Troubleshooting Custom loop CPU performance

[TheRedBaron_94]

Hey Guys and Galls , 

i'm currently troubleshooting some weird performance issues of my new custom loop. 
The GPU is doing Fine, went from about 95c to 50c ish under load ( while boosting significantly higher ) 

However the CPU is the problem child in this story , when using my old AIO ( Arctic 420 mm ) it would max out at about 75c ish ( running stock ) when running cinebench R23 on loop 
and at about 90 ish  when OC'ed to 5.6 all core at +0.020v Vcore ( e-core at stock ) 

With the custom loop the CPU goes up to 90c stock, with the pump on full , which i find to be weird seeing as i have 4 360 mm Rads and flow seems fine ... 

PC specs that matter in this story 

• i7 13700kf
• RX 6800XT

My custom loop consists of 

• Bykski XPR-C-i CPU block 
• Bykski A-MS6900-Trio GPU block 
• BARROW Dabel-20b 360 mm ( X4) 
• BARROW SPB17 DDC pump ( with res ) 
• 3/8 ID 1/2 OD tubing 
• all arctic p12 fans 

Loop order : Pump -> GPU -> RAD -> RAD -> CPU -> RAD -> RAD -> pump 

 

What i already tried : 
Remounting the CPU block (3 times)  , new thermal paste ( noctua ) , updating bios ( cuz you never know ) 


Any ideas ? 
 

[TheRedBaron_94]

while typing the post above , i noticed some really rather minor debris in the CPU block , could this be the cause ? 

 

[TheRandomness]

When removing the CPU block, what does the thermal paste contact look like?

Can we have a picture showing the whole system? From the side, such that we can see where components are plumbed and all that.

[TheRedBaron_94]

Thermal paste spread seems to be fine … nothing out of the ordinary 

[TheRedBaron_94]

To clarify the external rads Will be mounted under the desk hence the upside down rads and the fans are sandwiched in between the rads

[NorKris]

insane setup bro.. 

 

  lack of mounting pressure  or bad thermal paste job  

[TheRedBaron_94]

On 10/12/2024 at 9:58 AM, NorKris said:

insane setup bro.. 

 

  lack of mounting pressure  or bad thermal paste job  

Thanks for the compliment  

i'm gonna take of the block once more ( clean it just to be sure ) and add a washer on each corner to improve ( / increase ) the mounting pressure 
i'll keep you posted for the results

 

[TheRedBaron_94]

So as stated , i cleaned the block , added washers and remounted / repasted , this does seem to have improved the situation somewhat. 

Idles are in the mid 20's now (ambient at 21c) 
load temps still a bit higher then i'd like / expect in the mid 80's 

Might also just be the case that the block just isn't really all that great 

[NorKris]

1 hour ago, TheRedBaron_94 said:

load temps still a bit higher then i'd like / expect in the mid 80's 
 

what is "load"

[TheRedBaron_94]

Cinebench r23 on loop with furmark in the background ( or space marine 2 … kinda achieves the same goal)

 

[NorKris]

1 hour ago, TheRedBaron_94 said:

Cinebench r23 on loop with furmark in the background ( or space marine 2 … kinda achieves the same goal)

 

ok, any Hwinfo64 pic from the cine run?

 

was the 420 AIO intake or exhaust?

what about the top rad? 

 

maybe its so many elements too the loop that the extra rad capacity is lost to "the pump cant keep the pressure up "

[TheRedBaron_94]

[NorKris]

lack of mounting pressure or bad block  then  

[TheRedBaron_94]

my guess would be the latter , added the washers so mounting pressure shouldn't be the issue 

when i remounted i added quick disconnects on the block ... so i'm gonna change it out in the near future ( wont be as much of a chore  ) 

 

 

[NorKris]

35 minutes ago, TheRedBaron_94 said:

my guess would be the latter , added the washers so mounting pressure shouldn't be the issue 

when i remounted i added quick disconnects on the block ... so i'm gonna change it out in the near future ( wont be as much of a chore  ) 

 

 

if u had a water Temp sensor   my loop is similar to what u have and i have 4  

[TheRedBaron_94]

i saw a tip in a video linus made recently ... a shame i didn't think of it earlier , just touch the radiator end tanks ... they're cool to the touch

 

[DoctorNick]

On 10/15/2024 at 5:16 PM, TheRedBaron_94 said:

i saw a tip in a video linus made recently ... a shame i didn't think of it earlier , just touch the radiator end tanks ... they're cool to the touch

 

Did you solve it. I find it weird people need so many rads. Just finished my loop with 7900XT and 5800X3D running a single 420 rad and D5 pump. Max fan speed 800 RPM. Running rad as top exhaust. CPU temp max 81C, GPU: 53C (hotspot 75C).  Ambient ~ 20-25C. 

