in dual 240 radiator setup, the only exhaust fan should be the top slim 120mm fan
-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
Log SMALLEST ATX CASE : CERBERUS-X (Complete build log)
- Thread starter MarcParis
- Start date
Here are some news of my setup..
Currently I don't have a lot of times to optimize setup, however, since my last benchmarks, I made some interesting discoveries :
WHEA :
Windows Hardware Error are a critical point to check while overclocking (or even not). I've read a lot of type of errors, and I'm far from behind alone to face that on Ryzen. Root cause is coming from CPU micro code and Motherboard bios. Due to "marketing" purpose (or if I'm gentle, bios maturity on new Ryzen platform), motherboards seem to have unlock / badly setup some paramters resulting in those WHEA (facing them heavy cpu load, ie intel burn test with linpack is pretty quick to generate whea, cinebench, video encoding...). I don't really if it was done by purpose (hoping not), it's resulting simply that OC a ryzen cpu is more difficult now compared to earlier bios. Ryzen cpu needs more juice than before, resulting to a decrease of OC at same vcore...
On bios 1403 of Crosshair Hero VI (first to implement agesa 10.0.06a) i was able a stable 4.0ghz @1,52v or a 3.9Ghz @1,4375v back to july-august 2017)...however it was only supercially stable as I've redone all my testings...and I was never able to avoid WHEA even on 3.9Ghz...I even got some WHEA @3.8Ghz with this Bios 1403...holy crap.
I guess that the reason why Asus is continuing to publish new bios (ie 1701) with a vague note of "improving stability"...indeed with this bios 1701, I'm able to run an OC 3,85Ghz @1,4375v smooth and stable, WHEA free..
I'm pretty sure also that's why Asus and others motherboards manufacturers are not putting in place new agesa 10.0.0.6b or even 10.0.0.7 (introducing new architecture of micro code...and new ryzen-vega apu).
COLD START :
Ryzen is still struggling to properlly train RAM on cold start. Except a specific beta bios that was focused to help b-die ram kit on crosshair, most of the time, ram support on ryzen needs improvement.
However a nice feature was introduced to simply specify numbers of retries when booting. In my personal case I just put 3 times, and computer is booting, rebooting till everything is not ok. As a result, my setup is booting on second attempt as it required only a slight warm ram..
In final this retry tweak is taking as long as "specific" training from beta bios 9920...so fine for me.
COOLING CAPACITY :
I'll praise again combo Cerberus-X and Noctua C14S with dual 140 fans...I didn't not realize that cooling power since I saw some benchmarks between noctua coolers vs asetek kit on Threadripper...with a surprising victory from Noctua U14S over nzxt X62...
What I can witness with my Ryzen setup is that Noctua C14S is able to handle a OC Ryzen 7 (summit ridge) @3,9Ghz, @1,4375v (around 150W), bios 1701 Asus C6H (creating some whea by the way) stable @76°C (room around 20°C). Versus a asetek 240mm AIO solution, it should be around 5°C warmer maximum (after temperature balance, that is taking 20-40 min on watercooling solution)
The only way to be sure will be to put again my Apogee drive II in position..
No matter I'll try to optimise case fan flow (rear and top) and I'll let you know..
I'm eager to see other builds in Cerberus-X as everybody has their own solutions/thoughts..
Currently I don't have a lot of times to optimize setup, however, since my last benchmarks, I made some interesting discoveries :
WHEA :
Windows Hardware Error are a critical point to check while overclocking (or even not). I've read a lot of type of errors, and I'm far from behind alone to face that on Ryzen. Root cause is coming from CPU micro code and Motherboard bios. Due to "marketing" purpose (or if I'm gentle, bios maturity on new Ryzen platform), motherboards seem to have unlock / badly setup some paramters resulting in those WHEA (facing them heavy cpu load, ie intel burn test with linpack is pretty quick to generate whea, cinebench, video encoding...). I don't really if it was done by purpose (hoping not), it's resulting simply that OC a ryzen cpu is more difficult now compared to earlier bios. Ryzen cpu needs more juice than before, resulting to a decrease of OC at same vcore...
On bios 1403 of Crosshair Hero VI (first to implement agesa 10.0.06a) i was able a stable 4.0ghz @1,52v or a 3.9Ghz @1,4375v back to july-august 2017)...however it was only supercially stable as I've redone all my testings...and I was never able to avoid WHEA even on 3.9Ghz...I even got some WHEA @3.8Ghz with this Bios 1403...holy crap.
I guess that the reason why Asus is continuing to publish new bios (ie 1701) with a vague note of "improving stability"...indeed with this bios 1701, I'm able to run an OC 3,85Ghz @1,4375v smooth and stable, WHEA free..
