SSD flash: TLC vs. MLC vs. SLC

All things storage here: hard drives, DVD RW drives, little wicker baskets.

Moderators: morphine, Steel

SSD flash: TLC vs. MLC vs. SLC

Postposted on Mon Mar 03, 2014 9:23 am

Moderator's note: This thread was split from a new member's build thread, here:
viewtopic.php?f=33&t=93031



iamjsmith83 wrote:SSD: SAMSUNG 840 EVO MZ-7TE250BW 2.5" 250GB SATA 6Gb/s MLC Internal Solid State Drive (SSD) $190


Samsung EVO 840 uses crappy TLC chips and because it's not cheaper than MLC parts, it's impossible to recommend.

That product description lies, Samsung 840 EVO is TLC, not MLC.

iamjsmith83 wrote:HDD WD BLACK SERIES WD2003FZEX 2TB 7200 RPM 64MB Cache SATA 6.0Gb/s 3.5" Internal Hard Drive $150


WD Black is noisy and hot, it's up to you if it's problem or not.
Last edited by JustAnEngineer on Mon Mar 03, 2014 10:01 pm, edited 2 times in total.
Reason: Split from original thread
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: First Build

Postposted on Mon Mar 03, 2014 9:50 am

Wicked Mystic wrote:Samsung EVO 840 uses crappy TLC chips and because it's not cheaper than MLC parts, it's impossible to recommend.

What? I would suggest you read TR's SSD endurance test. Keep in mind that 600TB is WAY beyond the warranty periods for any of those drives. The Samsung 840 (non-pro) has TLC nand like the EVO, and yes, you can see endurance takes a hit. However, the 840 EVO has a SLC cache that relieves write/erase cycles from the TLC. I don't think even the heaviest consumer users would wear out an EVO.

To the OP: Do you live near a Microcenter? They have the best CPU+Mobo combo deals anywhere. If you do, I would suggest signing up for their newsletter. They sell 4770K's for $250 periodically (just had a 3-day sale starting 2/28, so they might not be $250 for another month or two).

Also, a good price on a high-capacity HDD (you would be fine with a 5400-5900 rpm drive) is between $30-40 /TB.
i5-3570K, ASRock Z77 Pro4-m, Asus GTX660 TOP, 120 GB Vertex 3 Max IOPS, 2 TB Samsung EcoGreen F4, 8GB G-Skill @1.25V, Silverstone PS07B
DPete27
Gerbil Jedi
Silver subscriber
 
 
Posts: 1656
Joined: Wed Jan 26, 2011 12:50 pm
Location: Madison, Wisconsin

First Build

Postposted on Mon Mar 03, 2014 10:17 am

DPete27 wrote:
Wicked Mystic wrote:Samsung EVO 840 uses crappy TLC chips and because it's not cheaper than MLC parts, it's impossible to recommend.

What? I would suggest you read TR's SSD endurance test. Keep in mind that 600TB is WAY beyond the warranty periods for any of those drives. The Samsung 840 (non-pro) has TLC nand like the EVO, and yes, you can see endurance takes a hit. However, the 840 EVO has a SLC cache that relieves write/erase cycles from the TLC. I don't think even the heaviest consumer users would wear out an EVO.


Endurance tests are useless. It's not how many times you can write. It's about how long drive can hold data. Yeah, maybe you can write to SSD 1000 times but if data is lost by an hour, write durability is useless.

You can get MLC drive cheaper, so you should buy TLC because.... ?

DPete27 wrote:Also, a good price on a high-capacity HDD (you would be fine with a 5400-5900 rpm drive) is between $30-40 /TB.


I recommend 7200 RPM drive, because they are much faster than those power saving drives.
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: First Build

Postposted on Mon Mar 03, 2014 10:30 am

Wicked Mystic wrote:Endurance tests are useless. It's not how many times you can write. It's about how long drive can hold data. Yeah, maybe you can write to SSD 1000 times but if data is lost by an hour, write durability is useless.

????

Are we suggesting that certain SSDs have a problem with just dropping data? Because I have never heard of that. There are a couple brands with common reliability problems, but that's about drives dying, not data losses.

DPete27 wrote:Also, a good price on a high-capacity HDD (you would be fine with a 5400-5900 rpm drive) is between $30-40 /TB.

