der8auer Ryzen 3000 Custom Mounting Frame Performance

[nick name]

Ok I wanna apologize because I am not gonna post all the numbers I recorded. After going through everything it just doesn't seem worth while inputting all of it here.  I can sum up the thermal difference I've seen: CCD 1 seeing a max difference of 2*C ~ 3*C and CCD2 maybe 1*C ~ 2*C drop.  Those improvements were when the nuggets were setup to simply move the block lower down the IHS and not laterally to favor either CCD.  There were times where CCD2 seemed to drop in temp.  That wasn't consistent in CB20 runs, but was observable in short Prime95 runs.  (It needs to be noted that I also previously found the optimal orientation of my block on the IHS and that is with the tubes at the bottom of the IHS, using a Fractal Design S36.)

 

- Ryzen 3900X.  Running under a Fractal Design S36 with liquid metal.

- All runs were CB20.  All runs were after periods of idle or light CPU usage.  No back-to-back runs.
- All runs were using per CCD overclocking.  CCD1 at 4.4GHz and CCD2 at 4.25GHz.  CPU Core Voltage set to 1.2625V with a droop to 1.244V (per HWiNFO) during the run.  
    - Without der8auer kit: 25.5*C ambient room temp CCD1 @ 67.5*C and CCD2 @ 69.3*C
    - With der8auer kit: 25.5*C ambient room temp CCD1 @ 65.3*C and CCD2 @ 69*C
    - Without der8auer kit: 28.3*C ambient room temp CCD1 @ 71.3*C and CCD2 @ 73.3*C
    - With der8auer kit: 28*C ambient room temp CCD1 @ 68.3*C and CCD2 @ 71.5*C

 

- Prime95 Small FFTs for 3 minutes with same CPU settings as above (AVX not disabled). 
    - Without der8auer kit: 26.7*C ambient room temp CCD1 @ 79*C and CCD2 @ 79.8*C
    - With der8auer kit: 27.4*C ambient room temp CCD1 @78*C and CCD2 @ 78.3*C

 

One absolute benefit of the der8auer custom frame is that it seems to make mounting the block onto to the CPU much, much, much more consistent when using liquid metal.  I didn't realize it at the time but every mount after shifting the nuggets was successfully done on the first try (ie I never saw high temps due to bad mount while using it).  When using the Asetek mounting kit it takes several attempts to get a good mount with enough pressure.  I'm talking like at least four and as many as ten.  With the der8auer kit every mount was just right in terms of pressure.  

 

Another note is that using the mounting screws with the spring and bottom thumb nut proved unnecessary and burdensome.  Using only the top threaded thumb nut made mounting easier and assured you were using enough pressure.  Now I didn't have trouble applying enough pressure while using the bottom thumb nut and spring -- it was just annoying as you end up compressing the spring completely between the top and bottom thumb nuts.  As such the spring seemed useless.  

 

A quick install tip -- if you don't have access to the CPU backplate through your motherboard tray be sure not to remove all four Asetek posts at the same time as the backplate will fall loose and you'll have to dismount your motherboard.  Remove one or two posts at a time and replace with the der8auer nuggets.  If the stock AMD mounting hardware is in place then only remove one side at a time and then replace with the der8auer nuggets.

[nick name]

I went and averaged my numbers.

 

CB20 Runs:

- Without kit: average ΔT CCD1 42.24*C and CCD2 43.68*C

 

- With nuggets centered: average ΔT CCD1 39.76*C and CCD2 43.35*C

 

- With nuggets favoring CCD1: average ΔT CCD1 39.97*C and CCD2 44.21*C

 

- With nuggets favoring CCD2: average ΔT CCD1 41.26*C and CCD2 43.04*C

 

[Guest]

Thank you so very much for this mini-review! 

 

I have determined from reading this, that purchasing this product would be a gross waste of time and money based on your results.

 

Critique the testing__________ 

 

 Prime95 Small FFTs for 3 minutes with same CPU settings as above (AVX not disabled). 
    - Without der8auer kit: 26.7*C ambient room temp CCD1 @ 79*C and CCD2 @ 79.8*C
    - With der8auer kit: 27.4*C ambient room temp CCD1 @78*C and CCD2 @ 78.3*C

 

The 3 minute test is way too short. We aren't hitting a stabilized temperature. The gradient should fluctuate to a higher number and settle as the cooler efficiency levels out. 

