drea.drechsler :
mrseauw :
drea.drechsler :
mrseauw :
drea.drechsler :
Have you determined what voltage you CPU is operating at? 'Standard. 1.45v' doesn't make much sense: 1.45 volt is extremely high and really not safe for continuous operation. What you need to do is monitor CPU Core Voltage during a stability test to determine what the actual core voltage is.
If it really IS 1.45V, that's WAY to high for 3.8G. The processor is doubtless getting hot, especially with the stock cooler, and that's affecting stability in a bad way so you also need to monitor temperature.
I can't change core voltage in my BIOS only offset or auto, and auto puts it at 1.45v, so i offset it with - 0.05v so it doesn't use more than 1.4v. Before i tried OC'ing the default CPU core voltage was at 1.45 (idk why)
I always monitor my temps, my idle temp is at 35-45 C and at max load ~60 C, when it gets above 60 i lower the ambient temp to get below 60 always. Ambient temp is always between 20-25 C or less if turn down the ac temp more (my ac is almost always on)
Apply an offset to bring CPU core voltage to about 1.325 under load at 3.8Ghz and check stability.
What utility do you use to read VCore voltage? What utility are you using to load the processor for stability testing?
I use Realbench to stresstest (15 min using 8gb ram) and HWinfo to monitor.
I just tested 3 times:
"stock" cpu 3600mhz max 1.4v -> passed; temp max 61 C
cpu 3800mhz max 1.325v -> froze at 3 min
cpu 3700mhz max 1.325v -> passed; temp max 61 C
I logged all 3 times via HWinfo, you need the excel files?
I'm wondering if i should change my memory timings , everything is on auto, i OC'd to 2666mhz @ 1.35v (since day 1)
current (auto) timings 16-18-19-44 trc: 63 trfc: 467
Is this the limit?
I read that 60 C is pretty low temp for max stress test
I'm still a noob at OC-ing tho, trying to learn
since you're using HWinfo. At 3.8G what is the "CPU Core Voltage (STI2 TFN)" reading. Every motherboard reads VCore voltages a little differently, but STI2 TFN voltage is the best one since it's the actual core voltage the processor reports using telemetry. It's should be quite a bit lower than the VCore readings under load and just slightly lower when idle.
If it crashed at 1.325, increase it a couple notches and try again. Keep doing that until you get up to the voltage it needs at that frequency. At 1.425-1.45V you really need to consider going down in frequency unless you're OK potentially reducing the life of your processor.
Don't try overclocking memory at all until you've dialed in a stable overclock on the CPU. You'll never figure out what's going on to get things stable.
Results:
- 3600 @ 1.4v: 1.369v(max) 1.322v(avg) CPU SV2 TFN | 77.6(max) 72.8(avg) CPU(tvtl/tdie) | 61(max) 58.2(avg) CPU(MCS*)
- 3700 @ 1.325v: 1.312v(max) 1.29v(avg) CPU SV2 TFN | 77.5(max) 71.5(avg) CPU(tvtl/tdie) | 62(max) 58.5(avg) CPU(MCS*)
- 3800 @ 1.325v: froze at 3 min
- 3800 @ 1.350v: froze at 5 min
- 3800 @ 1.375v: 1.369v(max) 1.338v(avg) CPU SV2 TFN | 87.3(max) 80.9(avg) CPU(tvtl/tdie) | 67(max) 64.5(avg) CPU(MCS*)
All results @ 100% cpu usage
MCS = motherboard cpu sensor
Should I monitor CPU(tvtl/tdie), or the motherboard cpu sensor?
Ryzenmaster shows temp from motherboard cpu sensor
I think i'll stay at this speed, maybe.
Thanks for helping btw