So I finally got my beta LOBO last week (international shipping over Christmas during Covid is no fun!), and I'm in the middle of testing it currently. Results so far: very promising compared to the OG Aquanaut.
First, a note: I'm testing with a 5800X on a B550 board, and I'm
not testing at fixed clocks or voltages. I want to see how this works in my real-world setup, but that means these measurements have added variability from a bunch of factors: auto boost/thermal regulation behaviour in both the CPU and motherboard, variance in background software, etc. This is also run under a completely unoptimized Windows install - my daily one, which includes a bunch of game launchers and other stuff running in the background. Still, my findings (so far) are consistent enough that I don't see this as a problem.
Due to my specific configuration (both HW and settings) I'm running two sets of tests: at a custom power profile for the 5800X (120/95/120) and at stock (142/95/140), and at three different fan/pump settings: 100% fans/45% pump, 65% fans/45% pump and 65% fans/20% pump (fan speeds are ~1600rpm and ~1100rpm; pump speeds ~3400rpm and ~1200rpm). I'm not done testing yet - I've run the stock power profile tests on both, but I'm just getting around to the limited profile testing with the LOBO.
I test using OCCT, 15m test runs, large data set, normal mode, variable load. (I should probably have set it to "steady", but I didn't spot that setting until I had swapped the cooler, so ... it is what it is). Auto thread count and instruction set, which means AVX2.
One thing worth pointing out here: the reason for the limited power profile is that the Aquanaut can't really handle the 5800X at stock, at least not without really cranking the pump speed, which gets noisy. Even then it struggles. This means that at stock, OCCT started throwing errors a bit into testing as the CPU got hot (nearing 80 degrees in various measurements) in anything but the 100% fan speed test. I didn't run the 20% pump speed test at full power with the Aquanaut simply because it was throwing tons of errors after just a few minutes at the 65%/45% setting. Throughout testing with the LOBO, I so far haven't seen a single OCCT error, even at 20% pump speed. That, regardless of absolute temperature changes, is a
major difference. (And yes, I did try to eliminate those errors previously, with no luck, so I'm reasonably sure they were caused by thermals.)
I've got HWI64 screenshots for every test, plus data logs from there that I need to find some way of making sense of (any tips on that are much appreciated), ideally I'll be able to post some graphs for comparison. For now, here are HWI screenshots from the 100% and 65% fan speed runs at stock power.
As you can see there are clear thermal improvements with the LOBO, in the 2-5°C range mostly. I was surprised to find that the "CPU Die (Average)" reading is improved nearly as much as the hottest core temperature, which speaks to the effectiveness of the large microfin array in the LOBO and its flow characteristics. Also, you can see that all cores boost 20-25MHz higher in peak effective clocks, with the average change over the run being a bit smaller at ~15MHz. These are obviously tiny and immaterial changes, but they're still consistent throughout the run and indicate better overall thermals for the LOBO.
100%:
65%:
When I'm done testing hopefully I'll be able to give you something a bit more legible than a huge HWI screenshot
I'll also include a full hardware config + relevant BIOS settings etc.
