Overclock.net banner

1 - 20 of 58 Posts

·
Not new to Overclock.net
Joined
·
78,989 Posts
Discussion Starter · #1 · (Edited)
The gist: Basically, I want to see if I did a good job installing the awesome and beautiful Noctua NH-D15S and if I did a good job applying its stock thermal compound. So I want to run some recommended tests and get feedback on my temperatures regarding if it's like, I dunno, very bad, bad, not bad, decent, good, very good, damn good, better than expected, how the hell are you getting those temps (lol), etc. :)


I haven't really done any testing yet because I don't know yet what's recommended for testing thermals. The closest I came to any kind of thermal testing was playing with Prime95 v30.3. For a few minutes, I ran Small FFTs and Smallest FFTs with AVX all the way off and my temps were just over 70 most of the time, maybe 75 peak (but again, it was only for a few short minutes). Regardless, I'm curious: when I start doing some real testing for the sake of testing my thermal performance going forward, what temps should I see? Or did I do a good enough test? If I did, what do you think?

I'll be replacing my cooler Master CM 690 later today with the Fractal Design Define 7 when it's delivered. It's the white one with the clear TG glass window. :)

Here are what I hope are all of the setting you need to see to help me:
Screenshot 1.png Screenshot 2.png Screenshot 3.png Screenshot 4.png screenshot 5.png Screenshot 6.png Screenshot 7.png

Screenshot 8.png Screenshot 9.png Screenshot 10.png

Screenshot 11.png Screenshot 12.png Screenshot 13.png


Edit: Oh, I'm using the thermal paste that came with the NH-D15S.
 

·
H₂O Aficionado
Joined
·
4,406 Posts
Define "real" testing.

If you want to run a real test, run a workload you would normally... run? Have HWiNFO running in the background and log the data. That will let you know what thermals you're going to be contending with.

I typically run CB R20 to get a quick snapshot of performance or thermals on the CPU, but for "stability" when adjusting clocks, I prefer AIDA64. The free trial includes the system stress test functionality. As well as a suite of other tests, like the "Cache* and Memory Benchmark".

Prime95 is quite extreme and not representative any workload that I am aware of.

I've been in the deep end lately logging everything from CPU clock frequencies vs power consumption, CPU thermals in games, and CPU thermals in AIDA64 vs. fan rpm. 90% of the fun I have is just playing with spreadsheets in my spare time... :)

Some of my recent examples...:

 

·
Not new to Overclock.net
Joined
·
78,989 Posts
Discussion Starter · #3 ·
That was a good question. I attempted to redefine what I meant. I edited my original post.

Basically, I want to see if I did a good job installing the awesome and beautiful Noctua NH-D15S and if I did a good job applying its stock thermal compound. So I want to run some recommended tests and get feedback on my temperatures regarding if it's like, I dunno, very bad, bad, not bad, decent, good, very good, damn good, better than expected, how the hell are you getting those temps (lol), etc. :)
 

·
Registered
Joined
·
225 Posts
At one point someone on reddit was trying to look at temperatures and boost of different 5800X's in the community and made a spreadsheet for Cinebench 20 results. The original might have disappeared but there are still a few copies, like this one. You can try to run CB20 and compare your results to others with a D15.
 

·
Not new to Overclock.net
Joined
·
78,989 Posts
Discussion Starter · #5 ·
At one point someone on reddit was trying to look at temperatures and boost of different 5800X's in the community and made a spreadsheet for Cinebench 20 results. The original might have disappeared but there are still a few copies, like this one. You can try to run CB20 and compare your results to others with a D15.
Oh wow. Dude. Thank you.
 

·
H₂O Aficionado
Joined
·
4,406 Posts
That's an interesting sheet! Thanks for sharing hazium223 :)

@OP if you use the temp filters, you can sort through the D15 results.

Idle (Desktop) TempAmbient TempSC T DeltaMC T Delta
32273349
36283450
31204163
26203057
42.0244161
32204260
34214357
 

·
Not new to Overclock.net
Joined
·
78,989 Posts
Discussion Starter · #7 ·
I have the NH-D15S. It's almost the same performance-wise though.
 

·
H₂O Aficionado
Joined
·
4,406 Posts
No other data will be good as your own. A lot of variables including binning, ambient temp, case air flow, cooler mounting, thermal paste etc.

At least you have somewhat of a record of other users :) But from what I can see, you're not far off with the DH-15. Best to start optimizing with your cooler with your own testing if you ask me :)
 

·
Registered
Joined
·
145 Posts
Since every cpu is different and everyone has different settings that will impact power draw, and therefore heat, the best way to do this would be to compare temps for a given workload while also recording cpu power use.

