Ryzen 5 1600 (non-"x") OC settings reality check

[frizzlebyte]

Hey guys,

Just tried overclocking my Ryzen 5 1600 proc for the first time today. I am using an ASRock B350 Pro4 with the latest BIOS, and DDR-2400 RAM.

I have been deliberately conservative with my settings, and have achieved 3.7ghz on the default voltage setting, 1.2375v. I tested Prime95 for an hour, and things seem to be going well so far. Granted, I should run Prime95 for longer, but this is more stable than I expected for a first attempt.

My question is really related to the voltage, namely, does that seem like a reasonable overclock to get out of the default voltage? It surprised me that I didn't have to increase the voltage AT ALL to get a fairly stable system, and the highest temps in Prime95 were 77.4 degrees Celsius.

Is this reasonable, or is there something I should be concerned about? This is my first overclock ever, so I want to make sure I get some input from more experienced OC'ers before i commit to this.

Thanks, all.

 

[Darkbreeze]

I'm assuming you're referring to the "beat it to death" responses above? Because if not I just want to add that there is nothing inherently "wrong" with overclocking, and that when properly configured and operating within the tolerances outlined for a given CPU and chipset by the manufacturer (Which is usually much broader than the stock configuration of the CPU itself), there should be no lessening of lifespan at all.

As long as you are not beyond the voltage and thermal spec for a given chipset and CPU it should last a reasonably long amount of time. Maybe not AS long as a stock configuration COULD last, but certainly the longevity isn't specifically shortened simply because it is overclocked. Usually when a CPU gives out early it's due to abuse or operating the system outside recommended thermal or voltage specifications continuously.

 

[Rogue Leader]

I'm assuming you're referring to the "beat it to death" responses above? Because if not I just want to add that there is nothing inherently "wrong" with overclocking, and that when properly configured and operating within the tolerances outlined for a given CPU and chipset by the manufacturer (Which is usually much broader than the stock configuration of the CPU itself), there should be no lessening of lifespan at all.

As long as you are not beyond the voltage and thermal spec for a given chipset and CPU it should last a reasonably long amount of time. Maybe not AS long as a stock configuration COULD last, but certainly the longevity isn't specifically shortened simply because it is overclocked. Usually when a CPU gives out early it's due to abuse or operating the system outside recommended thermal or voltage specifications continuously.

Yes, none of what I was saying was in regard to proper overclocking. Only specifically what noblspls was saying about just cranking the voltage and clocks to the moon and watching the fireworks.

 

[Darkbreeze]

Actually, I was referring to this.

Well you do you, I like to keep my CPUs from degrading in 4 years.

 

[frizzlebyte]

Sounds like you have realistic expectations are are willing to do what is needed. It does take time to configure and validate a stable overclock. The majority of people don't bother. That's why they are typically back here in about six months whining about "why am I getting all these errors all of a sudden?", "I think my motherboard is bad", "I think my graphics card is failing?", "I think I got a crappy CPU".

No, you have an operating system full of micro-errors due to not taking the time actually required to validate an overclocked configuration.

But then they do a clean install of Windows and now they are sure it was just a bad update.

No, you most probably corrupted the operating system with errors. This happens with unstable CPU configurations more than often enough to mention. And if you fiddle with your memory configuration, even setting it to XMP settings, regardless that a lot of people think you don't need to take the time to check stability and run tests afterwards, you do, and you should.

Anything above what the CPU and motherboard supports by default is still considered an overclock and the timings chosen by the DRAM manufacturer are no different than the settings used by motherboard manufacturers for automatic CPU overclock profiles. Yes, in both cases they WILL probably work. Also yes, there is every possibility that from module to module and motherboard to motherboard, there might be something that does not agree with the memory controller or any of ten other potential factors that could cause what seems to be a perfectly normal XMP overclock to be creating miniscule errors, a zero here where a one should be, two ones where there should have been a zero.

Before long at all you have a system that will begin giving you problems. It might not be blue screening, and it might, but there will be indicators of some kind. Simpler just to validate your memory XMP settings with one time runs (Hopefully) of Memtest86 and Prime95 blend mode. If you want to be even more thorough, run Memtest86 a second time after running Blend, and choose only test 6 and 13. This will check for row hammer which can be a common, but relatively unheard of issue on modern memory modules, especially at very high frequencies and very low latencies.

Run Realbench for 8 hours. If there are errors, the system is not stable and you need to make a change. It might be reducing the multiplier. It might be increasing the voltage. It might be that you enabled the XMP profile and your error is not actually due to the CPU overclock, but due to something memory related. Always leave the memory at the default configuration until you are completely through validating your CPU overclock settings and testing.

Also, ditch that version of Prime95. Download Prime95 version 26.6.