I would not recommend this unless you know that you can get all of your programs on the main drive. Yes, a "green" (5400 RPM) drive will save you some money, but a "normal" 7200 RPM drive can store a lot without taking a significant performance drop.
Damage wrote:Don't try to game the requirements by posting everywhere, guys, or I'll nuke you from space.

-Probably the best Damage quote ever.
superjawes
Graphmaster Gerbil
Gold subscriber
 
 
Posts: 1085
Joined: Thu May 28, 2009 9:49 am

Re: First Build

Postposted on Mon Mar 03, 2014 10:44 am

superjawes wrote:
Wicked Mystic wrote:Endurance tests are useless. It's not how many times you can write. It's about how long drive can hold data. Yeah, maybe you can write to SSD 1000 times but if data is lost by an hour, write durability is useless.

????

Are we suggesting that certain SSDs have a problem with just dropping data? Because I have never heard of that. There are a couple brands with common reliability problems, but that's about drives dying, not data losses.


I'm suggesting that while drives on those endurance tests may last many writes, they hold data so short time that nobody is interested how many times data can be written. Of course they did not test that one. They just tested how many times data can be written. That is why those endurance tests are nearly useless.

superjawes wrote:I would not recommend this unless you know that you can get all of your programs on the main drive. Yes, a "green" (5400 RPM) drive will save you some money, but a "normal" 7200 RPM drive can store a lot without taking a significant performance drop.


Save money? At same price you can get 7200 RPM or power saving model. So choice is quite easy.

http://www.newegg.com/Product/Product.a ... 6822148834
http://www.newegg.com/Product/Product.a ... 6822236404
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: First Build

Postposted on Mon Mar 03, 2014 11:07 am

Wicked Mystic wrote:They just tested how many times data can be written. That is why those endurance tests are nearly useless.

Again, I suggest you actually READ the article(s). TR has been doing data retention tests at every milestone.
i5-3570K, ASRock Z77 Pro4-m, Asus GTX660 TOP, 120 GB Vertex 3 Max IOPS, 2 TB Samsung EcoGreen F4, 8GB G-Skill @1.25V, Silverstone PS07B
DPete27
Gerbil Jedi
Silver subscriber
 
 
Posts: 1656
Joined: Wed Jan 26, 2011 12:50 pm
Location: Madison, Wisconsin

Re: First Build

Postposted on Mon Mar 03, 2014 12:24 pm

DPete27 wrote:
Wicked Mystic wrote:They just tested how many times data can be written. That is why those endurance tests are nearly useless.

Again, I suggest you actually READ the article(s). TR has been doing data retention tests at every milestone.


- They used way too short period between retention tests. Check what Jedec says about this.
- They always wrote whole disk, IRL disk is never empty.
- Test period was way too short.

This is something that need to be tested over years, not months.

Anyway, because MLC drive costs less than TLC drive, I cannot see any reason to pay more for worse.

Neutronbeam wrote:Re the poster that says WD Black hard drives are loud--then he must have ears like Superman--with some fans installed (or even without) the drive should be well within heat tolerances and virtually inaudible.


Some consider WD Black to bee too loud. Some people say WD Blue is too loud (compared to SSD). This depends on who you are asking.

I consider WD Black to be silent enough.
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: First Build

Postposted on Mon Mar 03, 2014 1:13 pm

Wicked Mystic wrote:- They used way too short period between retention tests. Check what Jedec says about this.
- They always wrote whole disk, IRL disk is never empty.
- Test period was way too short.

This is something that need to be tested over years, not months.

Anyway, because MLC drive costs less than TLC drive, I cannot see any reason to pay more for worse.

You're talking about something that is unrealistic and cannot be tested effectively, so making a recommendation based on this is pointless.

You should choose a SSD based on price/performance compared to how much you are willing to spend on one. In that respect, the 840 EVO (the part mentioned by the OP) is a good recommendation.
Damage wrote:Don't try to game the requirements by posting everywhere, guys, or I'll nuke you from space.

-Probably the best Damage quote ever.
superjawes
Graphmaster Gerbil
Gold subscriber
 
 
Posts: 1085
Joined: Thu May 28, 2009 9:49 am

Re: First Build

Postposted on Mon Mar 03, 2014 1:58 pm

superjawes wrote:You're talking about something that is unrealistic and cannot be tested effectively, so making a recommendation based on this is pointless.

