Samsung 870 QVO 1 TB

[bug]

Well we could all dream, but then we'd need a 1Tb network to cope with the speed of the drives I'm not sure we are quite there yet....

I'll stick with HD's for the moment, they seem a little more down to earth and possible!

I can live without the NVMe sequential speeds, I was only after the small box

 

[phill]

I'd try one of these HP mini units, they are pretty decent and fit in 4 drives I think, should be enough for 10 minutes??

 

[milewski1015]

The original advice for buying a SSD still holds true: get the biggest drive that fits in your budget.
Unless your usage pattern really benefits from NVMe's higher sequential speeds, do not sacrifice capacity to go NVMe. I only got my NVMe drive because it was ~$10 more than the SATA equivalent at the time (sale or whatnot, I don't remember exactly). Still, I went 970 EVO, a 970 EVO Plus was much more expensive.

Exactly this. Everybody's hopping on the NVMe train when I'd guess 95% of people don't make use of those high sequential speeds. Pair those unused NVMe speeds with low-endurance QLC NAND (Intel 660p, Crucial P1, etc.) and everybody goes crazy for them! I don't get it tbh. The 1TB P1 is currently selling for 105 USD. Personally I'd much rather have a 1 TB M.2 MX500 for 10 USD more. Yeah I don't get the higher sequential speeds NVMe offers, but the endurance on TLC is better and if I ever do anything requiring high sequential speeds, I'd rather it last instead of tanking to lower than spinning rust speeds.

 

[Th3pwn3r]

I just ordered an ADATA XPG SX8200 1tb nvme drive. It's for my HTPC where it's cramped already and I have my Sata SSD crammed in there but I'm going from mitx to an Atx board so won't have as much space to work with. I do also game with the machine so I want decently fast load times, I don't store movies anymore either so I don't need lots of space.

 

[danbert2000]

If I had to take a guess as to why Samsung is so clueless about the QVO and SSD pricing in general, I would say that it is due to their previous lead in SATA SSDs that allowed them to take a premium from gamers and enthusiasts for years. Their 850 Evo was all around the best choice at the time, especially with other drives being not much cheaper but definitely slower. Since then, every other SSD maker has caught up on how to make a decent SATA SSD, so Samsung is stuck fighting for the NVMe performance crown in order to get people to buy SSDs with way too much profit margin.

But they still need to sell SATA SSDs, as there is a market and they don't want to be missing from it. So instead of doing the sensible thing and selling their EVO SATA drives for a competitive price (which they would still profit from greatly due to using all of their own components), they keep that former cash cow at a 20% premium to the market, and come out with this QVO turd. Now the executives can say they're competing at all levels! They have two "premium" options with their 860 Pro and Evo, and then they have this new albatross to shore up the price comparisons.

I think they want people to run out of space on their 840 and 850s, look at Samsung for 1TB+ because "they're the best," see that the QVO is priced in line with all the other drives, and then buy that without even looking at reviews, or say "hey it's only $30" and buy the EVO for more than it's worth. Of course, they are losing good will from people that actually care about performance, and also probably the people who buy this QLC piece of crap and then run into slow performance and associate that with their Samsung purchase.

Two years ago, five years ago, I would have said stick to Samsung, or maybe Crucial. You don't want your drive to fail and these are the fastest. That's the world Samsung is pretending we're still in. They likely won't stop because as long as people still buy these drives, they'd probably rather do more profit margin than volume in an attempt to preserve their premium status.

 

[Makaveli]

Don't get me wrong, sequential numbers are nice if you're reading/writing big files in batches but a majority of users are going to be reading a whole lot of small files that are placed all over the place in terms of the NAND flash memory. My Samsung 970 EVO has a RND4K Q1T1 of somewhere around 57 MB/s. That right there folks is why SSDs result a system feeling like you just strapped a JATO rocket to the back of it and let it rip.

That is a great point.

Here is the numbers from my PCIe 4.0 drive



vs my Samsung 860Evo's in Raid 0



Huge difference in RND4K Q1T1 between them.

In my opinion the only reason to get an NVME ssd is the size , less cables and slightly improved air flow in some cases. The price difference makes them not worth it for any other reason.

In alot of newer cases SSD aren't really in the way compared to older cases. In my build both my SSD's are mounted behind the motherboard tray and not even visible.

 

[trparky]

In alot of newer cases SSD aren't really in the way compared to older cases.

Yeah. My one SATA SSD is mounted flat against the side of the case. The power cable comes from behind the metal that’s covered by the side panel meanwhile I have a very short SATA data cable that’s just long enough to go from the motherboard to the SSD (it does have some slack).

 

[Totally]

Are we looking at the same review? QD1 has always been there? Left-most point on the charts.


The drive comes in at 13 cents per GB, which is a low price price per GB if you consider nothing but "price per GB". But if you take into account performance offered etc, the whole package, then it's a terrible deal and virtually every SSD on the market is either much faster, or substantially cheaper per GB


No idea, don't have any data for 860 QVO. Just buy the MX500 or Seagate 120 for the same price, you even get +2 years warranty

ok gotcha, the more accurate phrasing is would be 'poor value.'

 

[bug]

That is a great point.

Here is the numbers from my PCIe 4.0 drive



vs my Samsung 860Evo's in Raid 0



Huge difference in RND4K Q1T1 between them.



In alot of newer cases SSD aren't really in the way compared to older cases. In my build both my SSD's are mounted behind the motherboard tray and not even visible.

This brings up another issue users seem to be oblivious about: RAID0 does squat for random access. It only improves sequential speeds which we already agreed aren't that important. It still doubles the chance of failure (not a biggie, if you keep an eye out for SMART reports).

 

[Makaveli]

This brings up another issue users seem to be oblivious about: RAID0 does squat for random access. It only improves sequential speeds which we already agreed aren't that important. It still doubles the chance of failure (not a biggie, if you keep an eye out for SMART reports).

I still have a Raid 0 array of two Intel 160GB G2 drives that is running 10 years later. So for me I'm not to concerned about the failure rate of SSD and even more so with those older drives since they have MLC memory. My Primary purpose for my current array is just game storage and I wanted to do that to a single drive not across multiple drives. My OS and everything else stays on the M2 drive. Those sequential speeds are nice when I'm transferring to the Array but that wasn't the primary purpose for me.

 

[ExplodingCaps]

I knew it, I knew QLC is a bad idea in the first place. But this, this takes disappointment to another new level.

 

[CheapMeat]

Nothing wrong with QLC except price. People here are just focused on having 1 or 2 drives in their system. I'm looking for entire HDD replacements on all my PCs for bulk data. I'm talking 20+ HDDs. QLC is totally fine for that tier. And I want SATA for the replacement, as all the cabling is there, my motherboard and cards are SATA focused. For OS and main programs I use NVMe OPTANE; yes people commented about random read & write QD1...well that's Optane; you get the lower latency from the NVMe protocol AND the inherent lower latency of Optane AND the fact it doesn't slow down the fuller it gets AND it has higher endurance than SLC. But barely anyone mentioned it. For my "hot" tier, I'll be uosing NVMe TLC SSDs with SLC cache. I'm 100% for QLC/PLC (5 states) for bulk data. The QLC SSDs over long sequestial operations as shown here slow down insanely, even below some more contemporary HDDs. But I feel like I can trust it more when doing something for hours at a time. This is why there's always been worry about reslivering (rebuilding) mirrors and paritiy and all that jazz because of the high chance of mechanical failure and then boom, it's gone. The QLC SSD will slow down like crazy but I think it would work even at 10 MB/s (insanely slow) for 24hrs straight, 3 days straight, whatever, without issue. And longevity, yeah it's QLC but as some tests have shown, most drives because of really good error correction and over provisioning can last way beyond their usual rated MTF; it'll likely out last any HDD at initial 1TB+ capacity.

 

[Vayra86]

Said before, getting it confirmed...QLC - not in my rig.

Happily paying a small premium for TLC and consistency. Even on a data drive, I'm not putting up with this shit.

 

[W1zzard]

Said before, getting it confirmed...QLC - not in my rig.

What if 1 TB QLC was $60 or $70 ? So basically 2 TB QLC SATA for the price of 1 TB TLC NVMe

 

[bug]

What if 1 TB QLC was $60 or $70 ? So basically 2 TB QLC SATA for the price of 1 TB TLC NVMe

Tempting, but leaves the question of durability. TLC is down to ~1,000 p/e cycles, QLC must be well below that. It's probably only suited for cold(ish) storage.

But it will never sell for so cheap. SLC->MLC was a doubling of capacity. MLC->TLC was another 50% increase. MLC->QLC is only a 33% increase. You just can't sell twice the capacity of TLC for the same price.

Edit: It seems my numbers were a bit off (manufacturing improved since I read about them last): https://www.computerweekly.com/feature/What-is-QLC-flash-and-what-workloads-it-is-good-for
But the idea is the same: QLC does not improve on TLC enough to warrant the downsides.

 

[Vayra86]

What if 1 TB QLC was $60 or $70 ? So basically 2 TB QLC SATA for the price of 1 TB TLC NVMe

Then we're talking because then we have a silent alternative for mechanical storage. That's a win. This right here, for sure is not.

So yes, 'if the price is right', certainly. But I have yet to spot a single QLC drive that does that. But regardless, if the drive is spotty in performance - inconsistent - and if filling it up makes things even worse... is this really something to pursue in the first place? I'm still on the fence. Its the same way I view high boost - low base clock Intel CPUs. Its opening the way for commercial trickery and I'm not a fan. Lots of new, supposed high performance parts are moving towards that way of working. I'm boycotting every single last one of them, because its not what I want to see in products and every time it hurts longevity, too.

It has yet to be decided whether a QLC drive is really a cheaper drive or whether the lower endurance will eliminate that advantage too. In a similar way, we have yet to see how these super bursty CPUs are holding up - and that isn't even Intel exclusive either, AMD is similarly pushing the envelope, albeit in a different way, but the vcore stories are known...

 

[Megatron]

I don't think this is a replacement for a hard drive, but I didn't see a test focussed on read performance. I see this as an SMR like SSD for archival purposes clearly has nothing to offer as an OS drive.

 

[bug]

I don't think this is a replacement for a hard drive, but I didn't see a test focussed on read performance. I see this as an SMR like SSD for archival purposes clearly has nothing to offer as an OS drive.

SSD is not a good choice for archival, it loses state when unpowered for long periods of time.

 

[trparky]

What if 1 TB QLC was $60 or $70 ? So basically 2 TB QLC SATA for the price of 1 TB TLC NVMe

I'd have to really see some numbers before I make that decision.

Tempting, but leaves the question of durability.

I'm with bug here. I lived through the days of the Samsung 840 and 840 EVO. Performance was shit when reading back data that was older than 90 to 120 days old. If that's going to be the case with QLC, the manufacturers can take QLC and shove it up their backsides. TLC would be the limit for me.

 

[Assimilator]

Sigh, once again people are getting on the "QLC is TEH EVIL!!!!!oneone" bandwagon again. OK geniuses, if QLC is so bad, how come Intel and Crucial can make QLC drives that perform decently?

The answer is simply NAND parallelism: Samsung only uses a single 1TB NAND die on this model. AnandTech did a review on the 1TB and 4TB models a while back https://www.anandtech.com/show/15887/the-samsung-870-qvo-1tb-4tb-ssd-review-qlc-refreshed and it quite clearly shows that the 4TB model of this drive is, while not exactly great, far better than this 1TB model.

The Samsung 870 QVO 1TB is a bad product, but bad products do not necessarily imply flawed technology. QLC is here to stay and will continue to improve regardless of how much Samsung f**ks up - don't be a luddite and don't spread FUD about the technology just because of them.

If I had to take a guess as to why Samsung is so clueless about the QVO and SSD pricing in general, I would say that it is due to their previous lead in SATA SSDs that allowed them to take a premium from gamers and enthusiasts for years. Their 850 Evo was all around the best choice at the time, especially with other drives being not much cheaper but definitely slower. Since then, every other SSD maker has caught up on how to make a decent SATA SSD, so Samsung is stuck fighting for the NVMe performance crown in order to get people to buy SSDs with way too much profit margin.

But they still need to sell SATA SSDs, as there is a market and they don't want to be missing from it. So instead of doing the sensible thing and selling their EVO SATA drives for a competitive price (which they would still profit from greatly due to using all of their own components), they keep that former cash cow at a 20% premium to the market, and come out with this QVO turd. Now the executives can say they're competing at all levels! They have two "premium" options with their 860 Pro and Evo, and then they have this new albatross to shore up the price comparisons.

I think they want people to run out of space on their 840 and 850s, look at Samsung for 1TB+ because "they're the best," see that the QVO is priced in line with all the other drives, and then buy that without even looking at reviews, or say "hey it's only $30" and buy the EVO for more than it's worth. Of course, they are losing good will from people that actually care about performance, and also probably the people who buy this QLC piece of crap and then run into slow performance and associate that with their Samsung purchase.

Two years ago, five years ago, I would have said stick to Samsung, or maybe Crucial. You don't want your drive to fail and these are the fastest. That's the world Samsung is pretending we're still in. They likely won't stop because as long as people still buy these drives, they'd probably rather do more profit margin than volume in an attempt to preserve their premium status.

