PC Processors

[robot468]

Config:

PSU 1kW Gigabyte

MB: gigabyte x670 aorus elite

CPU: 7900x3d 

Cooler: DEEPCOOL LE720

RAM: Kingston FURY Renegade Silver [KF564C32RSK2-32] 32 ГБ

I built a new system a week ago, updated the bios, installed the XMP-1 profile (6400-32), set the PBO to (80 Level 3) and started using it. It was stable.

After 3-4 days I noticed the first hangs, ran a stability test in OCCT. Cpu core errors were detected.

I did a full memtest86 test (with XMP-1 profile) - no errors.

Then I reset bios settings to initial values (Load optimized defaults) and started testing CPU cores.

It was found out that core #7 (the fastest) stably starts to generate errors after 5 minutes of load.

https://imgur.com/a/VIOYZjF

I added +4 to this core with Curve Optimizer. Now everything stable, even with XMP. 

Is my CPU degraded because of the PBO? Should i RMA this cpu?

[johnnyenglish]

Very unlikely, PBO is still bound to rules and limitations.

Unless you go all in with board tweaks and somewhat nullify the safeguards. Even so it would take quite a while to degrade if you know what your doing. Its not like you are putting over 1.5v constantly without knowing.

Monitor your CPU with some tools just to be sure and keep track of the averages and peaks.

Also, Curve Optimizer will undervolt as well.


With 6400 on that ram, be sure to check what voltage is being applied to CPU SoC.

Good Luck

[robot468]

Thanks for answer.

Core voltages not exceed 1.380v in any CPU loads.

SOC voltage is 1.250 with XMP and 1.025 with stock memory settings.

[AMID_of_IT_all]

RMA is of no use. You've probably not managed to ruin your CPU enough for it not to run in specs (that is without PBO and XMP) for a a RMA to be accepted. But keep tweaking and tinkering with the hardware long enough and you'll get there for sure. In the mean time backup your data while you still have the chance. Too bad for the peace of mind you now know the CPU is goner but you don't know when that will be. At least with a 1000w PSU the CPU has enough power to short circuit internally and go out in style. Over here the master himself explains his Silly-Con:  

https://www.youtube.com/watch?v=3WyfJyl9yfY&t=33s

Yep, those deep-cooling gadgets can give the customer a false impression of what your CPU is going through for sure. Gone are the happy days when a simple customer knew the hardware that was bought would function normally because the specifications for it to do so were made clear by the producer of IT.   

From:

https://www.youtube.com/channel/UCrwObTfqv8u1KO7Fgk-FXHQ/community?lb=UgwIDprhUtrJzFuhW0N4AaABCQ

4 years ago

What voltages were you using to get it to degrade that fast?

ActuallyHardcoreOverclocking

4 years ago

upper 1.3V range

4 years ago (edited)

My 1700 degraded after 2 weeks on 1.34v (3.8ghz) in 2017. 1600 in 2019 degraded after month on 1.38 (4.0). Soon 3500x comes, i think i will setup only 1.25-1.3 because of my experience and anyway it will degrade but little slower p.s. asus crosshair x370

Long story short: AI algorithms, like Intel's Turbo Boost Technology or AMD's PBO, already seem to burn up CPU's like candles on a winter's night. Meanwhile the costumers get more enthusiastic any time a new method, that lets them use their hardware out of specifications, is thrown at them. Soon they think they know better then using a lame AI algorithm for their overclocking. Essentially that AI makes them burn their own money. IMHO stupidity can not be more clearly defined in IT terms that make AI seem smart.