Question Frequent Long I/O Blocking

[Brett]

Hey folks,

I’m at my wits end with this one. I am getting frequent, long IO pauses on my machine. These pauses last for 30-45 seconds and happen every 1-2 minutes. I can trivially reproduce any time I run CrystalDiskMark. These IO pauses do not happen on my OS’s NVME SSD, only on the magnetic spinning drives connected via SATA. I’ve updated my bios and installed all the latest drivers and nothing fixes it.

OS: Happens in Windows 11 Pro and Windows Server 2022
Drives: 4xSeagate EXOS 16GB. I’ve tried them using mobo raid 10, individual disks, windows jbod, windows raid 1, and windows raid 0 and it has no effect.
Motherboard: Gigabyte Z690 Gaming DDR4
CPU: Intel i5-12600K
RAM: 32GB Crucial DDR4 3200

given what I’ve tried, I’m beginning to wonder if there’s an issue with my SATA controller.

I’ve attached IO graphs from windows performance monitor. As you can see, all IO queues up and then stops.

any help is very much appreciated!

 

[igor_kavinski]

Maybe you got all the HDDs from the same bad manufacturing batch? What is the nature of the I/O on these HDDs? If it involves writing/updating large number of small files, it could be that these drives are based on notorious SMR tech. There is no solution if that is the case, other than getting CMR drives.

 

[Brett]

Maybe you got all the HDDs from the same bad manufacturing batch? What is the nature of the I/O on these HDDs? If it involves writing/updating large number of small files, it could be that these drives are based on notorious SMR tech. There is no solution if that is the case, other than getting CMR drives.

These drives are CMR.
The IO on them is primarly Blue Iris (security camera recording) at around a total of 80MB/s.

This also happens with BI turned off and just running crystal disk being the only thing running - it’ll start reading or writing and then just stop for 30-45 seconds like something is blocking the IO operations, and then it will continue at full speed (500+ MB/s) until it blocks again. You’ll see all IO stop like and block, even when there’s only 1 reader/writer.

 

[igor_kavinski]

Install Speedfan. Please post a screenshot like the one in this post: https://forums.anandtech.com/thread...readings-the-first-time.2598149/post-40636677

I would also recommend a surface scan. Last but not the least, which PSU are you using? How old is it? Maybe the voltage rail of the PSU is providing a little more voltage than it should. Check in BIOS if the voltages for 3.3V, 5V and 12V are within spec.

 

[Brett]

I used CrystalDiskInfo, which is similar to speedfan and shows me the SMART info for the drives. There are 4 drives, and drive 3 tests bad. That being said, both drive 3 and 4 exhibit the long pause issue. Drives 1 and 2 do not. Since the data on these drives is replaceable, in the short term I've JBOD'd drives 1/2 together, and I'll see about replacements for drive 3 and 4, although it does strike me as really odd that 2 drives are bad -- is it possible this is a SATA controller issue? Check out the response time for one of the bad dries -- 60 seconds!

Voltages:
12V: 12.258V
5V: 5.182V
3.3V: 3.408V

PSU is a Super Flower Legion GX Gold PRO 650W 80+ Gold

All components here have been in service since November.

 

[igor_kavinski]

Your HDDs are exhibiting a high seek error rate. Something is wrong with the head or the platter surface is compromised. It could also be that the drives are so close together that their collective vibration is causing the heads to have trouble reading from the platters. Or it could just be that all four drives are from a defective batch.

Please rule out the PSU before you take any further steps. I used to have a hard drive (10GB, many years ago) that would freeze the entire PC if I downloaded any file from the internet. In the BIOS, it showed that the 12V rail had something like 12.15V or something. Bought a new PSU and never had a problem with that hard drive again. The new PSU's 12V rail showed 12.05V in BIOS.

EDIT: Sorry, I just noticed that you provided the voltage readings. Those are too high, man. Seriously, try a different PSU. If the issue is voltage related, it's only a matter of time before the other two drives also start exhibiting this issue.

 

[Brett]

I thought the ATX spec allowed for +/- 5%, so

12V should be between 11.4V and 12.6V
5V should be between 4.75 and 5.25V
3.3V should be between 3.135 and 3.465V

I'll try a different PSU and see if the behavior improves, but it'd really suck if these drives can't tolerate a 2.2%, 3.6%, and 3.3% delta in voltage when the specs allow for 5%

 

[igor_kavinski]

I thought the ATX spec allowed for +/- 5%,

You are right. However, 12.15V caused problems for my old drive, which incidentally, was also a Seagate. Guess Seagate uses cheaper components or their QA/QC is not good enough.