3900x/3950x cooling options and reviews

ik9000

ik9000

ik9000

Soldato
Joined
23 Nov 2019
Posts
3,340
Can anyone point me to some decent reviews/tests of coolers for 3900x and 3950x? Google keeps giving me generic lists of 10 options but no meaningful test comparisons.

Can I get away with high end air cooling or does it have to be aio or water cooled? My case can handle big air coolers but only has mounts for 2*120mm (240mm) size aio radiators (unless I mount externally when 360mm would be possible, albeit weird looking.)

Case is bitfenix shinobi XL.
I'll be using all the 5" and won't take the HDD bay out
Mobo MSi Ace x570
Ram 4*8GB trident 3600 cl16 non RGB. Not sure if they count as low profile or not.
 
Shinobi XL has 181mm CPU clearance.
Don't know of or find any 'Trident 3600' ..
Trident Z = 44mm,
Trident Y = 54mm w/ fin, 39mm w/o fin
Trident = 55mm​
Top of RAM is 5mm less than it's height measurement above top of CPU .. 55mm RAM stands 50mm above top of CPU.
MSI Ace X570 has 92mm center CPU to near side of PCIe socket but only 48.7mm center CPU to near side of RAM socket.

This means fan on front of single tower coolers (approx 52mm center CPU to front of fan) will put front of fan over nearest RAM socket. Even offset coolers like Thermalright ARO M14G / M14O with back offset have front of fan 52.2mm in front of center CPU. If you haven't already have RAM I would look for some that is not so tall. The PCB of DDR4 is only 31.35mm tall, so anything taller is not physically a part of the RAM module but just garbage stuck on with RGB lighing or fake heat shink labeling that doesn't really cool RAM at all. Some tests say RAM with heatsinks runs hotter than without. ;)
 
Last edited:
Wow that is good knowledge. It's gskill trident z 3600, 2 sets of F4-3600C16D-16GTZ to give 4*8GB
Two sticks have red tops two have black, neither are RGB and they are the older version not the neo ones.
 
basically most top end stuff is all the same.

If you run them on full auto they are pretty good temp wise. If you enable PBO to the max you'll see the temps sky rocket. manual overclocks will have lower temps.

I just leave my 3900x on full auto.

I don't think your set up with that case is suited to having a rad so better off going with a air cooler.
 
ik9000 said:
Wow that is good knowledge. It's gskill trident z 3600, 2 sets of F4-3600C16D-16GTZ to give 4*8GB
Two sticks have red tops two have black, neither are RGB and they are the older version not the neo ones.
Click to expand...
As others said, probably same specs for both.

In case you are interested (no pun) you might find the below link to basic guide about airflow and how to optimize it in our cases of interest.
https://forums.overclockers.co.uk/t...-i-put-my-temp-sensor.18564223/#post-26159770

No idea how good/bad your stock case fans are, but 230mm fans are almost always garbage because they have extremely low pressure ratings .. so can't overcome resistance to airflow from things like grill, filter, HDD cages, etc. 3x 120mm front intakes with HDD cage removed. PH-F120MP 2-pack £14.99 are quite good. Block any openings in fan mounting panel not covered by fans so air fan are pushing into case cannot leak back in front of them.
https://www.overclockers.co.uk/two-...120mm-fan-radiator-performance-bu-002-pt.html

Remove all PCIe back slot covers to increase rear vent area around GPU and thus improve front to back airflow for lower air temp into cooler and lower component temps.

My favorite cooler is Thermalright TRUE Spirit 140 Power for about 50 quid including delivery. OcUK do not stock them and I can't tell you alternate site to buy it from on open forum, but you want to know and can setup your Trust messaging (button in lower right of your post), send me a message to finde out where to get it. TRUE Spirit 140 Power has 6x 8mm heatpipes and is rated 360w TDP. It cools as well or better than any air cooler I know of with same airflow. I even tested it with TY-143 2500rpm hi-performance fans and while they make it quite loud it cooled about 9c better .. obviously case fans need to be similar so case airflow is also hi-performance to meet cooler needs.
 
Thats a really useful guide. The 230mm spectre fans (non pro) I have are 97.8cfm air flow, and 1.08mm H2O air pressure My old machine has two akasa Apache 120mm CPU fans supplying 57cfm each but at 2.6mm H2O. They will stay in that machine but I imagine the new cpu fans will be similar. So swapping the 230mm intake to two 120mm front fans would give similar air flow but over a smaller area and with much greater air pressure.
 
ik9000 said:
Thats a really useful guide. The 230mm spectre fans (non pro) I have are 97.8cfm air flow, and 1.08mm H2O air pressure My old machine has two akasa Apache 120mm CPU fans supplying 57cfm each but at 2.6mm H2O. They will stay in that machine but I imagine the new cpu fans will be similar. So swapping the 230mm intake to two 120mm front fans would give similar air flow but over a smaller area and with much greater air pressure.
Click to expand...
Glad guide was useful. Thanks for say so.