So these numbers you are reporting might actually be higher than expected after a duration. The numbers could actually be too low as well.

This will also tell you how well your cooler and case flow are handling an extended load period.

 

______________

 

 - Without der8auer kit: 25.5*C ambient room temp CCD1 @ 67.5*C and CCD2 @ 69.3*C
    - With der8auer kit: 25.5*C ambient room temp CCD1 @ 65.3*C and CCD2 @ 69*C
    - Without der8auer kit: 28.3*C ambient room temp CCD1 @ 71.3*C and CCD2 @ 73.3*C
    - With der8auer kit: 28*C ambient room temp CCD1 @ 68.3*C and CCD2 @ 71.5*C

 

- Prime95 Small FFTs for 3 minutes with same CPU settings as above (AVX not disabled). 
    - Without der8auer kit: 26.7*C ambient room temp CCD1 @ 79*C and CCD2 @ 79.8*C
    - With der8auer kit: 27.4*C ambient room temp CCD1 @78*C and CCD2 @ 78.3*C

 

These numbers are ok, but do not reflect the position of the cooler. 

I did see mention that you've found the optimal position for your processor.

If the mount and cooler where positioned else where, higher up, more to the left, ect... what where the results?

 

Please don't take offense here. I'm not trying to bash this at all. (the product yes, you no)

Really on the search for information at it's finest. In other words, the more data to look at the better.

If this was a straight on cooler design, where the position wouldn't have mattered, this would be said and done.

Not sure about others, but I'd like to see the data where the cooler is positioned all the way to the right, or left. And then some in between. 

 

So far, your testing doesn't have me interested enough to purchase this product.

 

IHS plate. Does it make a difference?______

 

And with the IHS plate, and knowing a few things here and there.... Is not a flat. The AMD IHS plates are concave to the center. Because of knowing this, positioning the cooler slightly one way or another is going to effect the thermal interface material thickness and position under the cooler's cold plate. Not measurable by the naked eye, the cooler may not actually sit flat as a result.

 

Since your testing isn't the first I've seen, I often wonder if the IHS plate does indeed play a major role in cooler positioning. I want to believe it does, however most people are not going to lap the IHS plate flat, and even less people to de-lid a cpu that's soldered.

 

Additional testing also could be done with different thermal interface materials.

If the cooler positioning really drops 1-2c in temps, it should be consistent to all the thermal interface materials tested despite the end temp.

 

That's my longer winded two cents here. Again, please don't be offended. That is not my intention at all, but rather thankful you have taken the time to test this product.

 

 

ŞɧʁɨmǷβʁɨmЄ™

[nick name]

@ShrimpBrime  To address your comments: 

 

The Prime95 runs were intentionally a short duration to prevent the AIO from reaching any sort of steady state.  A long run probably would have yielded the same results for with and without the kit so it would have been more a test of the AIO and not the mounting kit.  der8auer actually speaks to that in his second Q&A video.  

 

The posted results in OP were all with the nuggets positioned to move the block lower down the IHS without favoring either CCD.  I only posted those runs as an example of what more "obvious" runs looked like.  I posted all the averages for the tested positions below the first post.  

 

And with the block positioned with the tubes at the bottom you can see the best contact point/pressure is toward the bottom of the IHS where the CCDs are.  That can be confirmed when you remove the block and observe the TIM on the IHS.  Also by checking temps.  So while I can't speak to how deformed the IHS on my CPU is I can say I have accounted for it the best I could without lapping.  

 

And I certainly wasn't offended and appreciate your input.  

 

As far as any value existing in this product . . . maybe if you are satisfied dropping your temps about 3*C for short workloads like CB20.  Something undeniable though is that it absolutely made mounting the cooler while using liquid metal soooooo much more reliable and consistent.  

[Guest]

Indeed consistent cooler mounting is desirable for those that do it often. No argument there. Some coolers are just a pain in the butt to install.

 

A question I want to ask was how this fair under single core and 50% work loads.

Does SMT on and or Off seem to make any difference. 

How does this mount fair with different OC setups through the PBO for example vs per CCD clocking?

