Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 7,1 According to DriveDx, my 1TB Samsung 970 EVO has 10 media and data integrity errors. How worried should I be?
- Thread starter Parzival
- Start date
- Sort by reaction score
These are most likely to be transient issues, there are no major flags that would indicate impending drive failure here. DriveDX does give a comprehensive report that the vast majority of users would NEVER see. Ensure you have backups (as always). You should be fine.
Thanks Dave, that's great to hear.
If I had to make a guess as to what caused this, I think those errors started when the disk was really full, but I'm not entirely sure. Now I try to keep at least a gigabyte of free space on all my 970 EVOs.
If you click HEALTH INDICATORS in the pane in the upper left, there is a slightly more detailed report for specific health indicators available.
Generally the only time I have seen this come up in past is when SSDs are being used as some kind of server or being used in a developer/research situation where small sets of data are constantly being written and compiled.
I have never seen this on a general system boot drive, but unless you're using with MP5,1 or other machine, those system drives are generally Apple-issued these days with NVMe being "external" in some additional storage capacity.
Generally the only time I have seen this come up in past is when SSDs are being used as some kind of server or being used in a developer/research situation where small sets of data are constantly being written and compiled.
I have never seen this on a general system boot drive, but unless you're using with MP5,1 or other machine, those system drives are generally Apple-issued these days with NVMe being "external" in some additional storage capacity.
I'd keep an eye on it and make sure to have copy/clone of your data on this drive elsewhere. If it went from 1 > 10 (an actual +9) since last reboot, it would signal something potentially wrong with the blade itself and/or adapter.
So, its possible the adapters could be causing this? Shouldn't they either work or not (sorry if this is a stupid question)?
I've been trying to copy the contents of my four 1TB 970 EVOs and one 2TB 970 EVO Plus to a standard (spinning) hard drive and most of the files (like 99%) have copied just fine, but Carbon Copy Cloner has been unable to copy certain files from two of the 1TB drives:
Copying some (but not all) of these files in the Finder also doesn't work (error code 100083).
The strange thing is that some of the files actually WERE copied, but are registered by CCC as "An error occured while CCC was... Reading data from this file on the source".
DriveDX now says the following about drive one:
And this about drive two:
I've been trying to copy the contents of my four 1TB 970 EVOs and one 2TB 970 EVO Plus to a standard (spinning) hard drive and most of the files (like 99%) have copied just fine, but Carbon Copy Cloner has been unable to copy certain files from two of the 1TB drives:
Copying some (but not all) of these files in the Finder also doesn't work (error code 100083).
The strange thing is that some of the files actually WERE copied, but are registered by CCC as "An error occured while CCC was... Reading data from this file on the source".
DriveDX now says the following about drive one:
And this about drive two:
Last edited:
There is obviously some form of data corruption here. Would not rule out bad adapters, improper blade install, or failing NVMe.
Since it appears to be happening with two separate, are they separate PCIe slots? Two different adapters?
Approximately how much free space on each blade?
NVMe blades formatted APFS or HFS+?
Is the CCC clone from NVMe to spinning HDD with SINGLE PARTITION or does the HDD have MULTIPLE PARTITIONS?
Since it appears to be happening with two separate, are they separate PCIe slots? Two different adapters?
Approximately how much free space on each blade?
NVMe blades formatted APFS or HFS+?
Is the CCC clone from NVMe to spinning HDD with SINGLE PARTITION or does the HDD have MULTIPLE PARTITIONS?
Hi bsbeamer, thanks for your reply. I really appreciate people responding as I'm panicking here.
The NVME's are both installed on separate Aquacomputer kryoM.2 PCIe cards. I have a total of 5 of these cards in the MP, 2 of which are the EVO model. I'm not suspecting the EVO models are what is giving me trouble, because last month I installed the 5th card, and the blade on that is working perfectly.
To the best of my recollection, the blades are all installed in precisely the same way, and I've been really careful when installing the PCIe cards themselves.
One blade (the one with 22 Media and Data Integrity Errors) has about 4GB of free space, the other about 1GB.
All my blades are formatted APFS.
The CC clone is 1 8TB WD Gold drive, that has a single partition. Per the CCC insrtructions, I created folders on the WD drive for each of the blades and set CCC to copy the contents of the blades to the folders.
Thanks again.
The NVME's are both installed on separate Aquacomputer kryoM.2 PCIe cards. I have a total of 5 of these cards in the MP, 2 of which are the EVO model. I'm not suspecting the EVO models are what is giving me trouble, because last month I installed the 5th card, and the blade on that is working perfectly.
To the best of my recollection, the blades are all installed in precisely the same way, and I've been really careful when installing the PCIe cards themselves.
One blade (the one with 22 Media and Data Integrity Errors) has about 4GB of free space, the other about 1GB.
All my blades are formatted APFS.
The CC clone is 1 8TB WD Gold drive, that has a single partition. Per the CCC insrtructions, I created folders on the WD drive for each of the blades and set CCC to copy the contents of the blades to the folders.
Thanks again.
You never should use a SSD until full. SSDs need space to do the garbage collection, TRIM, wear levelling operations. If you use it almost full like you are doing, the empty sectors will be heavily used instead of the writes being spread over the NAND bank. This is extremely critical with SSDs that don't have SLC cache like the cheaper QLCs or entry level ones flooding the market that store the allocation table on the NAND itself.
Using a SSD until full it's the fastest way to kill it. Leave at least 7% (best if you can leave 10%) free, some people just shrink the partition and leave the space unallocated, seems the best fool-proof way to do it.
There are a TON of issues cloning APFS SSD to multiple partition spinning HDDs, especially when those HDDs are formatted APFS. Highly recommended to have single partition HDD if you need to use APFS on the HDD, or ideally use HFS+. You’ll lose some APFS linkage, but for non boot drive it make little to no difference.
The “clone” is not exactly a clone with the folder method in CCC, but your data is all backed up. Works best with HFS+. For non bootable, little to no noticeable difference. Just make sure HDD is single partition.
Your free space allocation on SDD is not enough. No other way to say that. That alone could be (very likely is) causing your data and corruption issues. You really need at least 10% of the drive capacity free for normal operations. I tend to suggest to keep this at least 20-25% whenever possible, especially if you’re looking for speed. DriveDx usually offers warnings on drive capacity as well, unless you disabled or ignored them.
The “clone” is not exactly a clone with the folder method in CCC, but your data is all backed up. Works best with HFS+. For non bootable, little to no noticeable difference. Just make sure HDD is single partition.
Your free space allocation on SDD is not enough. No other way to say that. That alone could be (very likely is) causing your data and corruption issues. You really need at least 10% of the drive capacity free for normal operations. I tend to suggest to keep this at least 20-25% whenever possible, especially if you’re looking for speed. DriveDx usually offers warnings on drive capacity as well, unless you disabled or ignored them.
Thanks tsialex and bsbeamer!
I should clarify, the large WD backup drive is formatted Mac OS Extended (Journaled), only the blades are formatted APFS.
The blades are being used as sample drives, so theres not a lot of writing happening, mostly reading.
So, just to clarify: I've got four 1TB 970 EVOs and one 2TB 970 EVO Plus.
For some reason, the DriveDX Free Space Monitoring was disabled on ALL my drives, and it turns out that the drives with Media and Data Integrity Errors are also the ones with the low disk space icon:
I've started freeing up space.
So, is it safe to say these errors started popping up because of the low disk space?
And has there been any permanent damage to my drives? A couple of corrupted files I can live with, but have my drives been fried?
I should clarify, the large WD backup drive is formatted Mac OS Extended (Journaled), only the blades are formatted APFS.
The blades are being used as sample drives, so theres not a lot of writing happening, mostly reading.
So, just to clarify: I've got four 1TB 970 EVOs and one 2TB 970 EVO Plus.
For some reason, the DriveDX Free Space Monitoring was disabled on ALL my drives, and it turns out that the drives with Media and Data Integrity Errors are also the ones with the low disk space icon:
I've started freeing up space.
So, is it safe to say these errors started popping up because of the low disk space?
And has there been any permanent damage to my drives? A couple of corrupted files I can live with, but have my drives been fried?
Even completely read only solid state drives needs space to do the wear levelling. SSDs that use multiple bit storage (MLC, TLC, QLC NAND) need to be refreshed from time to time, the content of the cell is written again over the time and data is moved during this process.
Never use a SSD completely full, without free space to spread the wear, the same cells are used over and over, to the point that it starts to die.
Ask Samsung if they can replace the drives, I wouldn't use them again for anything serious.
I would personally retire these drives for anything critical, but it is an individual risk assessment. Prior experience - once you have an issue, it is the first warning sign of things to come. Could it work for years without issues if you keep 20%+ free? Possibly, maybe even probably. But is that worth the risk and constant questioning or worry?
If you do not NEED the NVMe speed, a larger capacity SATA SSD for accessible storage might be worth the investment.
If you do not NEED the NVMe speed, a larger capacity SATA SSD for accessible storage might be worth the investment.
To add some detail that I'm sure that you understand, but didn't include....
Once data has been written to a region on an SSD, it cannot be "unwritten" and put in the free pool by shrinking the partition without a full erase or a "TRIM" command on the released space.
Better to never put the entire disk in partitions - but leave LOTS of free space not in any partition when initializing the disk for the first time.
If it's been written to full before the shrink - force a TRIM command on the free space (e.g. by making a filesystem on the free space, and then forcing the filesystem to TRIM free space, then deleting the filesystem) or verify that the shrink operation issues a TRIM.
Or, if you have an OWC SSD, ignore it because OWC claims to have some magic pixie dust that makes TRIM unnecessary.
Last edited: