Ryzen 3600x voltage - weird?

1_rick

Supreme [H]ardness
Joined
Feb 7, 2017
Messages
5,359
I'm running an Asus rog strix x470-i gaming with the most recent bios and I just dropped in a new 3600X. There seems to be a bit of a weird thing going on and I wonder if anyone else is seeing anything like it. The BIOS sets the voltage at 1.450, and HWInfo64 (latest beta, just downloaded today) seems to agree--at idle, the cores run around 4200MHz (bouncing roughly from 4000 to 4350, and sometimes dropping significantly lower, which is a really nice improvement over the first generation) with about 1.45V--the numbers fluctuate a bit over time.) But then I fire up a Prime95 Blend stress test, and it seems to settle in with around a 4100MHz with voltages of 1.332 or so. Now I know about voltage droop, but that seems like a big difference, and a pretty high idle voltage. Has anyone else observed anything like this? On my old CPU, a 1600X, I set it to 3900MHz and 1.375V and it pretty much stayed there.

Any thoughts? I haven't seen a lot of reviews yet but nobody seems to be talking about OC yet.
 
Which agesa is that? There is supposed to be a 1.0.0.3AB. many x570s have them but may not be released for the x470s yet.
 
I'm not sure. The bios download page doesn't say. I can reboot to check.

ETA: I didn't see the agesa version listed anywhere, but I'm running the 2406 bios released Friday.
 
Last edited:
Look at an x570 board bios notes. It lists the agesa version. I'm betting you don't have it. I asked the gigabyte guys about that and they said the x370 and x470 will all get the new agesa later. So I imagine it's the same for Asus.
 
Look at an x570 board bios notes. It lists the agesa version. I'm betting you don't have it. I asked the gigabyte guys about that and they said the x370 and x470 will all get the new agesa later. So I imagine it's the same for Asus.

Could be, and I'll check. I was considering buying the Asus X570 mini ITX, but that sold out first thing this morning.
 
The last gen would often hit 1.5v at idle and lower under load, my understanding is that it was to stabilize high boost clocks on cores with light activity. As long as it wasn't hitting 1.5v under full load it wasn't supposed to be an issue but it would be under load.
 
I see similar voltages on a 2700x. As long as you aren't over 1.4 on all core, and not going much over 1.5 on light load it looks fine.

I don't imagine overclocking will vary much from last gen other than getting a couple hundred mhz higher on PBO / manual overclock.
 
I see similar voltages on a 2700x. As long as you aren't over 1.4 on all core, and not going much over 1.5 on light load it looks fine.

I don't imagine overclocking will vary much from last gen other than getting a couple hundred mhz higher on PBO / manual overclock.

With HWInfo64, I'm seeing all the cores hover at 4.29 to 4.32 (except one slacker that likes to hang out around 3.45.)
 
You might want to check this review out, particularly section 7.

https://www.xanxogaming.com/reviews...w-english-dethroning-the-intel-core-i9-9900k/

Interesting!
I wondered why the boost clocks in all the reviews were lower than the published boost clock. At first I chalked it up to the wraith cooler being good, but not great.
Then I saw some 360mm custom water coolers having the same issue.

I wonder why this was not talked about more. This is when I miss Kyle and HardOCP.
 
I'm running an Asus rog strix x470-i gaming with the most recent bios and I just dropped in a new 3600X. There seems to be a bit of a weird thing going on and I wonder if anyone else is seeing anything like it. The BIOS sets the voltage at 1.450, and HWInfo64 (latest beta, just downloaded today) seems to agree--at idle, the cores run around 4200MHz (bouncing roughly from 4000 to 4350, and sometimes dropping significantly lower, which is a really nice improvement over the first generation) with about 1.45V--the numbers fluctuate a bit over time.) But then I fire up a Prime95 Blend stress test, and it seems to settle in with around a 4100MHz with voltages of 1.332 or so. Now I know about voltage droop, but that seems like a big difference, and a pretty high idle voltage. Has anyone else observed anything like this? On my old CPU, a 1600X, I set it to 3900MHz and 1.375V and it pretty much stayed there.

Any thoughts? I haven't seen a lot of reviews yet but nobody seems to be talking about OC yet.

Keep in mind that the voltage you set isn't what the CPU will always pull. Think of manual voltage settings as you setting the maximum CPU voltage. It will pull what it needs up to that point. There is some vDroop of course, but you can set your LLC to reduce it.
 
Keep in mind that the voltage you set isn't what the CPU will always pull. Think of manual voltage settings as you setting the maximum CPU voltage. It will pull what it needs up to that point. There is some vDroop of course, but you can set your LLC to reduce it.
All I can say is those numbers are what the bios chose and what hwinfo latest beta reported.
 
