I do not now if the power saving improved greatly, or if the temperature sensors are now wrong under linux but the CPU temp. plunges in intactivity period
see a graph showing light activity (one browser tab open, then closed)
the ambiant temp. is about 20°C
Activity (6 VM opened and checking for update) temp. look normal
Are you running a 2700X on this board? Thinking about upgrading and getting rid of a superfluous watercooler with the 1700X at the same time (too much junk hanging around).
I’m actually running the 2700 in my ITX system but I still have work to do on that one. I am running a 1700 @3.8 GHz with a fury but that rig is not at home and will stay like that. The second one is curruntly running a 1600X, I might replace that base system completely. And then there is the Taichi Ultimate that I am waiting for…
I think it comes down to whether the mainboard’s VRM are able to deliver enough power or not. (And the BIOS ofc ). As far as I can remember, the Taichi had the best VRMs when compared to the other X370 boards, with Asus boards following. Which is why you should be fine with the Prime. That being said, I haven’t tortured tested it or anything, it just noticed the clock speed going up the same amount as on X470 boards according to reviews.
The next thing I’ll do with the 2700X is undervolting it - once I have results I’ll create a corresponding thread on this forum.
I am running Linux OpenSuse Leap 42.3 with XFce windows manager and up-to-date “Kernel:Stable”
The pictures are snapshots of Ksysguard ( linux monitoring tool interface ), that I took with Gimp.
The “Sensors” and “monitoring-plugins-sensors” packages needs to be installed and configured for the it8628 chipset, it is not too complicated to do, althought not all the measuring points work
@Steinwerks thanks for the update notice, yet another one
As for the SMU update I would assume this would include the security modules firmware fixes for the issues raised in a spectacular maner by CTS and claimed “unfixable”
AMD released the fixes 3 weeks ago for the integrators
just read an update from phoronix, seems that the AGESA update benefits the Ryzen + 2700x performance, on Linux, nothing about our good “old 1700x” though
A pitty that Asus do not communicate better the release notes, keeps us guessing
Again almost thought that 3000MHz could be stable, but quickly gave up at first quirk, so I am back my prevous settings and to a stable 2933MHz
as for the power management I will need a few days to have a better view as the ambient temp is now 2°c up, still a nice idle low temp in the 28°C but for the moment it seems to drop a little bit slower
I know I wasn’t asked but this would be my answer.
If I can make the thing crash or hit limits just by running some programs, it’s not good enough. I am running mprime -t (prime95 kind of), memtest, a blender build from git using pacman (that was just to check for “segmentation fault” error), unigine valley… stuff like that to stress components. Is that realistic? No, mostly not. But that is the point. If the system can handle those unrealistic workloads, anything else will be running fine, even over long periods of time.
As a side note: I am currently running 4x8GB Kingston 2400MHz ECC DDR4 (micron stuff I think) without any problems on the prime and ECC seems to be working just fine. I am on BIOS 3805 on that one, I will update it shortly.
I just finished an ITX build and I tried to make a R9 nano work in that but it just gets too hot. Even modding the cooling didn’t help. So, a thermal limit would be an example.
Well, indefinitely is … a bit long. When I build a system generally I just let the first couple tests pass. Later with an installed OS I let it run at least until temps stop climbing, most times that means simply over night.
indeed for the temp I am impressed too by the noctua, I hesitated for the bigger one but this one already bearly left just enough room for the RAM installation
As for stability, as stated earlier in this thread I realized during my initial testing, that ryzen POST test were very sensitive, so instead of booting once and passing as much tests as possible to assess stability… I rather boot 5 to 10 times (soft and hard) BEFORE I do any other test.
This time it did not pass 5 consecutive boot cycles without a problem with the SPD overclock memory setting at 3000Mhz (even with cas at 16 instead of 15) hence the not-rock-stable status
Honnestly this methodology saves me quite some testing time and once passed it I never ( again on this platform) left me experiencing any instability in normal usage.