5800X + Thermals - What am I doing wrong?

[KeyperOS]

I just now built my new system:

5800X on a Gigabyte X570 Master R1.2 (yes, it is overkill but I liked the features at the price).

The RAM is a 2x16GB Crucial Ballistix kit @3600MHz, CL 16-18-18-38-84 (Single Rank).

The cooler is the Noctua NHU12A.

By all account a pretty OTT platform for this CPU.

 

I built the system, left all the CPU dials on Auto in the BIOS and installed Windows.

Even on idle, the CPU fans would constantly ramp up and down.

Ryzen Master showed CPU temps jumping between low 40s and mid 70s while doing practically nothing.
Running something semi-demanding would usually get the temperatures at mid 60s.

 

I finally tried to run the CTR 2.0 RC5's diagnostic but it would stop at the Cinebench test due to the CPU hitting 91% and would refuse to start the actual test!

Not knowing what else to do I went back into the BIOS, changed PBO from Auto to Advanced, set it to All Cores, Negative and 15.

 

That not only allowed Cinebench to complete but it also gave me the below unexpected result:

 

***ClockTuner for Ryzen 2.0 RC5 by 1usmus***
AMD Ryzen 7 5800X 8-Core Processor
GIGABYTE X570 AORUS MASTER
BIOS ver. F33j SMU ver. 56.50.00
TABLE ver. 3672325
DRAM speed 3600 MHz
05/11/2021 01:14:22


Phoenix - hello there 🙂
CTR completed profile creation in alternative mode
Last step was:
CCX1  4375 MHz  VID 1125  mV


Manual overclocking mode enabled

Penalties for the final profile: level 3
01:14:26: CCX1 (127): 4275 MHz, 1131 mV  OC+
Phoenix deactivated!
Cinebench R20 started
Cinebench R20 finished with result: 5581
Voltage: 1.131 V  PPT: 93.8 W  Temperature: 63.2°
Phoenix ready!
Cinebench R20 started
Cinebench R20 finished with result: 6045
Voltage: 1.328 V  PPT: 137 W  Temperature: 84.8°


01:17:32: Silicon FIT measurement started...
01:18:01: Silicon FIT measurement done!
01:18:01: Stress test stopped.


DEFAULT CURVE COEFFICIENTS
CORE#1  0  CPPC 146
CORE#2  21  CPPC 143
CORE#3  83  CPPC 150
CORE#4  1604  CPPC 135
CORE#5  1604  CPPC 150
CORE#6  1607  CPPC 139
CORE#7  1599  CPPC 131
CORE#8  1612  CPPC 127


AVX light mode
Cycle time: 120000 ms
Reference frequency: 4375MHz
Reference voltage: 1187 mV
Voltage step: 6 mV