Interesting!
I wondered why the boost clocks in all the reviews were lower than the published boost clock. At first I chalked it up to the wraith cooler being good, but not great.
Then I saw some 360mm custom water coolers having the same issue.

I wonder why this was not talked about more. This is when I miss Kyle and HardOCP.


All the [H] guys started another site, TheFPSReview.com
 
On my x470 board, I have discovered something new: If I lower the bios voltage (1.38v to start), it's ignored in Windows, where it's still 1.44-1.45. If I fire up Ryzen Master, I can't adjust the voltage unless I change it to "manual overclock", which (by default) locks the core clocks to the default speed (3.8GHz), meaning I lose the floating speed adjustments. On the positive side I saw temps drop noticeably.

None of this may apply to x570, but I can't get one of those right now.
 
However, when doing some light-middle gaming, I'm only seeing ~30W total package power at ~41.5GHz, according to HWInfo64.
 
  • Like
Reactions: Nobu
like this
Sigh. HWinfo64 says my idle temps bounce around 50+ degrees. Minecraft causes it to run 55-75C. I swapped out the TT HSF for the Wraith that came with the 3600X (and btw, dang, the springs on that could be used to lift cars. I thought the motherboard was going to break, I had to press down so hard!) and unfortunately it's no cooler. Definitely going to have to look for a 570 board, (or hope for an improved BIOS) or manually set the voltage and give up letting the CPU set the clock speed itself. :(

Unfortunately there are no mini-ITX boards to be had; the Asus one has even vanished from most sites (not "out of stock", there's no listing).
 
Sigh. HWinfo64 says my idle temps bounce around 50+ degrees. Minecraft causes it to run 55-75C. I swapped out the TT HSF for the Wraith that came with the 3600X (and btw, dang, the springs on that could be used to lift cars. I thought the motherboard was going to break, I had to press down so hard!) and unfortunately it's no cooler. Definitely going to have to look for a 570 board, (or hope for an improved BIOS) or manually set the voltage and give up letting the CPU set the clock speed itself. :(

Unfortunately there are no mini-ITX boards to be had; the Asus one has even vanished from most sites (not "out of stock", there's no listing).
Sounds like another asrock x370 Fatal1ty (replaced by the k4 and/or gaming x, iirc, and wiped from the history books)
 
Sounds like another asrock x370 Fatal1ty (replaced by the k4 and/or gaming x, iirc, and wiped from the history books)

Hah. I had that very board. What worked with it was nice, but ultimately is was too flaky, which makes me leery of trying them again.
 
Also, I should say, this board is the Asus X470 ITX one, and it's been rock-solid with my 1600X, so I suspect the most recent bios just isn't all that compatible with Zen 2 (yet).
 
Just replaced my motherboard with an Asus pro ws x570-ace, because that's all Micro Center had in stock today (other than that $700 board, which, no thanks.) So now I have a useless-to-me u.2 adapter. But at any rate, the CPU's running slightly faster, at slightly lower voltages. Temps haven't shown much improvement but I didn't see a case I liked so the motherboard is sitting out on my desk uncovered with nothing but the CPU fan until I can order a new case.
 
Just starting to skim it (and the referenced previous thread) but right off the bat there's some interesting stuff, like "downclock to 90% and then go to sleep". That's not exactly what I'm observing, but the next sentence is "Windows can't detect that", so fair enough. I'll have to check out Ryzen Master. My root concern (before reading the thread) is that temps seem to be high, even on X570. For example, I just woke the machine up, fired up HWinfo, and watched the thing sit at ~2% overall CPU utilizations, but the temps would hit 50C, then drop down to 40C over a period of a few seconds (aside: it'd be really cool if AMD could do something so that core temps would drop on idle the way Intel does; OCCT will show an Intel proc losing 20C in a couple of seconds when finishing a stress test, whereas Zen seems to not be able to shed more than maybe 3C a second), and then spike back up to 50C and drop again. It's probably safe but since I'm using a stock Wraith fan right now it's annoying because the fan is both loud, and even louder without a case.
 
Come to think of it I *have* briefly seen Ryzen Master show cores running at ~500MHz, which is actually pretty cool. But Windows, as he said, can't show that. I *do* see Task Manager showing multiple cores dropping down to like 2000-2500MHz, which Zen 1 wouldn't do without playing with Pstates.
 
LOL: "After investigating this issue further, it looks like monitoring tools with frequent monitoring intervals will cyclically wake every core in the system for a few milliseconds to probe their behavior. From the perspective of the CPU, this looks exactly like an application asking for frequency boost."

Oh G-d. That's messed up.
 
Back
Top