AMD 5950X what`s a safe voltage for single core processes?

Leb · Mar 23, 2023

Hey all,
I`m relatively new to Overclocking and there is a ton of mixed information regarding it,
I'm trying to find out what is a good/safe voltage for single-core processes.
I`ve tweaked my system a bit and I'm currently using the dynamic switcher:

All core processes 4.6 mhz @ 1.32v
Single core process 4.9 / 5.02 mhz @ 1.43v

Is 1.43v a safe voltage for a single core and as a daily driver?
some insight to this would be helpful thanks

cheers.

Font Technology Screenshot Darkness Terrestrial plant

JohnnyFlash · Mar 23, 2023

AMD has said that 1.35v for manual all-core is the safe limit. They also said that it boosts about that for small amounts of time for light loads only, which has since been proven incorrect.

Personally, I've always thought they pushed too hard to win benchmarks and each AGESA update seems to be a little more conservative. Using the 140 EDC or higher you have there limits the voltage to 1.425v. If you set the EDC to less than 140, it will go all the way to 1.5v, which is for sure too much.

I think the most important point is that the 5950X is a multitasking CPU and has a sweet spot for voltage - single core and multi core. If you play with curve optimizer and the boost limit, you can get the best overall real-world performance by lowering the boost limit a little and increasing curve optimizer. I'm about 800 hours into testing this on mine, and I'm going to end up between 4725 and 4775 it looks like.

Sir Beregond · Mar 23, 2023

Its not really "single core". Dynamic OC switcher is basically the best of a static all-core OC and PBO/CO for more single threaded work.

Generally you wouldn't want to run a static voltage above 1.35v for daily driving. Voltage above that is really for short boosts/benchmark runs/etc.

JohnnyFlash · Mar 23, 2023

Sir Beregond said:
Voltage above that is really for short boosts/benchmark runs/etc.

This was the marketing, but it's not accurate.

If you play Dragon Age for 5 hours and check the log, it will sit near that voltage the entire time if too many other cores aren't active. If a couple other cores are in use, the voltage will stay the same and clockspeeds will drop a little until you hit the EDC.

Sir Beregond · Mar 23, 2023

JohnnyFlash said:
This was the marketing, but it's not accurate.

If you play Dragon Age for 5 hours and check the log, it will sit near that voltage the entire time if too many other cores aren't active. If a couple other cores are in use, the voltage will stay the same and clockspeeds will drop a little until you hit the EDC.

Oh yeah if saying that voltage under more single threaded workflow should be ok. Wouldn't run that for all-core past bench runs though.

gvansly1_8906 · Mar 23, 2023

JohnnyFlash said:
Personally, I've always thought they pushed too hard to win benchmarks and each AGESA update seems to be a little more conservative. Using the 140 EDC or higher you have there limits the voltage to 1.425v. If you set the EDC to less than 140, it will go all the way to 1.5v, which is for sure too much.

Limits which voltage? CPU? Thanks in advance.

JohnnyFlash · Mar 23, 2023

EDC setting of 140 or higher limits vid to 1.425v on most updated boards. Depending on LLC settings and board power, that can be between 1.4-1.435v on the SVI2 sensor.

My personal opinion is that you should use 140 or higher if you plan on keeping the build for more than a couple years. Not to mention an EDC lower than 140 really hurts multithreaded performance on an 5950X.

Danny.ns · Mar 24, 2023

At default BIOS settings, that 99.9% of users of Ryzen 5000-series use every day, the CPU will use up to 1.5V during low current loads. This is per spec, and will be fine for many years to come. This is not the same thing as manually setting the voltage to 1.5V and running Prime95.

If you are using PBO then you need to avoid the EDC bug where setting it higher than default (e.g. >140) limits voltage to ~1.41V, which handicaps the CPU for things like gaming. If you use the CPU in workloads where 16 cores are at 100% constantly, then the EDC bug might not have a big effect since the voltage most likely would be much lower anyway.

It is really hard to talk about "safe voltages" because it depends on the current. All cores 4.6GHz playing a game or running CPU-z stress test is not the same as all cores 4.6GHz running AVX Prime95 or some other crazy load. But I do remember a few media outlets (e.g. Derbaur, Sweclockers etc) getting the info from AMD that when manually overclocking they can aim for 1.325-1.35V so I would guess that is fine.

Might be interesting;

JohnnyFlash · Mar 24, 2023

Danny.ns said:
At default BIOS settings, that 99.9% of users of Ryzen 5000-series use every day, the CPU will use up to 1.5V during low current loads. This is per spec, and will be fine for many years to come. This is not the same thing as manually setting the voltage to 1.5V and running Prime95.

So many people parrot this without testing themselves. That's their marketing info, but in reality it's not what happens at all.

Boost is governed by TDC , EDC and temp, in that order regardless of "load".

Open Prime95 right now, select 1 thread Small FFT AVX2 and tell me what voltage you get.

Owterspace · Mar 24, 2023

If I set edc to even 130 it bugs to 1.425. AGESA sucks.

gvansly1_8906 · Mar 25, 2023

@ JohnnyFlash,
I'm confused, I guess. I have the 5950X and the Gigabyte X570s Aorus Master with F1 BIOS so AGESA is something less than V2 1.2.0.3 which I thought was immune to the EDC bug.
But I did what you asked and here are the results to my surprise and disappointment:
Prime95 | small fft | 1 thread | AVX2
EDC 185 | (SV12 TFN) = 1.42
EDC 140 | (SV12 TFN) = 1.42
EDC 135 | (SV12 TFN) = 1.41

Rectangle Font Material property Parallel Screenshot

Font Rectangle Parallel Screenshot Number

JohnnyFlash · Mar 25, 2023

gvansly1_8906 said:
@ JohnnyFlash,
I'm confused, I guess. I have the 5950X and the Gigabyte X570s Aorus Master with F1 BIOS so AGESA is something less than V2 1.2.0.3 which I thought was immune to the EDC bug.
But I did what you asked and here are the results to my surprise and disappointment:
Prime95 | small fft | 1 thread | AVX2
EDC 185 | (SV12 TFN) = 1.42
EDC 140 | (SV12 TFN) = 1.42
EDC 135 | (SV12 TFN) = 1.41

Ya, it looks like it.

The next step is to start with curve optimizer to get some extra performance.

Leb · Mar 25, 2023

Owterspace said:
If I set edc to even 130 it bugs to 1.425. AGESA sucks.

Sorry, but what is this EDC bug that keeps being mentioned? I haven't heard of this.
I'm using a Darkhero VIII motherboard with 3600mzh 128gb g-skill ram (set at 3200mhz for stability).
the only issue I'm encountering is fine-tuning my CO to get the best low volts with high boosts for single and multi-core.

Owterspace · Mar 26, 2023

Leb said:
Sorry, but what is this EDC bug that keeps being mentioned? I haven't heard of this.

Font Technology Screenshot Electronic device Software

Edit:

Its a TDC bug, and its not present with 1203.

AMD 5950X what`s a safe voltage for single core processes?

Leb

JohnnyFlash

Sir Beregond

JohnnyFlash

Sir Beregond

gvansly1_8906

JohnnyFlash

Danny.ns

JohnnyFlash

Owterspace

gvansly1_8906

JohnnyFlash

Leb

Owterspace

Top Contributors this Month

Recommended Communities