I'm trying to overclock a 1800x to 4Ghz, and it's proving to be a nightmare on this board.

When I set the voltage to 1.45v on Ryzen Master and start stress testing with AIDA64, the voltage drops to an average of 1.378v, and the system crashes after a couple of minutes. Monitoring is done using HWINFO64.

I tried setting the voltage to 1.4625v and later 1.46875v; this allowed the average voltage under load to stay around 1.4v, keeping the system stable for around 15min, when i stopped stress testing.

The issue is that when idling, it idles to this extremely high voltage.

There is no Load Line Calibration option in the BIOS; it was discussed in a topic here that it was removed from newer versions of the  X370 Killer SLI.

I tried OCing from the BIOS, with no success; setting the voltage to 1.45v and the clock to 3900 or 4000mhz doesn't post.

I'm running out of options here, and really regretted getting this board. Any advice would be really appreciated!

I'm aware of the safe voltage for 24/7 usage, but an AMD representative disclosed a couple of months ago that 1.45v was the maximum voltage that should be used for more extreme overclocks.

zlobster, I'm using the latest UEFI. This motherboard's newer revison like the one I have does not have LLC options in the UEFI, unlike the first ones had.

Petrol, the PSU comes from my previous build, and is barely 1 year old, and I'm running a single 1070 now.

VRM average temps for the last hour is 84c.

I'm pretty sure it can do 4Ghz stable, since it's over an hour stress testing @ 3.9 1.38v, no problem.

I sent AsRock an email describing the issue. Do you know of any other means to get in touch with AsRock regarding this issue?

Does any AsRock technical personnel frequent this forum?

Thank you!!

Exactly, if I had LLC options to keep the voltage under load as close as possible to the defined voltage, things would go way more smoothly.

I sent the guys a private message, and am hoping they take a look at this topic!

That is the problem. I feel extremely mislead.

The 0.07v drop is killing the overclockability past this point.

I'll wait for AsRock's support to respond; maybe enabling LLC on this version could be possible in newer BIOS versions. I remember someone saying it was not necessary due to VRMs being replaced, but no one said it was impossible.

I have the X370 Taichi and while I was initially able to get 4.05ghz stable it did not remain stable for long on my 1600x. I had to tune it down to 4.0 then eventually 3.9ghz for daily use. Ryzen CPUs do not seem to hit 4.0 and beyond as easily or as often as early reviews seem to indicate without pumping unsafe voltages into them. With a 240mm AIO on my system I find my CPU temps get a little too high with a vcore of anything over 1.395v, my VRMs also get hotter than I am comfortable with for 24/7 use.

If you are getting a lot of voltage droop, try positioning a fan over your VRM section for some extra cooling. VRM controllers are a lot more advanced these days and actually do throttle when overheated/loaded. My system has 2 140mm fans and 6 120mm fans installed with about a 40% to 60% exhaust to intake ratio. Just changing my fan profiles can improve stability at 4.0 on my system. I can technically run my system at 4.0 but temps are not what I consider ideal (I like my system fairly quiet). 

The Killer SLI, Gaming X and Pro series boards are not designed to give you maximum overclocking potential. Even my Gaming K4 can't live up to my Taichi. 3.9ghz is not a huge jump down from 4.0 anyway. I find gaming at 3.9 on my system is no different to 4.0 for example and the only place I see a more noteworthy performance increase is in synthetics like cinebench. 

Don't let the internet and "expected performance" figures ruin your enjoyment of your system. All too often it is easy to get caught up in getting X or Y overclock and A or B RAM speeds stable when the reality is, you won't even notice the difference between your overclocks vs the "ideal" 3200 and 4.0 everyone seems to think is somehow essential to even use the system..... I most often use my systems at stock after tinkering with OCs when the system is fresh and new. Once I have my benchmarks all buttoned down I just load defaults and enjoy my build. With Ryzen I do notice that 3.8+ seems to make a difference so I keep my build at 3.9 and my RAM at 3200 because it's stable there for me. Don't get caught up in the who's is bigger nonsense, enjoy what you purchased/built. Also bare in mind that AMD's AGESA is still evolving and will continue to do so, each update brings change, not always good it seems. Once AM4 is more mature things will settle and actual expected figures/OCs can be established. 

