Log SMALLEST ATX CASE : CERBERUS-X (Complete build log)

threestripevida

Airflow Optimizer
Mar 28, 2017
277
466
Definitely happy with my choice to get the 645lt. It doesn't make much sense considering you can fit bigger coolers in this case, but since I have the window panel I can't put the radiator at the front without the psu right in front of my build. GPU temps dropped on average 10c, going from 75-77c to 65c max. Noise was amazing as well. My build is pretty damn quiet when gaming and all of the fans were going at 1000-1100rpm max. I'm having to use the top exhaust and front intake fan together with a fan splitter. I was thinking about getting a Corsair commander to give myself more control over the bottom and front intake fans for the gpu instead of the fan curve being based on the cpu thermals. Overall I am very happy with the 645lt.
 
  • Love
Reactions: MarcParis

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
Definitely happy with my choice to get the 645lt. It doesn't make much sense considering you can fit bigger coolers in this case, but since I have the window panel I can't put the radiator at the front without the psu right in front of my build. GPU temps dropped on average 10c, going from 75-77c to 65c max. Noise was amazing as well. My build is pretty damn quiet when gaming and all of the fans were going at 1000-1100rpm max. I'm having to use the top exhaust and front intake fan together with a fan splitter. I was thinking about getting a Corsair commander to give myself more control over the bottom and front intake fans for the gpu instead of the fan curve being based on the cpu thermals. Overall I am very happy with the 645lt.
So glad this move to 645LT was so successful!
To increase cpu cooling you can still put slim 92mm fan between case and radiator. You should gain 2-3°c.

Regarding your fans setup, i'm not sure front fan is still useful. Otherwise Corsair fan controller remains the benchmark.
 
Last edited:

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
-------------------------------------------------------------------------------------------------------------------------------
Watercooling 92mm inside Cerberus-X : Pushing the limit
-------------------------------------------------------------------------------------------------------------------------------

After checking again xtremerigs review of Hardware labs Nemesis GTX 140, I just realize that I should have noticeable gain by moving from push only to push/pull fan setup on my Nemesis GTX 92 radiator :

From this study you can also conclude the following :
  • Double row (approx 60mm thick) radiator in push (one fan) are performing similarly than single row (approx 30mm thick) in push/pull : result I find again as my Nemesis 92mm with 1 fan was performing just 1°C better than push/pull on Asetek 645LT
  • Push-Pull fans setup are bringing better results on double row radiator (+23,3%) than on single row (+16,6%)
  • Hardware labs have best copper radiator for both single and double row..:D

Thus I decided to test push/pull on my 92mm watercooling setup.

First I tried using Noctua NF-A9x14 (2500rpm and 2200rpm variant) fan between rear panel and radiator. It was working..BUT, BUT, Noctua NF-A9x14 was awfully loud. I figured out that clearance was too small between panel and propeller/blades, creating some suction effect. Noise I was hearing was coming from motor, not from air turbulence. Thus placing NF-A9x14 (blades) too close to panel is creating higher air resistance that fan motor will struggle to face peacefully.
In a first conclusion : NEVER USE NF-A9x14 as PULL against a too obtructive surface. Grid for Asetek 645LT or Grid from Corsair PSU are completely OK. But case mesh/perforated panel could create this effect.

Then I decided to try my good old Noctua NF-B9 Redux fan...not the best fan choice for radiator, but it could work. I also used only 2 screws for each fan...not a big issue. And it worked as intended! No excessive noise..:D


I was expecting 2-3°C of improvement versus pull only...but I got 4°C of improvment!! My Ryzen 7 3800X overclocked to 4.3Ghz @1,3125v was sitting at 78°C maximum on Intel Burn Test 2.54 (Very High profile) (vs 82°C with 1 fan in pull)


On addition I'm very pleased with my barbed fittings and hose clamps as they are preventing any rotation of tubing (thanks hose clamps in contact with fan or screws..;)) Mayhems tubing 10-16mm were pretty easy to bend also.

