The most reasonable application for this SSD seems to be consumer laptop with limited storage expansion, and at a more as approachable price than previous 8TB M.2 device

At those prices it had better be TLC not QLC...

New SSD's with older model controller got price of 16.25¢/GB.
2020 year of madness prices. I really hope new devices with much more layers and better nanometer process to correct this!

They need to get that 8TB price to $1000 considering a 2TB can be had for under $250

lexluthermiesterAt those prices it had better be TLC not QLC...

Definitely QLC, you can't get that kind of density with TLC. It wouldn't fit on the PCB. The pricing also roughly matches Sabrent's and Corsair's pricing for similar drives. 8TB QLC drives can be pretty good though. In AnandTech's testing the 8TB Sabrent Rocket Q nearly matches the 2TB 970 Evo Plus in performance, and when empty has a whopping 2TB SLC cache(!). It also manages to sustain a decent 330MB/s sequential write speed should you manage to exceed 2TB of writes in a single go. Here's hoping this has a similarly generous SLC cache system, and that we can get some price competition at the top end too!

AnarchoPrimitivThey need to get that 8TB price to $1000 considering a 2TB can be had for under $250

I agree, though it's unlikely to happen - the extra density makes it more expensive. 1-2TB is the definitive sweet spot for SSDs currently.

ValantarDefinitely QLC, you can't get that kind of density with TLC.

You can with the newest 176 layer 3D-TLC nand.

Valantar8TB QLC drives can be pretty good though. In AnandTech's testing the 8TB Sabrent Rocket Q nearly matches the 2TB 970 Evo Plus in performance, and when empty has a whopping 2TB SLC cache(!). Here's hoping this has a similarly generous SLC cache system.

Unless such a drive can be configured to run exclusively in SLC or MLC mode, the durability is going to be a very serious problem, even with the new wear-leveling schemes.

lexluthermiesterYou can with the newest 176 layer 3D-TLC nand.

Sure, but has that even reached mass production? And even if it has, aren't all early NAND runs pretty much exclusively low-grade flash destined for USB sticks and phones, to ensure that they can sell every single die?

Edit: nvm, I hadn't caught that Micron apparently beat Hynix to the punch with 176-layer NAND a month or so ago. Still, it's extremely unlikely that that's what this drive uses.

Before a week Samsung reading for 256 layers( 2 stac or 128 layers ) but not sure for timeline maybe 2021/2022 or? In Techpowerup has not specific article with this info.

TumbleGeorgeBefore a week Samsung reading for 256 layers( 2 stac or 128 layers ) but not sure for timeline maybe 2021/2022 or? In Techpowerup has not specific article with this info.

To quote AnandTech's article on the recent announcement of SK Hynix 176-layer 3D NAND:

AnandTechSamsung's 128L NAND started shipping a few months ago in the 980 PRO. While they have not officially announced specs for their next generation, it is expected to go into production next spring with a layer count in the neighborhood of 176L, and to be Samsung's first generation to use string stacking—a technique their competitors all had to adopt while layer counts were in the 64-96L range.

That they will be moving to >200 layers at some point is a given - so will all NAND manufacturers with the R&D budgets to keep up - but those layer counts are still a long, long way off, and Samsung definitely aren't at the forefront in that regard.

lexluthermiesterUnless such a drive can be configured to run exclusively in SLC or MLC mode, the durability is going to be a very serious problem, even with the new wear-leveling schemes.

