Question Ryzen 7 3700x Stuck on 1.4Volts and 4.4ghz on idle

[NeutronG]

Hi guys this is my first post on this and also my first ryzen build.

I have just built my first ryzen build;
Ryzen 7 3700x
MSI b450 Gaming Pro Carbon AC
Gskill Trident Z Neo 3600mhz cl16 (2 x 16gb)
NZXT Karken x53
MSI 2060 Super Gaming X

Ive noticed that on idle i am stuck at 4.4ghz and 1.4volts which was concerning and my idle temps were around 45-50 degrees. Under load it drops the clock and volatge which i assume is normal but the idle clock and manily the cpu voltage has me really concerned. After alot of ready and research i can to the conclusion that this was not good. Ive literally done everything and still have the same issue.

All the thread that i have read;
https://forums.tomshardware.com/threads/ryzen-7-3700x-4-4ghz-all-core-oc-safe.3559320/
View: https://www.reddit.com/r/Amd/comments/cbls9g/the_final_word_on_idle_voltages_for_3rd_gen_ryzen/?utm_source=amp&utm_medium=&utm_content=post_body

and many others i cant find anymore lol.

So i decided to undervolt and keep my clock at 4.4ghz and cpu core voltage at 1.375. This produced better thermals (idle 35-45) and a better cinebench score. I then dropped to 1.35 volts which was was fine but cinebench threw an application error, so after a reinstall and restart it was all good. I them decided to drop it to 1.325 volts which was all good and now thermals was 35-40 degrees on idle but cinebench threw the same application error. I reinstalled and restarted still same error. I dont know if that means its not stable because i ran aida 64 for about 15ish minutes and it was fine no issues.

I am now currently at 4.4ghz at 1.275volts with no crashes or bluescreens but i cant run cinebench because it keeps throwing the application error. Again, i dont know if this is an instability error or just the program.

If anyone knows whats going on and how to fix this i would greatly appreciate it.

 

[NeutronG]

UPDATE - I found out that icue was preventing my cpu from dropping voltage during idle. So i went back to a stock bios setting with xmp enabled and tested the performance.

Cinebench loaded and ran fine with no errors but my score wasnt as high. The same goes with gaming. I noticed around a 10-15 frame drop from the 4.4ghz at 1.35volts.

Will continue testing but if anyone has a better idea please let me know.

 

[angvil]

UPDATE - I found out that icue was preventing my cpu from dropping voltage during idle. So i went back to a stock bios setting with xmp enabled and tested the performance.

Cinebench loaded and ran fine with no errors but my score wasnt as high. The same goes with gaming. I noticed around a 10-15 frame drop from the 4.4ghz at 1.35volts.

Will continue testing but if anyone has a better idea please let me know.

Any updates bud? I am facing literally the same situation you did and I did exactly as you did and I still feel the pc kinda wonky did you run into any solution?

 

[SteveRX4]

My understanding that at idle, the CPU really only runs on one or two cores and so a high voltage results in a low wattage and therefore low temps. As soon as you run cinebench all cores get stressed. If the voltage is too low, it will crash.
I mostly run PBO+200Mhz with XMP. I only manually OC the CPU for benching. I test the manual OCs with CB20. If it crashes, I up the voltage - to a max of 1.4V. I run 4.4GHz at 1.38V.

 

[drea.drechsler]

Any updates bud? I am facing literally the same situation you did and I did exactly as you did and I still feel the pc kinda wonky did you run into any solution?

First, just a bit of advice: this is an old thread you really should have started your own with the same question.

If you're also running at 4.4G and 1.375V fixed, that's a dangerous setup for a 3700X with the 7nm manufacturing process. It will degrade early, as many over-clockers have found.

The best (easy) thing for Ryzen is to run with voltage and clocks in AUTO, letting the CPU manage itself. It will keep itself safe and it will still give very high performance when needed. In effect, it overclocks the CPU at every opportunity and then pulls the overclock back when it gets too hot or it thinks motherboard could be lacking available power.

Ryzen's boosting algorithm is very well refined, it may use a high voltage (up to 1.5V even) but only in lightly threaded bursty workloads to boost a single core up to max boost clocks...and very briefly. It will pull back the voltage, and the boost clocks, as workload gets increasingly heavy. Just let it do it's thing, it's actually safer. And fully warrantied--we really should stop trying to second-guess the army of very smart engineers (watched over by hordes of nervous accountants worrying over warranty claims) that make these things.

Overclockers are trying to run with a fixed clock and as low a voltage as possible...recommended is no more than 1.2V, but some are using 1.275 as the limit. Newer 3700X's should be game-stable at about 4.3Ghz at around 1.275V. That's game stable, not Prime 95 small FFT stable.

4.3Ghz is actually a pretty good overclock when you consider the BASE clock for a 3700X is 3.6Ghz...that's the rated speed at an extreme heavy load. But in use, if you run benchmarks (Cinebench 20 is preferred, repeatable and easy) of both multi-threaded and single threaded you'll probably find that it gains very little, or even loses, performance vs. just letting the algorithm control the whole thing. You'll also find that heat output and power consumption rises markedly since it's always at a high clock, even when just idling about doing nothing.\

That said, the best, not so easy, thing to do is get much better cooling for CPU, enable PBO and tweak it. PBO still uses the processor's boosting algorithm to keep the processor safe. What it does is tweak the boosting parameters so that it will keep the processor boosting, longer, at higher clocks so long as temperature allows. The much better cooling means it will have a lot more thermal margin to work with (the algorithm pulls clocks when it see temperatures rising) so why not use it?