Run version 26.6 and choose the Small FFT option. If it remains within the thermal specifications for your CPU for 15 minutes, then it is thermally compliant at it's current overclock configuration. Do not confuse thermally compliant with stable. They are different. Thermally compliant just means that your CPU, motherboard and cooling system are capable of withstanding the voltage and multiplier that are currently assigned and that you are not going to have to worry about thermal conditions unless you are running an unrealistic AVX workload like in versions of Prime95 newer than version 26.6. Actual applications and utilities that use AVX instruction sets do not use them in the same way that Prime versions that use them do.

Prime versions newer than 26.6 do not present realistic workloads. You would never see those temps in any practical application or game environment, ever. Trust the thermal compliance check using version 26.6 for 15 minutes.

So, Prime for 15 minutes. Thermally compliant, for THAT particular setting. If you change the multiplier or voltage, you need to do this step all over again, and you will likely do that during the process of determining where your overclock has reached a plateau.

Then, Realbench on the stress test using the setting for 50% of your RAM, for 8 hours. Do not cut corners here. If you cannot run Realbench for 8 hours, your overclock is unstable. There is no two ways about this.

Next, when you have determined that everything is good, THEN, and only THEN, at least in my opinion, should you enable your memory XMP settings or do any custom configuration of memory speed, voltage and/or timings. Once you begin doing that, you need to run Memtest86 for at least two passes after ANY change you make to the memory configuration to establish that the settings are at least basically stable. Follow that up by running Prime95 version 26.6 and choose the Blend option or the Custom option. I prefer the Custom option. If you choose the custom option, input approximately 75% of the amount of RAM that is currently NOT in use, and change the Minimum FFT size to 512(k).

Run that, either on Blend or on the custom option, for a minimum of 8 hours. If there were problems passing the Memtest validation then there is a problem with either the memory itself, or the configuration. You can try adding a small, minimal amount of additional voltage if they refuse to run at XMP settings. If you cannot gain stability at XMP settings without adding more than .020v to the stock XMP voltage value, then I'd look at whether there are issues with the modules or not.

Sorry it's taken a bit of time to get back on the forum. Thanks for the comprehensive suggestions, darkbreeze. What surprises me most of all is that the older Prime95 is best for testing. I never would have expected that.

Anyway, this is definitely helpful. Much appreciated.

 

[Darkbreeze]

You are, of course, most welcome.

I'm working on a beginner to intermediate level overclocking tutorial. Might be something in there useful to you.

http://www.tomshardware.com/faq/id-3761568/cpu-overclocking-guide-tutorial-beginners-work-progress.html

 

[larrycumming]

nobspl,

Well well. Since the advant of Adaptive Mode overclocking, I don't see how you can "burn out" a processor in 4 years unless you run a rendering shop or play games 24/7.

For me I use it for web browsing and youtube 70% of the time, so I expect my CPUs to last a long time. I will sell them before their time is up tho.

I get my CPUs to a good overclock and then turn on Adaptive Mode and leave all C-states on. The last couple of 0.2/0.4Ghz usually takes the greatest toll on voltage/heat increase, so I generally avoid going that far with my overclocks. It really doesn't matter to me if my CPU run sat 5.2Ghz or 4.8Ghz. Stability and longevity is of greater value to me

 

[TechyInAZ]

nobspl,

Well well. Since the advant of Adaptive Mode overclocking, I don't see how you can "burn out" a processor in 4 years unless you run a rendering shop or play games 24/7.

For me I use it for web browsing and youtube 70% of the time, so I expect my CPUs to last a long time. I will sell them before their time is up tho.

I get my CPUs to a good overclock and then turn on Adaptive Mode and leave all C-states on. The last couple of 0.2/0.4Ghz usually takes the greatest toll on voltage/heat increase, so I generally avoid going that far with my overclocks. It really doesn't matter to me if my CPU run sat 5.2Ghz or 4.8Ghz. Stability and longevity is of greater value to me

Plus, CPUs only degrade when there's a lot of load on it. So running 5ghz at 1.4v without any load on the processor 24/7 won't degrade it at all. Similar to how there's less strain on a car engine running at redline in neutral vs actually driving the car at redline.

 

[larrycumming]

nobspl,

Well well. Since the advant of Adaptive Mode overclocking, I don't see how you can "burn out" a processor in 4 years unless you run a rendering shop or play games 24/7.

For me I use it for web browsing and youtube 70% of the time, so I expect my CPUs to last a long time. I will sell them before their time is up tho.