Thankfully that's not true. Not even close. What, exactly, are you using that 8TB drive for? If it's a torrent drive getting hammered, I guess that's rather strenuous, but ... still not a problem. The Rocket Q that I mentioned above is rated for 0.13DWPD. That's more than 1TB of writes every single day for its five-year warranty period. If you have a workload like that, you really should be looking at enterprise-level products. For any consumer, even a power user, the endurance of these drives is perfectly fine. The obvious use for a drive like this is a game drive - there's not much else that combination of capacity and performance is useful for. Of course there will also be wealthy enthusiasts (well, anyone buying this drive likely qualifies for that designation) buying a drive like this for an all-flash NAS, but again, what are they supposed to be doing with their NAS that subjects it to >1TB of writes per day? Most consumer workloads are damn near WORM. My 500GB 960 Evo OS+software+some games drive (from late 2017) has 26TB of writes on it - that's about 22.5GB/day. I've by no means hammered it with writes, but I don't make any effort to spare it either, and it hosts my pagefile.sys and hiberfile.sys. I'm still literally at 1/44th the required load to wear out that Sabrent Rocket Q within its warranty period. Even the 1TB Rocket Q with its measly 260TBW rating is 10x the amount of writes I've managed in three years. (Of course the performance of a 1TB QLC drive is likely to be bad enough that you couldn't feasibly manage that kind of load, but that's besides the point here.)

ValantarWhat, exactly, are you using that 8TB drive for?

Boot drive. If I'm spending that much money on an SSD, I'm booting from it and it damn well better be durable. Now perhaps I should clarify, when I said "configured to run exclusively in SLC or MLC mode" it is understood that this will come at the expense of reduced capacity and I'm ok with that. Otherwise there is no way I'm using QLC for a boot drive.

ValantarThe Rocket Q that I mentioned above is rated for 0.13DWPD. That's more than 1TB of writes every single day for its five-year warranty period.

We're not talking about a RocketQ as it is not the subject of this article.

ValantarIf you have a workload like that, you really should be looking at enterprise-level products. For any consumer, even a power user, the endurance of these drives is perfectly fine.

That is a matter of opinion, one I do not agree with either on a personal level or as a professional. I expect technology to continue working well passed it's warranty period, by a factor of at least 3. QLC does not meet that requirement. Not even close.

ValantarMost consumer workloads are damn near WORM.

Also an opinion with no merit to back it up. I know many who have experienced high levels of wear on SSDs. Those are TLC based drives.

Valantar(Of course the performance of a 1TB QLC drive is likely to be bad enough that you couldn't feasibly manage that kind of load, but that's besides the point here.)

No, that's exactly the point. QLC does not and physically can not meet the durability demands required from a drive expected to be a boot drive.

If the above drive in this article is QLC it is not suitable for use as a bootable OS drive. It is suitable ONLY as mass storage drive that experiences infrequent incidental writes. The fact that it has only a 3 year warranty is not a reassuring sign..

Ignoring the price, is anyone else still really happy at how much capacity NVME drives seem to have? Mech drives were so damn stagnant for capacity, and NVME M.2 is going WHOOOOSH

lexluthermiesterBoot drive. If I'm spending that much money on an SSD, I'm booting from it and it damn well better be durable. Now perhaps I should clarify, when I said "configured to run exclusively in SLC or MLC mode" it is understood that this will come at the expense of reduced capacity and I'm ok with that. Otherwise there is no way I'm using QLC for a boot drive.

We're not talking about a RocketQ as it is not the subject of this article.

That is a matter of opinion, one I do not agree with either on a personal level or as a professional. I expect technology to continue working well passed it's warranty period, by a factor of at least 3. QLC does not meet that requirement. Not even close.

Also an opinion with no merit to back it up. I know many who have experienced high levels of wear on SSDs. Those are TLC based drives.

No, that's exactly the point. QLC does not and physically can not meet the durability demands required from a drive expected to be a boot drive.

If the above drive in this article is QLC it is not suitable for use as a bootable OS drive. It is suitable ONLY as mass storage drive that experiences infrequent incidental writes. The fact that it has only a 3 year warranty is not a reassuring sign..

You're entirely missing the point here. I'm using the Rocket Q as an example as it's likely an identical design to this drive - Phison E12S controller, QLC using 64 1Tb dice, etc. It's likely a Phison reference design.