I'm pretty sure also that's why Asus and others motherboards manufacturers are not putting in place new agesa 10.0.0.6b or even 10.0.0.7 (introducing new architecture of micro code...and new ryzen-vega apu).
COLD START :
Ryzen is still struggling to properlly train RAM on cold start. Except a specific beta bios that was focused to help b-die ram kit on crosshair, most of the time, ram support on ryzen needs improvement.
However a nice feature was introduced to simply specify numbers of retries when booting. In my personal case I just put 3 times, and computer is booting, rebooting till everything is not ok. As a result, my setup is booting on second attempt as it required only a slight warm ram..
In final this retry tweak is taking as long as "specific" training from beta bios 9920...so fine for me.
COOLING CAPACITY :
I'll praise again combo Cerberus-X and Noctua C14S with dual 140 fans...I didn't not realize that cooling power since I saw some benchmarks between noctua coolers vs asetek kit on Threadripper...with a surprising victory from Noctua U14S over nzxt X62...
What I can witness with my Ryzen setup is that Noctua C14S is able to handle a OC Ryzen 7 (summit ridge) @3,9Ghz, @1,4375v (around 150W), bios 1701 Asus C6H (creating some whea by the way) stable @76°C (room around 20°C). Versus a asetek 240mm AIO solution, it should be around 5°C warmer maximum (after temperature balance, that is taking 20-40 min on watercooling solution)
The only way to be sure will be to put again my Apogee drive II in position..
No matter I'll try to optimise case fan flow (rear and top) and I'll let you know..
I'm eager to see other builds in Cerberus-X as everybody has their own solutions/thoughts..
-------------------------------------------------------------------------------------------------------------------------------
Update 19th November 2017 : Back to watercooling!
-------------------------------------------------------------------------------------------------------------------------------
Well, well, well...
You already know that I love with my Apogee Drive II. Somehow it was a pain to keep it in a carton box as I struggled a lot to get it and make it work...
I had also the feedling that Cerberus-X could offer a great cooling capability...and I felt satisfied with my noctua C14S but not satiated...
I also wanted to know if my customed AIO was really a AIO..
Serial release of Cerberus is inspiring me..
For all these good and bad reasons, i went back to watercooling on CPU..
This time I will fight hard to keep GPU temperature and fan speed at lowest as possible whereas I'm using a watercooled CPU...with a radiator located at the bottom.
Let's call this watercooling 2.0 as I applied quite a lot of changes.
But first here some pictures :
As you can see, airflow setup vs my first attempt is really different.
COMMON SETUP
It's too early to have to have all benchmarks ran...but initial results were interesting..
What I'm sure of, is that setup is far quieter in CPU stress than my previous aircooling solution..
I'll keep you informed asap..
Here is Imgur Album :
Update 19th November 2017 : Back to watercooling!
-------------------------------------------------------------------------------------------------------------------------------
Well, well, well...
You already know that I love with my Apogee Drive II. Somehow it was a pain to keep it in a carton box as I struggled a lot to get it and make it work...
I had also the feedling that Cerberus-X could offer a great cooling capability...and I felt satisfied with my noctua C14S but not satiated...
I also wanted to know if my customed AIO was really a AIO..
Serial release of Cerberus is inspiring me..
For all these good and bad reasons, i went back to watercooling on CPU..
This time I will fight hard to keep GPU temperature and fan speed at lowest as possible whereas I'm using a watercooled CPU...with a radiator located at the bottom.
Let's call this watercooling 2.0 as I applied quite a lot of changes.
But first here some pictures :
As you can see, airflow setup vs my first attempt is really different.
COMMON SETUP
- CPU Waterblock/Pump : Swiftech Apogee Drive II (using AM3/AM4 bracket). Combination of Apogee HD waterblock + full laing DDC pump (MCP35X)..
- Radiator/Reservoir : Swiftech MCR H220 radiator, 240mm radiatoir used in H220 AIO (back to 2013). You can use Swiftech MCRx20-QP as alternative
- Radiator fans (Intake) : Dual LEGENDARY Scythe Gentle Typhoon AP15 (1850rpm, managed through DC)
- Top Fan : Noctua a12x15 (slim 120mm fan) placed in exhaust
- Rear fan : Noctua NF-A9x14 PWM placed in intake (with filter)
- Side Panel : Vents oriented on the top (vs bottom on first setup)
- Side Fan : Noctua a12x15 (slim 120mm fan) placed in exhaust (vs intake vs plain panel...)
- Front panel : Addition of a 120mm LEGENDARY Scythe Gentle Typhoon AP15 (1850rpm, managed through DC). This addition is completely new!
- Corsair Commander mini : It is now located on the motherboard panel, just below bottom radiator and front fan
-
- I also add a UPS : Eaton Ellipse Eco 1200 (1200VA/750W)..very useful while updating bios and/or gaming..
-
It's too early to have to have all benchmarks ran...but initial results were interesting..
What I'm sure of, is that setup is far quieter in CPU stress than my previous aircooling solution..
I'll keep you informed asap..
Here is Imgur Album :
Result in gaming is here...and it's tremendous..
Back to 4Ghz OC @1.525v (or maybe little bit higher). Ok I've some WHEA on intel burn test...but nothing while gaming..
Look at GPU temperature and fan speed...HUGE progress from original watercooling setup.
Thus all my changes of cooling (side panel orientation, front fan, side exhaust fan) are pretty effective..
Back to 4Ghz OC @1.525v (or maybe little bit higher). Ok I've some WHEA on intel burn test...but nothing while gaming..
Look at GPU temperature and fan speed...HUGE progress from original watercooling setup.
Thus all my changes of cooling (side panel orientation, front fan, side exhaust fan) are pretty effective..
As gaming temperature and noise are completely fine now, here are more serious CPU OC benchmarks..
For the fun, here is result of 4Ghz OC @1.525v vcore, RAM 3333 stitls fast settings :
Stable but with tons of WHEA..
Finally I found my sweet OC spot with 3900Mhz OC @1.4875v...with no WHEA, even after good warm up..
Nice and sweet..
I'll be able to come back @4Ghz with any benchmark activity (hwbot, etc...)
I love this new setup!!
For the fun, here is result of 4Ghz OC @1.525v vcore, RAM 3333 stitls fast settings :
Stable but with tons of WHEA..
Finally I found my sweet OC spot with 3900Mhz OC @1.4875v...with no WHEA, even after good warm up..
Nice and sweet..
I'll be able to come back @4Ghz with any benchmark activity (hwbot, etc...)
I love this new setup!!
Last edited:
Well i should be around 7-8 builds already inside cerberus-x..Again a full cooling rebuild ? You must love building in this case ! Stop it
I’m taking my pioneer role in cerberus-x very seriously..
I must admit that my last build has raised cooling (and quiet) to a next level...it’s a succesful build..
Hope this will help future owner of cerberus-x..
Never unfortunately as any waterblock will be too tall for cerberus-x.
And to be honest, watercooling a pascal gpu vs a good aftermarket aircooler is useless as you won’t be able to overclock more..
My next step is to replace my front 120mm fan. My aim will be more airflow and quietness. I’m thinking of be quiet silent wings 3 pwm (normal speed).
But i’m still inquiring on it.
And to be honest, watercooling a pascal gpu vs a good aftermarket aircooler is useless as you won’t be able to overclock more..
My next step is to replace my front 120mm fan. My aim will be more airflow and quietness. I’m thinking of be quiet silent wings 3 pwm (normal speed).
But i’m still inquiring on it.
In my personal setup, radiator 29mm and fans 25mm.
After that’s only space taken by radiator. You need to take into account space used by tubing.
I guess the lowest slot could be only used for half size card.
I am about to place an order for a couple of magicool g2 240mm slim rads, they measure at 27mm and have a nice amount of fpi (between 14 and 16).
Couple that with some super expensive noctua 15mm fans (they run anywhere from $35 to $50 aud).
Will give them a dry run before before I commit the rads to a case.
Couple that with some super expensive noctua 15mm fans (they run anywhere from $35 to $50 aud).
Will give them a dry run before before I commit the rads to a case.
sure magicool G2 240mm is a great choice for 27mm/slim radiator. They are better than alphacool and EK for this thickness.
However, king of single row radiator is hardware labs GTS 240 variant (details here)
Width 133 mm
Height 278 mm
Thickness 29.60 mm
Only advice with them is to make sure to clean them with distilled water before usage..
However, king of single row radiator is hardware labs GTS 240 variant (details here)
Width 133 mm
Height 278 mm
Thickness 29.60 mm
Only advice with them is to make sure to clean them with distilled water before usage..
Or the darkside lp240. hardware labs tend to be wider than most, I'd prefer to keep as close to 240mm as I can to keep it clear of any connectors at the bottom of the board.
I was thinking about the xspc ion pump/res, 3/8-1/2 flexible tubing.
Still undecided about the waterblock, as we don't know how much space we have with the SFX adapter.
Am also tinkering with the idea of a crossover rad in the bottom for tube management.
Still undecided about the waterblock, as we don't know how much space we have with the SFX adapter.
Am also tinkering with the idea of a crossover rad in the bottom for tube management.
