UPDATE - Thermal Review - C4-SFX 2021
1. Preamble:
After a week of working with the first prototype I am now done with the thermal test of the C4-SFX 2021. It was a lot of work because of the many possibilities with the case. I learned a lot about what configuration does work and what configuration will result in higher temperatures. At first, I like to talk about some basics so everyone is on the same level to understand this thermal review.
2. How does modern hardware works:
Modern hardware is very intelligent when it comes to fan speed, power consumption and clocks. This is valid for modern NVIDIA and AMD GPUs and also for Intel and AMD CPUs. These components try to clock as high as possible until they reached clock, temp or power limits. This means if the components are not cooled proper, they will increase in the first step the fan speed. If this does not help, they will lower clocks. If temps are low enough the components will clock as high as possible until they reached the power or turbo clock limit. Every review that only show the temps without the clocks and fan speed is not made thoroughly.
3. Basic principles what affect temperatures:
There are a few principles that needs to be known to understand how components can be cooled more efficient or how components affect each other.
3.1 Fan distance to flat surface:
A fan that is close to a flat solid surface like a table will not get enough air or can’t exhaust enough air to work on 100% performance. A case where the GPU is on the bottom needs higher case feet’s so the GPU fan will get enough air. This is why a radiator on the bottom of the case with slim case feet’s is very sub optimal. This is also valid for a situation where cables or other internal components are close to the intake zone of a fan. A basic min. distance rule is fan height. For the C4 I used 15mm feets in height. This is a nice balance of performance an aesthetics.
3.2 Fan distance to vented surface:
A fan that is very close to a vented surface will recycle less air from the inner surrounding. For example, in a sandwich case that has support 3 Slot GPU, a axial fan GPU with 2 Slot will work much worse than in a sandwich case that is made for only 2 Slot cards. It is easier for the GPU to pull the surrounded air. This has to do with the restrict level of the vent hole surface. For example, a mesh is much more restrictive than some bigger vent holes because it will work like a filter/more solid surface. For the C4 you can use the side mounted bracket in sandwich mode to attach two 140mm fans to provide enough air for a 2 slot gpu.
3.3 Heat origin and routing:
A heat location that is close to a vent panel, with fans on it that moves air outside the case, will result in much better overall temperatures. On many configurations I see a radiator for CPU cooling at the front, side or top of the case with fans setup to intake. These will result in hot air moving inside the case. This hot air will be recycled by the GPU. While it is easy to eliminate this situation for the CPU with an AIO it is more complex for the GPU where the fans are setup for intake. Only a de-shrouded GPU fan with custom fans will solve this problem. So, for every case the biggest rival is hot air inside, that will be recycled by components.
3.4 Recycled hot air:
A component its intake and exhaust are not ducted to an outer vented surface of the case will recycle hot air from itself or other components. In worse case it is easier for the component to suck in hot than fresh air through a restricted vent hole panel. This will happen in an ITX cases where a 2 Slot GPU is installed in a possible 3 Slot area. This problem is prominent for axial fan GPUs. Another solution could be a case fan under the GPU in the unused area that will work like a duct for intake and push hot exhaust air away from the intake (because case fans are bigger as the GPU itself).
Another problem of recycled hot air is a buildup thermal situation. For short loads this is not a problem but under constant load the heat will increase and increase so it will take very long for the system to have a balanced temperature that will be much higher. So, uncover this effect it is important to run thermal test very long.
3.5 Not balanced loads:
Fan speeds are not controlled by a single component. This means if a GPU is under heavy load the CPU fan will not crank up. This can result in a situation where under heavy combined load the overall temperature can be better as on single constant load if one component pushes hot air into the case. For example, while gaming the GPU is under heavy load and force hot air into the case. Depending on the game the CPU has a much lower load and the fan speed on the radiator is lower. These fans will move less air outside the case and the GPU will recycle much more hot air. So sometimes it could be better to have a CPU fan profile that increases the speed also medium temperatures. I know for Ryzen this is not so easy because if its idle temp peaks.
3.6: Fan size and speed:
In some situation it could be better to user slimmer fans that requires less intake zones for pull air for 100% performance. All of the explained points can be even worse if the fan spin on a slower level because vent holes or heatsink surface become more restrictive.
3.7: The perfect world
In a perfect world all fans are working against a very less restrictive surface, heat will be moved away so it cannot be recycled by other components and no fan is close to solid surfaces.