Two key points:
1: With this kind of capacity (at least with Sabrent's approach to SLC caching), performance is more than fine for any consumer usage. Unless you also think a 970 Evo Plus is too slow? That is a key difference between 1TB QLC drives and these 8TB ones. I sincerely doubt Sabrent has done anything particularly innovative with their drive, and thus expect competitors with similar designs to perform similarly. The sheer amount of flash makes drives this large inherently faster.
2: Again, with the rated write endurance of drives like these, 3x the warranty period should be perfectly fine. This Mushkin drive is rated for the same 1800TBW as the Sabrent, meaning you could write more than 500GB/day for nine years before reaching the limit of the drive. Of course, the 4TB is "only" 250GB/day. What a shame - who doesn't download 3-4 AAA games every single day? Or spend every minute of every day of their life rendering video? If you know of a consumer workload besides that causing >250GB/day of writes, every single day, then please share? (The Sabrent has a 5-year warranty, meaning >300GB/day for 15 years for the 8TB, or >150TB/day for the 4TB - though I'd be shocked if any m.2 SSD could last 15 years of daily use without something besides the flash failing.)

Of course there are some people who cause massive wear on their SSDs. And as I said above: those people clearly aren't the target market for a product like this, and should be looking at enterprise products, as their workloads clearly border on enterprise levels of wear. If you think it's too much to ask that people actually buy equipment suited for their needs, well, that's your problem. As you clearly like to say, that's an opinion with no merit to back it up. Who says extreme edge case power users should get to decide which products are viable for the majority of users? Who says products for the average user should be designed around the needs of edge cases? You're not doing anything but spouting massively entitled opinions yourself. It's of course unfortunate that people with these extreme workloads are also likely to be most interested in high capacity SSDs and to some extent also to have the money to afford a drive like this, but again: it's not meant for them. And that's fine. If you have a problem with that, well, that's your opinion.

As of now, the AnandTech review of the Rocket Q is the closest thing we have to actual data on the viability of a drive like this, and unlike previous, lower capacity QLC offerings, performance looks pretty good, and combined with massive SLC caching and write endurance ratings far beyond what can be expected for even most power users this looks perfectly suitable for a game drive or flash-only NAS. I still wouldn't touch a 1TB QLC drive due to the massive latency spikes and general sluggishness of those drives, but one of these? I wouldn't spend $1300 on one, but I sure wouldn't mind owning one.

MusselsIgnoring the price, is anyone else still really happy at how much capacity NVME drives seem to have? Mech drives were so damn stagnant for capacity, and NVME M.2 is going WHOOOOSH

Sure, but at what comparable cost? I can get a high quality 8TB HDD for between $175 to $200 that will likely last 10 years or longer. 8TB SSD? $1000 and it will not last nearly as long. And durability is getting worse with PLC on the horizon that has less than 800 P/E cycles. And when you take into consideration that many makers have released SSD's above 10TB, it's not really that exciting.

When someone releases a high capacity, high durability SSD that is available to the general consumer, that'll be exciting... We're not getting an SSD product like that until a very serious advancement is made in the manufacturing process. As is, SSD's are literally designed to fail and that is simply unacceptable.

ValantarYou're entirely missing the point here.

Yup, I missed your point.

ValantarUnless you also think a 970 Evo Plus is too slow?

Performance is not a concern as every SSD on the market today delivers more bandwidth than even power users practically need. Durability is the most important consideration people need to pay attention to.

lexluthermiesterSure, but at what comparable cost? I can get a high quality 8TB HDD for between $175 to $200 that will likely last 10 years or longer. 8TB SSD? $1000 and it will not last nearly as long. And durability is getting worse with PLC on the horizon that has less than 800 P/E. And when you take into consideration that many makers have released SSD's above 10TB, it's not really that exciting.

When someone releases a high capacity, high durability SSD that is available to the general consumer, that'll be exciting... We're not getting an SSD product like that until a very serious advancement is made in the manufacturing process. As is, SSD's are literally designed to fail and that is simply unacceptable.

