Update on Asus Rog Stix X470 F Gaming Bios Adventure
Update on Asus Rog Stix X470 F Gaming Bios Adventure
Update
https://imgur.com/a/afyOOGr
View: https://imgur.com/a/afyOOGr
Running OCCT small data set in pic
Next step medium then large.
So far this room this unit is in isn't too bad but it's only 9:46 am mountain time
With a high to today forecasted for 92f.
The large data set may be run over night.
Lowered LLC to level 3
vddcr manual over/ride voltage @1.36250v
Set SOC voltage to 1.05
which is what the Ryzen dram calculator suggests
For 3200 fast for my b-dies.
Just monitoring this discussion to track your progress.
Hoping you achieve a solid stable starting point.
Your last update seems similar to mine regarding 4GHZ settings on the 2600 (non-x).
PS
I've ordered a 3700x, expected delivery today or tomorrow, so I might not be doing an OC at this time.
Oh ok
Ya I had to give a little more voltage a run on p95 2nd option small fft's all caches max heat resulted in a blank screens x2 about 2 hrs into the run.
So upped it to 1.368?? And now at 1.37500v
Rerunning now.
Temps 75c.
Ambient room temp 80f
Please Let me know how the 3700x goes.
A 3000 maybe my best option for this bios.
Have A Good Safe Holiday Weekend
I am going to apply Optimized Default configurations and then update my BIOS from version 1201 to 2704 in a sequential manner.
I intend to reset the CMOS after each update.
I will share my progress or outcomes here at some time tomorrow.
Good luck! 🤞
I will also share my RAM speed findings.
Just to note, it took me roughly a week to achieve stable OC/RAM OC performance for my Ryzen 5 2600 at 4 GHz.
Please let me know if you'd like me to capture screenshots of my BIOS settings before installing the 3700x.
I can't promise these settings will work for you, but they might be useful.
Updated settings show increased Soc voltage to 1.10v. Vddcr CPU override set at 1.36875v. P95 performance maintained for 4.5 hours under 72.8 max temperature. 3Dmark Firestrike results available at the provided links. Timespy data can be found on their site. Four consecutive runs of Cinnabench R15 ranged from 1410 to 1420, with the first run at 1410. P95 experienced a blank screen. Adjusting to 1.37000v and rerunning P95.
Have you turned off virtualisation in the bios (unless necessary)?
I've also turned off smt, which gave me sufficient thermal headroom to upgrade from [email protected] to [email protected] while using less power than with smt. It may slightly affect multithreaded performance but should improve most games and make browsing web on the desktop feel quicker.
I'm running the 2600 with the stealth cooler.
Hi
Thank you for your reply.
I haven’t experienced this before. I’ll remember it.
My temperatures are fluctuating between 71c and 76c per hwinfo64 in a room at 80°F. When running p95, the screen suddenly goes blank after about five hours or sooner. There was no shutdown or freeze—just complete blackout.
Looking through the RoG Strix X470 forum, I found reports of similar issues with p95 and monitoring software using the same sensors. Some users mentioned problems when trying to overclock.
When I return home on Wednesday, I plan to disable the startup option on my Corsair Link 4 and reset certain settings that previously worked at 4.125 GHz (until I stopped running Corsair in the background). I’ll try again then.
There were also reports of p95 failing while no other software was running realbench for hours. It seems my current settings are close to those mentioned—between 1.36 and 1.38 volts for 4.125 GHz.
One odd thing is how p95 behaves. No matter which option I use—small, large, custom—I always see all worker windows blank until I stop it, then they reappear. After resetting the same settings and restarting, the progress of all 12 workers appears. It might be that p95 itself crashes even though realbench passes an 8-hour stress test with 8GB on my 16GB drive.
If I stick to using p95 and it continues functioning normally, I’ll likely end up reverting to BIOS tweaks and changes whenever it crashes, hangs, or workers fail.
I’m considering using Occt for stress testing followed by aida64, then realbench.