Hey everyone,
I spent the last 3 weeks overclocking my entire system—well, more like a few hours of OC and the rest stress testing. I've seen a lot of posts here from people who think they're stable just because they passed a few minutes or hours of stress testing. You're probably not. Are you going to find out the hard way? Hopefully not. The reason I’m saying this is explained below...
My system HW:
Motherboard: Asus Rog Strix B650E-E
CPU: 9800X3D with Liquid Freezer II 360
RAM: 32GB GSkill 7200CL34
GPU: Gigabyte Eagle OC 5070Ti
Stress tests I ran:
Over 12 hours of VT3 y-cruncher, 24 hours of AIDA64 stress test, 24 hours of Karhu, 24 hours of Prime95, a few hours of Linpack, 24 hours of TM5 AntaX3D, and 24 hours of TM5 1usmus.
Testing memory wasn’t too difficult. I used nearly the same timings I used with my 7800X3D and was able to slightly tighten them, with 2200 FCLK. I could probably push them lower, but I doubt I’d see any performance gain. I get a max of 42°C on the sticks after slapping a fan on them. Actually, not just fans—I also used thermal putty and stuck some NVMe heatsinks on the side of the RAM where the chips are to help with heat transfer of crappy G.Skill heatsink. I tested it, and it reduced temps from 51°C to 47°C, and then to 42°C with the fan. Probably unnecessary, but I had about 10 of them lying around, so I figured, why not?
Once I was done with RAM and confirmed stability, I moved on to PBO. Always do RAM first—you want to make sure it’s stable, otherwise if you get an error during PBO stress testing, you won’t know if it’s the RAM or CPU. To my surprise, I was able to run -25 +200 on AIDA64 and Prime95, but it would freeze/crash on TM5 1usmus. At this point, I believe 1usmus is the best stress test for finding PBO instability.
I found I couldn't run more than -5, which surprised me. Then I did per-core OC and had good luck with negative 15,5,5,10,15,15,15,15—until I got an error 13 hours into 1usmus. After adjusting to 10,5,5,10,10,10,15,15, I passed 24 hours. I tried again with the earlier values and errored again around 12 hours in. Then I reverted to the previous stable values and passed again. I later tried 10,5,5,10,10,10,20,20 and got an error after 6 hours. Clearly, 1usmus is excellent for finding PBO instability.
Error later in test :
https://ibb.co/1Y5cRPjf
My PBO:
https://ibb.co/GQPthsSQ
Zen Timings:
https://ibb.co/6074ZwQL
I run 1/2/0 Nitro Mode, GDM: Disabled, and both sticks have tPHYRDL set to 35.
Anyway, here are my results:
AIDA64:
https://ibb.co/pj6KKBKS
Cinebench23: 23700 (I didn't bother with screenshot)
Before all of the above, I played with the GPU. I flashed a 400W BIOS, maxed everything out, did +375 on core with a cutoff at 1V (which gave me better scores than running higher voltage), and +3000 on memory.
This got me:
1st place in TimeSpy: https://www.3dmark.com/spy/56288943
1st place in TimeSpy Extreme: https://www.3dmark.com/spy/56289405
19th place in Speed Way: https://www.3dmark.com/sw/2281860
39th place in Steel Nomad: https://www.3dmark.com/sn/6435602
5th place in Fire Strike Ultra: https://www.3dmark.com/fs/33363039
13th place in Port Royal: https://www.3dmark.com/pr/3455901
For gaming I dont increase core voltage, same +375 +3000 but I cut it off at 0.97V
My system temps after playing Expedition 33 for nearly 6 hours on 2k Max settings :
https://ibb.co/ZzZ7hGn6
Thank you for reading!