You should choose a SSD based on price/performance compared to how much you are willing to spend on one. In that respect, the 840 EVO (the part mentioned by the OP) is a good recommendation.


Exactly. That cannot be tested. However we do know that MLC > TLC, that is a fact. So why spend more on worse TLC product? If TLC stuff is cheaper, it's OK. Now it's more expensive. So?

Chrispy_ wrote:Avoid the K-series! - They're great for gaming, but not for what you're doing.

You DO NOT want to overclock when dealing with audio creation software. Many wave/audio/midi/recording packages are EXCEPTIONALLY clock-sensitive. There are thousands of threads on hundreds of forums including this one about the issues, and whilst a lot of the problems are drivers, some of it is just plain old PCI latency that relies on a rock-stable 100MHz PCI-e system clock.


PCI latency? PCI devices are already legacy and chipset doesn't even support PCI. Cannot see problem in that.

Chrispy_ wrote:Also, some of the better audio software is OLD, you may want to run Windows XP as a VM, and thanks to Intel's stupid product segmentation the K-series processors lack VT-D so you can't get such direct access to the hardware for VMs. For audio softare, eliminating lag is absolutely essential so a K-series chip will probably spoil your week if you need to do this.


I recommend using newer software than any XP legacy. If overclocking is not in plans, I already offered alternative, that is Xeon 1230 v3. I see no reason to waste 209 dollars for motherboard if not overclocking however.
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: First Build

Postposted on Mon Mar 03, 2014 3:41 pm

DPete27 wrote:I'll go ahead and say it since Wicked Mystic has been called out for this on other threads. If you're going to make an argument (especially one that goes against what others are saying) please provide references.
(@Wicked Mystic) Nobody is trying to personally attack you or try to shut you up. We value everyone's opinion, but we need to make sure to provide proven facts.


OK.

Chosen motherboard MOTHERBOARD: ASUS Z87-PRO LGA 1150 Intel Z87 HDMI SATA 6Gb/s USB 3.0 ATX Intel Motherboard $209

Then suggestion is to buy i7-4770?

As for Samsung Evo, MLC > TLC. So why pay more for TLC? Have not received answer for this one.

I already recommended Xeon 1230 v3 over i7-4770, no comments.

You can get 7200 RPM drive at same price as 5400 RPM drive, 5400 RPM is fine. Really?

Can somebody explain these using so called facts?
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: First Build

Postposted on Mon Mar 03, 2014 3:51 pm

Wicked Mystic wrote:As for Samsung Evo, MLC > TLC. So why pay more for TLC? Have not received answer for this one.

The Samsung EVO 250GB runs for $153 right now. Do you have an MLC suggestion with:

1. Similar Capacity?
2. Similar Price Tag?
3. Similar Reliatbility?

You made the assertion that "TLC is crap," but have not provided an example where an MLC drive offers better value over the Samsung EVO, or even an example where a TLC drive is more expensive than an MLC one.

Wicked Mystic wrote:You can get 7200 RPM drive at same price as 5400 RPM drive, 5400 RPM is fine. Really?

The price comparison you gave was not between the same brands. It might not be a great reason, but brand loyalty matters, especially when reliability is a factor. Not everyone trusts Seagate drives.

And it's a mixed bag when it comes to a secondary drive anyway...most of the speed comes from having a SSD, so if you don't have to run programs from the mechanical drive, a 5400 RPM model won't be a terrible burden on performance.
Damage wrote:Don't try to game the requirements by posting everywhere, guys, or I'll nuke you from space.

-Probably the best Damage quote ever.
superjawes
Graphmaster Gerbil
Gold subscriber
 
 
Posts: 1085
Joined: Thu May 28, 2009 9:49 am

Re: First Build

Postposted on Mon Mar 03, 2014 4:03 pm

superjawes wrote:
Wicked Mystic wrote:As for Samsung Evo, MLC > TLC. So why pay more for TLC? Have not received answer for this one.

The Samsung EVO 250GB runs for $153 right now. Do you have an MLC suggestion with:

1. Similar Capacity?
2. Similar Price Tag?
3. Similar Reliatbility?

You made the assertion that "TLC is crap," but have not provided an example where an MLC drive offers better value over the Samsung EVO, or even an example where a TLC drive is more expensive than an MLC one.