Xaltar, thanks a lot for the comprehensive explanation!

Even though the VRMs do not reach more than 80 something degrees, I'll try cooling them to see if it helps with the vdroop!

Thanks for the heads up regarding letting high expectations ruin my enjoyment.. I was really frustrated, and started regretting building a new PC.

If anything changes by cooling the VRMs, I'll let you guys know!

3.95Ghz seems stable enough at 1.4v, but I'm not sure it is worth it over 3.9Ghz at 1.35625v. 

I can say, if the newer revision Killer SLI boards utilize the same power design as the Gaming X then LLC will not be returned. It is physically not supported. However, if ASRock has chosen to use this design on another board (not just the Gaming X) then it has clearly passed their stringent in house testing and proven itself reliable for the segment it has been implemented for. Like I said, overclocking is not a selling point for boards in this range. The same is true of similarly priced boards from all manufacturers unless they have skimped in other areas to focus on overclocking. 

3.9 is perfectly acceptable with Ryzen, using something like a 1080ti the 100 mhz jump to 4.0 would only net you in the realms of 3 - 4fps at 1080p on high settings in most new AAA titles. Considering at this point you are well over 100fps anyway you are not going to even notice the difference. On a GTX 1080 or lower the difference would be practically non existent. Bare in mind that reviews are, by design, set up to force the bottleneck onto the component under review. In the case of a CPU this means GPU and RAM are maximized to increase workloads for the CPU. Unless you are using obscenely expensive GPU/s and RAM, you will not see the same benefits shown on a review system. 

Like I said before, the "whose is bigger" issue is a real problem today. Most reviewers do not outline the information I just did. Manufacturers sending them products want users to see the best possible performance that can be extracted not tell users that the lesser model is probably all you need. Some reviewers do break down value now which is a step in the right direction. 

The R7 1700 vs R7 1800x for example are physically identical save a few bits of code in the microcode changing frequencies and name (yes, the difference is in the firmware of the CPU, nowhere else). While in an ideal world the 1800x would be better binned and achieve higher clocks at lower voltages the reality is usually far from that. Demand will force higher binned chips to be branded with lower monikers, especially when manufacturing processes mature and defect rates are decreased. 

From my experience and reading many user accounts:

3.8ghz paired with a GTX 1080 is sufficient to get the most out of your system 

3.6ghz paired with a GTX 1070

Stock with anything less.

Also bare in mind this is for 1080p, as soon as you jump up to higher resolutions the GPU becomes the bottleneck in almost every instance. 

I think I did some stickies breaking this info down a while ago, I need to find them and update them again and maybe repost them in a few more sections. 

The only significant quirk of Ryzen is the RAM speed issue, it is not as severe as it has been made out to be but it is valid. Generally 2666 should be the go to speed for RAM on systems using a 1070 (or equivalent) or lower and 2933 for a GTX 1080 or higher. 3200 is neat but the tiny performance bump is not worth the additional cost vs an overclockable 2666 kit. 

Don't get caught up in getting specific figures, focus on the best balance of temps vs performance. Overclocking is about finding the balance between frequency and temps not breaking records and burning out your parts in half the time  Shooting for records or setting goals is not a game for the casual overclocker, a pro will go through a dozen motherboards, CPUs and GPUs for a record attempt. Many hardcore overclockers will purchase rebinned chips (tested by a 3rd party) that have been tested at high overclocks in store and cherry picked to deliver high clocks. These CPUs sell for hundreds of $$ more than the retail value. Stores like these will actually burn out a significant portion of the CPUs they test to find the golden silicon they then sell, the added cost is to cover this expense.

Honestly, the problem is how easy overclocking seems today with such simple options in the UEFI. Anyone can have a crack at it and sometimes you get lucky, after all, the CPUs sold for hundreds of $$ more still come from stores. Every casual overclocker dreams of getting one of these "lucky" chips or as the saying goes in the discipline "winning the silicon lottery". The term is quite apt  

So, the TLDR:

Anything over 3.8 is a bonus and 3.8 should be enough to extract the most out of your GPU. If you want to challenge yourself and have some fun with your overclocking, try and see how low you can get your voltages at ~3.8/3.9 while still being stable. This is the most often overlooked conclusion of a proper overclocking session, finding the sweet spot for 24/7 use 

Thanks again for the input, Xaltar!

