New Ryzen workstation

LOL, my mind just went “ugh, prbably windows” because I never change that… xD

Yeah, me too… at least usually. My old CPU (Core i7-7700K) had a really bad time with ondemand governor so I just forced the performance with cpupower.

I would print out a dot template of a pattern, picture or etc, and drill out the front panel and possibly top of the case for more airflow. I imagine this isn’t the kind of thing everyone would want to do/be comfortable doing, but it could be done tastefully.

That would definitely be nice but I don’t have the skill nor the tools to do such a job. I am probably going to look for another case soon enough and mount the old PC in the Evolv.

I am playing around with some settings. I run geekbench on both performance and schedutil frequency scaling governors. You can find the results here [1]. The temps are 10-15 degrees Celsius lower with schedutil.

EDIT: Also, I switched to PWM mode on the Celsius S36 so the pump/fans speed is now controlled by the motherboard. After customizing the CPU fan curve, I managed to top @ 80 degC with the performance frequency governor. A bit noisy but I prefer the CPU to stay cooler.

Strangely enough cpupower shows that boost is active but there are no boosting frequencies… WTH?!?

[1] https://browser.geekbench.com/user/303610

Kinda just guessing here since I don’t know the thermal limits, but I wouldn’t expect much if any boost at 80c. You could try running open air/no dust filter front and top to see what kind of temps/boost you get that way. I run a mesh front case without the dust cover - helps significantly - but I need to blow everything out with a compressor every month or so. I’m using an air cooler though - dunno if it would help you as much.

I’d say your temperatures are fine. You’re system is not going to throttle unless it’s under some heavy load, in which it should.

Take it from me. I tested with 4 radiators, a standard Heat-Sink/Fan, and with 1.5 radiators. And the performance was the same for all three.

Unless you have a chiller, you’re not going to see any gains and if that, they will be marginal.

If you want to tinker, that’s cool, but I don’t think a change is necessary, at all.

As far as I can see, core frequencies spike @ 4.5GHz just fine (in one case zenmonitor recorded 5GHz on a single core) on normal workflow but not sustainably: if I run a CPU benchmark which takes all 16 threads to 100% usage, die temperature raises up to 80 degC but frequency sticks around 3.9-4GHz on all cores. This actually matches what I read on reddit and AMD forums from other Ryzen 7 owners: boost is allowed only if core temps are below 65 degC.

I just did that and honestly, it did not helped much. I mean, die temperature stays around 79 degC (on CPU stress test) but with everything on it maxes out at 83 degC. On normal workflow, the difference is barely noticeable. I tinkered a bit with the fans configuration in the UEFI settings and I am quite happy with them: I run fans full speed only if it reaches 85 degC and below that I set up a exponential curve in order to keep it quiet.

Funny thing I noticed: when the CPU is maxed out, it draws out around 65 W [1] but this CPU actually should have a TDP of 105 W. I bought a Gigabyte AORUS X570 Elite because of the good reviews about its VRM and power delivery and I am power limited?

[1] I calculated this by summing up the CPU core and SOC power readings from zenmonitor; I don’t really know if that is correct though.

Uhm… at what temperature should it start throttling? I mean @ 80 degC it would not boost for sure but still the CPU frequencies are 3.9-4GHz rock solid during benchmarks.

This thread is just like axelgenus’s computer case. Things are really starting to heat up in here.

Just to close this discussion with a happy ending:

Solution: my AIO fans were not spinning at all! Probably my Celsius S36 could not draw enough power from the CPU fan connector. I connected the fans to the CPU fan connector using a Y-cable and the AIO pump to CPU-opt. Now temps are 30-40 degC in idle and they top at 70-75 degC on synthetic benchmarks.

2 Likes

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.