Question SK hynix Platinum P41 2TB SSD Write speed dropped by half

[Super Spartan]

Only about 4TB written to the SSD so far, when I first got it, the performance was:



Now, the write has dropped by about half:



the only difference which I think was different is I was on Windows 11 21H1 when I benchmarked it the first time and now I am on the latest 22H2. The read speeds are the same. What gives?

PS: I even tried using the Samsung NVMe modded drive but that didn't change the write speeds.
Antivirus is disabled while taking the test

Laptop specs are in my signature

 

[Buck Dancer]

A quick update on my refund. I was very hesitant to send my drive back to them. The communication was a little spotty, they didn't confirm how much my refund would be (I had something else in the same Amazon order). When I asked them how they would process the refund, they just said "we will proceed the refund process through Amazon system". That didn't fill me with confidence but the drive got to them on Friday and Tuesday night they emailed me saying they issued the refund. I see it pending on my credit card for the purchase amount + tax. I had the option to get the SN850X for $5 cheaper when I bought the P41, now I paid $60 more.

 

[Shmee]

I just ran crystal disk bench again, it seems the writes are getting pretty low.

This should still be under the 5 year warranty, think I should contact them and see what they can do? Either a firmware fix or refund?

 

[igor_kavinski]

Either a firmware fix or refund?

Refund I guess. They've been unable to fix it and don't think they can or don't want to.

 

[spike_V_sky]

I honestly never thought that a company like hynix would have such bad support for a premium nvme.

 

[igor_kavinski]

Their key engineers must've felt underappreciated and left for greener pastures or just poached by the competition.

 

[BuddyWasTaken]

Is it potentially this?

Windows 11 August 2023 Update: How to Fix SSD Performance Bugs

Since March 2023, Windows 11 users have faced extensive issues with NVMe SSD performance that seemed to begin with the original Moment 2 update and many since then. As we've covered in prior articles, these bugs have been long-lasting through each update and causing drives to experience much...

store.patriotmemory.com

Can someone run the benchmark with the KB5029263 update installed?

 

[Shmee]

Is it potentially this?

Windows 11 August 2023 Update: How to Fix SSD Performance Bugs

Since March 2023, Windows 11 users have faced extensive issues with NVMe SSD performance that seemed to begin with the original Moment 2 update and many since then. As we've covered in prior articles, these bugs have been long-lasting through each update and causing drives to experience much...

store.patriotmemory.com

Can someone run the benchmark with the KB5029263 update installed?

I kinda doubt that was the issue for me, as I have been and I still am running Windows 10. Though I suppose it is possible the bug is in Windows 10 as well, and that it won't be fixed for Windows 10, but only on Windows 11. That said, I contacted SK Hynix and they had me return the SSD for a refund. I got notified today that the refund was initiated, and it is processing now.

 

[BuddyWasTaken]

I kinda doubt that was the issue for me, as I have been and I still am running Windows 10. Though I suppose it is possible the bug is in Windows 10 as well, and that it won't be fixed for Windows 10, but only on Windows 11. That said, I contacted SK Hynix and they had me return the SSD for a refund. I got notified today that the refund was initiated, and it is processing now.

How long did it take for you to have the degradation? I'm also on Windows 10 and just checked my drive today after using it for a bit over a year and it still looks good, so I'm a bit paranoid right now.

Do Windows 11 updates usually not come to Windows 10?

 

[Shmee]

I don't remember exactly how long until I noticed reduced writes in benchmarks, but I seem to recall it was after a couple months. As for Windows, my understanding is that they are separate update trains, and I do know that Windows 10 updates are less often now, and will stop with EOL fairly soon. So some issues with Windows 10 / features added to Windows 11 may never be fixed on Windows 10, at least not officially.

 

[igor_kavinski]

will stop with EOL fairly soon. So some issues with Windows 10 / features added to Windows 11 may never be fixed on Windows 10, at least not officially.

May still get security updates for quite a while. I'm on Windows 8.1 on my Thinkpad and my work PC is on Windows 7. No loss of functionality so far. I would consider an OS dead after it can no longer connect to the internet due to outdated security protocols.

TBH, Windows 10 in its current state seems a lot less buggy to me. Just recently used an Inspiron with Core i5-1235U with Windows 11. It wouldn't boost above 2.8 GHz. Replaced with Windows 10 fresh install and it went up to 4 GHz. I guess Dell and/or Intel wants enhanced battery life at the expense of performance for its users.

 

[BuddyWasTaken]

I don't remember exactly how long until I noticed reduced writes in benchmarks, but I seem to recall it was after a couple months. As for Windows, my understanding is that they are separate update trains, and I do know that Windows 10 updates are less often now, and will stop with EOL fairly soon. So some issues with Windows 10 / features added to Windows 11 may never be fixed on Windows 10, at least not officially.

Thanks for letting me know.

Hopefully this issue becomes more public so people can figure out what the issue is. I'm hoping it's just a Windows bug and not an ssd firmware problem.

 

[THUMP1er]

Is this also affecting the 1TB P41? Because I have the 1TB P41 I don't have proof now because I'm at work, but mine does not have a reduced write speeds currently after 1 year of use. Its only used as an OS drive, and is like 6% full. I will post back with a quick bench.

 

[Justinus]

Is this also affecting the 1TB P41? Because I have the 1TB P41 I don't have proof now because I'm at work, but mine does not have a reduced write speeds currently after 1 year of use. Its only used as an OS drive, and is like 6% full. I will post back with a quick bench.

It likely affects all the drives. I have seen 1 and 2tb models affected

 

[BuddyWasTaken]

https://www.reddit.com/r/buildapc/s/LiPFkLRzH9

Not sure if this can be the reason? Not too familiar with how ssds are supposed to behave so trying to rule out everything.

 

[THUMP1er]

This is what I came up with. Might have something to do with how much I've filled it up. Or not filled up in this case.

 

[igor_kavinski]

View attachment 95572

This is what I came up with. Might have something to do with how much I've filled it up. Or not filled up in this case.

Seems fine. May obviously change the more you fill it up.

 

[BuddyWasTaken]

Here's mine. In use for a bit over a year. 8TB written essentially and no issues.

 

[Minjin]

My initial test after setting up this new computer (5 months ago). Glad I am a data nerd...




And just now that I read about the issue on Reddit. I did trim and values didn't change much.

 

[BuddyWasTaken]

My initial test after setting up this new computer (5 months ago). Glad I am a data nerd...

View attachment 95967


And just now that I read about the issue on Reddit. I did trim and values didn't change much.

View attachment 95966

Read through this:

https://www.reddit.com/r/buildapcsales/comments/1boa3u9/ssd_sk_hynix_platinum_p41_2tb_pcie_nvme_gen4_m2

I think this is regular behavior for TLC drives. As long as your reads and health is good, don't worry too much about it.

 

[Shmee]

Read through this:

https://www.reddit.com/r/buildapcsales/comments/1boa3u9/ssd_sk_hynix_platinum_p41_2tb_pcie_nvme_gen4_m2

I think this is regular behavior for TLC drives. As long as your reads and health is good, don't worry too much about it.

I am not so sure. Keep in mind, most new SSDs in production these days are TLC, or if they are cheap, QLC which is worse. I don't think a lot of other NVMe drives are doing this, but I could be wrong, maybe we just haven't heard. I have heard, however, that this might be an issue with the pSLC caching used. I think it was mentioned earlier here?

 

[BuddyWasTaken]

I am not so sure. Keep in mind, most new SSDs in production these days are TLC, or if they are cheap, QLC which is worse. I don't think a lot of other NVMe drives are doing this, but I could be wrong, maybe we just haven't heard. I have heard, however, that this might be an issue with the pSLC caching used. I think it was mentioned earlier here?

Check this

https://www.reddit.com/r/pcmasterrace/comments/1140ycn/ssd_sequential_write_slowdowns

It's one of the comments I posted on that thread. It's a bit of a pain but go through the comments and click on the links if you can.

 

[Shmee]

Check this

https://www.reddit.com/r/pcmasterrace/comments/1140ycn/ssd_sequential_write_slowdowns

It's one of the comments I posted on that thread. It's a bit of a pain but go through the comments and click on the links if you can.

Interesting, I wasn't able to go through the thread, but I did read the testing explanation for the different SSDs shown, and it looks like the SN850X is relatively unaffected, which is good news for me.

 

[Justinus]

Interesting, I wasn't able to go through the thread, but I did read the testing explanation for the different SSDs shown, and it looks like the SN850X is relatively unaffected, which is good news for me.

Same here, my drives are SN850X's.

This issue is very interesting given today's climate for SSD's. There have been numerous drives with issues clearing pSLC cache, and many of them have been fixed with firmware updates. It's surprising to me to see this seems to be affecting most modern drives (at least the ones tested here) to some extent.

To be clear, how, when, and how quickly a drive empties it's pSLC cache is a black box and may be different for every drive. For some period following a lot of writing, further write performance will be degraded until the pSLC cache is completely emptied. The concerning part here is " After 2 hours of idle time and a system restart the degradation remained." Granted, they don't say they performed a trim cycle, and 2 hours may not be enough for a full pSLC cache to migrate all the data to TLC. I think longer term testing would be necessary to observe if this degredation is permanent or simply takes more time to resolve than was allowed in their test methodology.

What is definitely not expected, correct behavior is the pSLC cache never completely emptying, resulting in permanently degraded performance that can only be fixed with a secure erase (or not at all). This is what we are observing with the P41 Platinum.

 

[far77]

I ran into a similar problem with my work laptop, it's a Dell Precision 5470. The NVMe inside is an SK Hynix PC801, which I think is the OEM version of the P41. Initially, the read/write speeds were amazing, hovering around 6000-7000. But a while back, I noticed the write speed had dropped to about 2000-2500, though the read speed was still above 6500. The funny thing is, I found this out purely by accident. I wanted to run CrystalDiskMark on an external drive but ended up running it on the NVMe by mistake! Honestly, I probably wouldn't have noticed otherwise. There were no error messages or warnings, and everything else about the daily work on the laptop seemed normal.

I started looking around for any information on what could be wrong and found this forum, among other places, but didn't find any answers. And since it’s a work computer, RMA'ing it was off the table —not to mention how silly it would sound asking our IT department to look into it just because the NVMe drive doesn't perform well in benchmark!

But then things took a turn yesterday. I was using BitLocker on an external drive and saw a message saying, “Device encryption is temporarily suspended. Encryption will resume automatically the next time you restart this device.” Trying to fix this issue (of course restart did nothing), I ended up turning off the device encryption altogether. Then it hit me—could encryption be messing with the write speed? So, I ran CrystalDiskMark again and, bingo! the write speed was back to normal, over 6000.

Now, I’m left with a new puzzle: I can’t re-enable device encryption. Still, I wanted to share this in case it helps anyone out there.

 

[Super Spartan]

I ran into a similar problem with my work laptop, it's a Dell Precision 5470. The NVMe inside is an SK Hynix PC801, which I think is the OEM version of the P41. Initially, the read/write speeds were amazing, hovering around 6000-7000. But a while back, I noticed the write speed had dropped to about 2000-2500, though the read speed was still above 6500. The funny thing is, I found this out purely by accident. I wanted to run CrystalDiskMark on an external drive but ended up running it on the NVMe by mistake! Honestly, I probably wouldn't have noticed otherwise. There were no error messages or warnings, and everything else about the daily work on the laptop seemed normal.

I started looking around for any information on what could be wrong and found this forum, among other places, but didn't find any answers. And since it’s a work computer, RMA'ing it was off the table —not to mention how silly it would sound asking our IT department to look into it just because the NVMe drive doesn't perform well in benchmark!

But then things took a turn yesterday. I was using BitLocker on an external drive and saw a message saying, “Device encryption is temporarily suspended. Encryption will resume automatically the next time you restart this device.” Trying to fix this issue (of course restart did nothing), I ended up turning off the device encryption altogether. Then it hit me—could encryption be messing with the write speed? So, I ran CrystalDiskMark again and, bingo! the write speed was back to normal, over 6000.

Now, I’m left with a new puzzle: I can’t re-enable device encryption. Still, I wanted to share this in case it helps anyone out there.

Encrypting your disk has a big impact on performance by up to 45% ( see: https://www.ghacks.net/2023/10/20/report-bitlocker-slowing-down-ssds-on-windows-11-by-up-to-45/ )