Ryzen 3000/5000 Overclocks

[Prickly007]

AMD has stated that seeing those 1.4+V numbers is fine though.

I spent most of last night researching a safe voltage for Rzyen, auto voltage on my Asus C6H seemed rather high. But never did find any consensus as the answer seemed to vary depending on whether one was talking about Zen, Zen+, Zen 2, Zen 3, etc.

I opted for 1.35V on my Ryzen 2700; in part, because my cooling sucks. The case fans that came with the Define R5 are suppose to be 1,000 rpm, but neither come anywhere close. The better one barely does 800 rpm; plus one or more of them make that weird clicking noise like a cable is touch the blades.

 

[JD]

I think it's a different methodology and people just have to forget the ways of Intel and start from scratch. Each generation behaves differently too and even using old monitoring software can cause you to see incorrect values. There's a major software element here, notably having to install chipset drivers and power profiles, to ensure full performance. That's not something you typically needed to do with Intel and didn't cause any adverse effects.

 

[Inik]

Once I finally install that DH-15, I'll fiddle around to OC my 3800x. I'm looking at PSU deals right now, because my XFX 750W is being noisy.

 

[Prickly007]

I think it's a different methodology and people just have to forget the ways of Intel and start from scratch. Each generation behaves differently too and even using old monitoring software can cause you to see incorrect values. There's a major software element here, notably having to install chipset drivers and power profiles, to ensure full performance. That's not something you typically needed to do with Intel and didn't cause any adverse effects.

Not to mention that AMD is rather tight-lipped on what the voltage limits actually are, especially for Zen+ from what I read.

 

[lowfat]

Some interesting info on AGESA 1.0.0.4B. Seems that EDC is bugged and it you set it to a really low value, it ignores the value lol. So the CPU will boost higher. In my case my 3900x will maintain ~4.4GHz on my fastest CCD. Although Windows still tries to juggle it to a slower CCD, so you may need to set affinity on an per app basis. CPU will also do ~4.15GHz all core turbo, as its hitting the 95C limit. This is significantly lower overall than I can get w/ a manual OC. But at least this way all CPUs don't get 1.3V 24/7.

EDC = 1, PBO TURBO BOOST

EDC = 1, PBO TURBO BOOST bug* *bug = *feature*. Some posters over at the X570 Crosshair VIII thread came upon a way to get PBO to boost like never before. LINK Shorthand info: Set PBO into manual mode: PPT 0 TDC 0 EDC 1 c-states disabled For better tweaked results: PPT [150]~[300+] TDC...

www.overclock.net

 

[JD]

Far more detailed thread than what people were posting in the Gigabyte ones that I've read through. There people thought it was a bug with Gigabytes BIOS but I guess it's an AGESA bug.

I typically get 4.2GHz all-core though with all default settings on my 3900X.

 

[Izerous]

I spent most of last night researching a safe voltage for Rzyen, auto voltage on my Asus C6H seemed rather high. But never did find any consensus as the answer seemed to vary depending on whether one was talking about Zen, Zen+, Zen 2, Zen 3, etc.

I opted for 1.35V on my Ryzen 2700; in part, because my cooling sucks. The case fans that came with the Define R5 are suppose to be 1,000 rpm, but neither come anywhere close. The better one barely does 800 rpm; plus one or more of them make that weird clicking noise like a cable is touch the blades.

Sounds like that usually suggest bad bearings, or a bad blade that is making contact with the frame. Have had that happen with a few cheaper fans where the blades actually stretched out over time and ceased against the frame completely.

I don't recall my 2700x voltage but it sits at an allcore of 4.1Ghz full time 4.2Ghz wasn't perfectly reliable.

 

[lowfat]

Hmm. So when I did CCD overclocking, I used 1.3V. Peak and average voltages never dropped according to RM. They stayed ~ 1.3V always. Moved to PBO yesterday, and average voltage drops down, to around 1.1 to 1.2 while I'm watching RM. Should in theory be way lower when idling. However it had absolutely zero impact on idle power draw. UPS shows ~160VA overnight when display, speakers, lights, etc are off and computer is idling. Exactly the same as the CCD overclock. I honestly thought it would have had a bigger impact.

Doing a heavy all-core workload, power consumption is probably considerly higher than my CCD overclocking too, since it uses way more voltage.

 

[lowfat]

Came across this today on Reddit. Even 1.325v may cause degradation w/ some CPUs.

https://www.reddit.com/r/overclocking/comments/ejd5c9

 

[rjbarker]

With the latest Windows scheduler changes (in 1909) along with the Ryzen Power Plan, it should prefer the "faster" cores and that generally appears to be the case. Usually my CCD1 is sleeping (6 cores) in normal usage as they are the slower bunch.

So yes, having my CCX0 on CCD0 be able to run at 4.45GHz should mean most lightly threaded apps will go to those 3 "fast" cores.

Overclocking based on the lowest common denominator is the old way of thinking, and I would say that's mostly what you're doing with Intel today. I don't believe Intel offers any way to OC cores separately.

The only option in the BIOS to OC individual cores (only physical along with virtual core, so 8 only groups), is by changing the multiplier for each vs "sync mode" which changes the multiplier for all Cores....so you can "quazi oc" each core, but with base clock and vcore limitations....
Going back thru this thread (OC'ing will always grab me).....looks like your able to change frequency and/or multiplier with physical and virtual cores?...but Voltage will be the same across all when you set vcore....?
Are you able to sync all cores and find highest OC? Is there Load Line Calibration or something similar?....

curious is all....