Correct me if I'm wrong, as of today there is no cooler for the Epyc 9965 that was launched 8 months ago in October 2024. There are liquid-cooled-servers, but these come as a whole. Let me know If you have managed to use Epyc 9965 with a cooler this includes both any new cooler in the future, or even if you have used a non TDP 500 capable cooler what were the results.
is there a Cooler for Epyc 9965 (Default TDP 500W)
- Thread starter fatherboard
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
dynatron makes a few:
www.dynatron.co
Server Air Coolers | High-Performance Heat Sink Solutions | Dynatron
Explore Dynatron’s full range of air cooling solutions designed for servers, AI platforms, and industrial systems—offering optimized airflow and thermal reliability across various form factors.
www.dynatron.co
try SilverStone XE360-SP5
they provide no TDP rating as Noctua e.g. because TDP is nothing saying today.
with big silicon chips it is more reasy to cool as with small hotspot ones.
i would also try a 12cm Cooler and swap the FANs with higher RPM ones.
do not buy low min RPM FANs, this results in overheating og the Fins in IDLE, and on load the FANs must ramp to max. for longer time,
that is more
Last edited:
J12 is the candidate here thought.dynatron makes a few:
Server Air Coolers | High-Performance Heat Sink Solutions | Dynatron
Explore Dynatron’s full range of air cooling solutions designed for servers, AI platforms, and industrial systems—offering optimized airflow and thermal reliability across various form factors.
rated 580W at an ambient temperature of 20°C
so you have to have air conditioning in your room.
edit:
max. 8000 RPM, tell your neighbors beforehand.
I have them, but not for the Zen 5 Epyc 9965 500TDP, just to be used with Zen 4 Epyc 9754 400TDP.
I contacted silverstone directly and they said they have nothing for 500TDP.
These 9965 will be HPC-tortured for weeks in a row and will get only a few hours rest before another round starts, so I don't know if I should risk these expensive 500TDP CPUs with a 400TDP cooler
What worries me is where they say "CPU Socket: Genoa®, Socket SP5"
All cpu coolers were designed for Genoa, non for Turin.
Last edited:
asked for 600W ?
they stop clocking high at some temp.
you may overthinking this.
modern desktop CPUs like core 14th gen TDP=65W can draw up to 200watts for some time,
but server processors habe more strict TDP borders.
have you thought about VRM cooling ? i guess that will be your more critical problem soon.
Stop this cooler non-sense and do yourself a favor and think about a liquid cooling solution that fits your project. Tons out there ...
PS: If you really go full throttle on a 500W TDP CPU, you better think of 600W cooling performance considering the hot airflow you probably never manage to get rid of with a basic airflow setup in a dense environment.
PS: If you really go full throttle on a 500W TDP CPU, you better think of 600W cooling performance considering the hot airflow you probably never manage to get rid of with a basic airflow setup in a dense environment.
If we're talking about air then It's all about (physics/thermodynamics):
1) heat transfer between CPU contact surface and delivery (pipes)
to
2) heat exchange space (fins) = total surface area
and
3) airflow (+input temperature) over that surface to move heat energy away.
Normally most premium coolers can dissipate wattage way above their declared specification if you provide sufficient extra airflow (given all else being constant or if not then even supplying lower inflow temp) without the need to increase total heat exchange area (fin mods).
One of paths for this conceptual solution was already mentioned above - larger diameter/higher RPM (higher rotational speed) cooler, although there are other ways to achieve the same (e.g. multiple fans, blade shape and count, etc. whatever is needed to grant higher CFM {air volume per time} / higher static pressure).
1) heat transfer between CPU contact surface and delivery (pipes)
to
2) heat exchange space (fins) = total surface area
and
3) airflow (+input temperature) over that surface to move heat energy away.
Normally most premium coolers can dissipate wattage way above their declared specification if you provide sufficient extra airflow (given all else being constant or if not then even supplying lower inflow temp) without the need to increase total heat exchange area (fin mods).
One of paths for this conceptual solution was already mentioned above - larger diameter/higher RPM (higher rotational speed) cooler, although there are other ways to achieve the same (e.g. multiple fans, blade shape and count, etc. whatever is needed to grant higher CFM {air volume per time} / higher static pressure).
some light reading from OP's other posts
Last edited: