EVGA X299 DARK Motherboard Thermal Review: Someone Finally Gets It

Recapping our previous X299 VRM thermal coverage, we found the ASUS X299 Rampage Extreme motherboard to operate against its throttle point when pushing higher overclocks (>4GHz) on the i9-7980XE CPU. The conclusion of that content was, ultimately, that ASUS wasn’t necessarily at fault, but that we must ask whether it is reasonable to assume such a board can take the 500-600W throughput of an overclocked 7980XE CPU. EVGA has now arrived on the scene with its X299 DARK motherboard, which is seemingly the first motherboard of this year to use a fully finned VRM heatsink in a non-WS board. Our EVGA X299 DARK review will initially look at temperatures and VRM throttling on the board, and ultimately look into how much the heatsink design impacts performance.

EVGA went crazy with its X299 DARK motherboard. The craziest thing they did, evidently, was add a real heatsink to it: The heatsink has actual fins, through which a heatpipe routes toward the IO and into another large aluminum block, which is decidedly less finned. The tiny fans on top of the board look a little silly, but we also found them to be unnecessary in most use cases: Just having a real heatsink gets the board far enough, it turns out, and the brilliance of the PCH fan is that it pushes air through M.2 slots and the heatsink near the IO.

EVGA’s X299 DARK motherboard uses some brilliant designs, but also stuff that’s pretty basic. A heatsink with fins, for one, is about as obvious as it gets: More surface area means more spread of heat, and also means fans can more readily dissipate that heat. The extra four phases on the motherboard further support EVGA in dissipating heat over a wider area. EVGA individually places thermal pads on each MOSFET rather than use a large strip, which is mostly just good attention to detail; theoretically, this does improve the cooling performance, but it is not necessarily measurable. Two fans sit atop the heatsink and run upwards of 10,000RPM, with a third, larger fan located over the PCH. The PCH only consumes a few watts and has no need for active cooling, but the fan is located in such a way that (A) it’s larger, and therefore quieter and more effective, and (B) it can push air down the M.2 chamber for active cooling, then force that air into the IO shroud. A second half of the VRM heatsink (connected via heatpipe to the finned sink) is hidden under the shroud, through which the airflow from the PCH fan may flow. That’s exhausted out of the IO shield. Making a 90-degree turn does mean losing about 30% pressure, and the heatsink is far away from the PCH, but it’s enough to get heat out of the hotbox that the shroud creates.

Here's an example of what clock throttling looks like when encountering VRM temperature limits, as demonstrated in our Rampage VI Extreme content:

For testing, we’re using multiple K-Type thermocouples scattered across VRM components, including one located centrally on the capacitor bank, one centrally on the inductor line, and one centrally on the MOSFET line. These central components should run the hottest, as they’re flanked by other high-heat components (and the CPU). The thermocouples measure 1/100th of an inch thick and do not meaningfully impact temperature measurements of the components, nor do they meaningfully impact the thermal transfer capabilities of the heatsinks. We are using a Kraken X62 at max RPMs and pump speeds, with G.Skill Trident Z Black 3600MHz memory in four sockets. The PSU is an EVGA SuperNova T2 1600W Titanium PSU.

Initial testing is being done on an open-air bench. This means that there’s limited airflow, and will stand between where most cases do. Before anyone chimes in on this approach, note that our case testing has proven that half of the most popular, recent cases would produce worse VRM thermals than an open-air bench, with the other half producing slightly to significantly better results. It depends on the case.

EVGA’s BIOS does not let us supply less than 1.0V to the CPU, which means that we can’t test the 0.992V / 3.9GHz clocks that we previously tested.

ASUS vs. EVGA VRM Cooling on X299

Let’s start with some perspective by comparing VRM component temperatures between EVGA’s X299 DARK and the ASUS Rampage VI Extreme. This chart looks only at the results when at 4.5GHz and 1.20V.

With these settings, the ASUS Rampage motherboard operated at throttle point when no direct cooling was added, nearing 100 degrees Celsius on our thermocouple measurements of the chokes and MOSFETs. Keep in mind that these are external case measurements, and that the internal sensor would read higher temperatures, thus resulting in the observed throttling. Adding a 140mm case fan brought it down to 65 degrees, which is well within spec.

