with the latest UEFI version does not have any of these problems at all.

My explanation seems to be supported by this post:

sgoemans wrote: Thanks for your support. Unfortunately, the new BIOS 2.30 didn't fix the issue, although I can see that VCCIO is now preset to 1.150V. If I alter the value to 1.050V, everything is fine again. Summary: 2.10 default - Not OK 2.10 VCCIO from 0.95 to 1.050 -> OK 2.30 default - Not OK 2.30 VCCIO from 1.150 to 1.050 -> OK Please note that I always set the DRAM profile to XMP 2.0 Profile 1 before I started testing.

If the extent of your interest in this situation is either it is "bugged" or not "bugged" for your memory model, then feel free to call it bugged.

IMO, bugged means broken under all circumstances. The cause of the failure is a voltage value that is wrong in any circumstance, like a CPU voltage set to 0.001V, for example.

I'm beginning to think ASRock  made a mistake attempting to enhance memory compatibility with UEFI updates for their Z170 boards. The new modifications are causing other problems with memory that was compatible before the updates.

Frankly, IMO the XMP profiles in some DDR4 memory are wrong, or contain bad data. XMP profiles for Skylake and X99 HEDT systems, the only ones that use DDR4 memory, are not 100% compatible, in reality. Some memory manufactures don't like that, and try to make their memory one profile fits all. Then the mother board manufactures get to deal with it... as usual. 

Dear Customer:
This is ASRock TSD Bryan.

For this issue, please update to BIOS P2.30 to try. Download link is as follow:
http://www.asrock.com/mb/Intel/Fatal1ty%20Z170%20Gaming-ITXac/?cat=Download&os=BIOS

If you have any further question, please contact us:
http://event.asrock.com/tsd.asp?ln=en

Thank you.
Yours truly,
ASRock TSD 

Hello, I seem to be another customer with exactly the same issue. My PC specs are also identical:

Last night I did a few some restarts, shutdown/boots and sleep/wakes. So far it works 4.0.

gusviper wrote: Dear Snipeuout, we have almost the same specs, and 100% the same problem. my specs I7-6700k Corsair vegennce 3200 16-18-18-35 2x16gb Corsair RM 550 PSU Samsung M.2 500GB 850 EVO UPS CyberPower Adaptive Sinewave CP900EPFCLCD The 1.8 Bios version fixes the problem. But if u want to use the latest version (2.10 ) without sleep mode problems, u have to raise from the bios settings, the CPU VCCIO Voltage from 0.95 up to 1.00 (1.00v-1.20v is safe). Give it a try, this worked for me.But i hope Asrock support give a permanent solution.

First I want to thank you for posting this information. If it is true (sorry but I cannot verify it myself, I don't use this board. I'm not saying you are wrong or giving false information, I simply do not know) it is useful for helping others with this issue. But this situation is not necessarily a problem with the 2.10 UEFI, as I will explain.

Your post implies that the default, Auto VCCIO voltage was changed (reduced) after UEFI version 1.80, in version 2.10.

UEFI 2.10 includes an "Improve DRAM compatibility" change, so that is possible.

Is that true? Do you know what the default or Auto VCCIO voltage was with UEFI 1.80? The VCCIO voltage can be different for several reasons.

Or do you think something else is happening?

VCCIO voltage is usually increased when using DDR4 memory at over clocked speeds, or the PC won't POST/boot. That can be done manually, or is set by the memory's XMP profile.

My ASRock Z170 Extreme7+ board, used with my G.SKILL 3200 memory, sets the VCCIO voltage (and VCCSA) automatically to much higher values when the memory speed is set to 3200. That is necessary for the higher memory speed, any speed beyond 2133 is an OC. The voltages were higher than necessary, and I was able to reduce them without any problem.

Since it is difficult to know what VCCIO and VCCSA voltage is needed for every model of DDR4 memory, and for the amount of memory being used, high values for these voltages are used to insure compatibility.

Your memory seems to be at 3200, obviously an OC beyond the standard DDR4 speed of 2133.

Are you using the XMP profile to set your memory at 3200, or did you change the memory settings yourself?

It is possible that the 2.10 UEFI includes reduced automatic VCCIO (and VCCSA?) voltage values used with memory at over clocked speeds. It seems that is what has happened.

If your memory at 3200 does not work perfectly (wake from Sleep problem) with the different, reduced value for the VCCIO voltage (assuming that is happening with UEFI 2.10), you need to adjust it, which you have done.

But that does not mean the 2.10 UEFI has an issue in general. UEFI 2.10 might be setting the VCCIO voltage to low in some cases, but that will take some research to verify.

I hope you understand that when we OC our memory or CPU, we are responsible for finding and setting the necessary voltages that provide stability. That is the only solution that is realistic. What other permanent solution is possible?

Yes it is frustrating that after a UEFI update, we find an issue occurring with our PC. You found the solution, but IMO we cannot expect the UEFI used in our boards to provide automatic voltage values that work for the over clocked memory we are using. Those voltage values will not work for all memory models. ASRock does provide memory OC settings for some models of memory, for some of their boards.

If the VCCIO voltage is being set to low at 2133 memory speed, then I would agree that should be fixed.

As I said above, all of this should be verified before we come to any conclusions.