Manual overclocking mode enabled
1:18:01 πμ: Saving temporary settings...
1:18:05 πμ: CCX1 (127): 4375 MHz, 1187 mV
1:18:05 πμ: Step# 1. Diagnostic VID: 1187 mV
1:18:06 πμ: Stress test 1 started...
1:19:12 πμ: Stress test stopped.
1:19:13 πμ: Stress test 2 started...
1:20:19 πμ: Stress test stopped.
1:20:19 πμ: Step# 2. Diagnostic VID: 1181 mV
1:20:20 πμ: Stress test 1 started...
1:21:25 πμ: Stress test stopped.
1:21:27 πμ: Stress test 2 started...
1:22:32 πμ: Stress test stopped.
1:22:32 πμ: Step# 3. Diagnostic VID: 1175 mV
1:22:33 πμ: Stress test 1 started...
1:23:38 πμ: Stress test stopped.
1:23:40 πμ: Stress test 2 started...
1:24:45 πμ: Stress test stopped.
1:24:46 πμ: Step# 4. Diagnostic VID: 1169 mV
1:24:47 πμ: Stress test 1 started...
1:25:52 πμ: Stress test stopped.
1:25:54 πμ: Stress test 2 started...
1:26:59 πμ: Stress test stopped.
1:26:59 πμ: Step# 5. Diagnostic VID: 1163 mV
1:27:00 πμ: Stress test 1 started...
1:28:05 πμ: Stress test stopped.
1:28:07 πμ: Stress test 2 started...
1:29:13 πμ: Stress test stopped.
1:29:13 πμ: Step# 6. Diagnostic VID: 1157 mV
1:29:14 πμ: Stress test 1 started...
1:30:19 πμ: Stress test stopped.
1:30:21 πμ: Stress test 2 started...
1:31:26 πμ: Stress test stopped.
1:31:27 πμ: Step# 7. Diagnostic VID: 1151 mV
1:31:28 πμ: Stress test 1 started...
1:32:33 πμ: Stress test stopped.
1:32:34 πμ: Stress test 2 started...
1:33:39 πμ: Stress test stopped.
1:33:40 πμ: Step# 8. Diagnostic VID: 1145 mV
1:33:41 πμ: Stress test 1 started...
1:34:47 πμ: Stress test stopped.
1:34:48 πμ: Stress test 2 started...
1:35:54 πμ: Stress test stopped.
1:35:54 πμ: Step# 9. Diagnostic VID: 1139 mV
1:35:55 πμ: Stress test 1 started...
1:37:01 πμ: Stress test stopped.
1:37:02 πμ: Stress test 2 started...
1:38:08 πμ: Stress test stopped.
1:38:08 πμ: Step# 10. Diagnostic VID: 1133 mV
1:38:09 πμ: Stress test 1 started...
1:39:14 πμ: Stress test stopped.
1:39:16 πμ: Stress test 2 started...
1:40:21 πμ: Stress test stopped.
1:40:22 πμ: Step# 11. Diagnostic VID: 1127 mV
1:40:23 πμ: Stress test 1 started...
1:41:28 πμ: Stress test stopped.
1:41:30 πμ: Stress test 2 started...
1:42:35 πμ: Stress test stopped.
1:42:36 πμ: Step# 12. Diagnostic VID: 1121 mV
1:42:37 πμ: Stress test 1 started...
1:43:42 πμ: Stress test stopped.
1:43:44 πμ: Stress test 2 started...
1:44:49 πμ: Stress test stopped.
1:44:49 πμ: Step# 13. Diagnostic VID: 1115 mV
1:44:51 πμ: Stress test 1 started...
1:45:56 πμ: Stress test stopped.
1:45:57 πμ: Stress test 2 started...
1:47:02 πμ: Stress test stopped.
1:47:03 πμ: Step# 14. Diagnostic VID: 1109 mV
1:47:04 πμ: Stress test 1 started...
1:48:09 πμ: Stress test stopped.
1:48:10 πμ: Stress test 2 started...
1:49:16 πμ: Stress test stopped.
1:49:16 πμ: Step# 15. Diagnostic VID: 1103 mV
1:49:17 πμ: Stress test 1 started...
1:50:23 πμ: Stress test stopped.
1:50:24 πμ: Stress test 2 started...
1:51:29 πμ: Stress test stopped.
1:51:30 πμ: Step# 16. Diagnostic VID: 1097 mV
1:51:31 πμ: Stress test 1 started...
1:52:36 πμ: Stress test stopped.
1:52:38 πμ: Stress test 2 started...
1:53:43 πμ: Stress test stopped.
1:53:43 πμ: Step# 17. Diagnostic VID: 1091 mV
1:53:44 πμ: Stress test 1 started...
1:54:49 πμ: Stress test stopped.
1:54:51 πμ: Stress test 2 started...
1:55:56 πμ: Stress test stopped.
1:55:56 πμ: Step# 18. Diagnostic VID: 1085 mV
1:55:57 πμ: Stress test 1 started...
1:57:03 πμ: Stress test stopped.
1:57:04 πμ: Stress test 2 started...
1:58:10 πμ: Stress test stopped.
1:58:10 πμ: Step# 19. Diagnostic VID: 1079 mV
1:58:11 πμ: Stress test 1 started...
1:59:17 πμ: Stress test stopped.
1:59:18 πμ: Stress test 2 started...
2:00:23 πμ: Stress test stopped.
2:00:24 πμ: Step# 20. Diagnostic VID: 1073 mV
2:00:25 πμ: Stress test 1 started...
2:01:23 πμ: Thread# 4 fall down!
2:01:23 πμ: Stress test stopped.
2:01:24 πμ: Step# 21. Diagnostic VID: 1079 mV


