Ryzen 5 3600x AIO watercooler vs. air cooler test

Recently i upgraded my system from a Ryzen 7 1700 to a Ryzen 5 3600x and appart from setting up the memory and reinstalling the OS and drivers, i kept it in all stock settings.
The hardware i’m using in direct relation with the CPU is an ASRock Fatal1ty X370 Professional Gaming motherboard (BIOS version 5.80), a Team Group Xtreem 3600mhz 16gb RAM kit and a Thermaltake Water 3.0 280mm watercooler kit.
I always like to benchmark my system and monitor it for any flaws i might encounter and from the get go i noticed i noticed that the CPU temperature was a bit out of whack, not only there were +15ºc spikes between the system being on idle and performing minor operations, when i decided to run Cinebench R20 the CPU temperature quickly jumped to around 76ºc and it stayed there for the duration of the benchmark, while not at all a daily use percentage, it did concern me as i hadn’t seem such high temperatures even on my OCed 1700.
I checked the voltage spikes and the value doesn’t seem to be abnormal, 1.372v was the maximum voltage spike registered by both HWiNFO64 and Ryzen Master during the benchmarks i ran, which is a tad lower than the 1.375v i constantly ran on my 1700, and yet the temperature was consistently higher all around, i even tried different thermal compounds (up to this point i was using Cooler Master’s Mastergel Maker Nano) and application methods, but there was no significant drop in temperature.
Talking with the members on the Ryzen 3000 & Navi Megathread we came up with a theory that the cooler could’ve been the culprit, so i ordered a Scythe Ninja 5 air cooler to test the theory.
Last friday the cooler arrived and i set out for testing.
The methodology was:

  • All stock settings on the BIOS, appart from the RAM;
  • Both coolers would use the same Cooler Master Mastergel Maker Nano thermal compound in the same “3 dot method” application;
  • All system coolers/pumps at full speed
  • Both coolers would be tested in the same case, an InWin 805 with closed panels and removed front panel (for mildly increased airflow);
  • When testing the Scythe Ninja 5, i also installed a pair of Akasa Apache 140mm fans on the same spot as the watercooler was installed;
  • Test 1 would be the temperature after 5 minutes of system idle;
  • Test 2 would be the temperature after 20 minutes of AIDA64 system stability test.

As simple as it gets really, so let’s see how it all went down.
For the record, the temperature outside was as follows, and i had my air-con set at 24ºc for the entirety of the test.

Temperature outside

Weather

Let’s start with the Thermaltake Water 3.0 280mm watercooler, which is the same as any other Asetek based 280mm cooler, with the added flair of a pair of RGB Riing Radiator fans.
Here’s the temperature after 5 minutes of idle:

And here’s the temperature after 20 minutes of stress test, which is what i’ve been seeing since i got the CPU:

And now for the Scythe Ninja 5, which is a large, 6 heatpipes, almost silent air cooler,
Here’s the temperature after 5 minutes of idle:

And here’s the temperature after 20 minutes of stress test:

Immediately after seeing the results i was at a loss, i re-tested both coolers with the case completely opened but didn’t record the data as there was no significant improvement in the temperatures, proving as well that the airflow wasn’t necessarily the culprit in this situation.
So there we go, i’m left with more questions than answers, what should i try next, the sit and wait game for a new BIOS update, should i contact AMD and tell them about this issue, try undervolting? Maybe i’m overthinking this and the system is just normal, you guys tell me.

2 Likes

I did see something on twitter this morning that some guy has fixed the voltage table for Ryzen 3000 and is getting better performance and lower temps all round on a variety of cpus

1 Like

Yeah not really sure what the issue is with voltages and temps on the new ryzen3000 series,
but once everything is set to auto, it seems like people getting rediculous temps, and sometimes also weird high voltage spikes.

My guess that there is still some sort of uefi bug.
Unfortunatelly i don´t have a Ryzen 3000 setup myself,
to do some tests and research with adding some voltage measuring wires at the back of the socket,
to see what is actually getting pumped into the cpu,
compaired to what the various software programm´s read out.
But yeah that will generally also vary per board.

