2TB MSI SPATIUM M461 M.2 2280 PCIe 4.0 NVMe QLC Solid State Drive

lopezej

Feb 3, 2025 10:50 PM

61 Comments

Joined Nov 2014

lopezejFeb 3, 2025 10:50 PM

61 Comments

Quote from dmbase

:

TBW is such a silly thing to get hung up on when choosing an SSD. At 450 TBW, you could write 150 GB per week for over 57 years before it would reach that warranty. How many people can say that they seriously write 150 GB per week on a regular basis? You are far more likely to have a controller fail than you will have a write failure.

I do, but 50+ years is good enough for me 😄

1

CTRL-ALT-DEALS

Feb 3, 2025 10:53 PM

7.6K Comments

Joined Sep 2014

CTRL-ALT-DEALSFeb 3, 2025 10:53 PM

7.6K Comments

Our community has rated this post as helpful. If you agree, why not thank CTRL-ALT-DEALS

Quote from RmmmRs

:

Why are you comparing QLC to TLC? Different technologies, different purposes...

Not different purposes.

Different performance tiers.

QLC < TLC < MLC

MLC (Multi-Level Cell) → 2 bits per cell (best durability, expensive)

TLC (Triple-Level Cell) → 3 bits per cell (balance of cost, speed, and endurance)

QLC (Quad-Level Cell) → 4 bits per cell (cheapest, but slower and wears out faster)

7

3

MWink

Feb 3, 2025 11:25 PM

2.4K Comments

Joined Oct 2011

MWinkFeb 3, 2025 11:25 PM

2.4K Comments

Quote from RyanL

:

LOL, nobody does that and that's not how they typically fail. A more realistic scenario is when the drive is kept at near capacity and windows continuously writes to the same sectors over and over. Then wear issues start to present themselves. I don't like the wear #s they throw out there for SSDs and are about as useless as synthetic benchmarks.

Wear Leveling helps mitigate that.

Quote from mfour

:

AFAIK that's not something reported by the drive, but rather an arbitrary calculation by each specific SMART app. So I don't consider it a very informative piece of info.

On most drives (especially NVMe ones) it explicitly is something reported by the drive. However, there is at least one popular program out there that does do its own arbitrary calculations, though I think it's mostly with SATA drives.

Quote from spiffytexan

:

I had a QLC NAND drive from Team Group (a moderately reputable brand) that was a scam. It was previously TLC that they swapped to QLC without telling anyone.

Write speed dropped to 5-10 MEGABYTES while I was setting up my PC, installing programs, and transferring files.

It was going to take multiple DAYS for all the data to transfer. Programs were stuck installing for HOURS.

QLC may be fine for grandma, but for any kind of a moderate power user, no. Its not worth the $20 savings to get a much much worse QLC drive. Pony up the few extra bucks for TLC or buy used.

What model? As far as cheap brands go, it's seemed like Team Group is a little better about not quietly changing from TLC to QLC. They usually release a slightly different model with QLC (such as the MP44Q vs the MP44/MP44L). However, they do like to swap between different brands/models of controllers and NAND. Then again, so does almost everyone else.

I do agree that QLC drives aren't priced attractively enough to make them worthwhile.

Quote from rosish

:

Get the M482 for a few bucks more or the same price with the non-retail boxed version. Read/write speeds are meh at 5000/3000 r/w speeds but still better than the cheaper 3000 w/r speeds alternatives.

The M482 is rated 7300/6400 MB/s, not that these numbers are meaningful to the average user.

Quote from trajanmcgill

:

| SSDs do wear leveling. So regardless of which logical sector you write, the actual writes are spread over the whole drive.

Not if the whole drive isn't available. If 90% of the drive is full, then all your writes can only be spread out within the remaining 10%.

Any decent implementation of wear leveling will try to even out the amount of writes by shifting old data around.

Quote from SavingsSamurai

:

Not different purposes.

Different performance tiers.

QLC < TLC < MLC

MLC (Multi-Level Cell) → 2 bits per cell (best durability, expensive)

TLC (Triple-Level Cell) → 3 bits per cell (balance of cost, speed, and endurance)

QLC (Quad-Level Cell) → 4 bits per cell (cheapest, but slower and wears out faster)

You forgot SLC, which is the best. Also, MLC isn't an option on consumer drives anymore. TLC is the best you can still get.

spiffytexan

Feb 3, 2025 11:28 PM

540 Comments

Joined Sep 2015

spiffytexanFeb 3, 2025 11:28 PM

540 Comments

Quote from MWink

:

Wear Leveling helps mitigate that.

On most drives (especially NVMe ones) it explicitly is something reported by the drive. However, there is at least one popular program out there that does do its own arbitrary calculations, though I think it's mostly with SATA drives.