4. Test scenario
For all tests inside the case I used a combined test of Valley Benchmark in 4k and Cinebench R20 Multithread. I used Cinebench because the load is extreme but less enough to give some performance for the GPU to run Valley. A too hard CPU test would result in less GPU load in Valley Benchmark. Valley benchmark is good because it requires less CPU load running in 4K mode (it does not work very well on lower resolutions). There are scenarios where maybe one or the other component could be a bit hotter but not in combined load. Making a thermal test only for one or the other component will not show a balanced temperature. Furthermore, it is important to disable VSYNC because otherwise a FPS cap will lower GPU load. Every test run was made with a room temperature of 22°C and the duration was as long it reached a level where the thermals not changed for 10min.
For the monitoring part I record the CPU/GPU Temp (not hotspot), the CPU/GPU fan speed and the CPU/GPU Clock. I used the newest version of GPU-Z and CoreTemp.
For testing I used the following hardware:
- Ryzen 9 3900X (more heat than 3950X because of worse binning) (145W TDP)
- 32 GB DDR4
- Gigabyte X570I Aorus Pro
- Nvidia RTX 3080 FE or RTX 3090 FE
- Corsair SF600 Platinum
- EKWB 280 AIO + 2x Noctua A12x25 with Noctua 120 to 140 adapter or 2x Noctua R14S
- Noctua U9S + 2x A9-PWM
I ran every hardware in default configuration so no undervolting or disabled Turbo.
I made no changes to the GPU fan profile. The CPU fan profile was setup in this way, that between 60-80°C the fan speed is increased from 30% to 100%. As long the CPU does stay under 80°C this does have the nice affect that the fans will balance itself between best RPM for CPU temp. For the AIO The AIO pump fan profile was setup in this way, that between 60-80°C the speed is increased from 70% to 100%. So in idle and light load the system was inaudible.
Here are the relevant test scenarios I picked out for the review. I did a lot of more testing but I do not thing these are relevant and will make the results even harder to read:
- Classis-Layout – 180° flipped – 280AIO (push out) 2xR14S /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Classis-Layout – 180° flipped – 280AIO (pull in) 2xR14S /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Classis-Layout – default – 280AIO (pushl out) A12x25 with 120-140 adapter /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Classis-Layout – default – 280AIO (pull in) A12x25 with 120-140 adapter /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Classis-Layout – 180° flipped – 280AIO (pushl out) A12x25 with 120-140 adapter /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Classis-Layout – 180° flipped – 280AIO (pull in) A12x25 with 120-140 adapter /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Classis-Layout – default – U9S Dual FAN, R14S push out /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Classis-Layout – 180° flipped – U9S Dual FAN, R14S push out /RTX 3080 FE + 2x P12 bottom fans (pull in)
- Sandwich Layout – default – 280AIO (pushl out) A12x25 with 120-140 adapter /RTX 3090 FE
- Sandwich Layout –180° flipped – 280AIO (push out) A12x25 with 120-140 adapter /RTX 3090 FE
- Sandwich Layout – default – 280AIO (pull in) A12x25 with 120-140 adapter /RTX 3090 FE
- Sandwich Layout –180° flipped – 280AIO (pull in) A12x25 with 120-140 adapter /RTX 3090 FE
5: Test results
The test results are separated by classic and sandwich ordered by the overall temperature (CPU + GPU) starting with the lowest.
6. Review of the configurations:
6.1 Classis-Layout – 280AIO
The classic layout with AIO is an easy basic configuration that result in very good temps. The A12x25 does perform better than the R14S Redux fans. This does show the power of these well designed 120mm fans. If the fans are configured to pull in you will get higher internal temperatures duo the fact of sourcing fresh air on all heat exchangers you will get the best temperatures on CPU and GPU.
6.2 Classis-Layout – default – U9S
This configuration is the "weakest" inside the C4-SFX but on the other side it surprised me a lot. I setup it in this way the U9S pull fresh air from the back of the case. On a board with not covered I/O area you can even improve this by adding a 92mm fan at the back of the case. The U9S and the back fan of the RTX 3080FE release a lot of hot air near the front of the case where a side mounted 140mm fan move it outside the case. The results are very impressive for such a small air cooler.
6.3 Sandwich Layout –280 AIO
This is maybe the easiest well-balanced configuration without deshrouding the GPU. But even this is possible with two fans attached to the AIO bracket with a different GPU. The GPU sources fresh air from the side and the AIO sources fresh air from outside the case. If you use this in 180° rotation the temperatures will be even better.
You have to know that it is impossible to mount default 140mm fans in this configuration because it will collide with the side panel clips. This is why I used the very powerful A12x25 with the very genius Noctua SFMA1 Adapter on the 280 radiator.
Extra Notes: I did some tests with a EKWB 240AIO and A12x25. It performs 5°C worse than the 280AIO and also have higher fans speeds.
7. Final Thought:
That it is. Product validation is done and I am very happy with the results. We are now leaving the prototype stage and moving forward to prepare production state.
Thank you for reading my wall of text
Regards and stay safe
Daniel
