With regard to your CPU I can only echo what Spock used to say, "May it live long and prosper"
If AMD were to update Ryzen Master such that it would load the previously set profile on the next boot (that is include an option to do this) I would be very happy with the solution I came up with.
As it is, it is a bit of a PITA having to apply the profile manually at each boot.
Not that I care all too much, because my systems run 24/7 so the annoyance is somewhat abated.
Im having issues getting this 3900x stable. Im not even concerned with overclocking as much as I'm concerned with stability within my DAWs. Any chance you can go over my my settings and make sure everything is ok? I got a 7500 score with cinebench r20 using your first guide with Ryzen Master but Studio One, Pro Tools and FL Studio are each suffering horrible popping that my prior i5 6600k did not exhibit. I tried to use your new guide here but I cannot seem to locate some of the sections as they may have different terms within ASRock. I have a few days before my return window closes and with all the hype surrounding the 3900x I would really for it to work out but I think I may be going the 9900k route.
OOF, I have never worked with an ASRock board.
What I have tried to do is keep the configuration of the CPU to an absolute minimum in the BIOS.
I had some experience with an MSI board and the LoadLine Calibration was borked on that, to the extent that setting it to anything other than auto caused Windows to go into repair mode.
I will gladly help you and the easiest way would be to join the Discord server where I am mod and you can then talk to me on a voicechat, and I can go through the steps to getting your system stable.
You should be able to get a stable all-core of 4.275 GHz but a lot of things depend on what kind of RAM you are using.
That would be important to know, because that can cause problems.
I found for instance that on my GigaByte motherboards (X470 AURUS Gaming 7 WiFi Rev 1.1 and X570 AURUS XTREME) that setting the XMP profile put me into a world of pain, whereas on an ASUS motherboard that I helped configure, I had to enable D.O.C.P (AMD version of XMP).
I don't get my stuff for free, so I could only experiment with what I have (or that people let me get access to) so just from that standpoint I would be interested in working with you and your ASRock motherboard.
Here is the link to the Discord channel where I am a mod, and my name on the channel is "Michael Nager".
I don't really know the software you are working with and having problems with, but one thing you might want to check into is if those programs have problems above a certain thread count.
What I mean is that it might be getting bogged down with task-switching between threads or thrashing on threads and causing you to have problems. This is easily fixed by turning off SMT and running the 3900X as a straight 12-Core CPU (which is also a lot better for gaming and you can add about 100 MHz onto the clock of the highest value you can reach with 24 Threads).
Here are the stable clockspeeds which you should easily be able to achieve and the Intel equivalent in GHz:
1) 12 Core/24 Thread
a) Ryzen all core clockspeed 4.275 GHz
b) Equivalent Intel all core clockspeed 4.831 GHz
2) 12 Core/12 Thread (SMT Off)
a) Ryzen all core clockspeed 4.375 GHz
b) Equivalent Intel all core clockspeed 4.944 GHz
Without however knowing a bit more about your system I am making educated guesses.
Hi nec_v20, your guide says to set the loadline calibrations to the third highest values. azjazz keeps saying he's using the "third" value. Just to be clear, if there are 8 options (Mode 1 through Mode 8), then the third highest value should be Mode 6, correct?
Also, when you said here "1.425 Volts is way too high", the plan is to change that in Ryzen Master, NOT in the BIOS, correct?
Thank you for all your effort!
dude EVERY MoBo has a different LLC
here a quick LLC-Guide:
set a fixed voltage for your CPU like 1.25V
run CB20 or Prime95
check the drop of voltage - like from 1.25V to 1.06V
set LLC in BIOS
check again drop of voltage - now maybe from 1.25V to 1.12V
set higher LLC in BIOS
check again drop of voltage - now maybe 1.25V to 1.18V
the voltage-drop with high load (should be AVX) should be MINIMAL
BUT voltage should not OVERSHOOT - like from 1.25V to 1.33V etc
Hey NEC, thanks so much for putting this together. I was following along, but like another user with a MSI MEG ACE board, there are a couple things i was unable to find. as Azjazz pointed out, the MSI boards are not labelled the same. Based on his questions to you, i dont see that there was ever a definite response to the point regarding which mode to use. Below is a screen shot to be, as you said use the 3rd highest, but he said he used the 3rd. 3rd highest on the graph is indeed 3, but 3rd highest out of 8 is 6... so some clarity there would be helpful. Im guessing its Mode 3 as azjazz used.
Also, as azjazz mentioned, we have no PWM Phase Control... so im assuming this is not an issue?
I noticed you requested setting the FCLK in 2 places. This was indeed intentional, correct?
Im also curious about why alot of these settings are being made in Ryzen Master, and not directly in bios. You mentioned needing to reload the settings every time windows boots. This could get tedious for some. Are there not settings in the bios for everything set in RM? Isnt RM just a GUI in windows for what bios is doing? Im only 4 days new to Ryzen and this board, but im quite sure I saw ram adjustments and timings in there to be adjusted. The only thing i did specifically go hunting for that were in RM and never found in bios were the following...
... I can find a couple similar fields, but not all of them, none of them written like that, as they appear in RM. Here are some screenshots of all the fields in our bios, if you care to look. Perhaps theres something im overlooking...
I highlighted in yellow all the settings you outlined above.
Also, regarding RM, for the ram settings, which i toggled from Auto to enter the values you suggested from my ram kit, i noticed they were all already set. I dont have XMP enabled, so how did it know those #s...? Or do my #s just happen to be what is considered normal...? Im using G.Skill Ripjaws 3200 64gb. The suggested numbers for the kit were 16-18-18-38.... and those are the #s that were already there.