DIAGNOSTIC RESULTS
AMD Ryzen 7 5800X 8-Core Processor
CPU VID: 1079
CPU TEL: 1044
Max temperature: 68°
Energy efficient: 4,19
Your CPU is GOLDEN SAMPLE
Recomended values for overclocking (P1 profile):
Reference voltage: 1250 mV
Reference frequency: 4625 MHz
Recomended values for overclocking (P2 profile):
Reference voltage: 1325 mV
Reference frequency: 4725 MHz
Recomended values for undervolting:
Reference voltage: 1125 mV
Reference frequency: 4375 MHz

Phoenix deactivated!

 

Now, this is my very first system in 10 years (literally, the last one was a 1st gen i7-975) and I have no experience with either overclocking or undervolting but I am coming to the conclusion that my M/B was auto-OCing my CPU without me realizing it (Ryzen Master would casually show the CPU clocks jumping between 4700 and 4800 MHz but I just thought it was the CPU's all-core turbo function).

 

Am I correct in thinking that? If so what would be the "proper" way to go about this?

BTW, I am not really interested in OC-ing this system but I would be happy with an UV that would give me stock performance at lower thermals and/or power consumption. However my primary concern is rock-solid stability.

 

Thank you in advance!

[Ripred]

Thermal paste?

[Deslumo]

Check to see if anything is running in the background that's hogging your system resources. If there's nothing, check to see if you have thermal paste under your CPU cooler. Make sure it covers the entire IHS. 

[jones177]

Not enough cooling.

 

I built a 5800x /Aorus Master setup as well but mine uses a 360mm AIO.

My idle temps are low 30s but they do jump up to the low 40s some times.

It games in the 60s and benches in the 70s.

 

On air I could only recommend a Dark Rock Pro 4 or Noctua nh-d15. 

 

[Chris Pratt]

What you did in the BIOS was an undervolt using PBO2's curve optimizer. Each step (the 15 you entered) represents 3-5mV, so a 75mV undervolt to be exact.

 

This is actually the same thing that CTR does, only via hardware, rather than software. As such, the two are incompatible. For Zen 3 (which the curve optimizer is exclusive to), the BIOS approach is superior, so you should stick with that. You can also go up to 30 steps, and most Zen 3 chips can handle that.

[Areyn13]

As a 5800x owner, I can confirm that it's indeed a hot boi cpu. It doesn't hurt to go overkill on cooling for this one, I'm using the dark rock 4 with idle temps around 40, gaming around 70 and stress tests can even creep into the mid 80's. I thought I was doing something wrong, but quickly google "5800x temps" and you'll see just how common this is for this particular cpu. More so than any of the CPUs in the Zen 3 lineup, it just runs hotter by nature

[KeyperOS]

For paste I used MX4 (2019) and I applied a REALLY thin layer that covered the entire IHS.

 

I guess I could re-apply it but I am finding it hard to believe that the U12A (which apparently has about 92% the performance of the D15) is inadequate cooling.

 

Before buying it some people right here even suggested that to be an overkill, saying the 212 would suffice...

[Alex Atkin UK]

39 minutes ago, jones177 said:

Not enough cooling.

 

I built a 5800x /Aorus Master setup as well but mine uses a 360mm AIO.

My idle temps are low 30s but they do jump up to the low 40s some times.

It games in the 60s and benches in the 70s.

 

On air I could only recommend a Dark Rock Pro 4 or Noctua nh-d15. 

 

I get the same idle tempts on a 5950X with a Deepcool AS500.  I mean sure it climbs to 85C on all-core overclock, but that's kinda to be expected given how many cores it has and they sustain 4.4Ghz in Cinebench.  A 5800X should run far cooler.

 

14 minutes ago, KeyperOS said:

For paste I used MX4 (2019) and I applied a REALLY thin layer that covered the entire IHS.

 

I guess I could re-apply it but I am finding it hard to believe that the U12A (which apparently has about 92% the performance of the D15) is inadequate cooling.

 

Before buying it some people right here even suggested that to be an overkill, saying the 212 would suffice...

Don't forget the CPU cooler is only as good as your case airflow.

I did the five dots when installing my paste as it was recommended by the cooler manufacturer.