I get my CPUs to a good overclock and then turn on Adaptive Mode and leave all C-states on. The last couple of 0.2/0.4Ghz usually takes the greatest toll on voltage/heat increase, so I generally avoid going that far with my overclocks. It really doesn't matter to me if my CPU run sat 5.2Ghz or 4.8Ghz. Stability and longevity is of greater value to me

Plus, CPUs only degrade when there's a lot of load on it. So running 5ghz at 1.4v without any load on the processor 24/7 won't degrade it at all. Similar to how there's less strain on a car engine running at redline in neutral vs actually driving the car at redline.

Well, heat is also a factor in depreciating a processor's lifespan, and your idle temperature @1.4V will be higher than idling @0.8V provided you use the same cooler under same ambient environments.

As CPU's degrade, it will require more and more voltage to run at the same frequencies. Thus, your stable overclock from 3 years back may need a bit more voltage to be kept stable today if you were to "burn it out" or "put it to the test."

 

[frizzlebyte]

You are, of course, most welcome.

I'm working on a beginner to intermediate level overclocking tutorial. Might be something in there useful to you.

http://www.tomshardware.com/faq/id-3761568/cpu-overclocking-guide-tutorial-beginners-work-progress.html

Looks good, darkbreeze. Very nice presentation.

So, I clocked my CPU to 3.7ghz and 1.2375v, and ran P95 26.6 Small FFT for 15 minutes, reaching a max of 74 degrees Celsius (idle temps are around 25 degrees). Then I ran Realbench for 8 hours at 8GB, 50% of my total RAM, hovering between 69.9 and 72 degrees. There were no errors of any kind. I didn't try to overclock RAM, but ran Memtest86 for a couple passes anyway, and it checked out fine.

Do you think this OC is probably stable, then? Seems a bit too easy, but as a beginner maybe I'm just suspicious. My next question would be, should I try to find the lowest stable voltage for this clock, try boosting the speed more, or leave it as-is and call myself lucky?

 

[Darkbreeze]

Yes. If you passed 8 hours of Realbench with no errors, then the CPU is almost certainly stable.

I would not try reducing voltage. You are already suspiciously low on the voltage as is. Is that the actual Vcore setting in the bios or is that the core voltage you are seeing in the monitoring software? They are rarely the same.

If you want to do some additional testing of the CPU to further validate stability, you can run Prime95 26.6 (Or the latest version of Prime95 with the local.txt file found in the Prime95 folder edited so that the line in the file reads CpuSupportsAVX=0 (instead of CpuSupportsAVX=1) which will prevent it from running unrealistic AVX workloads, but will allow the rest of the program to offer better support for CPUs that did not exist when version 26.6 was created.

Choose the Custom option (Instead of Small FFT, Large FFT or Blend). Input 50% of your installed memory amount where it says "Memory to use". Leave the minimum and maximum FFT sizes as they are. This will allow the program to run the full gamut of FFT sizes from small to large. Leave the time to run each FFT size on 15 minutes. Run it for 24 hours. It should run through all the relevant FFT sizes within that time. This is not particularly essential, but if you feel like Realbench was not enough of a stability test this will be.

 

[frizzlebyte]

Yes. If you passed 8 hours of Realbench with no errors, then the CPU is almost certainly stable.

I would not try reducing voltage. You are already suspiciously low on the voltage as is. Is that the actual Vcore setting in the bios or is that the core voltage you are seeing in the monitoring software? They are rarely the same.

If you want to do some additional testing of the CPU to further validate stability, you can run Prime95 26.6 (Or the latest version of Prime95 with the local.txt file found in the Prime95 folder edited so that the line in the file reads CpuSupportsAVX=0 (instead of CpuSupportsAVX=1) which will prevent it from running unrealistic AVX workloads, but will allow the rest of the program to offer better support for CPUs that did not exist when version 26.6 was created.

Choose the Custom option (Instead of Small FFT, Large FFT or Blend). Input 50% of your installed memory amount where it says "Memory to use". Leave the minimum and maximum FFT sizes as they are. This will allow the program to run the full gamut of FFT sizes from small to large. Leave the time to run each FFT size on 15 minutes. Run it for 24 hours. It should run through all the relevant FFT sizes within that time. This is not particularly essential, but if you feel like Realbench was not enough of a stability test this will be.

It is the actual voltage setting in the bios, but you're right, it isn't always the same in the monitoring software, especially the vcore voltage, which sometimes dips way down (0.4v). Could that be due to Ryzen shadow states? I uploaded some HWiNFO screenshots to imgur so that you can see for yourself what I mean: https://imgur.com/a/q9wAA0t

 

[Darkbreeze]

I am aware of what you mean. When cores are not under a load, if you have the power saving features enabled (Recommended) such as Cool N Quiet and the various C-states, then core frequency and voltage will drop considerably. This is normal.