I will replace my NF-B9 by another NF-A9 (brown and 25mm)...but I don't expect any noticeable temperature gain (max 1°C)

Overall my 92mm watercooling experience is really positive, even better than my expectations. It also just reminded me AIO influence on watercooling market with their 240mm slim & aluminium radiator whereas a double row & copper120mm radiator in push/pull can easily do the job...:D
Here is a picture of one of my previous setups (late 2009, Core i7 920 OC 4Ghz + chipset X58 + 2xGTX 285 fully loaded on Folding@home...just cooled with 2x120mm double row radiator, one in push-pull and one in push on Gentle Typhoon 1850rpm)

Now I'm sure even R9 3900X or 3950X could be easily cooled by 92mm watercooling...:D

Guess who is happy now, as he can lay peacefully on Cerberus-X?...:D
 
Last edited:

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
I'm a bit shocked now with my 92mm watercooling.

I just compared CPU temperature with my Noctua C14S versus my customed 92mm watercooling in push/pull.

On Intel Burn Test 2.54 (very high for watercooling, max for C14s) :
  • Asus WMI monitoring :
    • Noctua C14S (exhaust, Push/pull) : 84°C (link to post)
    • Asetek 645LT (Push/Pull) : 81°C (link to post), Gap vs Noctua C14S : 3.0°C (IBT 2.54 add. runs showed about 83°C at the end)
    • Customed WC 92mm Pull : 82°C (link to post), Gap vs Noctua C14S : 2.0°C
    • Customed WC 92mm Push/pull : 78°C (link to post), Gap vs Noctua C14S : 6.0°C
  • AMD CPU monitoring :
    • Noctua C14S (exhaust, Push/pull) : 89.0°C (link to post)
    • Asetek 645LT (Push/Pull) : 81.8°C (link to post), Gap vs Noctua C14S : 7.2°C (IBT 2.54 add. runs showed about 84°C at the end)
    • Customed WC 92mm Pull : 84,9°C (link to post), Gap vs Noctua C14S : 4.1°C
    • Customed WC 92mm Push/pull : 79.5°C (link to post), Gap vs Noctua C14S : 9.5°C
I noted interesting gap between Asus and AMD monitoring. I will assume AMD monitoring is more precise..:D

Without saying that Noctua C14S cooling performance is largely impacted when GPU is loaded, whereas it's not the case on customed WC 92mm.

Watercooling 92mm (customed or AIO) is clearly ahead of Noctua C14S. Pretty sure it's mainly related to case impact on air cooler.
What do you think?
 

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
As written I updated my good old NF-B9 by NF-A9 in push/pull system.




As expected no real performance win, except I was able to reach again my overclocking of 4.3Ghz and 3800MTS on RAM (1900Mhz FCLK) with no temperature increase...RAM/FCLK OC is bringing something like 5W-6W in addition.



OC R7 3800X is reaching 79.5°C (AMD monitoring) or 78°C (Asus monitoring) as max & stable temperature on 100% load of IBT 2.54 (very high).
Like this all my testings are now at same level of OC.

Overall I'm really happy with this 92mm customed watercooling : simple, space efficient & quite powerful..:D
 
  • Like
Reactions: dealda

ywori

Trash Compacter
Jul 6, 2018
38
30
I think that hardwarelabs 92mm is the same one that I am using. Because of its thickness, I think it covers roughly the same volume as a normal 140mm RAD. Hence the performance you are seeing.
 

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
I think that hardwarelabs 92mm is the same one that I am using. Because of its thickness, I think it covers roughly the same volume as a normal 140mm RAD. Hence the performance you are seeing.
Yes similar performance of slim 140mm rad..:)
 

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
Your water cooling solution is great. I think I might try something similar in the future. Can you provide a parts breakdown for the CPU cooler you have setup?

Cheers

John H
Thanks a lot.

No issue, here is list of material :
  • CPU wtaerblock/pump : Alphacool eisbaer LT
  • 2x 90° barbed fitting 10mm (alphacool)
  • 2x straight barbed fiiting (XSPC)
  • 4x hose clamp 16mm
  • Mayhems tubing 10-16mm transparent
  • Liquid : EK cryoful red blood+distilled water
  • Radiator : Hardware labs Nemesis M92 GTX (be careful of IN/OUT, it matters on double row radiators)
  • M4x28mm screws (provided with radiator)
 
Last edited:

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
Thanks a lot.

