Predictions: 120GB TLC SSD under $30? 240GB TLC under $50?

[VirtualLarry]

I think that the second already happened, with an AR price for an AData 240GB TLC SP550 drive. But I'm talking without rebate, just straight-up prices.

Looking for some (possibly educated) predictions.

I was pleasantly surprised to see Kingston 240GB V300 MLC SSDs for under $60 at Newegg's ebay store recently. (Still $56.99 FS)

I prefer MLC drives. (In fact, I rather loath TLC. But considering some of them implement SLC caching, perhaps they might have similar longevity to a straight-up MLC drive, due to the improved WA due to the caching, and average desktop write workloads?)

Edit: Also, what is the smallest market- and mfg-viable SSD size these days, for both MLC and TLC? It seems like the smallest TLC drive I've seen is 120GB, no 60GB TLC yet. And it seems like the 60GB MLC drives are drying up somewhat.

 

[Elixer]

I don't see them going that low from any reputable OEM, since it just isn't worth it for them to make them, unless they use bottom of the barrel NAND.
This is the problem with the race to the bottom, they keep using the lowest quality crap they can find, and stick a "low" price on them.

 

[cbn]

I am thinking the Phison S11 drives with 3D TLC NAND will probably hit $30 the soonest for 120GB .

They will compete against planar TLC drives with SM2256 and Phison S10 controllers.

Tough to say what will be a better value? (re: the planar TLC drives will likely have closeout pricing whereas the Phison S11/3D TLC drives will have fresh release pricing)

 

[cbn]

Edit: Also, what is the smallest market- and mfg-viable SSD size these days, for both MLC and TLC? It seems like the smallest TLC drive I've seen is 120GB, no 60GB TLC yet. And it seems like the 60GB MLC drives are drying up somewhat.

I don't think anyone will make 60GB, but maybe 90GB via two Micron 3D TLC 384Gbit NAND dies?

 

[hojnikb]

i dont think anyone is dumb enough to make 60GB 2d tlc drive. Speed would be horrible (due to lack of dies), endurance aswell (x2 less flash to write to).

Imo to hit 30$ for 120GB, we need sm2258 without dram. Or any other cheap controller, that can work without dram.
I suppose the bottom end for a sata ssd could be around 20-25$ (60GB chinese ssds sell for that kind of money nowdays).

 

[JimmiG]

At this point the NAND in 120GB drives makes up such a small part of the total cost (controller, case, manufacturing, logistics, support etc.) that they can't really drop the prices any more and still make a profit on them even if the NAND was free. Eventually 120GB drives will probably go away as a meaningful choice when 240GB drives reach the same price point. The same happened with 60/64 GB drives.

You could cut down on costs in other areas like the controller, but at that point it's just a glorified USB stick with terrible performance (write @ 10 - 15MB/s?)...

 

[Erithan13]

Kingston V300? That's the drive that was decent to begin with, then they started silently switching the flash to something slower (and presumably cheaper) after release. Race to the bottom indeed, that's the sort of shenanigans companies are going to be tempted by when they're chasing every last penny.

I don't see much of a place for 60GB SSDs anymore. I used one for many years and while the performance was just fine it became a massive hassle to keep enough space free on it. When you're desperately deleting <500mb temp folders/files all over the place that's a sign to move on. 120GB is much more viable for windows plus office/web applications and maybe a small game install or two. 250GB is still a great sweet spot although I could see 512GB becoming easier to recommend as prices continue to slide down.

 

[hojnikb]

At this point the NAND in 120GB drives makes up such a small part of the total cost (controller, case, manufacturing, logistics, support etc.) that they can't really drop the prices any more and still make a profit on them even if the NAND was free. Eventually 120GB drives will probably go away as a meaningful choice when 240GB drives reach the same price point. The same happened with 60/64 GB drives.

You could cut down on costs in other areas like the controller, but at that point it's just a glorified USB stick with terrible performance (write @ 10 - 15MB/s?)...

You can use dramless designs with a cheap and decent controller (like 2246XT) to hit pricepoints lower than 30$.

 

[cbn]

Just thinking a bit more about the utility of a 90GB drive vs. 60GB drive:

A 90GB drive with 25% capacity left as spare area would have 72 GB available (96GB x .75 = 72GB). So 30GB for Windows 7 leaves 42GB space available. Not bad for an office computer.

A 60GB drive with 25% capacity left as spare area would only have 48GB available (64GB x .75 = 48GB). So 30GB for Windows 7 only leaves 18GB space available

42GB vs. 18GB....that is a pretty big difference despite the total drive capacity only being 50% larger on the 90GB drive vs. the 60GB drive.

 

[BonzaiDuck]

I think that the second already happened, with an AR price for an AData 240GB TLC SP550 drive. But I'm talking without rebate, just straight-up prices.

Looking for some (possibly educated) predictions.

I was pleasantly surprised to see Kingston 240GB V300 MLC SSDs for under $60 at Newegg's ebay store recently. (Still $56.99 FS)

I prefer MLC drives. (In fact, I rather loath TLC. But considering some of them implement SLC caching, perhaps they might have similar longevity to a straight-up MLC drive, due to the improved WA due to the caching, and average desktop write workloads?)