Some models there:

http://www.newegg.com/Product/Product.a ... 6820178456
http://www.newegg.com/Product/Product.a ... 6820721108

superjawes wrote:The price comparison you gave was not between the same brands. It might not be a great reason, but brand loyalty matters, especially when reliability is a factor. Not everyone trusts Seagate drives.

And it's a mixed bag when it comes to a secondary drive anyway...most of the speed comes from having a SSD, so if you don't have to run programs from the mechanical drive, a 5400 RPM model won't be a terrible burden on performance.


Western Digital don't sell cheap 2TB 7200 RPM drives. Yes, some want WD, some want Seagate, some want Toshiba.

Moving files to 2TB 5400 RPM drive is much slower than moving files to 2TB 7200 RPM drive. I have both and difference is noticable.

How about processor and motherboard?
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: First Build

Postposted on Mon Mar 03, 2014 5:19 pm

Wicked Mystic wrote:As for Samsung Evo, MLC > TLC. So why pay more for TLC? Have not received answer for this one.

MLC is more durable than TLC. Always will be. No argument there. But that doesn't mean TLC SSDs are crap. Perhaps an equally important metric to discuss would be lithography. With new 19nm class NAND out there, more capacity per NAND chip, so 240GB doesn't fill all available lanes of the controller anymore. That's a much more tangible effect on performance that consumers will likely care about.
I could care less about TLC vs MLC. I buy whatever is cheapest, as long as it's from a reputable manufacturer. The rest is just speculation and splitting hairs IMO. Also, if you're storing your "sensitive data" on a mechanical hdd, why does TLC data retention matter anyway?

Wicked Mystic wrote:You can get 7200 RPM drive at same price as 5400 RPM drive, 5400 RPM is fine. Really?

Max data rates for 7200rpm drives (like the 2TB Seagate model you listed earlier) is 210MB/s while the 2TB WD Green is 147MB/s. No doubt faster, but necessary for a storage drive???? Maybe, maybe not. Most external USB3 hdds are 5400rpm (would control transfer speed) and you'd have to have a pretty big and pricey USB3 flash drive to write faster than ~100MB/s to it.
Keep in mind, platter density affects transfer rates also. If you're putting your OS on a mechanical hdd, yeah, 7200rpm without question, but response times will still feel like molasses compared to an SSD*. Drive noise is very difficult to compare since obviously not every mechanical hdd gets reviewed and/or tested for acoustics. As a general rule of thumb though, a slower spindle would create less noise (head seek noise is usually louder than the spindle motor in my experience, but I don't know if that noise relationship is related to spindle speed).

* It sounds to me like the OP will have plenty of space on a 240GB SSD for the OS and all programs....that's generally the goal when selecting SSD size anyway.
i5-3570K, ASRock Z77 Pro4-m, Asus GTX660 TOP, 120 GB Vertex 3 Max IOPS, 2 TB Samsung EcoGreen F4, 8GB G-Skill @1.25V, Silverstone PS07B
DPete27
Gerbil Jedi
Silver subscriber
 
 
Posts: 1656
Joined: Wed Jan 26, 2011 12:50 pm
Location: Madison, Wisconsin

Re: First Build

Postposted on Mon Mar 03, 2014 5:49 pm

DPete27 wrote:MLC is more durable than TLC. Always will be. No argument there. But that doesn't mean TLC SSDs are crap. Perhaps an equally important metric to discuss would be lithography. With new 19nm class NAND out there, more capacity per NAND chip, so 240GB doesn't fill all available lanes of the controller anymore. That's a much more tangible effect on performance that consumers will likely care about.


19nm class will probably be even less durable than predecessors.

There have been many faulty hard drive and also SSD series. Samsung Evo is only one that uses TLC chips. Why take risk with untested technogy?

DPete27 wrote:I could care less about TLC vs MLC. I buy whatever is cheapest, as long as it's from a reputable manufacturer. The rest is just speculation and splitting hairs IMO. Also, if you're storing your "sensitive data" on a mechanical hdd, why does TLC data retention matter anyway?


TLC is cheaper to manufacture and so TLC drives should be cheaper. If not, something is wrong. It's also worth thinking why Samsung is only big manufacturer that sells TLC drives for mid range. If there is nothing wrong with TLC and it's cheaper to manufacture, we already should have seen many TLC drives.