But with a 3600X i would assume that a fixed vcore of 1.325V ish should be fine technically.
So yeah first thing i would try is setting a said voltage and play around with llc,
and then see what happens.

2 Likes

I wouldn’t try doing much with that system until you nailed in the voltages.

There are two ways you can try this. I’ve seen some.luck with people using offsets.

Calculate the amount of voltage you are offset from say around 1.3-1.3375 volts. Then subtract or negative offset by that amount. Then go set LLC to the mid level. On Asus boards it’s level 3 and same with ASRock… idk what it is on others.

Or you can set stuff manually. Try to keep your SOC voltage .1 volt below your CPU and you should be good at whatever voltage you can get stuff stable to. Like 1.325 as mistery suggested.

Then I would play with LLC (level 3 is often ideal) to make sure the voltage doesn’t spike on idle or load.

3 Likes

Yeah this old school method is going to be very time consuming,
and normally i wouldn´t really recommend it to users,
who generally not know what they are doing.
But yeah, a friend of my recently build a 3600X with Asus X570 Tuf gaming board,
and he gets even voltage spikes around the 1.5V,
which tells me that atleast Asus still didn´t manage to fix their crap.

2 Likes

I suppose now is a better time than any two learn about load line calibration. And I have not expected much from Asus as of late so this does not surprise me.
The spike is technically okay at stock frequency under no load. However in reality what’s going on here is I am thinking that the voltage regulation module… isn’t compensating for the change and load fast enough and that’s why you’re seeing these spikes.

Actually I would be extremely curious to see if changing the response time to the fastest response time the vrm can support without any danger… Would maybe eliminate the spikes who knows?

it’s all down to the user and how much he or she wants to experiment as we said before the safe path is just a manual set and make sure that your LLC is set to the medium level… when you set it to the medium level if you get instability sometimes increasing the maximum allowed current helps

1 Like

Here’s the link to the tweet i was talking about

2 Likes

This would be an interesting thing to test.
Hoewever i would only really recommend when you have a decent highend board.
And again know what you are doing.
Because upping the switching frequency will dump a significant,
amount of more heat into the vrm.
And you definitelly don´t wanne do that when running on crap mosfets.

If that is what you mean. :wink:

1 Like

Ehh there limit is 120 C right :troll: :sweat_smile:

1 Like

I think i can do that, should i also set a fixed CPU clock?
I’m on an ASRock as well.

No don’t worry about the clock for now. Because it’s ASrock you should play around with the LLC a bit. Level 3 and 4 will have different droops.

Which ASRock motherboard again, sorry I have not had my coffee. We could play around with the VRM response time if you have cooling headroom and are down to do so

1 Like

Its the Fatal1ty X370 Professional gaming, which is the same as the X370 Taichi.

Oh you have solid af fricken VRMs don’t be afraid to push them.

I would say just go with the settings we spoke of and try both methods both fixed and offset. :wink:

1 Like

Fortunately yes, its a great mobo!
I’m not a home right now to test it, but i should be able to in the afternoon.

@PhaseLockedLoop i never actually knew what was the function of the LLC, i always left it on auto even when i had the 1700 at 3.8ghz.
What effect does it have, in general?

Funny you should ask ^ it affects the voltage drop or spikes during load to compensate for the power requirement of the CPU. Because a CPU is a complex load (capacitive and inductive in some cases) the voltage will change when pulling more current during load. This is because inherently more transistors are running and each has their own electrical properties. Without getting into the complexities the thing you need to know is switching them on and off and a power intensive task. The power needs to be clean and steady for proper function.

1 Like

Did you ever test it with the stock cooler to establish a baseline?

2 Likes

Typing that out yesterday i came to the realization that, in fact, i did not set a baseline with the stock cooler.
I’ll do that today as well.

1 Like

I think that, by default, i saw the mobo auto set it level 5 or level 3, i’ll have to check.
I remember Bryan from Tech YES City mentioning using Level 2 for OC on the first gen Ryzens.

Level 3 is by far more Consistent when manually setting voltages

1 Like