TR Forums

Waco wrote:

If you weren't using Ryzen Master to monitor those voltages, they're likely completely bogus (and even then, the measurements are pretty suspect given how quickly Zen 2 chips change state).
Have you done any long-term stability tests? Passing Cinebench is nowhere near real stability. I've had my rig encoding for 4 days straight (100% load) a few times on stock settings (PBO+AOC) without any issues whatsoever.

Yes, Ryzen Master, Agree that you can not trust many other programs. Was disappointed in Task Manager, it says 4.2Ghz and RM says 14 cores are asleep and the other two are running 300-600 Mhz (not Ghz)! CPUZ and I think HWInfo are no better.

I got the quick and easy benchmarks stable. At first low volt settings the pc would crash then just the benchmarks and now still getting a feel for what is normal before I get into other bios changes.

"Load Line Calibration" was a surprise...none of the lower settings seemed to have any effect at all on voltages,temps etc. Once things settle I'll set it on the lowest "normal?" and leave it there if all goes well.

"Long term stable" will have to wait till I get into a project and have a day or two of encoding to do.

W7 to W10, a 2011 intel 3930 6 core to a 2020 Ryzen 16 core. So many changes at one time...Ick. At least my temps are down now, that was the big one for me.
I am not interested in pushing for the bleeding edge anymore. "Just want things to work".

EDIT...
Oh, while I'm here, anyone know a desktop fan speed monitor program that will work with X370 chipset?
The one that came with the Gigabyte X370 Gaming 5 motherboard no longer works with this cpu and latest bios.
There is pretty much no chance Gigabyte will bother to update anything but the bios for this old a product.

I tried two others... SpeedFan and something else. No luck. I want to be able to change the fan settings as I test.
I can only monitor my fan speeds in the bios so can't test my cooling under working conditions.

Thanks.

just brew it! wrote:

Igor_Kavinski wrote:

jmc2 wrote:

Think I'm going play with reseating the heatsink a few more times.
If it is a fixable seating problem then it should be very obvious if I finally get it right.

My i7-3770 was going higher than 80 degrees celsius in stress testing the first time I booted it up. After reseating the heatsink, the difference was like night and day. I don't remember exactly but I think the temperature dropped like 20-25 degrees.

A 20-25 degree difference sounds like bad thermal compound application or an incorrectly mounted HSF. Hey, it happens...

Yep, Ivy Bridge and First-batch of Haswell chips were plagued with mounting issues at the factory since the fitting tools were originally set for soldered chips. This is the real reason behind whole Ivy Bridge ran "hotter" at load than Sandy Bridge. It was fixed by the time second batch of Haswells a.k.a Devils Canyon came out along with updated thermal paste.

FYI, 3950X are toasty chips when fully loaded. They are almost like a denser Threadripper 1950X running and boosting at a higher clockspeeds. I wouldn't fret too much about temperatures unless you are experiencing significant throttling or trying to find the golden spot for near-silence/performance without thermal pollution bleeding to your other components.

Krogoth wrote:

[quote= or trying to find the golden spot for near-silence/performance without thermal pollution bleeding to your other components.

Some results on my quest for low temps/noise

(4.25 Ghz @1.28125v)
In VideoReDo6 the temp dropped from 72C(stock) to 68C and gained 3-4fps in speed. Pass1 25%, Pass2 50% usage.
(X264 encode "Veryslow" Preset)
-----------
In TMPGEnc with contrast,color,contour,video denoise...
Stock was 11C hotter (82C/26Ambient) and 3fps slower.

4.25 Ghz @1.28125v was 71C/26Ambient (11C cooler) and 3fps faster... 86fps vs 83fps.

I did once run the cpu at 1.3v and 4.3Ghz and the CPUZ single core bench was equal to the stock single core boosting.
But that is getting toward "the bleeding edge". The temps were back in the 80Cs again.
Think I've read that you would be lucky to do 4.4Ghz.

Ryzen Master is the best program to monitor and set speeds/voltages for testing. I'll put the settings in the bios later.
Finally for my cpu (YMMV) I've settled on 4.25 Ghz @1.28125v.

I can not drop the voltage but I might be able to nudge the speed up to 4.275Ghz once I get a feel for what is normal.
Not that I would be able to "notice" such a small increase.

I'm on a Noctua D15 air cooler so temps are my biggest concern. The 11C drop in TMPGEnc was a nice surprise!
So a little faster and a little to a LOT cooler...Win,Win!

Thanks,
jmc2

Very interesting post about the observer effect when monitoring Ryzen voltage: https://www.reddit.com/r/Amd/comments/c ... gen_ryzen/

Igor_Kavinski wrote:

Very interesting post about the observer effect when monitoring Ryzen voltage: https://www.reddit.com/r/Amd/comments/c ... gen_ryzen/

Heh, Wow, very interesting, you're raised to keep temps and voltages low(as long as it functions).
And now that is all wrong? I'm going to have to go with lower volts and temps for my own mental health/stress levels. I can't sit there and see 85C and 1.5 volts and not cringe.

Thanks,
jmc2