Occasional System File Corruption 7950X

[Auraka]

Running W10, a 7950X on an ASUS ROG Strix X670E-E Gaming, PBO set with a mild -10 negative curve, buildzoid sub timings, and I occasionally experience some system file corruption that if left uncorrected will result in some video corruption, and possibly some suttery-like performance. Running Elevated CMD, then DISM.exe /Online /Cleanup-image /Restorehealth, then sfc /scannow usually corrects the issue for a while (weeks, sometimes many) until it starts acting a certain way. Memtest (purchased version) shows zero errors, I'm a bit at a loss with what appears to be generally a very stable system. Water cooled custom loop with a single large 180x360x45mm rad up front in a Fractal Torrent, don't believe heat is an issue at all.

 

[arestavo]

I'd run some stress tests that are a bit more....well, stressful. Memtest won't suss out deeper issues with the memory controller, as I found out for myself with a failing i9 10940X until it got to the point where I had multiple blue screens a day.

TestMem5, Aida64extreme, Prime95. I'm sure that other folks can chime in as well.

EDIT - there was also an issue with Windows 11 where an older version had, if I remember correctly, a bluetooth driver show as corrupted with every sfc /scannow. It was corrected with a later windows version update. Are you on the latest 24H2 build 26100.3476? If that HWiNFO64 software reading is accurate, you can disregard - I don't think that was an issue with windows 10 (and if it is, it's unlikely to get fixed).

 

[LigTasm]

Only things I see out of your screenshots if your tRAS is quite low for no real purpose, setting it tRCDRD+tRC might help. But honestly I would try running 2000FCLK and see if that fixes it. Some CPUs just do not like certain fabric speeds.

 

[Auraka]

I'd run some stress tests that are a bit more....well, stressful. Memtest won't suss out deeper issues with the memory controller, as I found out for myself with a failing i9 10940X until it got to the point where I had multiple blue screens a day.

TestMem5, Aida64extreme, Prime95. I'm sure that other folks can chime in as well.

I've done all that, or most of it, no failures. I'm suspecting either a memory controller, or PBO -10 curve when a core ramps up to 5850.

 

[cyberblogger]

I've done all that, or most of it, no failures. I'm suspecting either a memory controller, or PBO -10 curve when a core ramps up to 5850.

I would suspect the ram module as that is more of a symptom of memory overclocking failure on modules greater than 4Gb.

 

[Dreamerbydesign]

Another vote for a ram issue. Doubt this is a cpu issue.

 

[Auraka]

This is the fix. I can post the log file if someone knows what to look for.

Microsoft Windows [Version 10.0.19045.5679]
(c) Microsoft Corporation. All rights reserved.

C:\Windows\system32>DISM.exe /Online /Cleanup-image /Restorehealth

Deployment Image Servicing and Management tool
Version: 10.0.19041.3636

Image Version: 10.0.19045.5679

[==========================100.0%==========================] The restore operation completed successfully.
The operation completed successfully.

C:\Windows\system32>sfc /scannow

Beginning system scan. This process will take some time.

Beginning verification phase of system scan.
Verification 100% complete.

Windows Resource Protection found corrupt files and successfully repaired them.
For online repairs, details are included in the CBS log file located at
windir\Logs\CBS\CBS.log. For example C:\Windows\Logs\CBS\CBS.log. For offline
repairs, details are included in the log file provided by the /OFFLOGFILE flag.

C:\Windows\system32>

 

[Mr Evil]

How about the SSD? I had a Samsung 960 Pro that would randomly flip bits for no reason.

 

[chameleoneel]

Running W10, a 7950X on an ASUS ROG Strix X670E-E Gaming, PBO set with a mild -10 negative curve, buildzoid sub timings, and I occasionally experience some system file corruption that if left uncorrected will result in some video corruption, and possibly some suttery-like performance. Running Elevated CMD, then DISM.exe /Online /Cleanup-image /Restorehealth, then sfc /scannow usually corrects the issue for a while (weeks, sometimes many) until it starts acting a certain way. Memtest (purchased version) shows zero errors, I'm a bit at a loss with what appears to be generally a very stable system. Water cooled custom loop with a single large 180x360x45mm rad up front in a Fractal Torrent, don't believe heat is an issue at all.

View attachment 718598View attachment 718599

does it happen with no CO?

does it happen with XMP/EXPO, but not BZ timings?

 

[Auraka]

Only things I see out of your screenshots if your tRAS is quite low for no real purpose, setting it tRCDRD+tRC might help. But honestly I would try running 2000FCLK and see if that fixes it. Some CPUs just do not like certain fabric speeds.

