VRM Performance
Now when it comes to testing VRM performance, there’s multiple ways to do it including using the likes of HWiNFO to measure what the sensor tells you, or to use K-Type probes, or to get a better, well rounded picture, you can use both, which is exactly what we’ve done here today.
For testing, we used the 9950X to push the VRMs as far as we could, along with 32GB of XPG Lancer Blade 6000MHz CL30 memory, and an NVIDIA RTX 4090 Founders Edition graphics card. For our storage, we used a Seagate Firecuda 540 2TB Gen5 NVME drive and to keep everything cool, we used the NZXT Kraken 360mm AIO Liquid cooler to keep things under control and all of our testing was done inside an NZXT H7 Flow case with the side panels installed to simulate real world usage.
We took the ambient temperature of the room, which we control using a mobile air conditioning unit, which isn’t perfect, but it allows us to stay true to what we’re trying to do and to make our tests as fair as possible.
We then leave the system sitting on the desktop for 10 minutes to settle before taking our idle readings on both HWiNFO and the probes. After this, we wanted to simulate a worst case scenario using Prime95 with small FFTs to generate the most heat, as well as Cinebench 2024 to simulate more common, real-world usage. We run each test for an hour, along with letting the temperatures to settle down between each test.
The probes are fixed to the back of the board and we generally use one on the top row of phases, and the top phase of the column going down the board and position the probes a few millimetres away from the actual circuitry itself.
You will find that the software reading from the sensor will vary quite wildly in areas, and this is down to the sensor location for one, and if the sensor is reporting the internal temperature instead, which based on our testing and the sensor data generally reading higher, that is the case.
When we look at our table, we can see that every board did exceptionally well with nothing hitting over 80 degrees Celsius, with our highest result coming from the X870-I Gaming WiFi Mini-ITX board at 75 Degrees on the sensor temperature during Prime95, while our highest probe temperature was 68.8 Degrees from probe 1 on the same board, again during our Prime95 hour long run but as it is such a small board, that’s pretty much expected anyway.
Our best result on the probes came from the TUF Gaming X870-Plus WIFI at 39.4 Degrees during our Cinebench run, and only a little higher under Prime95 at 40.7 Degrees, though in both cases, probe 2 did come up slightly warmer, but not by anything of significant margin.
For the most part though, the average across the boards collectively was around 56 degrees, with some coming in a bit higher, and some a bit lower, but overall showing us that the cooling solutions were doing a great job at keeping things under control.
