ClockTuner for Ryzen (CTR) - and introduction guide and download

don't start the test at 1225, start the test at reference test voltage 1075 and it will only take 1-2 step to find out how low it can go for your chip. but it shouldn't restart on diagnostic with pbo turn off, you're not boosting past stock frequency

 

ram isnt the issue, I've tried auto 2133, XMP3200, 3600, 3800 they are all the same result. my ram is tested stable in 1susmus's dram calculator's memtest 200% 4 times. the program runs fine now and is able to finish stress test/START test/Diagnostic because I've finally figure out my FIT voltage, you can't really use this program and not know how to find that because you think its okay to pump 1.25v because the program said so, but my FIT is actually 1.212 on my 3900x so I have to use 1.225v as reference voltage to start test, and that with LLC3/Extreme power phase let me droop 1% to 1.196-1.212 which is exactly where my FIT is. I've finally fine tune my settings, manual OC works fine, program also works fine now. My ram is very fine tuned with low voltage, as I've taken this into consideration because 1susmus did say not to use seriously overclocked ram because of voltage instability, but then whats the point if manual OC let me use my ram to full potential but this program limits it? I've since done 6hour+ straight AIDA64 stress test that hits harder than the CTR stress test and its fine so far, but then I fail after like 10 runs of cinebench which makes no sense lol. None the less the issue is pretty much fixed so far so gotta figure out why cinebench is crashing after like the 10-11th run.
Aida64 test runs fine, no crash so system is definitely stable for gaming / everyday use. Currently have my 3900x running at 1.225v 4.3x CCX1/2 4.35x CCX3/4, avg 4.316ghz FLCK1900 1:1 spread spectrum on so its 99.8 BUS clock instead of 100, lowering the overall clock by couple % for more stability.

 

PBO was turned off of course. Test starts at 1250 mV, as this is the default value. I don't know what could be the problem.

 

I just said, do not start the diagnostic test at 1.25mv there is 2 reference number, the top one is for when you do "START" test, the bottom one is for CHECK REFERENCE VOLTAGE for Diagnostic. if you use 1.070 to start, it will tell you very fast if your core is stable at that voltage or not, if you restart means you got some other setting set wrong, I can't help you since you didn't list your entire bio's settings, can't really help unless we know all that, but every board is different so gotta find out the setting and change one at a time and to test. If you don't want to do that then gotta learn how to manually overclock and not rely on the program, which is harder. I've learn over the last 1.5 weeks while the program does a lot for you, it does not do everything for you, you still have to find out limitation of your chip and you can't just punch in the voltage it gives you and expect it to work. You still need to know about overclocking, if you don't already know the basics or some of the more advanced things, the program will not hold your hand, you will crash if you don't have the proper voltage settings. Your chip's FIT voltage, Max Idle core voltage, how much LLC you have, what droop %, will it droop to your FIT voltage at your current LLC with your core voltage, all these affect if the program will work for you or not.

 

Hey Yuri,
Found the issue which is causing random reboots on Gigabyte boards.
I have a Master and I'm suffering the same issue with my tool, even worse since I change it dynamically.

It's not about the VRM quality, it's a longtime bug in all GB BIOS.

The board just fails to set the correct VID; you change it in the BIOS and at first boot it's not going to be set, you have to reboot twice.
This bug is there from many years both on Intel and AMD platforms. They never fixed it.

Same happens when you set the VID on Ryzen via the SMU; the board is just not doing it 50% of the times.
Not sure if it's a stable fix but the workaround is to set twice the VID with 50ms sleep between the commands.

Probably not going to harm doing it with any other proper board but it's needed with Gigabyte.
Shameful.

 

You were right, but it looks like LLC 3 wasn't enough. I set it to LLC 4, and got instant results using Diagnostic, with no crashing.

 

Does CTR have to be loaded, in the tray (minimized) for it to work, or can I just tick the thing that says Load with OS?

 

This would certainly help explain a lot if it turns out to be true.

I have definitely seen issues with my gigiabyte mobo when changing settings in BIOS not taking properly. If I Make certain changes in BIOS and then do F10 to save and exit, after it does it's warm reboot I see performance loss if I log straight into Windows and run a test. If I do the F10 to save and exit, let it boot to logon screen but then shut the PC down and cold boot it I don't see the same issue.