[TheRedBaron_94]

On 10/17/2024 at 12:51 PM, DoctorNick said:

Did you solve it. I find it weird people need so many rads. Just finished my loop with 7900XT and 5800X3D running a single 420 rad and D5 pump. Max fan speed 800 RPM. Running rad as top exhaust. CPU temp max 81C, GPU: 53C (hotspot 75C).  Ambient ~ 20-25C. 

i didn't solve it entirely , I think I just have a less than ideal CPU block.  ( i did manage to find about an additional 5C by reversing the flow in the block though ) 
Also , i think our CPU's are a bit different , i'm running a 13700kf , which has a bit of a "hot boi", where the 5800x3d isn't really all that hot of a chip 
 

[DoctorNick]

10 hours ago, TheRedBaron_94 said:

i didn't solve it entirely , I think I just have a less than ideal CPU block.  ( i did manage to find about an additional 5C by reversing the flow in the block though ) 
Also , i think our CPU's are a bit different , i'm running a 13700kf , which has a bit of a "hot boi", where the 5800x3d isn't really all that hot of a chip 
 

True, but still, do you really need 4x 360 rads? 2 should be more than enough.  

Also this loop order instead: Pump -> GPU -> CPU -> Rad(s)

[NorKris]

On 10/18/2024 at 3:33 PM, TheRedBaron_94 said:

i didn't solve it entirely , I think I just have a less than ideal CPU block.  ( i did manage to find about an additional 5C by reversing the flow in the block though )

5c? thats more then i would have guessed  

15 hours ago, DoctorNick said:

True, but still, do you really need 4x 360 rads? 2 should be more than enough.  

Also this loop order instead: Pump -> GPU -> CPU -> Rad(s)

Rad between gives u 1c  

[thrasher_565]

its a balance between rads and flow speed the more fittings, longer tube and more rads slows the flow but as long as you got enough should be fine. if the flow is too slow you would need a second pump to speed it back up. but also depends on how much heat there is. if one or 2 rads can cool it enough then it wont matter. the jet plate slows the speed of the flow so that's the bottle neck but cools better then with out it.

 

the other problem is the chips on the cpu are harder to cool as there smaller so seeing a huge temp drop most likely wont happen.

 

its also importend to test apples to apples same setup as your aio build and then swap over to really see the difference.

 

air bubbles in the loop will also slow the flow witch it looks like you have.

 

as to why the aio is cooler then the loop i don't no/ the gpu is adding over all posably more heat then the aio but posable the aio mounted differently maybe an offset mount

Edited October 19, 2024 by thrasher_565

[NorKris]

45 minutes ago, thrasher_565 said:

as to why the aio is cooler then the loop i don't no/ the gpu is adding over all posably more heat then the aio but posable the aio mounted differently maybe an offset mount

my guess:
AIO has higher water temp but can mange to get more of the CPU-heat into the water
the Loop as colder water but the coldplate is so bad its not transferring the heat good enough 

[thrasher_565]

16 minutes ago, NorKris said:

my guess: AIO has higher water temp but can mange to get more of the CPU-heat into the water
the Loop as colder water but the coldplate is so bad its not transferring the heat good enough 

ya dont no. dont no what aio or if it uses a offset. so many unknowns. looks like 2 rads with 1 set of fans too as well as bubbles...

 

the aio would also have low around of restriction and probly faster flow too. way to test would be put one rand on a cpu only loop and see.

 

or maybe is dumping so much heat its increasing the room temp but it would have to be quit hi thow

 

i dont think any cpu blocks comes with offsets ?

Edited October 19, 2024 by thrasher_565

[NorKris]

17 minutes ago, thrasher_565 said:

ya dont no. dont no what aio or if it uses a offset. so many unknowns. looks like 2 rads with 1 set of fans too as well as bubbles...

 

the aio would also have low around of restriction and probly faster flow too. way to test would be put one rand on a cpu only loop and see.

 

i dont think any cpu blocks comes with offsets ?

he told us arctic 420 

and i dont think  custom-cpublocks need offsets 

 

18 minutes ago, thrasher_565 said:

 

or maybe is dumping so much heat its increasing the room temp but it would have to be quit hi thow

 

this would happen in both setups (the same lvl of heat) 

[thrasher_565]

17 minutes ago, NorKris said:

he told us arctic 420 

and i dont think  custom-cpublocks need offsets 

 

this would happen in both setups (the same lvl of heat) 

ya but is there an offset or not?

 

if the aio needs an offset so would a cpu block i would think? you need the block in the center too cool the best. and amd and intel have different mount and how the cooler should be mounted. one is more concave and other convex. myabe the aio had the proper mout for the cpu the block dose not. or poorly mounted.

 

the aio might not get all of the gpu heat thow

 

but don't no

 

contact is going to be really impotend. ever more so with an oc

Edited October 19, 2024 by thrasher_565