Im not super impressed with the CBr20 results I had after just adjusting clock and volts, before i found this thread... though Im sure by not adjusting the settings you suggested, thats why I was having problems. The best score i was able to run without a crash was 9533 @4.1ghz, 1.375v!!! That seems ridiculous that a)i could only hit 4.1, and b) i had to use such a high voltage to do so. Though, the temp never went above 70c. But, like alot of people, Im used to doing things the intel way, so this is all new to me. Im not looking to break records with an OC, i just want to know if i got a good chip, a bad chip, or an average chip, and know how much headroom i have in there.
Either way, if you could just confirm those few points, id like to be 100% sure the MSI version of this is ok before I go changing anything. Your comment on bricking the board if making a wrong setting kinda freaked me out!
Thanks again for your input and help!
Since i failed to mention above, ill just put my specs...
MSI MEG ACE
g.skill ripjaws 3200 64gb
I decided to go ahead and apply the settings, and assume that Azjazz's settings for the MEG were correct. However, i made a few changes as I inquired about above. I basically set the ram settings in bios. I did not use XMP, but set it all manually... 3200, dram to 1.35v, and then the timings, 16-18-18-38.
When in RM i set the all core to 42.5, and set the peak volts to 1.3. The stress test completed fine, and then i ran CBr20. I scored 9908, my highest yet, at 76c. So a definite improvement, though my hottest run by 6c, despite the fact that my previous best run was 4.1ghz @ 1.375 volts, 70c. I would have figured the much lower voltage on this run would have kept the temps more in line with 70c, but i guess the additional clock speed brought it up. 76c is still respectable though for an average h100i pro.
I did 2 additional tests. I set the RM peak voltage up just a tad, 1.31v, and 43ghz. That test failed. I thought that extra bit of voltage would get me that extra 50mhz, but i guess not. I then reset back into bios, and tried running a negative offset from the "auto" voltage. This is what iv been doing for the past week since getting the system to keep my temps more reasonable, and not idle at 45 or 50c. Anyway, i used a -0.15v offset which should have brought me to approximate 1.36v. Back in RM, i set it back to 42.5ghz, disabled peak voltage thinking it should now have more than enough voltage to complete the same test it just passed a few min earlier, but it also failed. So with nothing changed aside from having MORE voltage, it couldnt achieve the same 42.5ghz cinebench run.
So why would this be nec_v20?
Im not sure if my ram has anything to do with it, but maybe im at the limit for what i can do because im only using a 3200 kit? It seems most people who know a bit more about this than me are using 3600. I suppose i could run the tests again with all the ram settings back to auto in bios, and adjust them in RM instead, but since i didnt see that the ram speed was being set anywhere in RM, i thought it might fall back to its assumed slower speed of 2133. I figured this would hurt the results, which is why i set it manually in bios.
Im also trying to decide where to leave this for an everyday OC. My whole point with trying to figure this stuff out was to see how much headroom i have with this chip, and where i can leave it for day to day use. Its a workstation, so it needs to be reliable, but also be able to chew thru rendering, 3d modelling, etc. To avoid using RM every boot, could i not set the 42.5 in bios, and the 1.3v as well, and call it a day? However i know that when using voltage override (at least on the MEG), the voltage would be stuck at 1.3v, and not idle down. Perhaps this is why you're doing it in RM? So you can use the PEAK voltage limit, but still make use of the ability for the system to idle down? Im not sure why this is not built into bios. My old Asus intel board that just died last week had this feature, so i was able to dial in a higher OC, but still keep nice low idle.
azjazz, i noticed above you mentioned having to set some specific settings to STOP your idle temps from bouncing around all the time. I have the same issue, though after setting my negative offset, the values are more reasonable. As i said above, the first time i ran the system the temps were bouncing around between high 30s and mid 50s. Thats what started me on this "quest" to figure out what was up with the Ryzen. With my offset i now get high 20s to high 30s at idle, but as you said it bounces around like every few seconds. Did you find a way to settle this down a bit so the voltage is more stable?
What LLC are you running on your motherboard?
The problem that I found setting everything in the BIOS the way I had it configured in Ryzen Master resulted Windows coming up totally corrupted and my having to go into repair mode.
The reason for this is that the BIOS is just the motherboard manufacturers best guess at what the AGESA is doing. The AGESA - the actual BIOS is supplied by AMD as a binary, meaning that the motherboard manufacturers cannot adapt it to their own desires.
If you don't want to use Ryzen Master then the best of luck to you; but don't expect it to be stable.
Also on my motherboard, setting the values in BIOS will completely negate the CPU clocking down when not under load; meaning that whatever you set is what you get - all the time.
Using Ryzen Master, most of my cores go to sleep when I am on idle.
The bottom line is that if you don't want to us Ryzen Master then you have to be prepared to settle for foul compromises.
For me this meant that although using Ryzen Master was - and is - a PITA, is is far better than any alternatives I have tried out in my eight months of experimentation.
nec_v20, thanks for the reply.
as i said above, after reading azjazz's posts, i decided to go with LLC mode 3.... so the 3rd from the top of that little graph in the image i attached above.
thanks for the clarification on the CPU clocking down if its set in the bios. thats what i experienced as well, at least as far as setting the voltage there. thats realy unfortunate that setting it locks it in like that and doesnt let it adapt as the load requires. maybe that will be fixed in a future bios update?
i think for now ill just keep things as they are as per you BIOS setup instructions, along with my slight undervolt. im sure the CPU will be great anyway, compared to the one im coming from, so even if its not overclocked it will still be an improvement for me.
can you offer any insight to that various results and tests i made (noted above) and why higher voltages caused crashed at lower all-core settings? again, i think my best was 4.1 before i tried your method and got 42.5.