[jones177]

10 hours ago, Alex Atkin UK said:

I get the same idle tempts on a 5950X with a Deepcool AS500.  I mean sure it climbs to 85C on all-core overclock, but that's kinda to be expected given how many cores it has and they sustain 4.4Ghz in Cinebench.  A 5800X should run far cooler.

I don't have a lot of experience with AMD since I stopped using them back in 2007 and it has been Intel ever since, so when planning my first Ryzen built I noticed a lot of 5800x users complained about high temps.  

I decided if I was going to get a 5800x I would treat it like one of my Intels and give it a large cooler. The cooler I got is an EVGA 360mm AIO that cost $110 so $50 more than a Deepcool AS500 in the US.  I think it was money well spent just for the peace of mind.  

[Npiet1]

 

Thermals are normal. 85'c is a normal for spikes. As soon as you do something it spikes right up. Then dials back to what is needed.

I haven't had any success with cinebench with it either, it always crashes so I just use prime95 now to stress test.

 

If you want better temps some find under volting helps, however I found that changing VRM (to the pic below) is better and manually clocking it to 4.7ghz achieves best results. I went from 85'c to 60'c under stress test(Prime 95).

210418221500.BMP

[KeyperOS]

Ok, I have been checking and testing since I last posted here yesterday and I haven't for the life of me been able to keep it stable. It ALWAYS fails on the 10-minute mark of stress testing with Prime95.

 

I have cleared the motherboard CMOS and retested with everything stock.

I disabled XMP. I even disabled Core Performance Boost. The temps were at 45 degrees C.

It ALWAYS has a thread "fall down" around the 10 minute mark (last test, with everything disabled it was after just 7 and a half minutes).

 

Before I contact the store I bought both the motherboard and CPU from, is there something else I should first check?

 

Just FYI, I am using the F33j BIOS which incorporates AGESA ComboV2 1.2.0.2 and fixes the USB 2.0 devices stability/compatibility issues.

I did a Google search to see if there were any complaints but didn't find anything.

[Guest]

Quote

The RAM is a 2x16GB Crucial Ballistix kit @3600MHz, CL 16-18-18-38-84 (Single Rank).

Drop that to 3200mt/s and give the stability test another whirl.

[KeyperOS]

1 minute ago, ShrimpBrime said:

Drop that to 3200mt/s and give the stability test another whirl.

One of the first things that I did was disable XMP, it was running at 2666 in all subsequent runs...

[Guest]

1 minute ago, KeyperOS said:

One of the first things that I did was disable XMP, it was running at 2666 in all subsequent runs...

I did not thoroughly read your post above, my apologies. 

 

OK so what are the temps when you have this loaded with P95? 

If any core hits 90/95c, it'll probably drop it from the test.  

[Alex Atkin UK]

3 minutes ago, KeyperOS said:

One of the first things that I did was disable XMP, it was running at 2666 in all subsequent runs...

I admit that's concerning.  I had someone tell me they stick to Intel as they tried AMD and had a single core become defective, hopefully not what you have but possible.

[geo3]

21 hours ago, jones177 said:

Not enough cooling.

-

On air I could only recommend a Dark Rock Pro 4 or Noctua nh-d15. 

 

The NHU12A he's using has almost the same performance as the D15

[KeyperOS]

33 minutes ago, ShrimpBrime said:

I did not thoroughly read your post above, my apologies. 

 

OK so what are the temps when you have this loaded with P95? 

If any core hits 90/95c, it'll probably drop it from the test.  

After disabling XMP and CPB with nothing else being touched, it was at a mere 65 degrees, certainly nowhere near throttling.
Ryzen Master DID show EDC (CPU) at 100% of 140 A but I have actually no idea what that means as I couldn't find any information about it.

33 minutes ago, Alex Atkin UK said:

I admit that's concerning.  I had someone tell me they stick to Intel as they tried AMD and had a single core become defective, hopefully not what you have but possible.

I also believe to have a defective core as well as the thread that "falls off" is usually the same and at that same time I see Core 03 drop sharply in MHz...
Funny thing is Ryzen Master identifies core 03 as the "second fastest in the system"...

[jones177]

32 minutes ago, geo3 said:

The NHU12A he's using has almost the same performance as the D15

You are right.

I get them mixed up with the NHU12S.

 

 

[KeyperOS]

I went back in the BIOS and switched every single voltage dial from Auto to "Normal".

Now the EDC is stuck at 61% and the Peak and Average Core Voltage is at 1.025. Before it would fluctuate between 1.350 and 1.450.

 

Are these numbers normal? 0.325 to 0.425 seems like a huge voltage difference to me.

 

EDIT: First time the stability test reached 20 minutes...

[KeyperOS]

1 minute ago, jones177 said:

You are right.

I get them mixed up with the NHU12S.

 

 

D15 was the first cooler I looked at, but after I saw Noctua themselves say it had 92% the thermal capacity/performance of the D15 in a MUCH smaller package that fits in almost any case (and all the reviews that I saw corroborated that) I was sold.

[jones177]

Just now, KeyperOS said:

D15 was the first cooler I looked at, but after I saw Noctua themselves say it had 92% the thermal capacity/performance of the D15 in a MUCH smaller package that fits in almost any case (and all the reviews that I saw corroborated that) I was sold.

Noctua nh-d15s do have issues with most modern cases. Two of my computers are using one fan now but they both used 2 fans when they were in older cases. Fortunately they still stay cool with only one fan.

 

There is another active post with someone using a 5800x with a NHU12A. I am waiting to so how it turns out.

 

 

 

 

[Guest]

1 hour ago, KeyperOS said:

After disabling XMP and CPB with nothing else being touched, it was at a mere 65 degrees, certainly nowhere near throttling.
Ryzen Master DID show EDC (CPU) at 100% of 140 A but I have actually no idea what that means as I couldn't find any information about it.

I also believe to have a defective core as well as the thread that "falls off" is usually the same and at that same time I see Core 03 drop sharply in MHz...
Funny thing is Ryzen Master identifies core 03 as the "second fastest in the system"...

Interesting.

So it drops a worker from P95 then.... and that's about the extent of the worries? 

 

Or is there true instability when in actual use? 

Other Stress testing programs have issues as well?

[KeyperOS]

Pre-undervolting, Cinebench would crash as well.

I haven't actually tried any other stress testing programs.

 

Having Ryzen Master open I could actually see the moment the CPU Watchdog would "restart" a core.

The thing is, with 8 cores it's not hard for another to pick up the slack if one hangs and this being a single CCD CPU, I would think that there would be no issues with corrupt data either as all cores would have access to the same cache memories.
In short, no I haven't noticed any issues in day to day use, but I wasn't expecting to either in this particular scenario TBH.

[Guest]

19 minutes ago, KeyperOS said:

Pre-undervolting, Cinebench would crash as well.

I haven't actually tried any other stress testing programs.

 

Having Ryzen Master open I could actually see the moment the CPU Watchdog would "restart" a core.

The thing is, with 8 cores it's not hard for another to pick up the slack if one hangs and this being a single CCD CPU, I would think that there would be no issues with corrupt data either as all cores would have access to the same cache memories.
In short, no I haven't noticed any issues in day to day use, but I wasn't expecting to either in this particular scenario TBH.

So it then will pass benchmarking all threads like cinebench or anything meaningful aside from that?

Have you run OCCT and encountered the same core dropping out or throwing errors continously?

[KeyperOS]

So, after extensive testing, I can only get the system to remain stable if I SERIOUSLY cripple it by setting CPB, PBO2 and XMP OFF so that I can set ALL voltages to "Normal". If even one of these "advanced" functions is enabled and I set any voltages to "Normal" then the system will not boot.

 

In this scenario, with VCore stuck at 1.025 and all cores topping out at 3.800MHz and not a single MHz more, the system remains at 45 C under "full load" and also remains stable. BTW, this is the only case in which EDC doesn't jump to100% when stress testing but remains at a constant 61% instead.

 

Does anyone have an idea/experience that would suggest this to be a CPU issue, Motherboard (H/W or BIOS) or perhaps sth else?
I admit to not testing the RAM but seeing as with XMP off the stress test would still fail whereas with all voltages on "normal" it no londer does, I suspect the RAM to be fine.

Any help would be very appreciated.
Thank you in advance