Interesting to hear this is VID related, I hadn't yet bothered tying it down to one setting, as that would explain the perf drop is probably due to clock stretching because of low power to the CPU.

 

I feel stupid but after I successfully ran CTR and it created a profile - what do I do to activate this profile? When I check CPU-Z or play a game, all clocks are default.

 

Check this post by the author: https://twitter.com/1usmus/status/1318606670055002112

 

make sure you arent tightening your LLC too much compare to your core voltage, you want it to droop to or just 5mV above your FIT voltage while doing stress test or you might be above your FIT voltage and damage your chip(most likely won't at this low voltage its using to test, but I don't know I'm not an expert, I just know the safe values for chips)

 

Im also having problem with the profile not saving, tried everything, restarting computer does nothing, tried having it in a folder on the desktop, tried having it out of the folder with all the files on desktop nothing works lol. Profile > update > save > apply or apply > save > update > exit CTR does nothing

 

Just a word of caution guys. I have F’ed my CPU on this ver. I had perfectly functional 3 days old R-3600 on default setting (non OC) on MSI B550M mortar board. Ran beta 7 and while running during the diag stage had a complete black screen crash. Since then I have a red CPU debug light on mobo. I did follow all the instruction to the “T” including software and BISO settings. During the test nothing was alarming only ting I noticed that the CPU voltage was shooting 1.5+Volts but before anyting my CPU gave up it seems. Now I am going CPU shopping.

 

really wish you guys stop saying these things when you probably have no knowledge of overclock prior to using this program, and you probably killed your own chip because you don't know what the FIT voltage is.
Even then its kinda hard to kill a chip in 3 days if you only use this program for not even 1.
were you even tracking the program with HWinfo open at the same time for comparison?

 

So really silly question. Just got my first AMD Board, an MSI X570 Tomahawk, with a Ryzen 3600xt.
I went into the bios and set the LLC to Level 3, and double checked, but all my testes are always 2.5 -> 3% vDroop, and it always tells me to improve voltage compensation.

I thought the MSI Tomahawk was suppose to have good VRM's, but maybe I am messing something up here?

Edit; Seems the beta Bios fixed it.
Apparently a bug with MSI boards where LLC isn't always applied from BIOS correctly

 

So what do you recommend, what do I need to change and where? (Mobo is an MSI MPG X570 GAMING PLUS)

 

Hmm, I actually feel sorry for this guy.
There were a lot of complaints already anyways and for a simple diagnostic run, and/or some tests you shouldn't be damaging your hardware.
Well, at least never had any kind of problems so far with manual oc or with ryzen master and older stuff in the past.
With ctr unfortunately I had my share of hard resets a couple of times with the beta6 release (only).
In some cases people indeed should be more aware of what they are doing, but let's not forget that the apps should be safe as well, otherwise what's the point in using them.

 

Before going overdrive, clear CMOS and try again. It happened to me a few times my system would not post due to RAM issues and resetting BIOS to defaults helped. Perhaps you are lucky.

 

Concerning the red LED and the "dead" processor.
It is enough to reflash the BIOS to return the SMU integrity. In rare cases it is required to reflash the BIOS with afuefix and /CLRCFG key to reset all SMU settings to default.

For 7 months of persistent tests I had thousands of errors, had 1.55v in load and none of the processors has degraded. There's no reason to worry, otherwise I wouldn't have released CTR, as it's a huge responsibility.

You are absolutely right. Double VID "installation" is implemented in beta 7. This is the only chance to avoid setting the frequency for the VID in "transitional" state. However, it doesn't solve the problem or it does solve part of it. Low level frequency and voltage control contains a huge number of errors and does not contain any protection mechanisms (by the way, this is the main reason why the SDK with CLI version was withdrawn). Thank you very much for doing your experiments and telling me. I can't do it without you all.

 

I am happy to see enthusiasts in this topic. You are thinking in the right direction. I have ideas about FIT and diagnostics. In a few weeks I hope to get a satisfactory result to tell the public about it. Thank you.