Just completed a 13 hour run of AIDA64 with the following results:

Core clock: 3.9GHz

CPU voltage: 1.3625v

Core voltage under load: 1.29v

Average temp: 69c (air conditioning on drops it by 10c; Brazil is really, really hot)

VRM temps: 80c

The last attempt at 3.95GHz was @1.4125v (1.36 under load) unstable, so I gave up. Totally not worth the extra voltage and temps (around 5c).

Overall I'm really satisfied with the results! Next up will be a 12-hour run of Prime95 and OCCT to assess stability and then I'll work on tightening the memory timings or increasing the clocks; Samsung B-die should give me some working room!

Thank you guys a lot for the feedback! :D

I'll keep this in mind when time comes to purchase my next motherboard, or give advice to friends.

However, the price was extremely good at black friday, making it cheaper than lots of B350 boards, so I still feel like I got a lot o value for the money invested.

Looks are awesome too on my black and white build! :D

This is where ASRock shines, value for money  Get a good deal on top of that and you really score. 

On top of all that, there is no way of knowing if your CPU can even reach 4.0 to begin with, no matter the board, a lot of Ryzen CPUs can't. This is why I said overclocking should be focused on getting the most out of your particular system without sacrificing thermals, wherever you land, you land. In all my years of overclocking I have only ever gotten one good CPU that met (and exceeded) my expectations, all the others (I build PCs for a living so this is a LOT) have narrowly missed my goals or failed to come close at all. Eventually I wised up and realized that literally any overclock is free performance and more than I paid for. Change the way you look at it and you will no longer be disappointed and actually enjoy the process. 

I have been overclocking for a very long time, I have felt your frustration many times, especially when I started out. Eventually I mellowed out and realized how little that extra X or Y mhz really meant and how little performance it amounted to. Remember this, you have an 8 core 16 thread monster on your desk that kicks ass and takes names, even at stock 

Good luck with the RAM, I hate setting up RAM, it can be so tedious  I managed 3500 on my 3600 RAM with my Taichi then the next BIOS update killed that and I couldn't get 3200 stable then I got 3200 with the next update...... It wasn't my board or ASRock it was the damned AGESA code, I have never had such headaches with my RAM before. If you can get anything over 2666 you are doing good, 3200 or more and you are sitting pretty for sure  I pushed my system to the absolute limits for the review I wrote on the board, usually I don't go to so much effort.

Happy Holidays and congrats on your new beast  

Thanks for making me look at this situation from another, more positive, perspective!

I'm preparing my stomach to deal with the RAM later. Guess I won't even have the time, since I still have to complete runs of OCCT, Real Bench and some Battlefield 1 to consider it stable enough for my needs!

Happy Holidays to you guys, and I can't wait to get everything in order to finally play some Battlefield 1 on this machine. My i5 3570k @4.5Ghz was providing really poor .1% lows, even though the maximum reached between 90-100fps!

Cheers!

Thanks for the heads up! I'll install it when possible!

We are kind in the same boat then, I got this board at BF too and I am seeing the same thing. I am running my R7 1700 @3,6Ghz 1.1875V, but when stress testing, HWINfo shows a max Vcore of 1.131V. 0.044 is marginal, I know, but LLC would be put to some good use right now. I am using the stock cooler, so I am taking OC slow right now.

I upgraded but returned to 3.40. I had several issues with this new RAM. I was running my RAM micron @2933Mhz 1.275V, but with this new BIOS it responded with continuous shot beeps until I powered the system holding the power button. Also, if doing OC with p-states is your thing, you must chose the AMD CBS OC option. However, that closes the XMP options too and OCing the RAM gets complicated. And, the other situation I found is that the mobo gets the RAM voltage resetting to 1.20V (im my case), so it's not helping at all.