help understanding offset voltage with ryzen 5 1600

[serpentracer]

hey guys I have a newbie question and I'm sure it's been beaten to death so bare with me. but I'm having a issue (I think). I did search but came up without much to help me with my issue. first off I'm not new to computers but I am new to OCing one. I've never bothered to try one before and it's a lot of info to digest in a few days. so I gave up and decided to sign up and ask you guys. I've been trying to follow and understand a few guides on youtube from this site and some written articles on this site. but my pc doesn't seem to be responding the way the guides are showing. I think i just cant wrap my head around it all yet.

my motherboard is a asus ROG strix b350-f. and a ryzen 5 1600 I have it running at 3.8 Ghz. I'm running Linux Mint 18.3. so I know that throws a bit of complication in here for some of you. I cannot run the normal software tests you guys do. and I cannot run the overclocking software from Asus. I have to make all of the adjustments in the UEFI settings that I do not understand yet and none of the tutorials delve into. this mobo seems to have settings that the older videos and tutorials do not have at the time they were created.

so for testing, I found a site called CPUX.NET to run online cpu benchmark and stress tests. I'm not too concerned with my score as I am what happens with different settings. something isn't working right I don't think.

the guides show a recommended setting on the cpu voltage to be 1.3 to 1.4 for this processor. set to auto it is at 1.38 volts. but it runs a lot slower with that high voltage on that test site. I get a score on that site around 330 out of 669. a score of 12519 hashes processed at a speed of 39 hashes per second. but if I lower the cpu voltage to 1.243 and use + .00623 offset(the lowest this mobo allows) for a total 1.24923v It decreases (lower number = better results) my score to 79/667. my hashes score jumps to 30072 and a speed of 93 hashes per second. this is a significant increase in performance. if I raise the offset more and more the score drops and drops indicating a drop in cpu performance. it seems to be working backward to every tutorial on oc I can find. everything set to auto is also very slow and raises the score over 300.
the lower voltage also makes my psu(corsair cx750m) give out a squeal as it boots up. It almost sounds like a cd being read by a cd rom but much quieter. it sounds like coil whine? but if I increase the cpu voltage offset it goes away. I forgot to mention none of these tests increase my cpu temp beyond 136F (58C) and I stress tested it for 1.5 hrs at 64 threads and 100% power. while I was watching youtube videos at 1080p the entire time. never a hiccup or a glitch. temps remained around 134F (56 to 57C) I don't think that's too bad considering how hot some of my past computers ran at that were much slower.

I did run their stress test on many different voltage settings and it never crashed my pc.

so yeah, I have no idea if my results are correct or what, is that test site wrong possibly? please give me you opinions and or post your settings to try. I don't want to burn this thing up. and I'd like to not hear the psu squeal like that. and I didn't buy a ryzen 5 1600 to see it score lower than a athlon 64 X2 with the auto settings of the mobo. seriously, some of the cpu test software available for linux groups it with 15 year old celeron processors set on the auto settings. that's why I decided to give the OC stuff a try.

 

[trents]

Let me suggest installing Windows temporarily until you get your overclock ironed out. You can create a free Windows 10 USB installer by going to this website: https://www.microsoft.com/en-us/software-download/windows10

Format a 16gb USB stick in FAT 32, copy the Creation Tool executable file to the USB stick and execute the installer from the USB stick, rather from the download folder. This eliminates some problems some are currently having in getting the Creation Tool to work. You do not need a product key to install Windows 10. When you get to that point in the installation just click the choice that says "I don't have a product key". You just can't activate it, that's all and eventually you will get a water mark in the Windows 10 user interface. After you are done with it you can delete it from the hard drive. If you have a spare hard drive available to do the Windows installation you could swap in in so as not to have to wipe your Linux system drive.

It is just very difficult to overclock from Linux. Good tools we use and need are just missing. I'm also not sure that Linux monitoring tools report temps and voltage accurately with AMD CPUs. For that matter, some traditional Windows voltage/temp monitoring tools do not either. AMD's own "Ryzenmaster" is best for that but there is no Linux version.

 

[chrisjames61]

As Trents said, it is a lot easier to work with overclocking with a Windows boot disk even if it is only temporary. Plus it doesn't cost anything. You need CPU-Z, Hardware Monitor, maybe Thaiphoon Burner and Ryzen DRAM Calculator. Prime95. There are workarounds in Linux but it is much more difficult using the CLI.

 

[Mandrake4565]

I have to agree with the gents above. It certainly would make things easier.

 

[ShrimpBrime]

Firstly you may be running a test that is simply inconsistent and may give different results even if you didn't change anything. You also must be sure your monitoring is accurate to your settings in bios.

Any how a quick search shows you can OC from linux following this guide. https://forum.level1techs.com/t/overclock-your-ryzen-cpu-from-linux/126025

If your PC doesn't have issues at a lower voltage, then use the lower voltage. So if you can run 3.8ghz at 1.3v or less, then there should be no issues.

 

[Johan45]

If the OP only runs Linux and can monitor the temps then he can use Prime95 AFIK for stress testing.
I'm with Shrimy here, we know nothing about this benchmark you're using so we can't comment on the results but Prime95 is a very common stressing tool. You can run small fft at 128KB for testing the CPU and then custom blend with ~ 75% of you RAM capacity which will test the whole CPU and memory quite well.
Ryzen performance is also increased with RAM over 2666 MHz, for most if you can manage 3200 MHz is the sweet spot.
I would also recommend putting your rig specs in a signature so we know what you're working with be specific too.

