Overclock.net banner

16241 - 16260 of 16712 Posts

·
Registered
Joined
·
2,789 Posts
I am helping a friend troubleshoot his system he has the following specs

5950X
X570 Aorus Master (F33j)
G.Skill Trident Z Neo 2x16GB 3600Mhz CL18
Asus TUF 3090
2x Samsung 970 Evo Plus
Seasonic Prime Gold 1000Watts

He normally plays FS2020 and DCS and will get a random reboot at times and we've tested with PBO and Coreboost disabled, this has stopped any random reboots for the last few days.

We have not yet moved into enabling PBO/Coreboost yet but if the random reboots come back once PBO is enabled, does this mean that there is a problem with the CPU it self ?
 

·
Registered
Joined
·
581 Posts
Oh my... I've never seen this.
The factory/silicon order is the above, without CPPC.
The CPPC one is "adjusted" by AGESA.
Usually to avoid cores in CCD2 to have priority against those in CCD1.
Go in the BIOS and check you have both CPPC options on Enabled (not Auto).
Otherwise it's a BIOS issue; you may have to try another version.

I've spent all the morning fine tuning the CO counts... they changed how it works in AGESA 1.2.0.x
I'll post later how I did the tuning, maybe it can help you.
found the problem...

If you deactivate CnQ the order gets changed to that abomination.....guess its cool and quiet because load gets only applied to the cores that are better = less voltage needed.
 

·
Old crazy guy
Joined
·
3,175 Posts
found the problem...

If you deactivate CnQ the order gets changed to that abomination.....guess its cool and quiet because load gets only applied to the cores that are better = less voltage needed.
Guess it's messing with CPPC when you disable it.

I've posted the small guide to fine tune the cores


We have not yet moved into enabling PBO/Coreboost yet but if the random reboots come back once PBO is enabled, does this mean that there is a problem with the CPU it self ?
Did you enable the Curve Optimizer?
The most frequent reason for reboots is CO.
Is there a particular situation when it happens? During idle or high load?
 

·
Registered
Joined
·
76 Posts
Most important for checking stability with corecycler: no for- and the least possible background activities and maximize your cpu cooling while corecycler is running. This leads to maximum boost frequencies faster and longer.
 

·
Registered
Joined
·
581 Posts
I still don't get CO honestly...tested all cores out according to corecycler/OCCT and my 5th best core is now the one boosting highest while my two best ones are 50Mhz below, lol.

Maybe I need higher EDC dunno.
 

·
Old crazy guy
Joined
·
3,175 Posts
I still don't get CO honestly...tested all cores out according to corecycler/OCCT and my 5th best core is now the one boosting highest while my two best ones are 50Mhz below, lol.

Maybe I need higher EDC dunno.
The latest AGESA it's quite odd on how it boost with heavy load.
If you want to check the max absolute boost and how they behave with a light workload use boosttester:

CoreCycler uses Prime95, by default, it's much heavier than OCCT.
I had to raise VSOC to 1.175V and LLC to 2 (which is the level below the maximum on MSI) to pass it.
 

·
Registered
Joined
·
581 Posts
The latest AGESA it's quite odd on how it boost with heavy load.
If you want to check the max absolute boost and how they behave with a light workload use boosttester:

CoreCycler uses Prime95, by default, it's much heavier than OCCT.
I had to raise VSOC to 1.175V and LLC to 2 (which is the level below the maximum on MSI) to pass it.
I was talking about light load boosts, didnt care much about the p95 boosts honestly (somewhere between 4.9 and 5 i guess). Super weird...core cycler forced me to -8 on the good cores, which killed their boosts but my fifth strongest core is still passing at -30 ....... and thus boosting 50Mhz over my two best core...*** is happening

Maybe my power limits are ****ing me over...210/144/166...maybe i should test the cores out without limits.
 

·
Old crazy guy
Joined
·
3,175 Posts
I was talking about light load boosts, didnt care much about the p95 boosts honestly (somewhere between 4.9 and 5 i guess). Super weird...core cycler forced me to -8 on the good cores, which killed their boosts but my fifth strongest core is still passing at -30 ....... and thus boosting 50Mhz over my two best core...*** is happening

Maybe my power limits are ****ing me over...210/144/166...maybe i should test the cores out without limits.
I think it's the latest AGESA being weird with Prime95 workload.
You can also forget Prime95 if you are not picky...
If you pass OCCT SSE/Large/Extreme should be enough.
I've been a while not passing both and didn't even notice it.
 

·
Gaming enthusiast
Joined
·
873 Posts
del
 

·
Registered
Joined
·
581 Posts
I think it's the latest AGESA being weird with Prime95 workload.
You can also forget Prime95 if you are not picky...
If you pass OCCT SSE/Large/Extreme should be enough.
I've been a while not passing both and didn't even notice it.
Well, I dont need 12hours prime95 stability but 3 iterations of cycler should be ok as a target....but those results... lol
 

·
Registered
Joined
·
356 Posts
I think it's the latest AGESA being weird with Prime95 workload.
You can also forget Prime95 if you are not picky...
If you pass OCCT SSE/Large/Extreme should be enough.
I've been a while not passing both and didn't even notice it.
Prime95 has changed from a super usable stability benchmark to a very non-real-world one. I've used its 2.92 version for a long time, then realized it doesn't really serve as a stability tool anymore since v3. Since then I've been using x264 and x265 (AVX/AVX2) and OCCT to confirm daily stability.

Edit: Small and smallest are really non-usable anymore.
Edit 2: Verfifying CTR 2.1 stability with Prime is pretty much impossible nowadays. Even though its one-click OC is stable in all real-world loads.
 

·
Gaming enthusiast
Joined
·
873 Posts
New ram and I cannot change 1t to 2t no matter what I try.....


2489675
 

·
Registered
Joined
·
108 Posts
Sorry for the silly question: did you check the actual BIOS version is still F33C?
Seems to me it failed to boot and switched to the backup BIOS.
If that's so; there's a dip switch to disable the dynamic Dual BIOS switch.
Just set main or backup with the other dip switch. It's safer as it'll only switch when you ask for it.
Yes I did check, it is definitely F33C still, at least the BIOS is reporting as much. The BIOS it shipped with, which presumably is on the #2 BIOS, is definitely not F33C.
 

·
Old crazy guy
Joined
·
3,175 Posts
Prime95 has changed from a super usable stability benchmark to a very non-real-world one. I've used its 2.92 version for a long time, then realized it doesn't really serve as a stability tool anymore since v3. Since then I've been using x264 and x265 (AVX/AVX2) and OCCT to confirm daily stability.
Note that the way Prime95 is ran by CoreCycler is using SSE instructions instead of AVX.
I've always preferred to have Prime95 Small AVX stable in any case but I agree is very extreme.
Prime95 using SSE worries me more than a bit if it's not stable.
It's very harsh but not so extreme.

Code:
WorkerThreads=1
CpuSupportsSSE=1
CpuSupportsSSE2=1
CpuSupportsAVX=0
CpuSupportsAVX2=0
CpuSupportsFMA3=0
 

·
Registered
Joined
·
2,789 Posts
Did you enable the Curve Optimizer?
The most frequent reason for reboots is CO.
Is there a particular situation when it happens? During idle or high load?
Due to the pandemic I am only helping my friend remotely, Is the curve optimizer enabled by default? What is the most safe settings for this as I do not have a Ryzen system I am unable to help him with this specific setting.


He only plays two games in particular which and I can only assume those games fall under medium load and probably only use 4 cores.
 

·
Registered
Joined
·
581 Posts
The only two things i can think of why my no. 5 and no. 8 rated core can do 20 iterations of core cycler at minus 30 and my two best ones are at -5 and -8 is a problem with set EDC limit or a problem with overdrive set to 200Mhz...really dont understand whats going on here lol
 

·
Old crazy guy
Joined
·
3,175 Posts
Due to the pandemic I am only helping my friend remotely, Is the curve optimizer enabled by default? What is the most safe settings for this as I do not have a Ryzen system I am unable to help him with this specific setting.


He only plays two games in particular which and I can only assume those games fall under medium load and probably only use 4 cores.
No, it's not enabled by default.
Most safe is without PBO but it's weird that at default it's random rebooting.
There have been specific CPU that couldn't handle even the default settings but it's rare.
More often is a symptom of something else wrong.
Check if PBO is Enabled in Auto. Maybe he did set Motherboard limits and that's an option that can cause reboots sometimes.

If your friend plugs in a USB stick formatted FAT32 he can save BIOS screenshots via F12 key.
It'd be much easier for you to help him.
Ask him to take a screenshot of every screen.

The only two things i can think of why my no. 5 and no. 8 rated core can do 20 iterations of core cycler at minus 30 and my two best ones are at -5 and -8 is a problem with set EDC limit or a problem with overdrive set to 200Mhz...really dont understand whats going on here lol
Did you check what's the speed?
I have also bad cores that can run be set at -30, they run at 4.89 GHz. The best one max -11 but runs at 5,03 GHz.
The second best can do max -20 but it runs at 5,12 GHz.

Yes I did check, it is definitely F33C still, at least the BIOS is reporting as much. The BIOS it shipped with, which presumably is on the #2 BIOS, is definitely not F33C.
It's weird indeed... try re-flash and if it doesn't work the only option I see is to use flashrom.
 

·
Registered
Joined
·
208 Posts
No, it's not enabled by default.
Most safe is without PBO but it's weird that at default it's random rebooting.
There have been specific CPU that couldn't handle even the default settings but it's rare.
It's not that rare, I had 2 5900x that rebooted at default settings and the 3rd is now stable, like I posted last week.
If it's an earlier sample chances are higher.
 

·
Registered
Joined
·
76 Posts
...
He normally plays FS2020 and DCS and will get a random reboot at times and we've tested with PBO and Coreboost disabled, this has stopped any random reboots for the last few days.

We have not yet moved into enabling PBO/Coreboost yet but if the random reboots come back once PBO is enabled, does this mean that there is a problem with the CPU it self ?
You got *faulty CPU: Replaced 3950X with 5950X = WHEA and reboots
(the thread is not only for 5950X CPUs, every Zen3 CPU could be faulty)

*faulty means: The cpu cannot operate reliably at AMD's default specifications for this Zen3 model. Set everything to Defaults and do a UEFI-Reset and then only activate PBO->advanced and put +15 all-core in curve optimizer. If this helps to troubleshoot your problems, do you a favour and rma the cpu.
 

·
Old crazy guy
Joined
·
3,175 Posts
It's not that rare, I had 2 5900x that rebooted at default settings and the 3rd is now stable, like I posted last week.
If it's an earlier sample chances are higher.
Gosh, I forgot.
He's rebooting only with PBO enabled so I guess there's still a 50% chance it's about the settings.

New ram and I cannot change 1t to 2t no matter what I try.....
Try this profile from Kedar:


The 19-21 was suggested to me by someone and I find it a bit easier to get stable and a bit better in benchmarks. Technically it should be 21-21 but I find 19-21 benches a bit better. I changed the SCLs to 4 and lowered the ProODT to 40 and the ClkDrStr to 40, this passes TM5, SCLs at 2 doesn't. 1.2v on SoC is the highest I'll go but I've seen people running as high as 1.25v.
Prime95 has changed from a super usable stability benchmark to a very non-real-world one. I've used its 2.92 version for a long time, then realized it doesn't really serve as a stability tool anymore since v3. Since then I've been using x264 and x265 (AVX/AVX2) and OCCT to confirm daily stability.

Edit: Small and smallest are really non-usable anymore.
Edit 2: Verfifying CTR 2.1 stability with Prime is pretty much impossible nowadays. Even though its one-click OC is stable in all real-world loads.
CoreCycler, by default, is using SSE. Much less strain than AVX/AVX2.
My settings that were working with the previous AGESA failed also on OCCT SSE/Large/Extreme.
It's still a bit extreme but it's not that much unrealistic as workload.
A game engine, especially if lightly threaded, can be very similar.
I'm bothered to live in the doubt that if a game crashes maybe it's the CO config.
I'll keep the previous config only for benchmarks but I'll use for daily something safe.
 
16241 - 16260 of 16712 Posts
Top