I had MLC drive for about 3 years. Based on host writes I calculated how TLC could handle that. Most probably TLC drive would have weared out. Glad that time only MLC was available on middle class. That matters for me.

DPete27 wrote:Max data rates for 7200rpm drives (like the 2TB Seagate model you listed earlier) is 210MB/s while the 2TB WD Green is 147MB/s. No doubt faster, but necessary for a storage drive???? Maybe, maybe not. Most external USB3 hdds are 5400rpm (would control transfer speed) and you'd have to have a pretty big and pricey USB3 flash drive to write faster than ~100MB/s to it.


What's max data rate? Data rate drive can transfer between SATA cable and itself? That's useless. In practice, 7200 RPM disk can reach sequential transfer speeds about 150MB/s, for 5400 drive it's somewhere around 100 MB/s or even less. Using USB 3.0 you can notice difference between 5400RPM and 7200 RPM drive.

DPete27 wrote:Keep in mind, platter density affects transfer rates also. If you're putting your OS on a mechanical hdd, yeah, 7200rpm without question, but response times will still feel like molasses compared to an SSD*. Drive noise is very difficult to compare since obviously not every mechanical hdd gets reviewed and/or tested for acoustics. As a general rule of thumb though, a slower spindle would create less noise (head seek noise is usually louder than the spindle motor in my experience, but I don't know if that noise relationship is related to spindle speed).

* It sounds to me like the OP will have plenty of space on a 240GB SSD for the OS and all programs....that's generally the goal when selecting SSD size anyway.


OS an programs to SSD of course.

I still cannot see why anybody should buy 5400 RPM. When talking about differences that cannot or can be noticed, 5400RPM HDD vs 7200RPM HDD easilly falls into can be noticed category. At same price point, choice is clear except if OP don't like Seagate.

OP even chose WD Black that is faster than Seagate 2TB basic model. So why discuss about 5400RPM models anyway?
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Mon Mar 03, 2014 9:09 pm

I split this discussion from a new member's build thread in the System Builder's Anonymous forum.
viewtopic.php?f=33&t=93031

Chrispy_ wrote:There seems to be a lot of FUD appearing in these system builder topics as of late (not mentioning any names). Am I alone in thinking this?
DPete27 wrote: I'll go ahead and say it since Wicked Mystic has been called out for this on other threads. If you're going to make an argument (especially one that goes against what others are saying) please provide references.
(@Wicked Mystic) Nobody is trying to personally attack you or try to shut you up. We value everyone's opinion, but we need to make sure to provide proven facts.


Please keep this discussion civil.

--JustAnEngineer, SBA forum moderator
i7-4770K, H70, Gryphon Z87, 16 GiB, R9-290, SSD, 2 HD, Blu-ray, SB ZX, TJ08-E, SS-660XP², 3007WFP+2001FP, RK-9000BR, MX518
JustAnEngineer
Gerbil God
Gold subscriber
 
 
Posts: 15426
Joined: Sat Jan 26, 2002 7:00 pm
Location: The Heart of Dixie

Re: First Build

Postposted on Mon Mar 03, 2014 10:26 pm

Wicked Mystic wrote:19nm class will probably be even less durable than predecessors.

There have been many faulty hard drive and also SSD series. Samsung Evo is only one that uses TLC chips. Why take risk with untested technogy?


I've been saying this for ages (usually to a massive chorus of downvotes).

Samsung claims 1000-3000 write cycles for its 2x nm TCL flash (original 840), yet the TR endurance test drive started wearing out blocks and accumulating errors after only 400 write cycles (average). This is 5 times worse than the median projected endurance! I'm surprised more eyebrows haven't been raised at this.

They also claim the same 1000-3000 write cycle endurance for the 19 nm 840 EVO, which I find even less believable seeing as smaller process usually leads to smaller electron wells and thus earlier denaturing. Granted, the 840 EVO has less write amplification, but I'm talking about cell-level endurance here, which would likely be worse.

While the average user will probably never hit the 100 TB write mark for their SSD, TLC is still a less reliable technology than MLC, and since there are equivalently performing MLC drives from $120-140 for 240-256 GB models, I just don't see why anyone would pay $153 for the 840 EVO.

Remember that when the 840 came out, it was slower *and* less reliable than the 830 that it replaced. This is Samsung being complacent and coasting on the reputation they had built up on past products (which, admittedly, have been excellent).

