Overclocking and Thermals
We connected the second EVGA power supply for our overclocking testing, but after attaining a reasonable overclock for a water-cooled system and measuring power consumption, we determined that the extra PSU wasn't necessary. We're told it really becomes a requirement for systems that push higher frequencies via exotic cooling solutions like water chillers and liquid nitrogen.
The Xeon W-3175X was surprisingly easy to overclock through Intel's eXtreme Tuning Utility, though we prefer to overclock in the BIOS.
An all-core 4.6 GHz overclock delivered a nice balance of performance and thermal output under our EKWB cooling system. AVX-based workloads did overwhelm the thermal solution, so we dialed in AVX offsets that throttled back to the stock 3.1 GHz during those taxing workloads. Temperatures rarely exceeded 75°C in either case.
We dialed in a 1.9V VCCIN and 1.17V Vcore to stabilize our 4.6 GHz overclock, then bumped up the VCCSA and VCCIO to 1.3V to help stabilize our memory overclock. We topped out at DDR4-3200 with the Corsair Vengeance RGB sticks. We also bumped the mesh frequency up to 3.2 GHz (1.3V Vmesh) to give us an extra performance boost via lower cache latencies.
Plotting temperatures revealed that we rarely exceeded 80°C during our AIDA stress test with AVX disabled. Switching over to an AVX-enabled Prime95 stress test pushed our thermals up to ~88°C, though the CPU merely ran at its stock 3.1 GHz AVX frequency. We did experiment with clock rates as high as 3.6 GHz in AVX-enabled apps. However, that pushed the processor up to 110°C within a few moments, tripping its thermal throttling algorithms.
It's noteworthy that we were able to attain a higher overclock with the W-3175X than we've seen from some of Intel's other HEDT models with fewer cores, like the Core i9-9980XE. The Xeon W-3175X's larger heat spreader likely helps to alleviate thermal density issues that constrain cooling efforts on the smaller chips. Still, solder TIM would have been a welcome addition.
Power consumption measurements are always a bit tricky. But as long as your 12V supply (EPS) readings, motherboard power supply sensor values, and voltage transformer losses plausibly coincide, everything is fine. Therefore, we're using pure package power to avoid possible influences from our motherboard. Results from the PWM controller are very reliable if you take them as averages over a few minutes.
The Dominus Extreme does present power measurement challenges, however. In order to sidestep the CPU's power limits, Asus offers a secondary power reporting option in the BIOS. Intel's recommended setting (default) reports current by dividing the value by 1.25x, meaning that we have to multiply the power values we receive from the sensor loop by 1.25x to calculate the final value. Overclocking requires a 4x divider, meaning the software reports 1/4 the actual power consumed to avoid tripping the chip's internal protection mechanisms. Simply multiplying the output by four gives us the correct value.
We expected high power consumption from the Xeon W-3175X, and Intel's workstation flagship didn't disappoint. At stock settings, we recorded 318W from an AVX-optimized workload at 3.1 GHz and 319W in a stress test that didn't use AVX instructions.
Those numbers skyrocketed when we began overclocking. At 4.6 GHz, we observed 676W during an SSE-optimized stress test and 792W with AVX instructions in play, despite the AVX offset.
As with the Skylake-X processors we already reviewed, current has a big impact on both performance and heat. It's even possible to generate higher performance scores in threaded benchmarks like Cinebench by using a higher VCCIN voltage setting (at a given frequency).
MORE: Best CPUs
MORE: All CPUs Content