What model? As far as cheap brands go, it's seemed like Team Group is a little better about not quietly changing from TLC to QLC. They usually release a slightly different model with QLC (such as the MP44Q vs the MP44/MP44L). However, they do like to swap between different brands/models of controllers and NAND. Then again, so does almost everyone else.

I do agree that QLC drives aren't priced attractively enough to make them worthwhile.

The M482 is rated 7300/6400 MB/s, not that these numbers are meaningful to the average user.

Any decent implementation of wear leveling will try to even out the amount of writes by shifting old data around.

You forgot SLC, which is the best. Also, MLC isn't an option on consumer drives anymore. TLC is the best you can still get.

Ohhh wait it was an Adata that did this. IMO they're about the same quality tier as Team.

"Kinda bad, but not garbage"

This was the one that sneakily put QLC in:

https://a.co/d/5hm3807

RmmmRs

Feb 3, 2025 11:35 PM

254 Comments

Joined Oct 2015

RmmmRsFeb 3, 2025 11:35 PM

254 Comments

Quote from SavingsSamurai

:

Not different purposes.

Different performance tiers.

QLC < TLC < MLC

MLC (Multi-Level Cell) → 2 bits per cell (best durability, expensive)

TLC (Triple-Level Cell) → 3 bits per cell (balance of cost, speed, and endurance)

QLC (Quad-Level Cell) → 4 bits per cell (cheapest, but slower and wears out faster)

Different performance tiers are for different purposes. For basic low usage QLC is just fine. For more speed/tbw TLC is needed

MWink

Feb 3, 2025 11:48 PM

2.4K Comments

Joined Oct 2011

MWinkFeb 3, 2025 11:48 PM

2.4K Comments

Quote from zippyzap

:

QLC can perform well or it can perform badly. It just needs to have a good controller and firmware. I think for the most part people with Samsung EVO SSDs with QLC are pretty happy with them. I'm sure you can easily find TLC drives that have horrible performance in comparison.

A good controller and firmware may be able to adequately hide the issues with QLC from the average user. If you write more than the pSLC cache can absorb, the speed absolutely craters. I was recently digging into this a bit. While this is an uncommon situation in the real world, the time it takes to completely fill a TLC and QLC drive tends to differ massively. For example, the 2TB TLC Team Group MP44L (rated up to 5000/4500MB/s) can be filled in approximately 20 minutes. The 2TB QLC Orico O7000 (rated up to 7000/6500MB/s) takes almost 200 minutes (3 hours 20 minutes) to fill. That's a difference of 10x! These drives are generally priced within $10 of each other. Yes, for most tasks the QLC drive may not be substantially worse, but for the edge cases the difference can be huge.

What Samsung EVO SSD are you referring to? I'm not aware of a single internal Samsung EVO SSD that uses QLC. The QVO line are the ones that use QLC, and I hear plenty of complaints about them. Also, keep in mind that they're one of the few QLC lines that does have DRAM.

MWink

Feb 3, 2025 11:57 PM

2.4K Comments

Joined Oct 2011

MWinkFeb 3, 2025 11:57 PM

2.4K Comments

Quote from spiffytexan

:

Ohhh wait it was an Adata that did this. IMO they're about the same quality tier as Team.

"Kinda bad, but not garbage"

This was the one that sneakily put QLC in:

https://a.co/d/5hm3807

That's not surprising. These days, most brands pull this. Ones that don't are becoming the minority. This is why I tell people to put absolutely zero faith in those SSD databases.

I'm not at all a Team Group fan but I have to give credit where it's due. Some of their SSDs aren't terrible, though their USB flash drives usually are.

AllenA8916

Feb 4, 2025 12:47 AM

193 Comments

Joined May 2013

AllenA8916Feb 4, 2025 12:47 AM

193 Comments

Quote from megablank

:

It's a warranty, if you beat on it, it will likely reach way past but they simply won't garrantee it is all.

Exactly, reminds me of Tommy Boy

"Hey, if you want me to take a dump in a box and mark it guaranteed, I will."

It's like saying a Kia is more reliable than a Toyota or Honda because the warranty is 3x longer.

1

snydertalon

Feb 4, 2025 1:36 AM

284 Comments

Joined Apr 2013

snydertalonFeb 4, 2025 1:36 AM

284 Comments

Quote from FormerAstronut

:

And this is the thing right here, 1TB/week is an extreme use case, like upper 1% of users do this. And if a user is actually in this "professional" (even if you're an amateur) level of usage, you need to buy equipment that suits your usage case, and not poo-poo and stir everyone up in a frenzy to boycott it.

Honestly this is a good price for what you get compared to what's available currently. It is a low end product, yes, but even those sketchy Orico brand drives aren't this cheap for 2TB.

