Testing methodology
The Era 2 case is an evolution of the original model (Era ITX), where Fractal Design also changed the approach to the ventilation holes. Be it in terms of the top panel or the side, larger perforations. How the Era 2 will make its mark and stack up against competing SFF cases remains to be seen. Despite the detailed analysis, this is still the first case to pass our new testing methodology.
Testing methodology
What effect does system cooling have on the overall temperature of the PC build? That’s the first thing we look at in our case tests. This at various noise levels ranging from one that is at the limit of audibility (“31 dBA” mode), through very low (“33 dBA”), still low (“36 dBA”), relatively lower (“39 dBA”) to modes for medium (“42 dBA”), higher (“45 dBA”) and maximum performance. At these settings, the noise level will already be higher, but may make sense to some with respect to achieving the highest possible cooling performance.
However, some cases may not come with low-speed fans, just as they may not come with high-speed fans, which will prevent them from getting into the high-performance modes.
In such cases, the measurements of the respective modes are naturally missing, as they cannot be adjusted for.
The noise is measured diagonally from the left side of the case, where the sensor of the noise meter (Reed R8080), or microphone (miniDSP UMIK-1) is 30 cm away from the left corner of the case and is at a 45-degree angle to the front and side panels, and at a 30-degree angle to the top of the case. To increase the resolution, we use a parabolic collar. Once the case is tuned to a given mode, the acoustic profile is also examined in the TrueRTA application at 1/24 octave resolution. Based on this, it is possible to specify “weak” spots that are making the case sound unpleasant with unwanted tonal peaks.
The noise level reading always includes only the sound of the system fans. The noise of the CPU and graphics card coolers are filtered out, or rather these components are inactive during the measurements and practically do not interfere with the measurements. The CPU coolers we use are the Noctua NH-L9i-17xx chromax.black and the NH-D12L chromax.black. We choose between the models based on which one physically fits in the tested case and which can also be considered as a more suitable option.
So, the smaller model (Noctua NH-L9i-17xx chromax.black) usually comes into play with many SFF cases, but most of the time there will be an effort to use the NH-D12L chromax.black cooler. This is also with a view to building as large a database as possible within which the results can be described as comparable. There are always more options as far as specific coolers are concerned, and a large number of cases are rather optimized for AIO type liquid coolers. This is often reflected in the default configuration of system fans, of which there are more at the inlet than at the outlet. The design of the case here, in short, assumes that a liquid cooler is used and its fans are engaged at the outlet of the system cooling. But we don’t do this because of the very high variability of aerodynamic liquid coolers and use the traditional coolers mentioned above. The thermal paste used is always the Thermal Hero Quantum, which we replace after some time under the coolers of both the CPU and the graphics card to avoid the natural change of its properties from potentially distorting the results. For the same reason, so that incomparable situations do not arise over time, all the coolers are kept as clean as possible – dust-free.
If the case has a predominance of intake fans, in addition to the default configuration, the case is also tested using a modified scheme that simulates the equilibrium flow or positive air pressure in the case, depending on the specific situation. This is always specified in the description of the charts, numerically in brackets following the designation. The first number expresses the number of intake fans and the second the number of outtake fans. For example, “1+2” means that the case has one intake fan to two outtake fans.
The main heating components are a Intel Core i9-13900K processor (with E-cores disabled) and an Nvidia GeForce RTX 4070 Super FE graphics card with the fans fixed at 12V. CPU coolers also have fixed speeds, with fans at 12 V. The PWM-enabled fans are controlled by a Noctua NA-FC1 controller, and the 3-pin models (regulated by DC/linear voltage only) are then controlled by a Gophert CPS-3205 laboratory PSU.
CPU performance is set to be constant at all times, with a good chance of being sustainable in any situation, even in cases with weaker cooling. For configurations with the NH-L9i-17xx chromax.black, this is approximately 75 W, and with the NH-D12L chromax.black, it is approximately 150 W. If, due to insufficient cooling performance, the power consumption drops below the specified level, the test in that mode is evaluated as a failure and its results are not recorded in the charts (the corresponding bar of the chart remains without a numerical value).
The stability of the power during the tests is monitored at the EPS of the PSU. The one we use, by the way, is a Seasonic Focus SGX-750 (with a direct 12VHPWR cable), SFX format. This is for consistency of measurements in all cases, even those in which you would be able to fit a PSU in the ATX format. For the same reasons, a Mini-ITX motherboard is also used – an Asus ROG Strix B760-I Gaming WiFi (BIOS v1663) will also fit into small cases.
The memory is DDR5, Patriot Viper Venom (2×16 GB, 6000 MHz/CL36) and the SSD is a Gigabyte Aorus Gen5 10000 (2 TB) with a finned heatsink. Also thanks to it, the SSD cooling will scale well with the intensity of the system cooling.
The SSD is loaded with CrystalDiskMark (in a maximum sequential read and write performance loop), – the CPU is loaded in Cinebench R23 and the graphics card in the Shadow of the Tomb Raider game menu. If the case allows vertical mounting of a PCI Express expansion card and at the same time its accessories include a PCIe ×16 riser, we also proceed to test the graphics card in the vertical position. All tests are run independently, going in sequence, with a duration of 900 seconds. There are 10-minute cooldown pauses between each mode, during which the component temperatures are brought back to default values.
From the temperatures of the internal sensors monitored by the HWiNFO application (v8.10-5520), we record the following in the charts: the average of the maximum values of the P-cores of the CPU, the VRM (Vcore and Thermistor), the chipset, the SSD, the GPU of the graphics card and its memory.