And lastly if there are any significant differences from running a high cooler temp gradient vs a lower one? (Delta)

 

 

[nick name]

37 minutes ago, ShrimpBrime said:

Indeed consistent cooler mounting is desirable for those that do it often. No argument there. Some coolers are just a pain in the butt to install.

 

A question I want to ask was how this fair under single core and 50% work loads.

Does SMT on and or Off seem to make any difference. 

How does this mount fair with different OC setups through the PBO for example vs per CCD clocking?

And lastly if there are any significant differences from running a high cooler temp gradient vs a lower one? (Delta)

 

 

Damn it.  I didn't test any light or single-core loads and I've returned back to the stock mount.  Most likely not going to keep the der8auer kit.  

 

As far as different OC setups -- I wouldn't have been able to accurately capture the data with PBO.  I wouldn't expect any boost in PBO performance with what I saw in temps though.

 

And I'm not sure what you mean with the last statement.    

[Guest]

2 hours ago, nick name said:

 

And I'm not sure what you mean with the last statement.    

The last statement -

 

Meaning high idle to load gradient. 

So the idle temps, say above 60c, and the load temps at 72c would be running a higher temp gradient than 40c to 72c which would be a lower temp gradient.

 

For a lot of coolers, they actually become more efficient in this way. 

It's been said that the temp "spike" impact is greatly reduced running a higher Delta (water/air temp) or high temp gradient.

 

I always aim for cooler temps at load vs worrying about temp gradients like higher water delta or higher case temps. It just seems counter productive to overclocking, but may yield some interesting results when tested thoroughly. 

 

[nick name]

Just now, ShrimpBrime said:

The last statement -

 

Meaning high idle to load gradient. 

So the idle temps say above 60c, and the load temps at 72c would be running a higher temp gradient than 40c to 72c which would be a lower temp gradient.

 

For a lot of coolers, they actually become more efficient in this way. 

It's been said that the temp "spike" impact is greatly reduced running a higher Delta (idle temp) or high temp gradient.

 

I always aim for cooler temps at load vs worrying about temp gradients like higher water delta or higher case temps. It just seems counter productive to overclocking, but may yield some interesting results when tested thoroughly. 

 

Ahhhh.  I've never considered that or really seen anyone speak about it.  

 

So how would one go about accomplishing this?  

[Guest]

1 minute ago, nick name said:

Ahhhh.  I've never considered that or really seen anyone speak about it.  

 

So how would one go about accomplishing this?  

Really? Wasn't too long ago some discussion back and forth. A good read I assure you.

https://www.techpowerup.com/forums/threads/why-intel-cpus-run-at-95°c-and-why-amds-should-also.263479/

Near the end it does get rowdy and the thread was closed, but I had already walked away from it thinking it was a silly idea to run chips with a higher temp gradient. Well at least in the fashion the original poster was trying to aim it.

 

I'd imagine fan curves could help a little bit. 

My end argument was that these chips run so cool at idle, it's actually difficult to run a high temp gradient. 

 

[nick name]

9 minutes ago, ShrimpBrime said:

Really? Wasn't too long ago some discussion back and forth. A good read I assure you.

https://www.techpowerup.com/forums/threads/why-intel-cpus-run-at-95°c-and-why-amds-should-also.263479/

Near the end it does get rowdy and the thread was closed, but I had already walked away from it thinking it was a silly idea to run chips with a higher temp gradient. Well at least in the fashion the original poster was trying to aim it.

 

I'd imagine fan curves could help a little bit. 

My end argument was that these chips run so cool at idle, it's actually difficult to run a high temp gradient. 

 

Interesting.  I'll give it a look.  

[Guest]

Was just looking at my Ryzen 5 1400. I recently pulled this from the HTPC and upgraded it.

 

Look at how the thermal paste let go when I removed the heatsink. The cpu never saw 70c in its life yet. Even under a full load. 

 

Can see the core layout. The heat must never had spread much across the IHS plate. The paste did a well enough job.

Thats  2017 to 2020 on a single pasting. 

 

Thats exactly what and where the core looks like underneath.

[nick name]

Interesting.  

[Guest]

29 minutes ago, nick name said:

Interesting.  

Aint it though??

 

A delid wouldnt be needed, but I may just cause I can..... It wont help thermals really. Would be like mounting a cooler 2 mm lower maybe get a couple c difference.