Overclock.net banner
1 - 7 of 7 Posts

highc1157

· Registered
Joined
·
37 Posts
Discussion starter · #1 ·
I've only encountered this issue since I upgraded to a Ryzen 5800x3d CPU. I've had the chip for 6 months but never started getting this freezing issue until last month. I'm not sure its a CPU issue, but that's what I'm assuming from other posts about idle freezing here and on other forums.

When idle freezes occur, CPU temps are around 48-63. I've had a couple instances of freezing during games, or games crash, or cpu restarts while gaming, but, that's probably 1 in 10 times. 9 of 10 times it freezes when idle, for instance, when I am actually afk from PC and walk back into room, PC can be frozen. I'm getting freezes almost daily for last month or so.

I've ran every combination of PBO undervolt offsets I can think of, ran OCCT for 8 hours today using a variety of tunings (-30 all cores, -25/-25 best cores (-30 rest), -15/-10 best cores (-30 rest), but still can't produce an error during stability testing. I've also used a bunch of different OCCT stability settings, still, no errors. I also ran Core Cycler for 7 hours overnight last night, which produced no issues on -30 all cores.

For PBO I use : 114 PPT / 75 TDC / 115 EDC

Any suggestions on where to go from here?

System: (amd chipset drivers up to date, bios up to date, windows up to date)

ryzen 7 5800x3d - cooling is peerless assasin 120 - room temp 75F
3070
32GB ddr4 3000
samsung ssds
650W Corsair
MSI x470 gaming plus
windows 10 - version 19045
 
Most likely your issue.
Personally iv never found a 5800X3D that does -30 outside of only running benchmarks.
Iv built 11 of them now with the best doing -25 on all but the best two cores which needed -20.
The rest of them fell between -10 to -20 with -15 being the starting point for stability testing I use then dialing back the 2 best cores if it has issues.
Full load will hardly ever crash and even single core full load is less likely to crash. All the crashes/issues iv ever seen with to much undervolt have been when playing games the game will freeze for 2-3 secs but not crash, And when the computer is idle it will randomly restart with no errors at all just like some one hit the restart button.
 
Competing instances of core cycler (though I recommend 720k FFTs for Prime), or just disabling the 100MHz C6 boost (by disabling C-states) and testing all-core loads, should reveal the overwhelming majority of CO related instability.

Both of my 5800X3D's do -30 CO with unconditional stability (e.g. weeks of uptime at a stretch in all sorts of loads and idle conditions, no mysterious behavior), but one needs C6 boost disabled to not fail core cycler.

Anyway, I'd suspect SoC/memory before idle CO problems. Actual idle CPU voltage (not VID, but the voltage being delivered to the chip as reported by SVI2) should only be barely lower than load voltage, unless the CPU is asleep.

Any errors in event viewer that could be relevant?
 
Most likely your issue.
Personally iv never found a 5800X3D that does -30 outside of only running benchmarks.
Iv built 11 of them now with the best doing -25 on all but the best two cores which needed -20.
The rest of them fell between -10 to -20 with -15 being the starting point for stability testing I use then dialing back the 2 best cores if it has issues.
Full load will hardly ever crash and even single core full load is less likely to crash. All the crashes/issues iv ever seen with to much undervolt have been when playing games the game will freeze for 2-3 secs but not crash, And when the computer is idle it will randomly restart with no errors at all just like some one hit the restart button.
I've got mine on all core -30 and ran double corecycler ( 5800X3D Owners ), OCCT, etc. 24 hours, zero issues.
I thought it was normal for most to achieve this?
 
Kind of. It's a lottery just like any other CPU in terms of overclocking / undervolting. I had 3 X3D's come through my hands and only 1 does -30 with ease and can go way lower even if it had more -CO. It can run with the 1.2.0.6 sleep voltage bug as low as 1.110v just fine which is way lower then -30 allows. The other -23 at best and the 3rd is trash and barely does -11.

-30 is not a given. Check event viewer for WHEA 18 errors related to core.
 
is cpu undervoltage
suggestion

step 1
: get an cores approximation
CoreCycler:
stressTestProgram = YCRUNCHER
[yCruncher] mode = 20-ZN3 ~ Yuzuki
runtimePerCore = 10m
maxIterations = 3
coreTestOrder = (single core)


step 2 : get core final value
CoreCycler:
stressTestProgram = PRIME95
[PRIME95] mode = SSE
[PRIME95] FFTSize = Large
runtimePerCore = auto
maxIterations = 10
coreTestOrder = (single core)
 
1 - 7 of 7 Posts