To this date, no one else has used TLC on their desktop SSDs other than Samsung, not even Toshiba, which fabs TLC NAND for other applications (EDIT: Toshiba also fabs toggle level NAND, which is not to be mistaken with TLC (Tri-level cell) NAND).
Last edited by Voldenuit on Mon Mar 03, 2014 10:37 pm, edited 2 times in total.
Wind, Sand and Stars.
Voldenuit
Minister of Gerbil Affairs
 
Posts: 2454
Joined: Sat Sep 03, 2005 11:10 pm

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Mon Mar 03, 2014 10:31 pm

When I looked at the failure rates of 20nm (or smaller) flash vs 25nm (or larger) flash MLC chips, I "stocked up" on the older technology. As TLC started to appear, I made a mental note to avoid it entirely.

TR's SSD Endurance Experiment articles mostly proved that I was overly cautious. But, in my defense, I really dislike (and if I can avoid it) never have the time for) throwing away dead system drives and reinstalling, then fully (re)setting up, a machine.

In my recent builds, I use SSDs for system/boot drives, and as a place to hold "stuff" where performance in critical. These days, that mostly means virtual machine images, code or documents I'm actively working on, etc.. I don't game (much). For less critical stuff I use 7200RPM drives (eg, WD Blacks). I use Green 5400RPM drives to back the SSDs and Blacks. or to store stuff like music, videos, ISOs, etc., were the streaming rate is plenty.

As a friend of mine (who does big data stuff) says: these days, treat RAM as SSDs, SSD's as mechanicals, and mechanicals as tape drives.
MarkG509
Gerbil First Class
Gold subscriber
 
 
Posts: 126
Joined: Thu Feb 21, 2013 6:51 pm

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Mon Mar 03, 2014 10:53 pm

MarkG509 wrote:As a friend of mine (who does big data stuff) says: these days, treat RAM as SSDs, SSD's as mechanicals, and mechanicals as tape drives.

And tape drives CD/DVD-ROM's, discs as papers, papers as stone tablets?

How much of a retention are we talking about? It really is up to the user. It seems like as we get higher into the performance/cost, "durability" suffers. But then you have to ask, are you really going to back to stone tablets?

BTW:
Wicked Mystic wrote:http://www.newegg.com/Product/Product.aspx?Item=N82E16820178456
http://www.newegg.com/Product/Product.a ... 6820721108

PNY? I am not sure if I am comfortable with these "off brands". The Kingston? Isn't there talk about them doing some bait and switch with some inferior memory chips in there? And that they are using the SandForce controller but not keeping up with the updates? Even Crucial has reported firmware issues that you need to keep on top of. Intel was not immune but they seemed to be the best. If I am running into controller issues this is way more serious than MLC vs TLC.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24422
Joined: Mon May 24, 2004 2:19 am

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Mon Mar 03, 2014 11:37 pm

Flying Fox wrote:papers as stone tablets?
:lol: Actually, yes. Although, paper may be making a come-back given privacy/surveillance issues. Even BitCoin allows/supports paper back-up.
MarkG509
Gerbil First Class
Gold subscriber
 
 
Posts: 126
Joined: Thu Feb 21, 2013 6:51 pm

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 12:20 am

Flying Fox wrote:And tape drives CD/DVD-ROM's, discs as papers, papers as stone tablets?


BD-XL's are pretty awesome at 128 GB/disc; sadly, the prices are crazy high right now ($40 a disc? Ouch). And with the new 'rocklike' recording substrates, I'm hoping they will last longer than the old CD/DVD-Rs that suffered from rot.
Wind, Sand and Stars.
Voldenuit
Minister of Gerbil Affairs
 
Posts: 2454
Joined: Sat Sep 03, 2005 11:10 pm

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 12:26 am

Flying Fox wrote:PNY? I am not sure if I am comfortable with these "off brands". The Kingston? Isn't there talk about them doing some bait and switch with some inferior memory chips in there? And that they are using the SandForce controller but not keeping up with the updates? Even Crucial has reported firmware issues that you need to keep on top of. Intel was not immune but they seemed to be the best. If I am running into controller issues this is way more serious than MLC vs TLC.