Switched to tRAS=tRCD(primary timing) + tRTP(secondary timing)36+12=48

 

[Auraka]

How about the SSD? I had a Samsung 960 Pro that would randomly flip bits for no reason.

Not sure how I'd even check that, Samsung tools?

 

[Mr Evil]

Not sure how I'd even check that, Samsung tools?

There will probably be an error count displayed in there somewhere that should indicate if something is amiss. Or, you can make a fair guess by looking at the modified/created dates of the files being corrupted. If it's old stuff that was fine before and has now suddenly become corrupt, then it could be the SSD itself corrupting data at rest. If it's files that were recently written to, then maybe the corruption is happening in RAM or in the CPU before it's written to the SSD.

 

[freeagentt]

tRFC is super tight for the voltage.

 

[Auraka]

Only things I see out of your screenshots if your tRAS is quite low for no real purpose, setting it tRCDRD+tRC might help. But honestly I would try running 2000FCLK and see if that fixes it. Some CPUs just do not like certain fabric speeds.

FCLK set to 2000 because there is likely no benefit with my current bios version. The only other variable that's easy is heat related to gaming. I've made my fan curves a tick more aggressive to help cool the dimms.

 

[Auraka]

Most recent record playing Doom Eternal 2K

 

[Dan_D]

Typically when you see this kind of thing it comes down to a faulty SSD. It may have intermittent issues. I wouldn't rule out memory issues either though I doubt this is hardware and more of a configuration issue. Sometimes some memory kits are really fickle on some specific motherboards. I've seen this very type of issue many times with different fixes being needed to resolve it. Namely, SSD replacement (I've also seen this with mechanical drives), memory module replacement, memory settings changes, and even motherboard problems. Though the latter is by far the most rare fix for an issue like this.

When I see issues like this, I ditch the overclock and any super tight memory settings. Run XMP/EXPO to configure and just set PBO and automatic CPU settings and see what happens.

 

[chameleoneel]

does it happen with no CO?

does it happen with XMP/EXPO, but not BZ timings?

Typically when you see this kind of thing it comes down to a faulty SSD. It may have intermittent issues. I wouldn't rule out memory issues either though I doubt this is hardware and more of a configuration issue. Sometimes some memory kits are really fickle on some specific motherboards. I've seen this very type of issue many times with different fixes being needed to resolve it. Namely, SSD replacement (I've also seen this with mechanical drives), memory module replacement, memory settings changes, and even motherboard problems. Though the latter is by far the most rare fix for an issue like this.

When I see issues like this, I ditch the overclock and any super tight memory settings. Run XMP/EXPO to configure and just set PBO and automatic CPU settings and see what happens.

Indeed, I tried to ask if there were issues with the stock timing profile and no undervolt/Curve Offset.

 

[Auraka]

Indeed, I tried to ask if there were issues with the stock timing profile and no undervolt/Curve Offset.

I didn't ignore you, I'm trying a suggestion (or three) at a time. The issue takes weeks to appear and I suspect might be heat related, it seems to start after a gaming session, though not right away after sfc. So I'm making changes slowly and seeing if it persists, so far it has not with three changes FLCK to 2000, tras to 48, and a bump in fans curves. I have noticed my 5V is low under load, but that may only be true at the point the chip is reading it, not at a supply 5V. The last image I posted includes DIMM temps.

 

[KonaKona040]

Run ycruncher VT3 on loop if you think it's a stability issue.

 

[emphy]

Another vote for checking the ssd and making sure its firmware is up to date. Both can be done with samsung's tools.

Additionally (likely unrelated, unless it's your main drive), definitely make sure that the 870evo's firmware is up to date, as that model is known to silently destroy itself on older ones.

 

[Auraka]

2025-04-15 20:35:11, Info CSI 00000007 StoreCorruptionRepair transaction begun. WcpVersion: [l:38]'10.0.19041.5676 (WinBuild.160101.0800)'.
2025-04-15 20:35:11, Info CSI 00000008@2025/4/16:00:35:11.552 Starting corruption detection (InnerFlags=2)
2025-04-15 20:35:11, Info CBS FLOW: Entering stage: CheckCsi
2025-04-15 20:36:17, Info CSI 00000009 Direct SIL provider: Number of files opened: 102588.
2025-04-15 20:36:17, Info CSI 0000000a@2025/4/16:00:36:17.056 Corruption detection complete. numCorruptions = 0, Disp = 1.
2025-04-15 20:36:17, Info CBS Repr: CSI meta data corruption found, will commit repair transaction if repair is asked.
2025-04-15 20:36:17, Info CBS Repr: CSI Store check completes
2025-04-15 20:36:17, Info CBS FLOW: Entering stage: CheckStagedPackages
2025-04-15 20:36:18, Info CBS Appl:Feature On Demand package without explicit comparator, using GE on build version
2025-04-15 20:36:18, Info CBS Appl:Feature On Demand package without explicit comparator, using GE on build version
2025-04-15 20:36:18, Info CBS Package Format: PSFX
2025-04-15 20:36:18, Info CBS Appl:LCU package and revision compare set to explicit
2025-04-15 20:36:18, Info CBS Package Format: PSFX
2025-04-15 20:36:18, Info CBS Delta Format: ForwardOnly
2025-04-15 20:36:18, Info CBS Generate playback status for m_fGeneratePlaybackDeltaWhenPackageHasReverseDeltas: 0
2025-04-15 20:36:18, Info CBS Generate playback status for m_fShouldGeneratePlaybackDeltas: 1
2025-04-15 20:36:18, Info CBS KIR Feature_Servicing_PlaybackDeltaGenerationV2 is enabled
2025-04-15 20:36:18, Info CBS Package Format: PSFX
2025-04-15 20:36:18, Info CBS Update.mum in package has 'PackageSupportedFeatures = 0x1
2025-04-15 20:36:18, Info CBS Repr: Staged package check completes
2025-04-15 20:36:18, Info CBS Exec:Only CSI metadata corruption found, Process CSI repair.
2025-04-15 20:36:18, Info CBS FLOW: Entering stage: RepairCsi
2025-04-15 20:36:18, Info CSI 0000000b Creating NT transaction (seq 1)
2025-04-15 20:36:18, Info CSI 0000000c Created NT transaction (seq 1) result 0x00000000, handle @0x25f40
2025-04-15 20:36:18, Info CSI 0000000d Transaction UoW: {318dc99d-1a5a-11f0-bcd5-c87f54cf94ce} for BeginTransaction
2025-04-15 20:36:18, Info CSI 0000000e@2025/4/16:00:36:18.237 Beginning NT transaction commit...
2025-04-15 20:36:18, Info CSI 0000000f Transaction UoW: {318dc99d-1a5a-11f0-bcd5-c87f54cf94ce} for CommitTransaction
2025-04-15 20:36:18, Info CSI 00000010@2025/4/16:00:36:18.252 CSI perf trace:
CSIPERF:TXCOMMIT;11023
2025-04-15 20:36:18, Info CSI 00000011 StoreCorruptionRepair transaction completed.
2025-04-15 20:36:18, Info CBS Repr: Finished repairing CSI store.

Can't say this is consistent with prior corruption, but after this I did update the 970 Pro firmware, so we'll see.

 

[Auraka]

SFC found this which seems like a nothing burger...

2025-04-29 21:16:06, Info CSI 000001e3 [SR] Repairing 1 components
2025-04-29 21:16:06, Info CSI 000001e4 [SR] Beginning Verify and Repair transaction
2025-04-29 21:16:06, Info CSI 000001e5 [SR] Repairing file \??\C:\Users\Default\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\\OneDrive.lnk from store
2025-04-29 21:16:06, Info CSI 000001e6 [SR] Repair complete

 

[ND40oz]

The onedrive.lnk getting borked makes it seem even more like an SSD issue. Replace the OS SSD and run off of it for a while to see if the issue persists.

 

[Auraka]

The onedrive.lnk getting borked makes it seem even more like an SSD issue. Replace the OS SSD and run off of it for a while to see if the issue persists.

Smells more like MS bullshit to me lol, I uninstalled it. My next drive is a 4TB 990 Pro.

 

[ND40oz]

Smells more like MS bullshit to me lol, I uninstalled it. My next drive is a 4TB 990 Pro.

You should never even have to run the system file checker, something is corrupting that data and it's most likely a bad SSD given everything you've experienced so far. What's the current SSD and how are the temps on it?

 

[Auraka]

You should never even have to run the system file checker, something is corrupting that data and it's most likely a bad SSD given everything you've experienced so far. What's the current SSD and how are the temps on it?

2TB 990 Pro, drive temps gaming attached, top box. HWinfo says drive temp 3 is usually the ASIC controller. It's installed on an ASUS X670E-E MB under the main drive heatsink. That OneDrive issue might be a bug. I've done several tweaks that may have eliminated real corruption, be it SSD or Memory Controller or DIMM related.