I'm just talking about capacity vs size, and how fast its approaching. it took decades for mech drives to go from single GB's to these multi terabyte drives, and SSD's are progressing at a much faster rate.

MusselsI'm just talking about capacity vs size, and how fast its approaching. it took decades for mech drives to go from single GB's to these multi terabyte drives, and SSD's are progressing at a much faster rate.

That's fair and a good point. The thing is, if makers went back to SATA 2.5" or 3.5" form-factors, capacities could easily got very high because of all the physical space inside the drive that's available. With M.2 you're limited in physical space available, so 8TB on an NVMe drive is somewhat impressive.

Musselshow much capacity NVME drives seem to have?

E18 Phison controller limited to 16TB. That mean no more of 16TB on one m.2 NVMe device before next future model of controller with name maybe E20 or something else when be ready next year or 2022?

TumbleGeorgeE18 Phison controller limited to 16TB. That mean no more of 16TB on one m.2 NVMe device before next future model of controller with name maybe E20 or something else when be ready next year or 2022?

It's highly unlikely we'll have dense enough flash to hit 16TB before then (which would require either 2Tb dice or 16-die packages). We'll get there, obviously, but likely not in the next two years. So to go beyond that you'd need a >8 channel controller (or use multiple packages per channel, which hurts performance). Besides, even if flash prices drop 30% per year, a 16TB SSD would then cost something like $1700. That would of course be great value compared to a $12-1300 8TB one, but still the price of a very good gaming PC...

lexluthermiesterThat's fair and a good point. The thing is, if makers went back to SATA 2.5" or 3.5" form-factors, capacities could easily got very high because of all the physical space inside the drive that's available. With M.2 you're limited in physical space available, so 8TB on an NVMe drive is somewhat impressive.

You see this a lot in enterprise drives - aren't there 100TB 3.5" datacenter SSDs out there? It doesn't make sense for consumers whatsoever though - the extra space is needed to fit more flash, which is what drives the price up, etc. I bet those massive SSDs are fantastic for their use though. Still, it doesn't make sense in terms of cost to use more space than what's available on an m.2 for consumer use - you'd be pricing yourself out of the consumer market by default.

I'm very sure that on CES '21 will be announced or presented m.2 NVMe SSD's with more than 8TB. For this year is too late for this. Hmm 30% drop of prices is just preferably, no guaranteed value. The percentage may be lower or higher. Covid-19 and it effects on market and production volumes and prices also security and interstate delivery time of materials and finished products from that were not foreseen in the development plans are already present on the scene.

ValantarIt's highly unlikely we'll have dense enough flash to hit 16TB before then

It's probably already here ~
SK hynix Unveils the Industry's Most Multilayered 176-Layer

ValantarYou see this a lot in enterprise drives - aren't there 100TB 3.5" datacenter SSDs out there? It doesn't make sense for consumers whatsoever though - the extra space is needed to fit more flash, which is what drives the price up, etc. I bet those massive SSDs are fantastic for their use though. Still, it doesn't make sense in terms of cost to use more space than what's available on an m.2 for consumer use - you'd be pricing yourself out of the consumer market by default.

That would depend. If makers used slightly slower but still reliable NAND, which is less expensive, they could get more into a single drive cost effectively. I would happily pay $200 for a 2TB SATA SSD that got between 250 to 300 MB per second performance with MLC. That is more than acceptable performance.

lexluthermiesterThat would depend. If makers used slightly slower but still reliable NAND, which is less expensive, they could get more into a single drive cost effectively. I would happily pay $200 for a 2TB SATA SSD that got between 250 to 300 MB per second performance with MLC. That is more than acceptable performance.