Agreed. I value my data so I write a copy to my RAID HDDs and a copy to my SSD, process it on the SSD, and then delete from my SSD as I no longer need the speed. I still don't trust the tech enough to use it as an archival medium. But at the same time, I have an old 60GB OCZ 2.5 SSD as an OS primary drive on my file server. Still runs flawlessly and passes any status of health check I throw at it...13 years after purchase.

2

booboloo

Feb 4, 2025 4:25 AM

3.3K Comments

Joined Jun 2010

boobolooFeb 4, 2025 4:25 AM

3.3K Comments

Quote from snydertalon

:

Agreed. I value my data so I write a copy to my RAID HDDs and a copy to my SSD, process it on the SSD, and then delete from my SSD as I no longer need the speed. I still don't trust the tech enough to use it as an archival medium. But at the same time, I have an old 60GB OCZ 2.5 SSD as an OS primary drive on my file server. Still runs flawlessly and passes any status of health check I throw at it...13 years after purchase.

The ones to get were the x25-e intel extremes, 50nm giant process slc, tank grade last forever, 100,000 p&e cycles.

HappyDome8591

Feb 4, 2025 9:43 AM

708 Comments

Joined Oct 2022

HappyDome8591Feb 4, 2025 9:43 AM

708 Comments

I'm not sure why we are getting into so much details about SSD durability but here we are.

Quote from BrightBelieve888

:

Most SSD also overprovision ~10% space to further spread out the writes to mitigate near capacity issues.

This is one of the cheapest QLC SSDs by a somewhat well-known manufacturer. It is cheaper than most QLC SSDs at this capacity. Since the amount of over-provisioning is not part of the spec sheet, I expect over-provisioning to be cut to the bone to make that price happen. There must be a reason why its endurance rating is lower than most QLC SSDs at this capacity.

Quote from MWink

:

Wear Leveling helps mitigate that.

I would say moderate but not mitigate. It helps to a degree but cannot completely get rid of it. If you fill a drive up to 90% then hammer it with write cycles until it fails, I can bet money that if you do that with an empty drive it can take 4x the writes.

Quote from MWink

:

Any decent implementation of wear leveling will try to even out the amount of writes by shifting old data around.

These firmwares are closed sourced, so who knows if they are half decent. Don't forget just a few years ago it is common for some drives to pretend to do encryption because nobody can tell if the drives are doing (or not doing) encryption.

Quote from MWink

:

If you write more than the pSLC cache can absorb, the speed absolutely craters.

pSLC mode uses free memory cells. So if you fill the drive past a certain point, it would no longer have enough free cells to put into pSLC mode and the speed will crater a lot faster.

For people who do not have demanding needs, a QLC is totally fine. But they should probably keep the caveats in mind. I recommend keeping the drive under 80% full for both performance and durability reasons, which is honestly not a hard thing to do. Otherwise, spend slightly more money for a better drive.

Last edited by HappyDome8591 February 4, 2025 at 01:52 AM.

Mr Jumpy

Feb 4, 2025 1:09 PM

153 Comments

Joined Jul 2004

Mr JumpyFeb 4, 2025 1:09 PM

153 Comments

Last time I ordered this from MSI they cancelled my order. No reason given.

mfour

Feb 4, 2025 2:52 PM

232 Comments

Joined Jun 2014

mfourFeb 4, 2025 2:52 PM

232 Comments

Quote from HappyDome8591

:

Since the amount of over-provisioning is not part of the spec sheet, I expect over-provisioning to be cut to the bone to make that price happen.

AFAIK, flash chips are power-of-2. So 1TB is 2^40 natively. Storage devices are sold in power-of-10 sizes. If so, theoretically overprovisioning should be (2^40) / (10^12) = 10%.
Were there any reports of partially-faulty flash chips, with bad cells marked internally as bad sectors, being a thing done in SSDs or simpler flash devices?

Last edited by mfour February 4, 2025 at 06:54 AM.

coolfilmaker

Feb 4, 2025 10:25 PM

4 Comments

Joined Feb 2017

coolfilmakerFeb 4, 2025 10:25 PM

4 Comments

Quote from pur

:

It might have been a slightly different url that copied the site so it looked the same. This deal is from msi.com

It's who I tried to buy from last time. The US MSI Store is not legitimate. They don't have a phone number to call, don't have an email address that responds, the legitimate MSI company can't contact them. But feel free to waste your time and possibly money.

3

MWink

Feb 5, 2025 1:15 AM

2.4K Comments

Joined Oct 2011

MWinkFeb 5, 2025 1:15 AM

2.4K Comments

Quote from mfour

:

AFAIK, flash chips are power-of-2. So 1TB is 2^40 natively. Storage devices are sold in power-of-10 sizes. If so, theoretically overprovisioning should be (2^40) / (10^12) = 10%.
Were there any reports of partially-faulty flash chips, with bad cells marked internally as bad sectors, being a thing done in SSDs or simpler flash devices?

