eikon - The MSI LLC Mode 3 seemed to fit (my interpretation of) nec_v20's LLC goal: The optimal CPU Vcore voltage is being set to 1.3 VDC Max, and the CPU Vcore LLC selection was to prevent the BIOS from modifying Vcore under CPU load.
If you look at the tiny graph on the right side, the Mode 3 graph is a straight line, meaning that the CPU voltage stays flat and doesn't get modified from its setting based on CPU loading. That's why I selected it.
There wasn't a similar graph for the SoC LLC, so it was my assumption that the SoC LLC Modes followed the CPU Vcore LLC curves.
For the LLC it would be Mode 3 which is the 3rd highest (it is the opposite way around on my board).
I set the FCLK because anything above setting the RAM speed to 1800 data rate (or 3600 clockspeed) will not automatically be changed with regard to the FCLK. If you stay at 1800 or below for RAM then the FCLK will automatically be adjusted.
For your RAM you should set it to 1600 in Ryzen Master.
The thing is that one cannot set the values that are set in Ryzen Master in the BIOS without corrupting your Windows installation.
I thought this was only my system, but in a video from JayzTwoCents he mentioned the same problem with various motherboards with different 3rd Gen Ryzen CPUs when he tried to clock them above 4.2 GHz or so in the BIOS.
I can clock my 3950X to 4.3 GHz all core at 1.3 Volts and my crappy 3600X (which I got at launch) would run at 4.225 GHz at 1.3 Volts.
On the MSI motherboard you have to set the XMP profile (which I didn't need to do on my motherboard).
These are minor variations and you should not set any voltages or anything else in the BIOS.
If you try to mix and match BIOS setting with Ryzen Master setting you will end up in a World of Pain.
thanks again for clearing up those points. i find it so odd that setting too high of a boost would corrupt a windows installation. im not saying your wrong (id rather not find out the hard way!), but they dont even seem related!
as i said above somewhere, i did input some times in bios. since running the test with your settings, i havent messed with it further. i think the only thing i tried to do was use the Ryzen DRAM calculator to input proper ram settings. I couldnt get it to boot with those #s.... but perhaps i failed to input a few. Havent tried since.
Its currently running stably with my offset voltage, which should be keeping it right around 1.3ish, and boost is just set to auto. I still do find it runs hot when i work (now that iv actually had time to put some of my normal work tasks through it), but not stupid hot. It will idle at 30 if im doing nothing, so thats great. But even minor loading over an extended time will cause it to get up to mid to high 40s, maybe even to 50. And when i say minor, i mean task manager is showing me 1-5% load, as i do my work.
Maybe once this project im working on is done, ill have another go at it, and try to more fully utilize RM as you suggested, but its not the kinda thing i want to mess around with, potentially adding system instability, while im in the middle of a paid project.
Do you think that these settings will ever make their way to the bios so we dont need to keep using RM? Even the auto profile loading in RM would be helpful. Im sure you follow this stuff much more than I do... has there been any talk of that?
Thanks again for the replies.
eikon - Your Windows installation can get corrupted by improper RAM or bus speed settings. For one thing, Windows isn't a read-only application, it is dynamic. There are Windows configuration settings that get written to every time you boot your PC (think, Windows registry here). If your RAM or data bus are running just slightly on the wrong side of right, data can start getting corrupted on your hard drive without your knowledge. You may not see a hard BSOD failure at first, but you could corrupt Windows without realizing it until it is not recoverable.
Case in point: Going back a decade or so, lots of people ran disk defragmenters. I never did. Disk defragmenters provided very minimal real-life performance improvements for even highly fragmented disks. However, whenever you ran a disk defragmenter, a large percentage of the files on your hard drive were passed through RAM before they were re-written to another block of your hard drive. If your RAM was marginally flaky back then (not uncommon), you were corrupting system files by running defragmenters. I had a friend who asked me to check their PC that was slowly getting less stable with more frequent BSODs. They had a daily defragmenter running in the background, and when I checked their RAM in MemTest86, it failed some of the memory tests. We replaced the RAM in his system, reinstalled Windows, and removed his daily defragmenter. I never heard any further complaints from him about his Windows failing.
hmm, interesting. i didnt know that, but i suppose it makes sense. i was also 1 of those people who avoided defragging. now that you've made me a bit paranoid, i think ill go build a windows disc image incase i need to restore
A small side-nugget, for anyone who is interested. I nailed down one application / service that was causing a lot of background processor churning. The app was causing my 3900X to run an average of 4 degC warmer during idle, even though I was never purposefully using the app/service.
That app? The Nvidia GeForce Experience. I don't use the streaming to my Nvidia Shield / main TV as much as I thought I would.
My 3900X idle temp was around 36-37 DegC with GeForce Experience installed, and my CPU temps are now back down to 32-33 DegC after removal.
You state PWM Power Phase Control, but in my Strix X570-E Gaming motherboard, I've got CPU and SOC, but nothing specifically stating PWM. I assume PWM Power Phase Control is in regards to CPU and not SOC?
That is something specific to my GigaByte motherboard.
My bad, sorry about that.
There might be something similar to this with ASUS motherboards with a different name.
Oh no, it's all good, man! Can't expect you to look up every single motherboard possibility. What you've already done is awesome as it is. I found PWM Power Phase Control specifically means CPU Power Phase, from what I could find anyway.
- Gigabyte x570 xTreme
- 64Gb G.Skill Trident Neo 3600
- Ryzen 9 3950x
- 2 x Aorus NVME M.2 1Tb SSD - slot 1 and 2
nec_v20 Hi Michael,
I have followed, to the T, your instructions for right-clocking the 3950x using Bios/Ryzen Master. We have an identical MB/CPU configuration.
I've tried several times but when Ryzen Master "Restarts" my system after "Apply/Test," my system doesn't post... zero, zilch, nada... just sits silent forever. The power switch doesn't even work with long-press.
So, I turn the system off with power supply switch, unplug it, press the Clear CMOS button, and reboot. It posts and boots windows. Then I restart Windows and go into the bios and reset things back to how they were before the experiment.
I then uninstall Ryzen Master and re-install it before trying again. But, nothing. Rinse and repeat.
Can you help me get this working? I've joined Discord and sent you a friend request. My profile name there is JB7#2231.
Thank you for your help.