No issue, here is list of material :
  • CPU wtaerblock/pump : Alphacool eisbaer LT
  • 2x 90° barbed fitting 10mm (alphacool)
  • 2x straight barbed fiiting (XSPC)
  • 4x hose clamp 16mm
  • Mayhems tubing 10-16mm transparent
  • Liquid : EK cryoful red blood+distilled water
  • Radiator : Hardware labs Nemesis M92 GTX (be careful of IN/OUT, it matters on double row radiators)
  • M4x28mm screws (provided with radiator)
I just added 4x M4x30mm to hold radiator on case side...;)
In total :
  • 4xM4-30mm to hold radiator through case
  • 4xM4-28mm to hold fan (provided with radiator)
 

sumone

What's an ITX?
Feb 13, 2020
1
1
Great write up, have been thinking about putting together a build based on the matx version. you might have convinced me to put the order in. thanks.
 
  • Love
Reactions: MarcParis

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
Great write up, have been thinking about putting together a build based on the matx version. you might have convinced me to put the order in. thanks.
Thanks for your comment.
I guess you won't regret it.

If you have any question, please let me know
 

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
-------------------------------------------------------------------------------------------------------------------------------
Off Topic : Cryorig C7+ (Aluninium fins with graphene coating + 4000rpm fan)
-------------------------------------------------------------------------------------------------------------------------------

After some drama about product change between order on reception (Drama C7 SFFTEC), I decided finally to make a short review about Cryorig C7+ (standard C7 with graphene coating + name pwm fan 4000rpm).

This C7 cooler is a customed version coming from China directly.
https://www.sfftec.com/product-page/cryorig-c7-plus

My original idea was to transfer my Cerberus-X setup inside Sliger SM550/560 case. Thus I was looking at tough and compact cpu cooler. On my HTPC (using SM550), I finally went for IS-60 as I bought compatible RAM and motherboard.
As leadtime to get, either IS-60 or Cyrorig C7 "graphene", is pretty long I bought them before Christmas. IS-60 arrived first and proved to be a very good choice.
https://smallformfactor.net/forum/threads/sliger-sm550-compact-quiet-gaming-htpc.11928/

I finally received C7 Graphene by end of January/early February.

After drama on technical specs changes between "original" Cryorig C7G (basically C7 copper version with graphene coating) and actual/received product (standard C7 aluminium fins with graphene coating), I decided to review this C7+ using my setup inside Cerberus-X.

Here is packaging with specs :


"Graphene" Coating looks nice

Compared to original C7 I appreciate to dedicated fans brackets, that allows to switch fan easily :

Speaking of the fan, included one is well spinning at 4000rpm (pwm control). To be honest, it was less noisy as expected at 100%. However at lower speed, friction noise was audible. But as stated, it's really easy to switch fan (dislike original C7)...
Please note that some glue spots are acting like noise dampeners. However you cannot use this fan with screw (blind holes)



Inside Cerberus-X, I revised slightly airflow to provide fresh air to short C7+..:)


As it was easy to switch fan, I also tested Noctua NF-A9x14 (2500rpm) and NF-A9x25 (2000rpm).



In terms of testing method, I cannot use with my standard method (R7 3800X overclocked 4.3Ghz with RAM 3800MTS) as I know it won't do it...(in fact I tested it...leading to CPU over temperature protection, black screen & reboot..:D).
I wanted to know if C7+ will be able to handle stock R7 3800X (ram 3200MTS) and what will be sustained speed at which temperature.

For information, at stock, Ryzen processors are monitoring their temperature and adjust its core speed to stay within safe temperature range.
Max operating temperature is 95°C for Zen 2 cpu.