Samsung has an 8TB 870 QVO (AT reviewed it together with the Rocket Q), and it's a lot cheaper than the 8TB NMVe drives, but ... it's still $900, or $0.11/GB, so still no cheaper than lower capacity drives. And of course the 2.5" enclosure still has plenty of room at that capacity. But still, even if you went with absolute bargain-basement SATA controllers and the cheapest flash you could find, it still wouldn't be cheap. Let's say they perform a minor miracle and manage to sell a 12TB drive at 8c/GB. (For reference, the cheapest SSDs on PCpartpicker are currently 7.8c/GB) That's still $960. At 7c/GB it would be $840. The vast majority of consumer SSD sales is in the $100 range, with some stretching to spend ~$200. I would be surprised if $300 drives sold .01% of $100 drives, frankly, let alone $800-1000 drives.

Also, where are you finding that "slightly slower but still reliable NAND"? Is it some form of low interface speed TLC? Most of that is likely going EOL as higher interface speed flash is no more expensive to produce per bit (mostly less expensive), and QLC is still going to be cheaper.

But I think we're talking past each other here. You were talking of using form factors bigger than m.2 to achieve higher capacities for cheaper, yet you're talking about 2TB, which can fit ... well, it might not fit on an m.2 2242, but 2280 single sided is doable, let alone double sided. More space doesn't make for any savings there; there's no real reason for a 2TB 2.5" SATA SSD to be any cheaper than a 2TB m.2 SATA SSD.

Plus, an MX500 2TB isn't too far from what you're asking for. I couldn't find any numbers for sustained non-cached writes alone, but Tom's showed it at ~400MB/s sustained mixed r/w. And it's $190. The only reason older drives like that are cheap is that they no longer have any non-BOM costs to amortize, allowing manufacturers to maintain margins at lower prices, plus they might have had the option to either get a late run of flash for cheap before the fab moved to a newer design, or they just moved to a newer and cheaper flash type altogether. They might also be getting the old controllers dirt cheap, but that doesn't make much of a difference. The controller, PCB, housing and other componentry contribute a base cost of $20(SATA)-50(high end NVMe)-ish, so especially at higher capacities this really doesn't make much of a difference (and it certainly doesn't explain the price difference between the 8TB QVO and Rocket Q!).

ValantarAlso, where are you finding that "slightly slower but still reliable NAND"? Is it some form of low interface speed TLC?

No. It's the physics of how NAND works. Higher speed requires more voltage to affect phase change. Lower the speed, lower the voltage, the NAND lasts longer. This is of course very simplified, but that's the general idea.

lexluthermiesterNo. It's the physics of how NAND works. Higher speed requires more voltage to affect phase change. Lower the speed, lower the voltage, the NAND lasts longer. This is of course very simplified, but that's the general idea.

Ah, that's understandable. Don't manufacturers generally rate their NAND for some operating voltage (range) from the factory though, and leave it at that? Or do they sell lower-speed bins of the same flash to maximise yields? That does sound smart (if it increases yields meaningfully), but given that increased longevity is valuable in various sectors I wouldn't be surprised if they instead just branded it as "increased longevity NAND" and sold it for the same price (if not more).

ValantarDon't manufacturers generally rate their NAND for some operating voltage (range) from the factory though, and leave it at that?

It varies from vendor to vendor. I think there is a lot of binning going on...

ValantarOr do they sell lower-speed bins of the same flash to maximise yields?

...which naturally leads to this.

AnarchoPrimitivThey need to get that 8TB price to $1000 considering a 2TB can be had for under $250

Much less than that. The prices of SSD's and NVME drives may have come down a lot since their initial launch but they've stagnated over the past few years. Aside from the odd deal on Black Friday, you're looking at having to pay £100 for just a 1TB, and the higher capacities tack a premium onto it.

Meanwhile mechanical drives are powering on with going 20+TB by 2021, likely all under £800 too. Unless you're looking for the performance, SSD manufacturers just keep shooting themselves in the foot, spinning rust remains king. Small SSD for boot up, large capacity drive for storage, that probably won't change for at least another decade the rate they're going.