Edit: Also, what is the smallest market- and mfg-viable SSD size these days, for both MLC and TLC? It seems like the smallest TLC drive I've seen is 120GB, no 60GB TLC yet. And it seems like the 60GB MLC drives are drying up somewhat.

I was going to post a thread asking a question about my new ADATA SP 550 -- a ~480-to-500GB TLC drive such as you've described.

I got it for $109. And there's a Terabyte version going for . . . hold on to your hats and socks! -- $199.

The small sample of customer-reviews suggested to me that it would not likely arrive DOA. It's got the same 3-year warranty offered by all the others. MTBF spec is 1,500,000 hours.

Just don't use the aggressive SSD-Tools software to "optimize" the drive. If the optimizations per Pre-fetch, SuperFetch etc. can be done under the OS, you can install the software (for what it's worth and that's next to nothing). But don't use the optimization feature. Any such software with an optimization feature should have a list of some 10 or 12 different items as does Samsung Magician. Otherwise -- it's a good price for a drive; the ANVIL performance results seem off the scale.

I'm happier than a pig in S***. When I bought it, I'd also picked up a Crucial MX200 250GB for $81. That was the first item in my cart, but I decided to just "try" the ADATA. I could've ordered two of those suckers!

 

[VirtualLarry]

I'm glad that you like your Adata SP550, BonzaiDuck. I almost ordered a few of the 120GB models to test out the other night.

I'm kind of curious what a "modern planar TLC" drive can do. Though, I'm still extremely wary of TLC. I guess I would feel more comfortable using a drive with 3D NAND, even if it's TLC.

Samsung's 850 EVO drives, though somewhat pricey, seem pretty rock-solid, performance and longevity-wise.

 

[cbn]

Samsung's 850 EVO drives, though somewhat pricey, seem pretty rock-solid, performance and longevity-wise.

Those EVOs are built on 40nm process.

So I have wondering how much difference in endurance compared to 15nm MLC?

 

[BonzaiDuck]

I'm glad that you like your Adata SP550, BonzaiDuck. I almost ordered a few of the 120GB models to test out the other night.

I'm kind of curious what a "modern planar TLC" drive can do. Though, I'm still extremely wary of TLC. I guess I would feel more comfortable using a drive with 3D NAND, even if it's TLC.

Samsung's 850 EVO drives, though somewhat pricey, seem pretty rock-solid, performance and longevity-wise.

I thought I'd read something of a basis for your worries about TLC, but I forgot what it was. I'm getting so old and still have so many distractions it's hard to keep up with this stuff.

What was it -- about TLC -- that you refer to? The ADATA for both the OS partitions uses a Primocache RAM cache of some 2TB. I ran ANVIL against it, and the results are off the charts. I even implemented deferred writes.

I guess as long as it's backed up nightly to my server, It will merely be a minor setback if something goes wrong with the drive.

 

[Glaring_Mistake]

I ran ANVIL against it, and the results are off the charts.

Okay, the ADATA SP550, like most SSDs using TLC NAND has a small cache of SLC NAND which is much faster than the TLC NAND.

And if you then use a benchmark which only writes like 1GB then you are going to get great results because you're just measuring the speed of that cache and not the drive in general.

But when they run out of that cache things are going to slow down considerably.
A slight exception being Samsung SSDs using TLC NAND whose performance doesn't tend to take as big a hit when their SLC-cache runs out as the rest of them.


And here you have some old test results of mine to give you an idea of how the SLC-cache affects results.

It is not as big as yours so it is a bit slower since bigger SSDs tend to be faster in general and yours has a larger SLC-cache but you get the picture.

 

[bononos]

I think that the second already happened, with an AR price for an AData 240GB TLC SP550 drive. But I'm talking without rebate, just straight-up prices.

Looking for some (possibly educated) predictions.
.......

There was supposed to be another 3d-nand model from Crucial which would've given indication on how much prices were going to move since it would be the 2nd company with 3d-nand ssds. But theres no news so far and the only leaked Crucial MX300 listing on Amazon has been removed.

 

[BonzaiDuck]

We've been down this road before, as it pertained to Samsung RAPID. I fully realize that benchmarks are synthetic.

I'm using Primocache for three of my systems -- a 3PC license. With the dual boot Win7/win10 systems, it's configured with consideration for the fact that I can't do SSD-caching for one OS and expect to use the same SSD for the other. But that's for a hard disk. I need to poke around in Primo some more to see if I can be successful in splitting the SSD between the two OS's.

So, without claiming any bragging rights, here's my own ANVIL benchmark:

I can't even remember how it compares to the Samsung EVO that I was using before I replaced it with the ADATA -- the ADATA could be marginally worse.

But I've done my own manual "optimization" for this drive in both OS'es, and it seems lightning-fast to me.

I suppose that's also why I've been so sluggish about replacing these old Sandy systems with a Skylake system.