Core voltage will also drop somewhat under a load, this is known as vdroop. Raising the Load line calibration setting will mitigate or eliminate that, depending on what you have it set to. It will also increase your full load thermals though. Higher LLC results in more stable under load, but also results in hotter core temps.

Some lower end boards don't have that setting, and of course looking back on this thread it seems yours is one of those. But if it's stable, and passes testing, I would not be too concerned.

 

[kerberos_20]

llc was present on first am4 boards, it got removed later with bios updates, new boards shouldnt have it anymore
cpu is taking care of it (ryzen)
there is just cpu soc voltage offset which works fine when u overclock too much and system freezing, that offset seems to be replacement for llc setting

cool n quite and c states are disabled if u touch vcore or cpu frequency
there are p states tho which u can use for power saving when u dont like cpu automanagement

 

[frizzlebyte]

I am aware of what you mean. When cores are not under a load, if you have the power saving features enabled (Recommended) such as Cool N Quiet and the various C-states, then core frequency and voltage will drop considerably. This is normal.

Core voltage will also drop somewhat under a load, this is known as vdroop. Raising the Load line calibration setting will mitigate or eliminate that, depending on what you have it set to. It will also increase your full load thermals though. Higher LLC results in more stable under load, but also results in hotter core temps.

Some lower end boards don't have that setting, and of course looking back on this thread it seems yours is one of those. But if it's stable, and passes testing, I would not be too concerned.

Great! I understand how you are suspicious of how low my voltage setting is, because I myself am finding it difficult to fathom how it remains as stable as it does on ~1.23 volts. It seems some other people need into the 1.3x volt range to make it work at 3.7ghz. Is this simply a matter of "winning" the silicon lottery, you think? And if so, would this chip be more likely to be able to comfortably push higher than this with a small boost in the voltage?

 

[Darkbreeze]

Gimme a screenshot of CoreTemp alongside Prime95 running Small FFT. If you don't mind.

 

[frizzlebyte]

Gimme a screenshot of CoreTemp alongside Prime95 running Small FFT. If you don't mind.

Welp, looks like the stability was a phantom, at least when ambient temps are higher, as they are today. While I was running P95 to get you the shot, my PC locked up.

EDIT: Thanks for your continued help, darkbreeze. This has been fun and interesting. I don't know, given that it can appear stable and then not be, whether I would run an OC on my daily driver PC all the time, since I'd rather not risk microerrors due to issues like this, where the thing seems stable but then isn't. However, it looks like my proc would be a good candidate for OC'ing, if I wanted to do it.

 

[Darkbreeze]

Well, I've never seen an overclock be stable, and then not be, unless something changed or something was overlooked on the initial run. One setting being different can change a variety of behaviors.

If you decide to start over, I'd wipe it all. Reset the CMOS to default settings and begin again. Make sure not to touch the memory settings until AFTER the CPU overclock process is finished and final, with validated stability. Then you can configure the memory to its XMP profile or whatever custom settings you decide to run it at. If you decide not to OC the CPU, then you can just configure the memory and test for memory stability.

Either way, good luck to you.

 

[larrycumming]

Well, remember your cpu's operating power is dependent on svid/pwm tables which take into consideration temperature as a factor. as your vrm temperature rises its power delivery efficiency decreases, so you always give it a bit of a head room after you reach a "stable overclock."

Gimme a screenshot of CoreTemp alongside Prime95 running Small FFT. If you don't mind.

Welp, looks like the stability was a phantom, at least when ambient temps are higher, as they are today. While I was running P95 to get you the shot, my PC locked up.

EDIT: Thanks for your continued help, darkbreeze. This has been fun and interesting. I don't know, given that it can appear stable and then not be, whether I would run an OC on my daily driver PC all the time, since I'd rather not risk microerrors due to issues like this, where the thing seems stable but then isn't. However, it looks like my proc would be a good candidate for OC'ing, if I wanted to do it.

 

[Darkbreeze]

Maybe you do, I don't. If a configuration is stable at a given clock and voltage setting, and I'm happy with both the clock speed and thermal value, that is where it stays. I do not test at a given clock speed just to reduce it after determining stability. If you're going to do that, you might as well do your testing at that speed and voltage too.

 

[larrycumming]

Well, my point is ambient temperatures are a factor. For example, if you did your stress testing in the evenings and during winter times (without house heating), you're stress tested against that particular ambient environment. In OP's case, when the ambient temperature rises, the overall system temperature is affected because a cooler's ability is more realistically measured in ambient temperature delta, not as an absolute value.


Mind you, not everyone live in a thermally controlled environment with central AC, and even if we do, ambient temperature varies according to four seasons and locale.

 

[Darkbreeze]

That is why they only issue overclocking licenses during the summer.