You're correct. This drive should have 2TiB of NAND, with 2TB user accessible. The overprovisioned area will be just south of 200GB. Note that some brands (like Team Group, Silicon Power, etc.) tend to go with 2048GB (not GiB), while others (like MSI) generally stick with 2000GB for a "2TB" drive. I prefer drives with more factory overprovisioning, especially when they use a static pSLC cache (which is confined to the overprovisioned area).

Yes, most NAND chips do have bad blocks. Ones present at manufacture will be mapped out at the factory, but more may develop as the drive is used. This is why you can see descriptions such as "bad blocks early/later." Some of the Flash ID utilities can report the number and even location of the bad blocks. On drives of this size, don't be surprised to see potentially dozens of GB worth of bad blocks. To calculate the capacity, you need to multiply the number of bad blocks by the number of pages per block (which varies, depending on the NAND type), and then the page size (often 16KiB).

Quote from HappyDome8591

:

This is one of the cheapest QLC SSDs by a somewhat well-known manufacturer. It is cheaper than most QLC SSDs at this capacity. Since the amount of over-provisioning is not part of the spec sheet, I expect over-provisioning to be cut to the bone to make that price happen. There must be a reason why its endurance rating is lower than most QLC SSDs at this capacity.

People put way too much stock in the TBW number. It's a number primarily for warranty purposes. Better brands are often more conservative. Cheaper brands commonly go with nice big numbers, likely counting on the fact that most will never hit those numbers, that NAND wear is not commonly the cause of SSD failure, and the fact that many people aren't going to bother jumping through all the hoops to RMA one of their cheap drives. Let's not forget that some brands have been known to require you to send the faulty drive to Taiwan/China, at your expense. More reputable brands are going to be held to a higher standard.

If TBW number alone was a direct indicator of quality, Crucial drives would be considered garbage, due to their often very low numbers. It's also important to remember that host writes are not directly correlated with NAND wear. Write amplification also has to be taken into account.

That said, I'm not disagreeing that this drive has a low TBW value, by MSI standards. It is a cheap drive and I wouldn't have high expectations for it. Most people won't come anywhere close to the TBW value during its useful lifespan anyway.

I've already addressed the fact that this drive likely has ~200GB worth of factory overprovisioning.

Quote :

I would say moderate but not mitigate. It helps to a degree but cannot completely get rid of it. If you fill a drive up to 90% then hammer it with write cycles until it fails, I can bet money that if you do that with an empty drive it can take 4x the writes.

Of course it's not going to completely mitigate it. I didn't claim otherwise. My point stands that most people aren't killing drives in this manner. I've seen a few older drives that were kept 90+% full the majority of their lives and didn't die.

Quote :

These firmwares are closed sourced, so who knows if they are half decent. Don't forget just a few years ago it is common for some drives to pretend to do encryption because nobody can tell if the drives are doing (or not doing) encryption.

Yes, there is a lot we don't know about how the firmware behaves. However, there are plenty of drives that do report stats such as min, max, and average block erases. From this, we can get at least some insight into their behavior.

I'm curious what you're referring to when you say that some drives pretended to do encryption? This drive may very well fall into that group, though I'm not sure if it's the same thing you're talking about. This drive features TCG Pyrite, which is essentially "fake" encryption. Before learning the details, I thought it was odd to name something after pyrite (fake/fools gold). Anyway, these days it's often recommended not to rely on the hardware encryption built into any drive. Several were found to have flawed implementations.

Quote :

pSLC mode uses free memory cells. So if you fill the drive past a certain point, it would no longer have enough free cells to put into pSLC mode and the speed will crater a lot faster.

This totally depends on the implementation of the pSLC cache. What you state can be true for an exclusively dynamic cache. A static cache, which is confined to the factory overprovisioned area, may not shrink at all. Of course, it will never be particularly large either. In my experience, a good number of drives with Phison controllers (which this drive claims to use) appear to implement a static pSLC cache. That observation includes the MSI M482. These days, many brands, especially the better ones (Samsung, Crucial, etc.), use a hybrid cache (static + dynamic).

Quote :

For people who do not have demanding needs, a QLC is totally fine. But they should probably keep the caveats in mind. I recommend keeping the drive under 80% full for both performance and durability reasons, which is honestly not a hard thing to do. Otherwise, spend slightly more money for a better drive.

I completely agree. I'd much rather spend the extra $17 for the Team Group MP44L, which uses TLC NAND. That drive can be completely filled in roughly a tenth the time (~20 minutes vs 200) of some common QLC drives.

2TB MSI SPATIUM M461 M.2 2280 PCIe 4.0 NVMe QLC Solid State Drive

$85

$100

Editor's Notes

Original Post

Editor's Notes

Original Post

Community Voting

Top Comments

Join The Conversation

72 Comments

Join The Conversation

Related Deals

Related Searches

Popular Deals

Trending Deals