I guess it all boils down to whether I can expect the ADATA to "last." But why wouldn't it?

But if it comes to shelling out for a 3D NAND Sammy Pro or a Crucial that doesn't seem to be on the market yet, I'll wait until I build a whole new system. For $110, the ADATA looks pretty good right now.

 

[cbn]

Earlier in the thread I was wondering about how Samsung's 3D TLC NAND (built on 40nm) compared to 15nm MLC NAND as far as endurance goes. I couldn't find that info, but I was able to find the following Toshiba marketing slide which claims the company's 3D TLC NAND has 3x the endurance of their 15nm MLC:

That's right - QLC (Quadruple Level Cell), which is also 4-bit MLC, has been mentioned by Toshiba. As you can see at the right of that slide, storing four bits in a single flash cell means there are *sixteen* very narrow voltage ranges representing the stored data. That is a very hard thing to do, and even harder to do with high performance (programming/writing would take a relatively long time as the circuitry nudges the voltages to such a precise level). This is why Toshiba pitched this flash as a low cost solution for archival purposes. You wouldn't want to use this type of flash in a device that was written constantly, since the channel materials wearing out would have a much more significant effect on endurance.[/b] Suiting this flash to be written only a few times would keep it in a 'newer' state that would be effective for solid state data archiving.

The 1x / 0.5x / 6x figures appearing in the slide are meant to compare relative endurance to Toshiba's own planar 15nm flash. The figures suggest that Toshiba's BiCS 3D flash is efficient enough to go to QLC (4-bit) levels and still maintain a higher margin than their current MLC (2-bit) 2D flash.

More to follow as we continue our Flash Memory Summit coverage!

 

[BonzaiDuck]

Earlier in the thread I was wondering about how Samsung's 3D TLC NAND (built on 40nm) compared to 15nm MLC NAND as far as endurance goes. I couldn't find that info, but I was able to find the following Toshiba marketing slide which claims the company's 3D TLC NAND has 3x the endurance of their 15nm MLC:

I won't hunt down the link unless you bother me, but earlier today I found an article testing several varieties of both TLC and MLC. The impression I got was that you could expect any drive to last through 500TB of writes. Was I wrong about this? The actual range of performance was between 600TB and 1.2TB.

So my only gripe with that arises from my need to hibernate my systems. The Hiberfil.sys on this system is something like 12GB. It can be shortened, but you can't get rid of it, and it has to reside on the boot-system disk. I saw all sorts of people hoping they could move it to another disk, but no cigar . . .

Hibernation is essential for the proper operation of my UPS, and my nightly client-system backups to my WHS-2011 server. And I expect to have the same arrangement when I move to Win Server 2012 R2 Essentials.

My brother's system, built by me and running upstairs, was equipped with a Samsung EVO, going between wake, sleep and then hibernate on a daily basis for the last year and 3 months. 15 months.

It's showing about 2.9 TB of writes on Magician. So in a year, if it's 6TB, or in three years it shows 10TB, it would show 20TB in 6 years.

Does this seem in the ballpark? That's nowhere near even 25% of a Petabyte, and you'd think you might still have SSD-life thereafter.

 

[Glaring_Mistake]

I fully realize that benchmarks are synthetic.

Not pointing out that benchmarks are synthetic.
They are but that's not the point.
The point is that you only measure the very fastest part of it by using the standard setting of 1GB with that benchmark.

For example mechanical harddrives may get a very fast burst at the start of a transfer and then slow down quickly.
But you wouldn't just look at how fast the cache is and determine that you have one of the fastest harddrives in the world now would you?
Yet that is all that you are measuring when using such light loads and thereby concluding that the results are ”off the chart”.

I can't even remember how it compares to the Samsung EVO that I was using before I replaced it with the ADATA -- the ADATA could be marginally worse.

If you had an 850 EVO at the same capacity as the SP550 the EVO would beat it in every aspect.
If you had an 840 EVO at the same capacity as the SP550 the EVO would probably still have the edge.

 

[Glaring_Mistake]

Earlier in the thread I was wondering about how Samsung's 3D TLC NAND (built on 40nm) compared to 15nm MLC NAND as far as endurance goes. I couldn't find that info, but I was able to find the following Toshiba marketing slide which claims the company's 3D TLC NAND has 3x the endurance of their 15nm MLC:

Hellhammer here seemed to believe that the number of write cycles would remain at 3000.

Given that I have seen claims of their 15nm TLC NAND capable of 1500 write cycles using BCH ECC then the double for their 15nm MLC does not seem like such an outlandish claim.


Now, about the marketing slide.

Samsung claimed an endurance improvement of ten times using V-NAND yet what we saw in 850 Pro/EVO was just a doubling of the number of write cycles.

IMFT has also claimed that their 3D NAND can reach 30000 write cycles (basically the same as Samsung).
Yet their flyer (which you likely have already seen here in the forums) says their 3D MLC will have 3000 write cycles while for their 3D TLC it will range from 500 to 1500 based on the type of ECC used.

So I think it would be wise to take any claims of extreme improvement in terms of endurance with a grain of salt.