Samsung has become the Apple of the SSD world. Their products aren't actually better than their competitors' (in fact they're often worse, as this review demonstrates) but they have a bunch of brand-loyal sheep who will buy their products regardless of how bad they are.

I hate people who are loyal to faceless corporates who only care about their money. How much effort is it to read a couple of review before buying, FFS?

 

[bug]

Sigh, once again people are getting on the "QLC is TEH EVIL!!!!!oneone" bandwagon again. OK geniuses, if QLC is so bad, how come Intel and Crucial can make QLC drives that perform decently?

A lot of things perform decently in the lab. And will never reach mainstream.

Adding another bit to the storage tech comes with these by default:
+ increased capacity
- lowered voltage difference between various levels; this translates into more difficult reads and less charge that can leak before reads can't discern levels properly
- more problematic writes (because of how SSDs write)

As more levels are added, the amount of capacity gained decreases and the negatives get harder to control. So without whining or anything, adding one more level automatically means you're getting a worse drive than before. In the absence of proof that the downsides have been mitigated somehow (like TLC did when it went vertical), we have every reason to at least not cut the new "hotness" any slack.

Case in point (yours): for $500, that 4TB 870QVO is still at 50-75% of its 860EVO predecessor's performance.

 

[Vayra86]

Sigh, once again people are getting on the "QLC is TEH EVIL!!!!!oneone" bandwagon again. OK geniuses, if QLC is so bad, how come Intel and Crucial can make QLC drives that perform decently?

The answer is simply NAND parallelism: Samsung only uses a single 1TB NAND die on this model. AnandTech did a review on the 1TB and 4TB models a while back https://www.anandtech.com/show/15887/the-samsung-870-qvo-1tb-4tb-ssd-review-qlc-refreshed and it quite clearly shows that the 4TB model of this drive is, while not exactly great, far better than this 1TB model.

The Samsung 870 QVO 1TB is a bad product, but bad products do not necessarily imply flawed technology. QLC is here to stay and will continue to improve regardless of how much Samsung f**ks up - don't be a luddite and don't spread FUD about the technology just because of them.



Samsung has become the Apple of the SSD world. Their products aren't actually better than their competitors' (in fact they're often worse, as this review demonstrates) but they have a bunch of brand-loyal sheep who will buy their products regardless of how bad they are.

I hate people who are loyal to faceless corporates who only care about their money. How much effort is it to read a couple of review before buying, FFS?

QLC is not 'teh evil' but the way it is being marketed certainly is. Companies hide in the spec sheet and technical minutia to serve you notably worse drives at nearly the same price.

So yes, in that case, QLC is 'teh evil'. It truly is inferior and priced too high. I don't know what else to make of it. The performance inconsistency speaks volumes that this tech is a bridge too far - and if its not, how about actually making it work properly to begin with? Even on drives that do not lose lots of performance along the way you do suffer major degradation and endurance loss. The lifetime is quite simply a lot shorter on a QLC drive. Why would we cheer for that? To save a few bucks? NOTY.

And its not a first for SSD technology and especially controllers either. Whole companies have sunk for releasing bad stuff on storage. Specifically for SSD there are numerous examples of sub par controllers. You'd think by the time we reach QLC they'd know better.

IMO, QLC belongs in ultra cheap drives you buy in broken plastic packaging at the dollar store. Its good for casual use if you don't have anything important but still lots of data, let's say, some 128kbps MP3 collection or some dank porn collection. Anything half serious doesn't belong on these drives, if you ask me. This is like what TN is for monitors, and yes they will put infinite amounts of lipstick on that pig too - its 2020 and they're still making TN more fantasticer, go figure.

Storage also has a major trust issue going on. Reliability but also endurance are massively more important than anything else. Speed is just a bonus really, and wasn't that initially the argument to move to SSD? So why switch to one that loses that speed all the time or in specific situations? Consistency matters, its something we rely on and that inspires faith = trust.

And then there is one final argument: the use case. QLC drives with their higher capacities kind of invite 'semi' cold storage. You know, that drive you got with all the family pics on it, bulk storage, stuff you want to keep on backup. But nooooo, gotta keep drive powered. Its not like that mechanical HDD you can drop off in a drawer and it'll (likely) work in five years time. So I seriously question the actual advantages of QLC above even mechanical drives.

 

[Assimilator]

QLC is not 'teh evil' but the way it is being marketed certainly is. Companies hide in the spec sheet and technical minutia to serve you notably worse drives at nearly the same price.

Companies have done that since the dawn of time and will do so until the end of time. Did you have a point?