Bitfenix Shinobi XL has 3x 120mm fan mounts in front, so I would use 3x 120mm intakes in the front like I suggested. No exhaust fans are needed with good pressure rated intakes. It's like push/pull on coolers or radiators, the temp to noise level is the same while at full speed temps are 1-3c lower at full load. The higher fan pressure rating is, the less the difference is. Keep in mind we want more case airflow than component airflow. So if case as 3x 120mm fans and CPU has 1x 140 with GPU having 2x-3x 80-90mm fans we need 4-5x 120mm intakes of flow a little more air than components to push heated air out and supply coolers with cool air. Ideally 3x front and 2x bottom 120mm with PCIe back slot covers removed. Adjust case fan curves to match cooler fan curves' airflow needs. The key to good cooling is supplying components with air as close to room temp as possible. This is where that cheap indoor/outdoor thermometer comes in very handy for monitoring air temp into coolers and then adjusting case fan speed up/down so air into coolers is always less than 5c above room temp. I usually get it to less than 3c, but 5c is pretty good. Just remember every degree warmer air into cooler is results in same degree hotter component will be at same component fan speed.
 
finally starting to get to build point now. The option for 3 front fans comes with removing some of the 5.25bays to get the third fan in. In the box I also have a spectre pro which is 156CFM and 1.81mm H2O so better airflow than 2 120mm fans and reasonable pressure. I'll stick with that. The bigger issue is the rather solid front panel limiting the air intake. I'll see what it's like on vs off and maybe mod the thing to be more porous if it needs it. Will also stick a 120mm fan at the bottom of the case for extra air intake over the chipset.

Now just debating Noctua DH-15 or AIO. I'm leaning DH15 for lower noise + cheaper, but the lack of weight on the mobo and tidier build of an AIO is appealing.
 
Don't forget to block any openings in fan mounting panel not covered by intake fans. We don't want the cool air fans push into case leaking back in front of these fans.

Air cooling is better all-around setup than AIO/CLC. Even best AIOs wear out much faster. On air coolers only the fan can wear out because heatpipes last almost forever .. of course as years go buy new mount spacing will eventually /potentially cause problems, but air cooler will last longer than you will likely use it.

I wouldn't limit myself to just NH-D15 or NH-D15S which has better PCIe socket clearance. There are many other coolers equally as good, most lower priced with some significantly lower priced like Phanteks PH-TC14PE for £52/99
https://www.overclockers.co.uk/phanteks-ph-tc14pe-cpu-cooler-black-hs-009-pt.html
 
lau604 said:
The Zalman CNPS20X did really well, even beat the NH-D15 by like 8c or 7c.
Click to expand...
What was the actual air temp entering cooler at time CPU temp was recorded?
CNPS20X is no better if even as good as D15, so if you are seeing test results claiming it is they at the least not accurate and possible fabricated to gain higher web viewing scores.

Most reviews do not use established tried and proven test procedures. Things like starting and ending test session with a previously tested cooler to be sure test results in current session are giving same results as previous testing did .. or monitoring air temp entering cooler and recording it each time CPU temp is recorded .. or doing tests at same noise level or fan rpm .. or re-mounting with new TIM and running tests again to be sure TIM print is transferring heat properly.

What was fan speed of each cooler when temp was recorded? Very few reviewers keep accurate air temp records while testing.
Or what was fan noise level of each when temps were taken? If one has higher fan speed / more airflow, then at cooler will score lower CPU temps. Example; Thermalright TRUE Spirit 140 Power with TY-143 fan is 8c cooler at 2500rpm than at 1200rpm (on i7 [email protected] with air temp into cooler the same).

All of above are important parts of testing, and if they are not the same with all coolers being tested resulting temps can easily be 2, 3, 4, even 5c from accurate.
 
doyll said:
What was the actual air temp entering cooler at time CPU temp was recorded?
CNPS20X is no better if even as good as D15, so if you are seeing test results claiming it is they at the least not accurate and possible fabricated to gain higher web viewing scores.

Most reviews do not use established tried and proven test procedures. Things like starting and ending test session with a previously tested cooler to be sure test results in current session are giving same results as previous testing did .. or monitoring air temp entering cooler and recording it each time CPU temp is recorded .. or doing tests at same noise level or fan rpm .. or re-mounting with new TIM and running tests again to be sure TIM print is transferring heat properly.

What was fan speed of each cooler when temp was recorded? Very few reviewers keep accurate air temp records while testing.
Or what was fan noise level of each when temps were taken? If one has higher fan speed / more airflow, then at cooler will score lower CPU temps. Example; Thermalright TRUE Spirit 140 Power with TY-143 fan is 8c cooler at 2500rpm than at 1200rpm (on i7 [email protected] with air temp into cooler the same).

All of above are important parts of testing, and if they are not the same with all coolers being tested resulting temps can easily be 2, 3, 4, even 5c from accurate.
Click to expand...
just had a quick glance of the review done on kitguru
 
I am sure that any mid to top end coolers would be fine. It depends on how you are using the cpu - ie. just for gaming (or even some video editing or rendering like adobe substance painter) in which case its under utilised or continual handbrake conversion (utilising all cores at 100%).
 
You can technically cooler a Threadripper 3000 on a basic air coolers so trust me, any old cheap air cooler will handle the 3950x just fine

however for me personally my 3950x is on a 360mm aio. Ocd to 4.3ghz, maxes out at 80c in prime 95
 
You must log in or register to reply here.
Back
Top Bottom