Here are results :
With included 4000rpm fan :


With Noctua NF-A9x14 @2500rpm fan :


With Noctua NF-A9x25 @2000rpm fan :


First surprise is coming from global performance :
  • On idle/low load CPU is sitting at around 55-65°C and is really quiet.
  • On Torture test, R7 3800X is reaching 89°C with core frequencies between 4.0ghz and 4.1Ghz (with some peaks to 4.2ghz). As comparison, my customed 92mm watercooling is providing 78°C @4.3ghz...completely different performance league.
Amongst 3 fans tested, included 4000rpm proved to be the coolest by small amount (1.8°C). Even in terms of noise, all fans tested were loud.

Conclusion time.
What can you do you this Cryorig C7+?
Well you can cool efficiently & quietly a Ryzen 7 3800X at stock settings (RAM at XMP). It will be audible on heavy load but will remain quiet otherwise.
It will be perfect for gaming and light editing machine (10-15min 4K movies).

My only reserve will be its actual performance on AM4 mini-ITX motherboards on which you cannot choose heatsink orientation. By default hot air will be exhausted towards RAM and rear I/O. On my ATX board, it is not an issue as clearance around CPU is larger than on mini ITX. But it should not change its performance from day to night.

I got similar feelings & results than with IS-60 but different constraints.
 

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
I’ll continue C7 experiment during 1 or 2 weeks...just to see how it behaves.

I’ll also try again c14s...:)

My main fear ? Judging C7 is just right for my use case...leading to sandwich case again...like sidearm or sm550/560...
 

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
Slowly but surely I'm continuing my testing of C7 with graphene coating.

I just simply applied an offset of -0.1v on vcore (stock R7 3800X otherwise) and provided huge gains (-6/7°c) on temperature :


I already tested C7 while gaming (Assassin's creed Odyssey, quite cpu intensive) with pure stock R7 3800X and it went really fine (frequencies & temperature).
I'll redo my testing using this undervolted R7 3800X...:D

Embarassing questions will be raised (common to Zen 2 cpu from R5 to R7) :
  • Do you still need to practice manual overclocking?
  • Why do you need beefier cpu cooler?
I'm sure headaches will come to me soon...:D
 
Last edited:
  • Like
Reactions: rfarmer

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
Well I guess I'm experiencing similar issues as this video :


I noticed some performance drop with vcore setup as offset..:D Apparently monitoring tools are not that accurate lol..;)

I'll let you posted
 

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
In fact I noticed something pretty interesting.

While on stock settings on CPU, this later is adapting itself to vcore, if set on offset mode (at least on my asus crosshair VI).

At vcore offset of -0.1v, I just noticed minimal to not noticeable performance drop (Intel Burn Test Very High completed in 976sec, completely in line to full stock performance).


However at vcore offset of -0.15v, I experienced significant performance drop (Intel Burn Test Very High completed in 1037sec, 6% gap versus stock).


Good point is coming from thermals..:)

To sum up :
  • Stock R7 3800X (C7 Graphene coating+NF A9x14 2500rpm) :
    • IBT 2.54 Very High : 978.25sec
    • Max temperature : 89.3°C
    • Max CPU Package Power : 115.962W
  • Stock R7 3800X (C7 Graphene coating+NF A9x14 2500rpm) (vcore -0.1v) :
    • IBT 2.54 Very High : 976.72sec (-0,15% vs stock)
    • Max temperature : 82.6°C (-6,7°C, ie -7.5%)
    • Max CPU Package Power : 107.170W (-7,5%)
  • Stock R7 3800X (C7 Graphene coating+NF A9x14 2500rpm) (vcore -0.15v) :
    • IBT 2.54 Very High : 1037.22sec (+6% vs stock)
    • Max temperature : 78.8°C (-10,5°C, ie -11.5%)
    • Max CPU Package Power : 92.788W (-19,98%)
  • Overclocked R7 3800X 4.3Ghz @1.3125v RAM @ 3800MTS (customed 92mm watercooling, push/pull) :
    • IBT 2.54 Very High : 917.12sec (-6%)
    • Max temperature : 79.5°C
    • Max CPU Package Power : 118.33W
Thus on stock settings, it's important to test performance to see if there is no unwanted/hidden underclock/underperformance..:D

Lowering vcore by -0.1v seems to be sweet spot, especially for Cryorig C7 with graphene coating, but also globally..:D Testing to be continued..:)


As reference point, I reminded score with my customed 92mm watercooling setup with R7 3800X at 4.3Ghz (Ram @3800Mhz). Strangely enough, each time I tried fixed OC with fixed vcore with C7, I finish to got reboot (cpu temperature too high stated in bios)....however while monitoring cpu was just at 83°C.
 