The EVGA X299 DARK motherboard operated at 73-74 degrees on the chokes when disconnecting the VRM heatsink fans, meaning that the EVGA heatsink alone – with zero fans, and with no case fans – is able to dissipate the heat far better. Part of this is the better VRM, which has four more phases than the Rampage, but the other part is that the heatsink is finned – as a heastsink should be, and similar to the old ASUS WS X99 designs. The finned heatsink then connects via heatpipe to another block of aluminum.

Enabling both the VRM and PCH fans brought us down to about 65 degrees on the choke, or about 68 on the MOSFET, with the uncooled capacitors at about 65 degrees. This is competitive with the thermal performance of the ASUS board when under a 140mm case fan.

EVGA X299 DARK VRM Thermals

Now that we’ve gotten the comparative data established, here’s a look at all the EVGA X299 DARK numbers, tested at various clocks. Under auto settings, the fan follows system temperature and lands at about 4200RPM, for which we’ll have noise charts in a moment, and the auto fan speeds stick to around 5900-6300 in heavier workloads. EVGA’s VRM fans try to keep the MOSFETs at around 60-65 degrees, but this is way overkill for VRM temperature. The components can take over 100 degrees Celsius, and the VRM heatsink is good enough that we’re only 10 degrees over the target – landing at about 73-74C – when both fans are completely disconnected. Up to 4.5GHz and 1.2V in open air, you could use this board without the fans active at all. They would help in a case, of course, but we’d recommend just running them at lower, quieter RPMs.

Thermal scaling has the 4.3GHz and 1.15V configuration at about 55-57 degrees, with the 4.5GHz and 1.18V test at nearing 61 degrees. 4.5GHz and 1.24VID, stressing the system harder, landed at 67 to 70 degrees on core components, or 60 degrees on the capacitors. This was also accompanied by a fan speed increase to 6300RPM, which increases our noise to somewhat noticeable levels.

Plotted over time, the EVGA X299 Dark motherboard gradually increases in temperature without any fan support, but eventually reaches steady state. It does take a long time to achieve steady state, which speaks to the mass of the heatsink, the passive dissipation abilities, and the VRM design. The 4.5GHz test with fans enabled reaches steady state quicker, and also keeps lower temperatures, somewhat obviously.

For comparison, here’s a chart of just the MOSFET temperatures between ASUS and EVGA when at the same clocks and voltages. The EVGA board, even with its fans disabled, does well to compete; unfortunately, EVGA’s BIOS at time of testing did not permit voltages below 1.0V, and so we couldn’t test 0.992V at 3.9GHz on the EVGA board.

EVGA X299 DARK Fan Noise

And here’s the noise chart. The fan tends to stay around 5900 to 6200RPM, which has us in a range of 34.4 to 35.8dBA of noise, measured at 20” distance. Of course, it’s not just the noise level, but the type of noise – tiny fans are whiny, and that means the type of noise is more noticeable and annoying than the lower whirs of bigger fans. Running the fans at 5900 to 6200RPM is nearly entirely unnecessary in our tested configurations, though your CPU cooler and case configuration will also dictate performance. We’d recommend operating closer to 3300RPM to 4200RPM, which measured at 28dBA to 30dBA, with a noise floor of 26dBA in the room.

The fans aren’t too terribly loud, but are annoying at the top-end of the RPM scale, where we measured 46.3dBA. That’s nearing the noise levels of some graphics cards at a moderate RPM for loaded cooling.

Conclusion: A VRM Heatsink That Isn’t Garbage

EVGA’s PCH fan placement, although limited in its impact to VRM thermals, is brilliant: The fan is able to provide some marginal amount of airflow to the secondary part of the VRM heatsink, ensuring that hot air doesn’t remain trapped under the IO shroud. Finning the VRM heatsink also means that the board is able to operate almost entirely without direct VRM cooling, which means no VRM throttling in our testing. We tested up to 4.5GHz at 1.24V in open air, and found that the heatsink alone was largely sufficient. Placing this into an enclosure would benefit from the VRM fans, but you could run them at lower (~4000) RPMs to help with noise management.

EVGA has done well with the X299 DARK. The BIOS still needs work, and we’re actively analyzing the PCB and VRM, but the cooling is best-in-class. EVGA has no competition right now, at least for VRM heatsinking. We wish EVGA would consider transplanting this board design over to Threadripper, but in a world of aesthetics-first heatsinks, we’ll take what we can get. Job well done on the heatsink, certainly.

Editorial, Testing: Steve Burke
Video: Andrew Coleman