I installed two new 4TB Extreme 3D SSD drives in my Supermicro editing workstation last month.
After install, the benchmark test was in the 500MB/sec range.
This week I started having stuttering playback of h.265 video in DaVinci Resolve Studio.
After some investigation, I ran a disc benchmark again and was shocked that my new SanDisk drive had slowed to near floppy disc speeds!
It was a long shot, but I tried forcing a TRIM operation in Windows defrag menu. Benchmarked the drive and it was up to 470MB/sec.
A day later, after shooting some video and copying it to the drive, I had speed problems AGAIN. So I ran TRIM again, and once again, got the speed back.
I have many older drives in this machine from 2011-2015 and none of them have this slowdown. All are benching close to 500MB/sec. Most are 95% full. These new drives (two of them) are nearly empty. The one I haven’t used yet still benchmarks as it did the day I Installed it last month. But the drive I’ve been using for video storage keeps slowing to a crawl.
I have never in my entire career of building editing workstations ever encountered an SSD that didn’t run at at least 465MB/sec R/W. All the new drives I purchased in the last year are having these problems. Bad RAM chips? Shoddy manufacturing? Poor quality control? What’s the cause?
I’m at the point where I’m about to shop for USED pre-2020 4TB SSDs just so I have something reliable. This is so ridiculous, considering I paid 8X more for these drives that drives I bought last summer.
Also…
I tried to install SanDisk Dashboard app, but after thirty minutes of launching it, it’s still at the logo screen with a spinning circle around it. How long does this disc utility take to launch?
If the registry key does not exist, creating a new DWORD value named DisableSelectiveSuspend and setting its value to 1 will force Windows to keep the ports fully powered at the system level. A full system reboot is required to initialize these updated configurations and restore your baseline transfer speeds.
I’ve disabled ALL sleep functions (this is a server motherboard anyway) as this workstation is used for realtime video editing tasks.
The standard troubleshooting hasn’t resolved the issue.
I’m beginning to think that nothing beats the old SSD drives with MLC chips. Seriously considering buying a bunch of used Samsung 860 EVO 4TB drives to match my older drives that have performed flawlessly since 2015.
Any pre-Covid-era SSD seems not to have issues like this. Suspect they are using really poor quality chips in these new drives.
I have exactly the same problem.
Got 3 4TB models and created a RAID5.
After a few Months the Performance completely tanked and I contacted Support.
Support blamed it on my Motherboard and power supply but after a lot of discussion I got 3 replacements.
I did not use them at first because of the previous experience but recently I connected one of them and started copying data there.
It went well until I hit about 75% capacity and now the Problem is back.
I am convinced this is just a limitation of whatever technology is used there as I have replicated it on 5 different disks now.
Last year (when I originally had the Problem) I did not find any info online about similar issues but now that I have found this Thread it seems the Problems start appearing for others as well.
I have not found any solution to this except formatting the disk and then not using more than 50% of the capacity (which is of course completely unacceptable for such expensive disks).
Thanks for the tip, I am not using them in a NAS, just my normal PC.
I am currently copying the data off the “broken” SSD and the maximum read speed is about 100MB/s. I would literally be better off by using my old 4TB HDDs, at least they get 120MB/s . This is obviously by design, as I have had the problem with 5 different !brand new! 4TB 3d drives and the fact that instead of addressing or acknowledging it, SanDisk recommends to “buy our more expensive drives” tells me everything I need to know.
So respectfully:
Recommend printing on the boxes “This SSD will not work correctly if you use it as intended and put a normal amount of data on it”
This way future buyers at least have a chance to avoid these problems.
Since I have two identical 3D Ultra drives (K: and L:), K: being in use and L: allocated for future use, I decided to see if the file writing operation affected this other drive the same way.
So I copied the entire contents of K: to L:. Took about 2 hours, but it completed with an average speed around 400MB/sec:
As can be seen here, the drive didn’t slow down nearly as much as the K: drive with an identical set of files on it.
Also interestingly, the K: drive slowed to a crawl again on WRITES after being copied FROM for the past 2 hours. I’m running TRIM on it again and it’s taking the longest I have ever seen TRIM take. About 4-1/2 minutes so far and 97% completed. Vs. 2.5 seconds for any other SSD in the system.
TRIM just completed, and I’m running SSD Benchmark on the K: drive again. Something different happened: the drive speed did not increase after TRIM as was the case before. The long read operation seems to have permanently slowed the writing on this drive. It is stuck at 1.43MB/sec now.
Not neccesarily, this is exactly what I see with my drives.
Your K: drive is “stuck” in some kind of activity.
If you check your Task Manager, you will likely see, that whenever you want to read/write from the K Drive (even just delete files), it has 100% activity and a very high response time and low speed. If you format the disk (which might also take forever) then the drive will return to more or less normal until it decides to “be 100% busy” again.
This is precisely what I have seen on 5 different units of the 3d SSD 4TB model.
I’m trying once more to replicate this and am currently copying data to 2 of these disks, one connected intenrally with SATA and one with USB3
Already during the copy I am encountering Problems.
My E: Drive has been the one “broken” yesterday and I have formatted it.
The Copy started with a max speed of ~400 MB/s then quickly changed to 120 MB/s for around 15 minutes.
Now I am copying to it with a max speed of 30 MB/s.
This is exactly what happened with the first batch of disks last year.
If I pause the copy for a few minutes (during the pause time the Task manager still shows that data is being written to the disk at 30 MB/s), the speed will go back to 120MB/s for a bit before going back down to 30.
Almost feels like the disk has 3 categories of “storage cells”.
Cells that can read/write at max (500MB/s)
Cells that can read/write at 120MB/s and
Cells that can read/write at 30 MB/s
and somehow the disk is shuffling data around and I end up with these low speeds.
Then once the disk is full enough again (or maybe also randomly) it goes into the “activity death spiral” again, where the write speed is like 1MB/s and the disk becomes completely unusable.
So the question is, how long will it work normally before I have to do this again and why do I have to do this every 3 weeks when I have drives that are 11 years old and needed no maintenance at all?
I’m going to wait and see. I’m not holding my breath that this is going to last though. But I really don’t want to deal with having to yank this system out, disconnect 20 cables and lift it onto a table to work on it. My back limits my ability to lift and I’m going for hernia surgery next week and won’t be able to lift or bend for several weeks, so I am praying this is a permanent fix for the problem, despite the odds.
