Discussion Ryzen 7 5800X + NH-D15s (two fans) High Temperatures, Overclocking, Settings

[aescat]

So, it seems I didn't win the silicon lottery with my new 5800X or I don't know how to overclock it.

My PC:
Aorus Elite b550 Rev 1.0
Corsair Vengeance PRO RGB 4x8GB (32GB) at 3200 Mhz. X.M.P enabled. 16 - 18 - 18 -18 - 36 and tRC 56 instead of 54
Ryzen 7 5800X
Noctua NH-D15s with a second Nf-a12x15 Pwm fan
MSI RTX 3080 Ventus 3x OC
Sentey SDP750-SS 750W 80+
Thermaltake V250 TG ARGB Case

So I started replacing my R7 3700X with this new CPU, without reinstalling windows or changing BIOS. I'd already f10 version that supports 5x series. I enabled X.M.P and tried Warzone. The next day I couldn't boot, no BIOS boot, so I removed and put again the CMOS battery to reset BIOS settings and I could boot again. I tried then the last BIOS version, F13a but after a few minutes in desktop, the temperature went very high (90 C) on idle... I tried to apply an undervolt changing only the VCORE but still getting high temps.

So I tried after that, the F12 bios version, it seems very stable, not high temps on idle (32 C). I get 90 C running prime just in the first 10 minutes and never goes down, the frequencies go down instead (4000 mhz).

So, stock frequencies, stock voltage, very high temps. I get 86 C playing Warzone for instance. For AMD this is normal, you can see their response in this reddit post.

The problem is when I want to overclock this thing. I tried a lot of combinations only changing VCORE (voltage), CPU Ratio (frequency) and PBO. Enabling or disabling PBO, voltage from 1.1 to 1.4v and CPU ratio from 4.6 to 4.8. I can't make it stable, I get errors running Prime95 on the first 10 minutes and the PC restarts after that, very unstable.

I tried 4.6 with 1.35v but I got very high temps, like 102 C in short time.

I have very good airflow, let me show you:

Front: Thermaltake case's fans fixed at 1000 rpm (3x120mm).
Top: 2x Noctua Nf-a14 Pwm Chromax.black.swap 140mm
Rear: Noctua Nf-a12x15 Pwm Chromax.black.swap 120mm
CPU: Noctua NH-D15s + Nf-a12x15 Pwm. I used the dot method for the thermal paste NT-H1 that came with the D15s

My room is at 24 C with air conditioner.

I don't know what to change to use this at maximum capacity and get normal high temps (80 C max)...
Will I get a lot more FPS if I run this CPU at 4.8? Is this even worthy?
Should I remove the D15s and check the thermal paste?
Should I reinstall windows? (for me this is the worst thing I can do, I believe I won't get different results)
I know I'm an early adopter too, there are drivers and bios versions involve indeed.

 

[Phaaze88]

1)Fans smashed up against a solid panel is always an ouch, even if they are Noctuas. * Looks at V250's front panel

2)The 'folds' on the sides of the NH-D15/S reduce the effectiveness of fans placed up top. This was done on purpose to promote more uniform front to back airflow.

3)Well, having 2 top exhaust would help the gpu more - not so much the NH-D15S, because of point 2.

4)With all that intake, you effectively only have a single exhaust. How hard it must be for the combined heat of that 3080 - which is dumping its heat in the chassis, mind you - and the 5800X to get out of there...

5)Prime95, what? The settings you set do matter. Small FFT, AVX off, AVX2 off, AVX-512 off(Ryzen doesn't support 512 anyway), is the recommended setting for temperature stress testing. Keep in mind, what clocks you see in P95 will be different to games/blender/Cinebench.

6)You already know about Ryzen 5000 boost algorithm, so that's already covered.

That's my 2 cents on the cooling end. Overclocking and the stability issues I'll leave to someone else; I'm not that familiar with Ryzen beyond what I've read.

 

[Sergei Tachenov]

Probably has nothing to do with this particular problem, but if I ever want to add top fans to my case, I'd put them into the exhaust position. Hot air goes up, so it only makes sense to help it go up. And besides, 5 fans in, 1 out looks like a terrible disbalance to me. I really don't want to start a positive-vs-negative pressure holy war, but whether positive or negative, there gotta be some balance, right? Although with this case, I can't figure out how to achieve positive pressure and have top fans set as exhaust at the same time. But I'd rather go for a balanced setting and make 3 exhaust (rear and top) and 3 intake (front).

 

[aescat]

2)The 'folds' on the sides of the NH-D15/S reduce the effectiveness of fans placed up top. This was done on purpose to promote more uniform front to back airflow.

I didn't understand sorry. What do you mean? Is this NH-D15 bad installed?

Probably has nothing to do with this particular problem, but if I ever want to add top fans to my case, I'd put them into the exhaust position. Hot air goes up, so it only makes sense to help it go up. And besides, 5 fans in, 1 out looks like a terrible disbalance to me. I really don't want to start a positive-vs-negative pressure holy war, but whether positive or negative, there gotta be some balance, right? Although with this case, I can't figure out how to achieve positive pressure and have top fans set as exhaust at the same time. But I'd rather go for a balanced setting and make 3 exhaust (rear and top) and 3 intake (front).

ok, I should remove the front panel I think. It has a space on the side but not the best:

and I will try with top fans set as exhaust.

 

[Sergei Tachenov]

With this top notch hardware you may want to replace the case itself. Because if you just remove the front panel you'll probably end up accumulating dust inside very quickly. My case (Fractal Design Define S2) isn't exactly a gaming one, but it has places for 9 140 mm fans (intake: 3 front, 2 bottom; exhaust: 3 top, 1 rear). I'm only using the stock 3 at the moment, but I'm not really into OC, neither is my hardware that good. I bought it purely for low noise.

My previous case was very old, no dedicated sections for PSU or cables, and to get better cooling and less vibrations I removed both the front and the side panels. But dust management was a nightmare. I had to clean between the plates of the CPU cooler radiator about once every month.

 

[Phaaze88]

I didn't understand sorry. What do you mean? Is this NH-D15 bad installed?

No, it's not a bad install, but I do suggest remounting it again. You see, the little in the middle paste application isn't the best for that cpu, due to it having multiple dies beneath it.
A full but thin spread is preferred. Perhaps you have an old unused credit card laying around - you can use that to spread the paste. You can even do the 5 dot method.

Spoiler

See those folds in the heatsink, like it's sealed off? This is what I'm talking about.
This promotes more front to back airflow, thus making it more effective with front to back fan setups, but because of that, fans placed above it don't really benefit it.

 

[aescat]

No, it's not a bad install, but I do suggest remounting it again. You see, the little in the middle paste application isn't the best for that cpu, due to it having multiple dies beneath it.
A full but thin spread is preferred. Perhaps you have an old unused credit card laying around - you can use that to spread the paste. You can even do the 5 dot method.

Spoiler

See those folds in the heatsink, like it's sealed off? This is what I'm talking about.
This promotes more front to back airflow, thus making it more effective with front to back fan setups, but because of that, fans placed above it don't really benefit it.

Understood. I will try to remount again, I will do 5 dot method and set the top fans to exhaust. I will post results here.

 

[aescat]

Ok, these changes got me way better results!
Never hits 82 C, maximum temp with cinebench R20 was 81 C. The scores are very different now:

Cinebench R15 CPU 2618 cb
Cinebench R15 CPU (single core ) 265 cb MP Ratio 9.85
R20 5995 pts (4077 was before!)

It has to be related to the D15s and thermal paste application. This was how it looked like when I unmounted (dot method):

I did the 5 dot method, I really don't know if I did it well, again, I'm thinking if these results can be improved again with another try. But I really don't know what is the best method or the specific amount of thermal paste I need to apply. I watched tons of videos but again, not sure. What I know is now the thermal paste covers all the surface in the D15 plate and CPU.

With these temperatures now I think I can play with voltages and frequencies again.

 

[Sergei Tachenov]

I don't know if that is of any help, or even isn't of any harm, but for what it's worth...

https://www.gamersnexus.net/guides/3346-thermal-paste-application-benchmark-too-much-thermal-paste

It agrees with my experience (but I've probably assembled less than 10 rigs in my life). No matter whether I was spreading it thin, or not-so-thin, or just put one big drop in the center, I got approximately the same results, maybe 1–2 degrees of a difference, and even that was when the paste itself was different. So apparently, as long as you don't put too little and the heatsink is pressed strongly enough by its mount (which is the case for any decent cooler), it will do its job and spread the paste while you're screwing it.

So if that benchmark is to be believed, and you say that the paste covers everything it needs to cover, you're probably good paste-wise.

 

[Phaaze88]

But I really don't know what is the best method or the specific amount of thermal paste I need to apply.

For Ryzen, it's best to ensure a full coverage of the IHS. There are multiple dies beneath that IHS that you can't see, and when using the 'little in the middle' application, there's the possibility that you won't yield the best heat transfer to the cooler because you 'missed a spot'.

Little in the middle still works for Intel cpus, but news has it that LGA 1700(Alder Lake) will be following the multi die design too.