With a properly tweaked PBO you should be able to get similar multi-threaded performance and better lightly threaded performance as a (reasonable) fixed voltage overclock without the risk of degradation. Remember, the algorithm is still working to protect the CPU as it heats up in use. And this setup, done right, will even be Prime95 small FFT stable.

 

[angvil]

First, just a bit of advice: this is an old thread you really should have started your own with the same question.

If you're also running at 4.4G and 1.375V fixed, that's a dangerous setup for a 3700X with the 7nm manufacturing process. It will degrade early, as many over-clockers have found.

The best (easy) thing for Ryzen is to run with voltage and clocks in AUTO, letting the CPU manage itself. It will keep itself safe and it will still give very high performance when needed. In effect, it overclocks the CPU at every opportunity and then pulls the overclock back when it gets too hot or it thinks motherboard could be lacking available power.

Ryzen's boosting algorithm is very well refined, it may use a high voltage (up to 1.5V even) but only in lightly threaded bursty workloads to boost a single core up to max boost clocks...and very briefly. It will pull back the voltage, and the boost clocks, as workload gets increasingly heavy. Just let it do it's thing, it's actually safer. And fully warrantied--we really should stop trying to second-guess the army of very smart engineers (watched over by hordes of nervous accountants worrying over warranty claims) that make these things.

Overclockers are trying to run with a fixed clock and as low a voltage as possible...recommended is no more than 1.2V, but some are using 1.275 as the limit. Newer 3700X's should be game-stable at about 4.3Ghz at around 1.275V. That's game stable, not Prime 95 small FFT stable.

4.3Ghz is actually a pretty good overclock when you consider the BASE clock for a 3700X is 3.6Ghz...that's the rated speed at an extreme heavy load. But in use, if you run benchmarks (Cinebench 20 is preferred, repeatable and easy) of both multi-threaded and single threaded you'll probably find that it gains very little, or even loses, performance vs. just letting the algorithm control the whole thing. You'll also find that heat output and power consumption rises markedly since it's always at a high clock, even when just idling about doing nothing.\

That said, the best, not so easy, thing to do is get much better cooling for CPU, enable PBO and tweak it. PBO still uses the processor's boosting algorithm to keep the processor safe. What it does is tweak the boosting parameters so that it will keep the processor boosting, longer, at higher clocks so long as temperature allows. The much better cooling means it will have a lot more thermal margin to work with (the algorithm pulls clocks when it see temperatures rising) so why not use it?

With a properly tweaked PBO you should be able to get similar multi-threaded performance and better lightly threaded performance as a (reasonable) fixed voltage overclock without the risk of degradation. Remember, the algorithm is still working to protect the CPU as it heats up in use. And this setup, done right, will even be Prime95 small FFT stable.

Thank you so much for all this information, Since I've posted here I have been doing the PBO and I managed to OC stable with prime95 small fft stability my findings below

VCORE 1.25
CPU Clock: 4.3
LLC level 3

and everything has been stable for 2 days of stress testing I think I can lower the Voltage to 1.180 on a 4.2 I've done that and stress tested it for 5 hours with no crashes and tbh I didn't notice a big difference in performance between them now my thermals never go above 60c and I'm using an X62 NZXT AIO with a well ventilated case + 5 fans not counting the radiator fans.

I also did some testing with the CTR program but it was just to get the most suitable clocks and volts and I manually tuned things in Ryzen Master afterwards


I hope everyone gets lucky as I did! I'll now tackle the long long long process of tightening my RAM cas any tips on that sir?

I am using HyperX Fury 16GB RGB 3733 MHz DDR4 Memory currently xmp'd to 3600 on CAS 18

 

[SteveRX4]

I hope you have better luck than I did with Typhoon and Ryzen Dram Calculator coming up with a substanially better result. I've got 3800/18 LPX and my CB20 score improved by only 15 points. Every now and then I got hiccups. So I ended up just sticking with XMP.

 

[CountMike]

I hope you have better luck than I did with Typhoon and Ryzen Dram Calculator coming up with a substanially better result. I've got 3800/18 LPX and my CB20 score improved by only 15 points. Every now and then I got hiccups. So I ended up just sticking with XMP.

CD20 is not influenced by RAM speed much.

 

[SteveRX4]

Buildzoid says to run Ryzen with PBO and XMP. I tend to agree. Less buggy and still works OK.

 

[drea.drechsler]

Buildzoid says to run Ryzen with PBO and XMP. I tend to agree. Less buggy and still works OK.

I'm running PBO on my 3700x and it works terrific. Most people set a low OC and voltage and run CB20 in multithread and think it's great...little or no loss of performance. But if they would take the time to run CB20 in singlethread they'd find they've badly hurt light threaded performance because the CPU can no longer boost to clocks that are a lot higher than what they settled on.

Light threaded performance is what's needed when gaming because even multi-threaded games only hit one thread heavily. In gaming action, one core is constantly boosting to 4400 on my system so being held back to 4200 or so (the best OC it can muster at reasonable voltage) would be a big FPS hit.

If the desire is just to 'undervolt' on the idea that will stress their processor less then even that's misguided. When set up right, even with PBO, the processor will pull clocks, and voltage, way back much further than their fixed clocks/voltage much of the time.

Just set up a graph and monitor voltages with HWInfo64 to prove it to yourself. I did and even when running a heavy rendering task I'm amazed to see how much of the time is spent at very low voltages as the algorithm finds time to 'sleep' the core. The average core voltage through that period is much less than I could set in any stable 'overclock'. Higher fixed voltage, even low ones, drive a higher core current all the time. It's that core current that will degrade the processor faster since it can't lower voltage...and core current...when it's managing itself. That's the way electron migration works.