We can use HWinfo to measure die and package power.

You will want to know room temp and the temp inside the case as well... since if my room is 10C hotter than yours my cpu will be 10C hotter. Same deal with the case temp, neither of which reflect on your cooler.

For approx case temp, you can use the "Motherboard" temp sensor on your B550-F.

Here are my results on a 5800x, B550-F, prime95,v30.3,build 6, Arctic Freezer ii 240 with the radiator top mounted in a case with 2x120+2x140mm intake fans, all configured to try and keep motherboard temp at 25C or below (so it remains pretty quiet in normal use).

Torture test settings, note "Custom" to have as little variation as possible.

2486623


After letting it run for a few min to stabilize temps, CPU is pulling ~150W at 85C when my motherboard (so cooler intake temp) is ~25C.
I have PBO enabled to mobo limits, but my cpu temp limit is set to 85, so this limits max power and temp.
It is possible to hit 85C with lower power as well if you use a high load on a smaller number of cores, so you do want a similar load to compare.

2486625


R20 MC hits 85C for me as well, at ~145W
2486627


R20 SC sits at ~65C and 66W. It did boost up as high as 86W and 68C before clocking down a bit.

2486628
 

Attachments

·
Registered
Joined
·
145 Posts
The gist: Basically, I want to see if I did a good job installing the awesome and beautiful Noctua NH-D15S and if I did a good job applying its stock thermal compound. So I want to run some recommended tests and get feedback on my temperatures regarding if it's like, I dunno, very bad, bad, not bad, decent, good, very good, damn good, better than expected, how the hell are you getting those temps (lol), etc. :)


I haven't really done any testing yet because I don't know yet what's recommended for testing thermals. The closest I came to any kind of thermal testing was playing with Prime95 v30.3. For a few minutes, I ran Small FFTs and Smallest FFTs with AVX all the way off and my temps were just over 70 most of the time, maybe 75 peak (but again, it was only for a few short minutes). Regardless, I'm curious: when I start doing some real testing for the sake of testing my thermal performance going forward, what temps should I see? Or did I do a good enough test? If I did, what do you think?

I'll be replacing my cooler Master CM 690 later today with the Fractal Design Define 7 when it's delivered. It's the white one with the clear TG glass window. :)

Here are what I hope are all of the setting you need to see to help me:
View attachment 2486568 View attachment 2486569 View attachment 2486570 View attachment 2486571 View attachment 2486572 View attachment 2486573 View attachment 2486574

View attachment 2486575 View attachment 2486576 View attachment 2486577

View attachment 2486578 View attachment 2486579 View attachment 2486580


Edit: Oh, I'm using the thermal paste that came with the NH-D15S.
Just looking at your BIOS pics. You should disable "PBO Fmax Enhancer" as this basically hurts both performance and power use.

I don't think you need Eco-Mode enabled. Probably it has no effect due to your other PBO settings over-riding it, but if it does you are giving up multi-core performance with that. if you want to keep max temp lower than 90C, then set Platform Thermal Throttle Limit under PBO settings to whatever max temp you want.
Same comment applies to PPT Limit, TDC Limit and EDC Limit. You'll get better performance with these on auto or motherboard and a Platform thermal throttle limit set if you are worried about temps...
You also need to be careful about which PBO setting is taking effect. Personally I do everything except "curve optimizer" settings under "Ai Tweaker\Precision Boost Overdrive.", and then leave
everything under AMD Overclocking\Precision Boost Overdrive at the defaults/auto, as anything set under Ai Tweaker seems to take priority. I have not tried it the other way around (leaving Ai Tweaker on Auto).

If you are concerned about power use or vrm temps, then you can leave "power phase control" and "power duty control" on optimized rather than extreme.
 

·
Registered
Joined
·
1,677 Posts
Just looking at your BIOS pics. You should disable "PBO Fmax Enhancer" as this basically hurts both performance and power use.

I don't think you need Eco-Mode enabled. Probably it has no effect due to your other PBO settings over-riding it, but if it does you are giving up multi-core performance with that. if you want to keep max temp lower than 90C, then set Platform Thermal Throttle Limit under PBO settings to whatever max temp you want.
Same comment applies to PPT Limit, TDC Limit and EDC Limit. You'll get better performance with these on auto or motherboard and a Platform thermal throttle limit set if you are worried about temps...
You also need to be careful about which PBO setting is taking effect. Personally I do everything except "curve optimizer" settings under "Ai Tweaker\Precision Boost Overdrive.", and then leave
everything under AMD Overclocking\Precision Boost Overdrive at the defaults/auto, as anything set under Ai Tweaker seems to take priority. I have not tried it the other way around (leaving Ai Tweaker on Auto).

If you are concerned about power use or vrm temps, then you can leave "power phase control" and "power duty control" on optimized rather than extreme.
In another thread I think someone told him to enable Eco-Mode to reduce heat output at idle or near-idle
 

·
Laptop Enthusiast
Joined
·
9,399 Posts
That was a good question. I attempted to redefine what I meant. I edited my original post.

Basically, I want to see if I did a good job installing the awesome and beautiful Noctua NH-D15S and if I did a good job applying its stock thermal compound. So I want to run some recommended tests and get feedback on my temperatures regarding if it's like, I dunno, very bad, bad, not bad, decent, good, very good, damn good, better than expected, how the hell are you getting those temps (lol), etc. :)
Run Cinebench R23. It runs for a guaranteed 10 minutes, and if it takes any longer than those ten minutes to normalize temps, you've got other issues to worry about.

I don't run the D15, just the Liquid Freezer II 120, and I hit like... 89C under sustained load. It's a LOT hotter than I'm used to, or comfortable with for that matter, but it handles lower loads just fine. I imagine 70s are the absolute hottest you'll see? Maybe 75C?
 
  • Rep+
Reactions: TwoCables

·
Registered
Joined
·
145 Posts
In another thread I think someone told him to enable Eco-Mode to reduce heat output at idle or near-idle
I haven't seen that post, but my understanding of this setting is that it limits the cpu to a max power of 65W. I don't think it will help with idle temps. Or if it does, you need to figure out why your idle was >65W to start with ;-)
 

·
Cheesebumps!!
Joined
·
2,421 Posts
I hit 83c but I am drawing 160w at that limit..my stock limit (145w) is around 66-70c range, my pumps are on 40% cycle (2 DDC3.2 pumps)..on AIR COOLING or AIO's, I expect it around 75c or more (depends on the cooler I guess), also Turn OFF the PBO max Enhancer..thats for E-PEEN purposes only and does nothing for daily normal usage, probably makes things worse since multi-core performance is messed up..

In another thread I think someone told him to enable Eco-Mode to reduce heat output at idle or near-idle
makes the TDP 65w and run around 3.9ghz - 4ghz, marginally performance is reduced but heat as well is greatly reduced, performance though isn't acceptable (for me at the least)

edit: @blu3dragon how did you get your PPT TDC and EDC display on HWINFO?? mine doesn't display those 3 on the newer version..

edit: NVM..sorted it out..AMD does too much things users have no idea about..even technical docs aren't there to help..***..surely next platform will be on Intel..
 

·
Registered
Joined
·
194 Posts
You really need to under volt the cpu. As long as it stays stable, decreased voltage will give you a very large thermal reduction. I initially set my core V to 1.3 Is now set to 1.275 MAX

First one is max gaming load. (still dialing in the core clock speeds, btw)

Untitled.png

Second is max CPU load

Untitled.png
 

·
Cheesebumps!!
Joined
·
2,421 Posts
You really need to under volt the cpu. As long as it stays stable, decreased voltage will give you a very large thermal reduction. I initially set my core V to 1.3 Is now set to 1.275 MAX

First one is max gaming load. (still dialing in the core clock speeds, btw)

View attachment 2486640

Second is max CPU load

View attachment 2486641
PBO2 + CO will give you much better results..than just offsetting..per-core fine tuned Undervolting plus you can decrease the overall TDP draw to your liking..
 

·
Registered
Joined
·
194 Posts
That's what I've done. Its very close to the best I can get it. Set hard limit, then adjust further through PBO.
 

·
Registered
Joined
·
145 Posts
For zen 3 it's normally better to just leave vcore on auto and tune core offsets to lower the voltage. With a vcore offset you very quickly get into clock stretching. You'll need to benchmark or look at effective clocks to detect this.
 

·
Cheesebumps!!
Joined
·
2,421 Posts
For zen 3 it's normally better to just leave vcore on auto and tune core offsets to lower the voltage. With a vcore offset you very quickly get into clock stretching. You'll need to benchmark or look at effective clocks to detect this.
on old firmware this is true..on newer it seems to have changed..(as far as I noticed..)
 
1 - 20 of 58 Posts
Top