Question Super slow nvme drive

[zephyrprime]

So I have this nvme drive that is a 4 year old Inland drive (1tb). I noticed recently that the read speed on it had become super slow. Windows was showing that the disk was at 100% utilization while reading at only ~30MB/sec. That's worse than pata speeds. I cloned it onto another drive while using a linux program called Clonezilla and the slow speed was the same. The new drive is now seated in the same slot the old one was using but its speed is normal. Anyone ever encounter this problem before? Is this some sort of controller bug? Is there anyway to fix this or is it just effed?

 

[Shmee]

It could be the Windows install/software issues, especially if you tried cloning to a new drive. Also, how full is the drive? And what are the specs of your system?

 

[mikeymikec]

Could you dig out a model number for the drive? I wonder if you have a lemon rather like the Samsung vanilla 840 model drive.

 

[coercitiv]

So I have this nvme drive that is a 4 year old Inland drive (1tb). I noticed recently that the read speed on it had become super slow. Windows was showing that the disk was at 100% utilization while reading at only ~30MB/sec. That's worse than pata speeds. I cloned it onto another drive while using a linux program called Clonezilla and the slow speed was the same. The new drive is now seated in the same slot the old one was using but its speed is normal. Anyone ever encounter this problem before? Is this some sort of controller bug? Is there anyway to fix this or is it just effed?

One possible issue is the one @mikeymikec alluded to, the first generations of TLC NAND had data retention issues, in the sense that cell voltage varied too quickly over time. This meant that files older than 1-2 months became increasingly harder to read, with read speeds that could drop down to 20-30MB/s. Cheaper drives/brands may have continued to use low quality NAND with similar behavior even to this day. One way to diagnose this issue is to copy a new big file to the Inland drive, to make sure it's freshly written. Then copy this file to another drive (or even to itself) and observe the speed. If the new file transfers nicely (both when writing it and reading it) while older files are too slow to read... then you know where the problem lies.

A quick search online reveals a case of someone going through this exact process with an Inland Performance Plus 4TB NVME. It's the exact behavior I have described above, which is funny considering I decided to search after writing the post

The user even posted screenshots of file copy speeds:

Spoiler: New file

Spoiler: 3 month old file

 

[zephyrprime]

OMFG coercitiv! That was the problem! The older files on the drive are slower to read than newer files.

I couldn't test in my original system unfortunately because my cloning of the drives produced 2 drives with the same disk id and even after I changed one of them Windows still didn't like it.

I had to put it on a usb3 enclosure and test on my laptop.

The new files can get about 500mb/s which is probably the limit of my cheap usb enclosure.

Files that are about 1.5 years old can get about 370mb/s.

Files that are more than 2.5years old just drop off a cliff in terms of speed. They get uneven transfer speeds of 1-50MB/s.

What the heck, I've never even heard of this problem before. I had noticed my install of windows being slow at times but I had thought the problem was somewhere in windows, not a hardware problem.

Wow remind me to never buy another inland drives. I can't believe they're still having the same sort of problem with a drive 4 years newer than mine.

 

[DavidC1]

OMFG coercitiv! That was the problem! The older files on the drive are slower to read than newer files.

There's a storage law that says the tradeoff of faster storage is lower retention rates.

Tape/Optical, they both have excellent retention rates. Hard Drives are ok for consumer usage, and unlike SSDs they fail gracefully.

SSDs have a definite finite retention time, and Optane SSDs were quite a bit worse with Intel saying 3 months.

 

[mikeymikec]

@DavidC1 I think if you had more experience with HDD failure you wouldn't be inclined to say they fail gracefully. Just like most complex electronics, they can go bang, they can be fine the night before and never work again for no obvious reason, and they can get slower until they maybe eventually die.

They can also make horrific noises, sometimes opening them up and physically shifting the head back to the resting position helps, sometimes the drive head has scratched the hell out of the platters, sometimes freezing them will give a few more minutes of life to allow data recovery... likely due to their highly rotational nature there are many weird and wonderful ways for them to go, and no-one should make the mistake of thinking that HDDs are predictable.