  • Like
Reactions: Jerdz

MarcParis

Spatial Philosopher
Original poster
Apr 1, 2016
3,663
2,770
-------------------------------------------------------------------------------------------------------------------------------
Return of the king : Noctua C14S
-------------------------------------------------------------------------------------------------------------------------------

Well I will surely feel ashamed in the below lines...but let's say, only people that are doing nothing do not make mistakes..:D

After pretty positive experience with Cryorig C7 (and behaviour of turbo boost of R7 3800X), I decided to test again Noctua C14S, this time in full intake mode, like 3 years ago when I start playing with Cerberus-X...:D

Now was perfect timing to test again Noctua C14S as
  • Proper protocol of testing CPU
  • Proper & recent benchmarks results based on comparable parameters
  • Similar weather conditions

For installation I tried to install Noctua C14S with heatpipes orientated towards top of the case



However, heatsink was slightly touching top fan (NF-A12x15)...so I was forced to stick to traditional orientation..;)



Here is fan setup :
  • Noctua C14S :
    • Top : NF-A12x25 intake, put on side bracket
    • Bottom NF-A14 intake
  • Top & Rear : NF-A12x15 & NF-A9 exhaust

Benchmarks : Well i will complete my previous list. All screenshots will be at the end of this post.
  • Overclocked R7 3800X 4.3Ghz @1.3125v RAM @ 3800MTS (Noctua C14S Push/Pull Intake) :
    • IBT 2.54 Very High : 915.80sec
    • Cinebench R20 (multi) : 5120
    • Max temperature : 73.9°C
    • Max CPU Package Power : 119.23W
  • Stock R7 3800X (Noctua C14S Push/Pull Intake) :
    • IBT 2.54 Very High : 950.25sec
    • Max temperature : 75.1°C
    • Max CPU Package Power : 122.362W
  • Stock R7 3800X (Noctua C14S Push/Pull Intake) (vcore -0.1v) :
    • IBT 2.54 Very High : 946.00sec
    • Max temperature : 69.9°C
    • Max CPU Package Power : 107.640W
  • Stock R7 3800X (C7 Graphene coating+NF A9x14 2500rpm) :
    • IBT 2.54 Very High : 978.25sec
    • Max temperature : 89.3°C
    • Max CPU Package Power : 115.962W
  • Stock R7 3800X (C7 Graphene coating+NF A9x14 2500rpm) (vcore -0.1v) :
    • IBT 2.54 Very High : 976.72sec
    • Max temperature : 82.6°C
    • Max CPU Package Power : 107.170W
  • Overclocked R7 3800X 4.3Ghz @1.3125v RAM @ 3800MTS (customed 92mm watercooling, push/pull) :
    • IBT 2.54 Very High : 917.12sec
    • Max temperature : 79.5°C
    • Max CPU Package Power : 118.33W

Conclusion time :
  • Noctua C14S Push/Pull (intake) is beating by 5°C my customed 92mm watercooling. My previous results with C14S (full exhaust) were simply detoriated by hot ambiant temperature. I did my former C14S during heatwave, important detail I forgot (my bad).
  • At stock settings, vcore offset of -0.1v proved again to be very useful on R7 3800X
  • I still need to check behaviour of Noctua C14S intake on GPU temperature. After months of testing I can say there is NO significant impact on GPU temperature using Noctua C14S with side intake.
  • Zen 2 boost likes bigger heatsinks as it allows higher boost clocks (Noctua C14S is beating Crorig C7+ by 3% at stock settings)
  • Cryorig C7+ remains impressive for its size. No issue at all to manage R7 3800X at stock or slightly undervolted. Undervolting on Zen 2 is really easy as boost is automatically it

Benchmarks screenshots :



 
Last edited: