Bought an SSD that ended up being QLC, should I keep it?

[sam_86314]

TLDR: To any current or former owners of a Crucial P3 Plus or Corsair MP600 Core XT, what do (or did) you think about them?

So, I recently bought a 4TB Fanxiang S660 to use as my game drive in my main system. I chose this drive because I had an excellent experience with the 2TB version.

Unfortunately, while the 2TB drive uses TLC NAND (YMTC 128-layer TLC), the 4TB uses QLC (Micron 176-layer QLC.

I've been led to believe that QLC is objectively terrible compared to TLC and can even be worse than hard drives in some cases.

I did a small test comparing the QLC drive to the RAID0 SATA SSD array that it was supposed to replace. I copied Baldur's Gate 3 from my boot drive (2TB WD SN850X) to the same folder on both drives.

The SATA array (4x 1TB Crucial MX500) started out at around 3.5GB/s and then settled in at around 1.8GB/s until I stopped the run at 50%. That's perfectly in line with what synthetic tests show on it.

...yeah, this pretty much speaks for itself. The QLC drive does 3.5GB/s for a bit before plummeting to around 200MB/s. That's absolutely terrible compared to synthetic results.

I also ran it on a WD SN570 (DRAMless TLC drive) I have in an external enclosure, and that drive settled around 450MB/s.

I know that sustained sequential performance isn't really real-world performance, but I do sometimes shuffle large files around on my system, and the SATA array handles that infinitely better.

The QLC drive performs way better in synthetic tests, but I don't know how relevant these are to real-world usage.

And, of course, there's my concern about the endurance of QLC. The specific drive I got claims 2400TBW, but given that it's virtually identical to the Crucial P3 Plus and Corsair MP600 Core XT (both advertise around 900TBW), I have my doubts. The MX500s may not have the best endurance either, but they're still better than this when combined.

 

[sam_86314]

I ended up returning the drive.

Yesterday, I went to copy Saints Row 2022 onto it. The transfer rate went up to 70MB/s (during a 7GB DLC file) and stayed around that. Completely unacceptable for an SSD.

Swapped the old SN570 into the same drive enclosure, and it went right up to 600MB/s with the same transfer (the SN570 was even more full than the S660, so it should've been at a disadvantage).

 

[lexluthermiester]

I ended up returning the drive.

Yesterday, I went to copy Saints Row 2022 onto it. The transfer rate went up to 70MB/s (during a 7GB DLC file) and stayed around that. Completely unacceptable for an SSD.

Swapped the old SN570 into the same drive enclosure, and it went right up to 600MB/s with the same transfer (the SN570 was even more full than the S660, so it should've been at a disadvantage).

Yeah, QLC is just pain for a lot of simple tasks.

 

[goms]

Swapped the old SN570 into the same drive enclosure, and it went right up to 600MB/s with the same transfer (the SN570 was even more full than the S660, so it should've been at a disadvantage).

The SN570's design (TLC and static SLC cache) favours consistency over peak performance/burst writes. A full SN570 is expected to perform better than a full S660 in my opinion.

 

[Assimilator]

The SN570's design (TLC and static SLC cache) favours consistency over peak performance/burst writes. A full SN570 is expected to perform better than a full S660 in my opinion.

This is because the Chinese drives optimise their firmware for synthetic benchmarks, so that they can appear better in marketing material. Whereas Western manufacturers have figured out that consumers actually read reviews and expect decent real-world performance.

 

[80251]

I still like my Samsung 4 TiB QVO better than the Toshiba 2 TiB spinner it replaced and the perf. is better across the board, especially in IOPs. My spinner had a ceiling of 200 MiB/s but it almost never hit this mark in actual use.

 

[lexluthermiester]

This is because the Chinese drives optimise their firmware for synthetic benchmarks, so that they can appear better in marketing material. Whereas Western manufacturers have figured out that consumers actually read reviews and expect decent real-world performance.

Pretty much this! The chinese bank on the ignorant.

I still like my Samsung 4 TiB QVO better than the Toshiba 2 TiB spinner it replaced and the perf. is better across the board, especially in IOPs. My spinner had a ceiling of 200 MiB/s but it almost never hit this mark in actual use.

The thing is, once someone gets over 4TB, HDDs are easily the better option. Storage space is more important than speed for anyone who needs the space.

 

[Mussels]

TLDR: To any current or former owners of a Crucial P3 Plus or Corsair MP600 Core XT, what do (or did) you think about them?

So, I recently bought a 4TB Fanxiang S660 to use as my game drive in my main system. I chose this drive because I had an excellent experience with the 2TB version.

Unfortunately, while the 2TB drive uses TLC NAND (YMTC 128-layer TLC), the 4TB uses QLC (Micron 176-layer QLC.

I've been led to believe that QLC is objectively terrible compared to TLC and can even be worse than hard drives in some cases.

I did a small test comparing the QLC drive to the RAID0 SATA SSD array that it was supposed to replace. I copied Baldur's Gate 3 from my boot drive (2TB WD SN850X) to the same folder on both drives.

View attachment 312605

The SATA array (4x 1TB Crucial MX500) started out at around 3.5GB/s and then settled in at around 1.8GB/s until I stopped the run at 50%. That's perfectly in line with what synthetic tests show on it.

View attachment 312606

...yeah, this pretty much speaks for itself. The QLC drive does 3.5GB/s for a bit before plummeting to around 200MB/s. That's absolutely terrible compared to synthetic results.

I also ran it on a WD SN570 (DRAMless TLC drive) I have in an external enclosure, and that drive settled around 450MB/s.

I know that sustained sequential performance isn't really real-world performance, but I do sometimes shuffle large files around on my system, and the SATA array handles that infinitely better.

The QLC drive performs way better in synthetic tests, but I don't know how relevant these are to real-world usage.

View attachment 312607

And, of course, there's my concern about the endurance of QLC. The specific drive I got claims 2400TBW, but given that it's virtually identical to the Crucial P3 Plus and Corsair MP600 Core XT (both advertise around 900TBW), I have my doubts. The MX500s may not have the best endurance either, but they're still better than this when combined.

QLC has slow writes, especially once full - but a lot of the modern ones have a very high TBW rating and are good for long term reliability.


The questions comes down to:
Does the write speeds negatively affect your use for the drive?
Is the TBW rating too low for your intended use for the drive?

I ended up returning the drive.

Yesterday, I went to copy Saints Row 2022 onto it. The transfer rate went up to 70MB/s (during a 7GB DLC file) and stayed around that. Completely unacceptable for an SSD.

Swapped the old SN570 into the same drive enclosure, and it went right up to 600MB/s with the same transfer (the SN570 was even more full than the S660, so it should've been at a disadvantage).

Shoulda checked page 2.

 

[chrcoluk]

TLDR: To any current or former owners of a Crucial P3 Plus or Corsair MP600 Core XT, what do (or did) you think about them?

So, I recently bought a 4TB Fanxiang S660 to use as my game drive in my main system. I chose this drive because I had an excellent experience with the 2TB version.

Unfortunately, while the 2TB drive uses TLC NAND (YMTC 128-layer TLC), the 4TB uses QLC (Micron 176-layer QLC.

I've been led to believe that QLC is objectively terrible compared to TLC and can even be worse than hard drives in some cases.

I did a small test comparing the QLC drive to the RAID0 SATA SSD array that it was supposed to replace. I copied Baldur's Gate 3 from my boot drive (2TB WD SN850X) to the same folder on both drives.

View attachment 312605

The SATA array (4x 1TB Crucial MX500) started out at around 3.5GB/s and then settled in at around 1.8GB/s until I stopped the run at 50%. That's perfectly in line with what synthetic tests show on it.

View attachment 312606

...yeah, this pretty much speaks for itself. The QLC drive does 3.5GB/s for a bit before plummeting to around 200MB/s. That's absolutely terrible compared to synthetic results.

I also ran it on a WD SN570 (DRAMless TLC drive) I have in an external enclosure, and that drive settled around 450MB/s.

I know that sustained sequential performance isn't really real-world performance, but I do sometimes shuffle large files around on my system, and the SATA array handles that infinitely better.

The QLC drive performs way better in synthetic tests, but I don't know how relevant these are to real-world usage.

View attachment 312607

And, of course, there's my concern about the endurance of QLC. The specific drive I got claims 2400TBW, but given that it's virtually identical to the Crucial P3 Plus and Corsair MP600 Core XT (both advertise around 900TBW), I have my doubts. The MX500s may not have the best endurance either, but they're still better than this when combined.

I suppose 200MB is progress when it used to be 50MB, but of course by todays standards its horrible.

They shouldnt be switching nand type just based on capacity on same model drive in my opinion.

 

[sam_86314]

I suppose 200MB is progress when it used to be 50MB, but of course by todays standards its horrible.

They shouldnt be switching nand type just based on capacity on same model drive in my opinion.

I personally think SSD makers should be required to list the NAND type and if they have a DRAM cache at the minimum.

 

[lexluthermiester]

I personally think SSD makers should be required to list the NAND type and if they have a DRAM cache at the minimum.

Agreed. It should be a law. Truth in advertising kind of thing.

 

[80251]

The Samsung 870 QVO 4 TiB is rated for more than twice the TBW of the old Samsung 850 Pro series.

 

[lexluthermiester]

The Samsung 870 QVO 4 TiB is rated for more than twice the TBW of the old Samsung 850 Pro series.

This is true and Samsung has improved their version of QLC greatly. While I still can not recommend it for a Boot/OS drive, they make for great additional/auxiliary/external storage drives.

 

[80251]

@Lex
That's what I'm using my QVO for after my Toshiba 2 TiB died w/basically no warning whatsoever, but strangely came back long enough for me to get most everything off it over several days of leaving it on.

My QVO is being cached via primocache to a SK Hynix P35 Gold NVME M2 drive (which was previously used to cache the dead Toshiba 2 TiB spinner).

 

[Mussels]

I personally think SSD makers should be required to list the NAND type and if they have a DRAM cache at the minimum.

I'm utterly furious at every company who does part swaps - corsair with their 'revisions' on RAM, should have a different product code. If they can print a revision label they can also label them S for samsung, etc. Same goes for mobile phones, where the "S22 ultra" was two entirely different phones with almost nothing in common depending on which model you received.


These companies think it's fine and treat it like the same mechanical component being made at two locations - like AMD CPU's being made in different factories

I can't find that CPU not from malaysia, but the concept is simple - even made elsewhere, it's got the same components and design. There are no intentional changes.


What they're doing is like when a company swaps out electrical components like asrock did with the x470 master/sli - List one set of components that are all high end, and swap them with "equivalent" cheap shit that makes it an entirely new product, but keep the same product name and ride on it's reputation.

It'd be like Apple or Samsung having review models with OLED displays and then swapping to TN LCD's for consumers because "same refresh rate and resolution, same specs right?"

 

[ViperXTR]

I personally think SSD makers should be required to list the NAND type and if they have a DRAM cache at the minimum.

reminds me of the Kingston NV2 ssd as well with its ever changing nand and controller

 

[Mussels]

@Lex
That's what I'm using my QVO for after my Toshiba 2 TiB died w/basically no warning whatsoever, but strangely came back long enough for me to get most everything off it over several days of leaving it on.

My QVO is being cached via primocache to a SK Hynix P35 Gold NVME M2 drive (which was previously used to cache the dead Toshiba 2 TiB spinner).

That sounds like a weak solder connection - if you still have the drive, you may be able to add some flux and a heatgun and fix that.