The air temperature at the inlet to the cases is monitored during the measurements and varies within a narrow range of 21.0–21.1 °C. This by a sensor of a properly calibrated UNI-T UT325 thermometer. Its second sensor is then at the outlet, from which we also read (and already record in the charts) the outlet air temperature.
Of the standard measurements that are no longer related to cooling or acoustics, what’s measured is the maximum sequential read and write speed via USB ports (connected to the motherboard Asus ROG Strix B760-I Gaming WiFi) on the front panel of cases. Again in CrystalDiskMark, the test device is a WD Black P50 external SSD. The resulting speed in MB/s is the arithmetic average of three passes. Faster than 10-gigabit ports are limited by the capabilities of the test motherboard with support up to USB 3.2 gen. 2.
Extra tests of vibrations…
During the tuning of the fans to the specified noise levels, a three-axis vibrometer is used to examine the intensity of vibrations at various locations in the cases. It should be noted that it is only on the outer parts where the vibrometer sensor can easily reach even when the case is covered. The locations monitored are usually the individual panels (including the front and rear, which often hold the fans) and also the areas around the pull-out dust filter. This is also an area where secondary rattling noise can occur under certain circumstances.
The output shows the velocity values (in mm/s) in each axis, but for simplicity we will always work with only one number, a computed 3D vector. However, we will not pick up the most subtle vibrations, as we can with fans, with cases. This is because the sensitivity of the technique we use to sense the vibrations is too low. This is also why the vibration information is interpreted only verbally, at the end of each chapter with graphs. This is a rough orientation which is useful mainly to detect possibly excessive vibrations.
… and HDD soundproofing
As an additional measurement, we monitor the degree of suppression of the noise of the WD Gold HDD (WD4004FRYZ) in cases supporting 3.5″ internal storage bays, both at idle (spinning platter noise) and during random access with 512B block writing (when HDDs tend to be noisier due to more aggressive movement of the reading/writing headers). The HDD is the only sound source in these measurements.
- Contents
- Exterior details...
- ... and interior details of the Fractal Design Era 2
- Testing methodology
- Results: Maximum performance
- Results: Higher performance (45 dBA)
- Results: Medium performance (42 dBA)
- Results: Lower noise level (39 dBA)
- Results: Low noise level (36 dBA)
- Results: Very low noise level (33 dBA)
- Results: Limit of audibility (31 dBA)
- Results: USB port speeds
- Conclusion
On testing methodology:
1. Going for L9 over L12x77 is certainly a better choice as the former is basically guaranteed to fit in any cases (if you must only use two coolers). Still, I think having more coolers in the arsenal would be more “fair” to the cases being tested. The inability to use larger air coolers like the L12S, or liquid coolers like what this case is designed for, heavily penalizes this case.
2. There’s a good reason that the stock fans should be used when provided, but I wonder what you would do if the case comes with no case fans but intend you to use some?
3. I think there should be additional noise and spectral analysis for when fans are pressed against the side panels and/or fan mounting hardware. Somewhat related, removing cooler from noise level may not be appropriate in some cases, where the intake panel is located close to the cooler.
4. Can you explain why E-cores are disabled?
Specific for this case:
1. “Both side panels are heavily perforated. Firstly on a relatively large area (approx. 50 %)”…
By 50%, are you referring to the area where the vents are present? If we’re talking about open area %, by my rough estimation it’s only around 25% at the most open areas.
2. The bottom dust filter looks interesting, seems similar to the metal filter you have tested. I’d imagine it being quite favourable to airflow.
3. Despite Fractal installing the bottom fans as intake at stock, it is in fact inferior to bottom exhaust, even when using liquid coolers on top. Machines and More on YouTube (https://youtu.be/vyNmPt6nBTI?si=3IF5SPVDaiIT2EmF&t=890) has extensively tested various fan layouts and load scenarios, and found that flipping the bottom fans to exhaust are always better (for both CPU and GPU). And this makes a lot of sense in sandwich layout cases – you’re aiming to pull air from the sides and exhaust hot air via other routes. When there’s airflow from the bottom, the case internal pressure increases, which means less air will be pulled in from the sides, leading to marginal gains or even negative effects.
Yes, the NH-L12Sx77 cooler does not fit in the Era 2 and we had to adjust the fan position in the case of the Mood case as well. The NH-L9i cooler has better compatibility due to its smaller dimensions, but at the price of lower cooling performance. Of course, I agree that in an ideal world, tests with liquid coolers would be great too. But these are already beyond the available capacity in terms of time.
In the future we will deal with spectral sound analyses with respect to different fan modifications in the framework of the evaluation of the level of psychoacoustic optimizations. Unfortunately, there is no time for this now due to too much pressure on tests of components from other categories.
The E-cores were disabled because of the higher heat flux achieved on the P-cores, which the coolers are also able to keep cool while maintaining a stable performance.
The 50% is not meant to be the “open area”, but the “total area covered by the perforation”. I accept that the wording could have been clearer. My head was already in other topics when writing the article, haha. And you probably know which ones… 🙂
On time requirements, a reasonable compromise would be to reduce the resolution of sound tests… that is, test fewer sound levels. I don’t think testing 6 sound levels so close together would lead to a different conclusion than, say, 3-4 sound levels spread wider apart (e.g. 31, 33, 39, 45 dBA). I’d say it’s the same for fans too, and the extra time gained could be better used for other things (for example by testing more fans, or by expanding the range of sound testing to something like 31/33/39/45/51 dBA etc.)