Well people, this has got a whole lot more interesting.
I got another drive from samsung, they exchanged it. Its a what I presume a refurbished 850 pro, given they no longer manufactured.
I put the drive in the system and crc errors spiral up.
I then test it in my AMD system, and is no crc errors, test it in my NUC, no crc errors.
Put it back in my intel machine, crc errors. I swap its port and cable with my 860 evo, still crc errors, but at the same the 860 evo gets 0 errors on the port and cable the 850 pro is on, so whats going on?
Just to clarify as well the crc errors had visible effects, such as 10 seconds to load paint, chkdsk taking 3x as long and a massive slowdown on read performance on benchmarks (no affect on write).
I then thought what if I have a bad cable on a good sata port and a good cable on a bad sata port.
So I swapped the cables and put the 850 pro again on the second port, and the crc errors stopped. Did lots of benchmarking, various tests no crc errors.
The explanation for the 860 evo getting no errors I think its down to the enhanced ECC controller, the 860 series SSD's have better error correction. Which as a side effect it seems can mitigate dodgy cables/ports. I tested the 830 from my laptop on the bad port and that also had spiralling crc errors during reads.
So I did the following as I didnt want any storage device on that bad port and I have no spare ports.
Optical drive moved to sata port 0 (bad port)
System SSD on sata port 1
Second SSD (860 evo) on sata port 2
Spindles moved round accordingly, one of them moved to the asmedia port that the optical drive was using.
However there is bad news, I have had 2 CRC errors since I did this setup in the space of about 30 hours uptime. Its never gone up during a benchmark, one came whilst not on pc, another came during a chkdsk, but I have done dozens of chkdsk's (trying to trigger it) without issue. So this other port is definitely in much better condition but there is still something not right, I havent checked any other ports, and I am ordering more cables.
It is looking like tho I am going to switch the board, probably go back to asus as on asus boards I have never had sata ports fail and especially when the board is as new as this one is.
Now I may be wrong but from my observations the CRC error counter on SSD's is related to failed read attempts, I believe failed write attempts wont increment this counter, with writes you either get what the controller thinks is a success but actually flipped bit (silent corruption), or a completely failed write which would increase SMART but I think would be a different variable. So my concern is silent data corruption at this point.
Also perhaps what is worth mentioning is the affect of the drivers. The system has been running on the msahci drivers, which I always considered the safer drivers to use. I observed during testing that (a) errors accrued faster on msahci vs iRST, and (b) when 2 more or more CRC errors occurred during a session on iRST then until the next reboot that device is throttled to sata 300 speeds, which killed the errors dead. Been throttled to half burst speed is preferable to having to retry reads and having silent data corruption. So I now consider intel drivers safer.
What I am doing going forward.
I think I am going to buy a 1tb 860 EVO to replace the 850 pro in this rig, having realised the value of better ECC. I will repurpose the 850 pro to something else. Currently the system is running with the 850 on the second sata port.
I will replace the board, damaged sata traces I think makes this board impractical to use and I can only see that going one way which is further decaying.
Asus Z390 ROG hero board to replace it alongside an asmedia sata card since that board only has 6 sata ports.
The board itself has been a nightmare if I am been honest.
Purchased the board, and originally it couldnt even post all the time just in XMP mode on two different sets of ram, even in jedec it wasnt 100%. Brought a 2nd board, and the posting issue was fixed, so first board was returned.
On the 2nd board which is this one in use now, it has aweful coil whine when cstates are enabled, in addition coil whine when NIC port is active regardless of cstates on/off. It also cannot manage 3200mhz ram speeds on the secondary dimm slots, and now has broken SATA ports. It is out of retail warranty although I will check if it has a manufacturer one.
I am considering the option of just throttling sata down to gen 2, as can be done in the bios.
The kingston SSD is like the 860 evo in that it can just eat up all the errors and correct them on the fly. Hence that was working ok.
final note, my hatred of m.2 has diminished, I now see a legit value of removing cables from the equation. Although I still consider the better solution been using pcie slots for nvme.
