OC a Ryzen 5 3600X on ASUS X370-Pro - help a returning OCer

[sagramore]

Hi everyone. Thanks for reading. I want to start off just by noting that I know I probably don't need to do any of this, but I was interested in what was possible so here we are.

I am running the following rig:

• AMD Ryzen 5 3600X cooled by Coolermaster Masterliquid 240 AIO watercooling
• Asus Prime X370-Pro (Socket AM4) DDR4 ATX (latest BIOS revision, 5220)
• Corsair Vengeance LPX 16GB (2x8GB) DDR4 PC4-25600C16 (3200MHz) @3200 MHz, 16-18-18-36-1T, 1.35 V (DOCP 3200)
• Zotac nVidia GeForce RTX 2080 8GB (I know... I couldn't afford a Ti when I built this)
• Windows 10 on Samsung SM961 Polaris 256GB M.2 PCI-e SSD

I wanted to see what I could achieve overclocking the CPU. At present it runs "stock" with what I believe is the precision boost, which means 3.8 GHz base with up to 4.4 GHz single core boost, 4.1 GHz all core boost. In this mode, looking at HWMonitor, I can see that under general daily use the min/max core clocks vary anywhere from 3.0 GHz to the boosted 4.4 GHz, and the individual core voltages drop as low as 0.2 V, while hitting max values of 1.475 V. For reference, under P95 small FFT torture, all cores boost to 4.1 GHz and the individual core voltages sit at ~1.35 V.

This all seems fine. In this mode I get cinebench scores of ~3640 and the system seems very stable under my regular use for a few months.

Venturing into using Ryzen Master to tweak the all-core clock speeds I seem to hit a hard limit at 4.25 GHz stable. Setting it to 4.3 GHz leads to a pretty rapid failure during P95 up to voltages that I'm prepared to try (more on this below).

At all core clocks of 4.25 GHz, with a "Peak Core(s) Voltage" setting of 1.325 V I seem pretty stable. Under full load in P95 of cinebench the cores sit drawing 1.29 V each and my temps peak around 75 C. Cinebench scores go up a bit, and I can get to just shy of 3800. However, here lies my first question. When I am idling in Windows with these settings, HWMonitor shows each core still drawing the max setting of 1.325V. I'm assuming this is usual, as the system is no longer using any kind of boosting and is just sitting at the values I've given it - is that a correct assumption?

If so, this leads me on to my more pertinent question - what's a value that I can put in the "Peak Core(s) Voltage" setting that I can be confident shouldn't lead to any major degradation in the lifespan of my CPU compared to the default boost option? I understand any increased voltage is bad for CPU life, but I imagine this is a distinctly non-linear trend and there must be a level of "safe compromise"?

All in all, if I can't go much above 4.25 GHz all-core, and I don't do a lot of multi-core operations, I imagine I'm best off sticking with default settings for most of my gaming needs. The single core boosts to even higher clock speeds are probably more beneficial?

Any thoughts or guidance gratefully received as I haven't done any serious overclocking since I ran a 1 GHz thunderbird at 1.4 GHz for years


EDIT: It turns out I had made some mistakes in the above, which I would like to clarify here:

I didn't have the latest p95 (there were no AVX settings anywhere!), so I've updated that. Secondly, I can't tell what the PBO scalar was during a "stock run" because I was using Ryzen Master rather than the BIOS directly to set it.

With those things in mind I have done a bunch of things to confirm my original results, i.e. enabled PBO 1X manually in the BIOS, get HWInfo to ensure I'm reading voltages correctly, and updated p95. This image shows both HWMonitor, HWInfo, and p95 running a small FFTs test with these settings, and ensuring that the "disable AVX2" box is not checked. https://imgur.com/sWk03QA

Unless I'm missing something I appear to be sitting at 1.31V, 4GHz all core boost, and the temps are notably higher than before - hitting 86 C.

My questions from the main post above, still stand! Thanks

 

[MentalSharps]

Interested to follow this as I am just preparing to pull the trigger on a Ryzen 3600 with interest in overclocking to get the most out of it. Have seen several recommendations to leave it up to the PBO as the best all-around solution

 

[roger_37]

watching with interest.

 

[sagramore]

Hey guys, just responding because two of you said you're watching!

Not great news I'm afraid After a few days of testing, I figured I could get an all core clock of about 4.2 GHz that seemed stable enough, the issue being at the cost of voltage. All in all I weighed my options and decided sticking with PBO was better - you still get single core boosts to 4.4 GHz when needed, and it'll still boost all cores to 4.1 GHz as well but when you're not using much it clocks down and sits with nice cool/low voltages which I feel is far better for the chip long term. Overall, yes, clockable, but no, not really worth it. For me that is.... YMMV.

 

[roger_37]

Hey guys, just responding because two of you said you're watching!

Not great news I'm afraid After a few days of testing, I figured I could get an all core clock of about 4.2 GHz that seemed stable enough, the issue being at the cost of voltage. All in all I weighed my options and decided sticking with PBO was better - you still get single core boosts to 4.4 GHz when needed, and it'll still boost all cores to 4.1 GHz as well but when you're not using much it clocks down and sits with nice cool/low voltages which I feel is far better for the chip long term. Overall, yes, clockable, but no, not really worth it. For me that is.... YMMV.

thanks for the info ..nice one.

 

[merlin6r]

Hi. First of all, Ryzens sometimes sit at high voltages when idle, but draw little current so don't worry about that. A friend of mine was concerned when his 3900x was sitting at over 1.4v on idle but researched this and discovered the above.

I think it's widely accepted that 1.4v is the maximum for 24/7 use. A little higher is ok for short term and/or testing.

 

[Joxeon]

My ryzen 5 3600 can do a stable all core 4.2@ 1.325v which considering it would only boost to 4.2 on 2 cores is a very good perf boost.

I personally wouldn't recomend over 1.35 especially if your doing lots of heavy all core workloads.

 

[luismenendez]

I decided also not to overclock my 3600. Running at default settings underwater also boost speeds to around 4100 all cores and a few cores can reach 4250Mhz often so what's the point of fixing the volts and void warranty?

AMD did a very good job creating a processor that auto overclocks to almost its limits depending on the cooling solution (it sounds boring for overclockers but we can focus on memory tuning, cooling, etc.

 

[PieEater]

Yup the 3000 series are pretty efficient at boosting within the tolerances AMD have set, overclocking them really doesn't reap great rewards and really is for enthusiasts who are happy to spend hours just to gain a few percent in benchmarks. IMHO The best things you can do to increase performance is to concentrate on the memory timings and keeping the CPU cool.

 

[jms198]

My 1600AF boosts to the max of 3.7 constantly when I undervolt. At stock it jumps around and settles at about 3.5. Does it help PBO?