So yes, in that case, QLC is 'teh evil'. It truly is inferior and priced too high. I don't know what else to make of it. The performance inconsistency speaks volumes that this tech is a bridge too far - and if its not, how about actually making it work properly to begin with? Even on drives that do not lose lots of performance along the way you do suffer major degradation and endurance loss. The lifetime is quite simply a lot shorter on a QLC drive. Why would we cheer for that? To save a few bucks? NOTY.

That performance inconsistency is inherent to all NAND flash, it is merely amplified the higher the layer count gets. The mitigation is NAND parallelism and having a large enough portion of your xLC NAND running in pseudo-SLC mode, neither of which this drive has.

NAND endurance remains a non-issue for anyone in a non-commercial scenario, regardless of the NAND type.

And its not a first for SSD technology and especially controllers either. Whole companies have sunk for releasing bad stuff on storage. Specifically for SSD there are numerous examples of sub par controllers. You'd think by the time we reach QLC they'd know better.

This has nothing to do with the controller.

Storage also has a major trust issue going on. Reliability but also endurance are massively more important than anything else. Speed is just a bonus really, and wasn't that initially the argument to move to SSD? So why switch to one that loses that speed all the time or in specific situations? Consistency matters, its something we rely on and that inspires faith = trust.

Any drive based on non-SLC NAND loses speed when its pseudo-SLC cache is exhausted. Drives that lack DRAM for the mapping tables are consistently slower than drives that do include DRAM. Different controllers and different firmware has different performance characteristics. None of these concerns are specific to QLC.

And then there is one final argument: the use case. QLC drives with their higher capacities kind of invite 'semi' cold storage. You know, that drive you got with all the family pics on it, bulk storage, stuff you want to keep on backup. But nooooo, gotta keep drive powered. Its not like that mechanical HDD you can drop off in a drawer and it'll (likely) work in five years time. So I seriously question the actual advantages of QLC above even mechanical drives.

There is only one reliable solution for cold storage, and that is the cloud. Physical devices, whether they are SSDs or HDDs or optical, are all susceptible to degradation.

As for SSD lifetime data retention, I suggest you read https://www.anandtech.com/show/9248/the-truth-about-ssd-data-retention - it was published half a decade ago, and even then the estimated retention was 10 years without power.

 

[bug]

As for SSD lifetime data retention, I suggest you read https://www.anandtech.com/show/9248/the-truth-about-ssd-data-retention - it was published half a decade ago, and even then the estimated retention was 10 years without power.

For a MLC drive discussed in that article, yes. TLC was worse. V-NAND TLC brought endurance back up to MLC levels (which is why nobody is complaining about current TLC drives). But QLC's endurance can't be as good as TLC's, physically.

 

[Vayra86]

Companies have done that since the dawn of time and will do so until the end of time. Did you have a point?



That performance inconsistency is inherent to all NAND flash, it is merely amplified the higher the layer count gets. The mitigation is NAND parallelism and having a large enough portion of your xLC NAND running in pseudo-SLC mode, neither of which this drive has.

NAND endurance remains a non-issue for anyone in a non-commercial scenario, regardless of the NAND type.



This has nothing to do with the controller.



Any drive based on non-SLC NAND loses speed when its pseudo-SLC cache is exhausted. Drives that lack DRAM for the mapping tables are consistently slower than drives that do include DRAM. Different controllers and different firmware has different performance characteristics. None of these concerns are specific to QLC.



There is only one reliable solution for cold storage, and that is the cloud. Physical devices, whether they are SSDs or HDDs or optical, are all susceptible to degradation.

As for SSD lifetime data retention, I suggest you read https://www.anandtech.com/show/9248/the-truth-about-ssd-data-retention - it was published half a decade ago, and even then the estimated retention was 10 years without power.

I don't care about theoreticals, I care about the practical reality in the market right now. That is what I'm buying into, not the promise of tomorrow. So far, QLC did not deliver and there were numerous other SSD related developments that similarly did not deliver. In all new things storage the wise approach is to wait and see how reliable it truly turns out to be. You're right that in theory you could make great QLC drives... but then the next question is, how expensive are those? If you have to put expensive memory and other hardware next to ultra cheap NAND, the net gain might be zero, and judging by the pricing, it apparently is.

But hey, to each his own, if you have faith in QLC, by all means, worship it.

The retention figure is interesting, I knew the article, but yes, all things considered 10 years is still pretty long in the consumer situation I suggested. Its not very likely a mechanical HDD will do that much better as far as recovering your data goes; it just runs a different sort of risk. So fair enough, maybe QLC is half decent for longer term mass storage (not necessarily cold storage ).