 

[serpentracer]

alright I'll install windows on another drive to give it a whirl. but I was more interested in why does the performance drop with a higher voltage when all tutorials say or imply it will increase in performance with more voltage. and it has to have more the higher the speed goes.
mine is going down.

the site is https://cpux.net/ try if out for yourselves and see what you think.

 

[Johan45]

alright I'll install windows on another drive to give it a whirl. but I was more interested in why does the performance drop with a higher voltage when all tutorials say or imply it will increase in performance with more voltage. and it has to have more the higher the speed goes.
mine is going down.

the site is https://cpux.net/ try if out for yourselves and see what you think.

As I said you don't need to install Windows, Linux should work just fine. I am curious about your monitoring and if it's picking up the right temps. If not your lower performance could be due to throttling.
If yes it's possible that left on auto with an offset it's going higher than the 1.38V you set. Ryzen CPUs have a very large V_Droop and if the CPU wasn't getting enough juice it will also affect performance.

 

[trents]

Throttling and inaccurate temp reporting was my concern as well. That's why I was suggesting in the first place to temporarily install Windows so he could use monitoring utilities that can be trusted.

 

[serpentracer]

As I said you don't need to install Windows, Linux should work just fine. I am curious about your monitoring and if it's picking up the right temps. If not your lower performance could be due to throttling.
If yes it's possible that left on auto with an offset it's going higher than the 1.38V you set. Ryzen CPUs have a very large V_Droop and if the CPU wasn't getting enough juice it will also affect performance.

maybe I'm not explaining it right. maybe it's linux I don't know. so I'll do the windows route before I get into the whats going on with this bleeping thing discussion.
as far as I know the linux kernel I'm using is appropriate for the Ryzen. I'm currently using 4.15.0-30 on a linux mint 18.3 system. I'm not a linux expert by a long shot. I couldn't understand the instructions on the link shrimpbrime posted. as far as installing prime95 it's a .tar file. and I just don't understand how to install those even with the instructions, it's a bit over my head. I am using Psensor for the temp monitoring and I-nex is a linux version of CPU-Z.

anywho, I'm getting ready to install win 10 after it downloads.

 

[serpentracer]

omg i'm about to vomit. I forgot how awful microsoft is. win 10 is way too invasive into privacy. needs me to create a microsoft account? wth? this is messed up! can't wait to wipe this drive clean

 

[trents]

No. You do not have to create a Microsoft account. Read the small print and create a local account instead.

 

[serpentracer]

ok guys I'm logged in from win 10 and I installed cpu-z some of the utilities from the cd that came with my mobo. what do you need from me? so far I can see everything being reported was the exact same as what linux reported. same temps, same voltages etc.

what tests should I run?

 

[serpentracer]

using that site I got the worst score to date using asus's tuning suite. score was 100 points higher (400) using the bios is way better. I can get it down to 89.
asus has it running at 4.1ghz with 1.8 v and it scored the lowest. I have it at 3.8 ghz and 1.2v and it's significantly faster.

I also ran the tests in cup-z with the same results. what is going on?

 

[serpentracer]

ok here are some official results. i ran cinebench in the top pic at what the Asus ROG OC software set my cpu at. 4.10 ghz and 1.38V.
the second screenshot is what I set in the bios. 3.8 ghz with an offset voltage of .00625. so it is running around 1.24V while testing. look at the incredible increase in performance. it's beating intel i7's. with ease. and it doesn't crash the system while never reaching over 55C temps with an air cooler. I don't know how this is possible. I either have the baddest ryzen 5 1600 ever made or something is wrong with the whole OC theories.

 

[serpentracer]

 

[trents]

Run the Realbench stress test. Looks like in the pic you are running the benchmark. If you can pass four hours of the Reabench stress test and three hours of OCCT you should be stable. Both have built-in temp monitors and OCCT also monitors voltages.

 

[Johan45]

Yeah, AISuite is not a great OC utility. That first CB15 pic is terrible. Almost as if the machine was throttling. The second pic looks right about where it should be.
A good monitoring SW is HWInfo64, they have been right on top of AMD CPU changes so quite reliable. I typiclly use Prime95 for my stress testing. Small FFT set to 128 KB for CPU and blend custom with ~ 75% of RAM selected.

 

[serpentracer]

so getting back to my original question since we can rule out linux being a hindrance.
are you guys familiar with the ROG Strix b350-f UEFI settings? there's quite a lot and my goal is to keep the cpu performance since it doesn't seem to crash the system and temps remain well controlled. but I'm concerned with the sound the power supply gives off on this setting.
I had an older no name brand power supply that made this noise. it never died but it's annoying when I have the tower on the desk next to me. I can hear it over the fans and it has all the panels installed. it's quiet with the power cranked up higher though. but the cpu performs slower.

you guys seem well versed in this stuff. I have no clue what I'm doing lol I just know one way seems to work great and one doesnt. but the one that doesn't seems to be the settings most people use.

 

[Johan45]

I don't think you're going to be able to avoid it. The reason one doesn't make the noise and performs worse is because it's not stressing the PSU. If it's that bad and fairly new (within warranty) see if you can get it exchanged.