Every single manufacturer, including Intel, has had more than one instance of major firmware bugs come to light. However given the majority of controllers have been on the market for a 2-3 years, worrying about firmware bugs shouldn't be a factor unless it's a new unproven controller design.

As for Kingston's V300 they are keeping up with firmware updates. They simply changed their NAND from toggle to asynchronous so performance took a big hit. http://www.anandtech.com/show/7763/an-u ... icron-nand
Kougar
Gerbil Team Leader
 
Posts: 298
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 12:33 am

Kougar wrote:As for Kingston's V300 they are keeping up with firmware updates. They simply changed their NAND from toggle to asynchronous so performance took a big hit. http://www.anandtech.com/show/7763/an-u ... icron-nand

And the "performance hit", would be categorized as "crap" as any broad strokes go.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24422
Joined: Mon May 24, 2004 2:19 am

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 5:33 am

Kougar wrote:
Flying Fox wrote:PNY? I am not sure if I am comfortable with these "off brands". The Kingston? Isn't there talk about them doing some bait and switch with some inferior memory chips in there? And that they are using the SandForce controller but not keeping up with the updates? Even Crucial has reported firmware issues that you need to keep on top of. Intel was not immune but they seemed to be the best. If I am running into controller issues this is way more serious than MLC vs TLC.


Every single manufacturer, including Intel, has had more than one instance of major firmware bugs come to light. However given the majority of controllers have been on the market for a 2-3 years, worrying about firmware bugs shouldn't be a factor unless it's a new unproven controller design.


Yes. Some controllers have been on market so long time that no major bugs are expected to appear. So there are many "SSD manufacturers" who just slap their own sticker on drive.

Kougar wrote:As for Kingston's V300 they are keeping up with firmware updates. They simply changed their NAND from toggle to asynchronous so performance took a big hit. http://www.anandtech.com/show/7763/an-u ... icron-nand


Taiwan models are good, China models suck. However Taiwan models are still available. It's just hard to know what version you will get.

Flying Fox wrote:And the "performance hit", would be categorized as "crap" as any broad strokes go.


Exactly. Just have to buy Taiwan drive and no problems with that one. At least that is still MLC and not untested TLC.
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 7:05 am

Wicked Mystic wrote:That product description lies, Samsung 840 EVO is TLC, not MLC.

A little misleading, perhaps; but definitely not a lie.

SLC stores one bit per cell. MLC (Multi Level Cell) is anything that can store more than one bit per cell. TLC stores 3 bits per cell. So TLC is actually a subset of MLC. The only reason it is potentially misleading is because historically the acronym MLC was used to refer to devices that stored 2 bits per cell, since 3-bit-per-cell devices didn't exist yet!

If we're going to nit-pick, TLC is really a misnomer anyhow as the cells don't have three voltage levels, they have eight. They should really be called TBC (Triple Bit Cell) or ELC (Eight Level Cell). :wink:

Regardless, the quality of the wear leveling and write amplification mitigation algorithms, and the degree of overprovisioning are likely to have a greater effect on longevity than 2-bit vs. 3-bit per cell.

An interesting thought: The physics that determines how quickly the stored charge leaks out of a worn cell should be reasonably well understood. X hours at elevated temperature T should be equivalent to Y days at normal operating temperatures. I'll bet you could simulate an extended data retention test by writing to a drive, then holding it at an elevated (but carefully controlled) temperature for a period of time. Someone with semiconductor design experience just needs to figure out what X and T need to be.
(this space intentionally left blank)
just brew it!
Administrator
Gold subscriber
 
 
Posts: 37705
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 7:34 am

just brew it! wrote:
Wicked Mystic wrote:That product description lies, Samsung 840 EVO is TLC, not MLC.

A little misleading, perhaps; but definitely not a lie.

SLC stores one bit per cell. MLC (Multi Level Cell) is anything that can store more than one bit per cell. TLC stores 3 bits per cell. So TLC is actually a subset of MLC. The only reason it is potentially misleading is because historically the acronym MLC was used to refer to devices that stored 2 bits per cell, since 3-bit-per-cell devices didn't exist yet!

If we're going to nit-pick, TLC is really a misnomer anyhow as the cells don't have three voltage levels, they have eight. They should really be called TBC (Triple Bit Cell) or ELC (Eight Level Cell). :wink:


I guessed someone checks this one. We need more expressions for this, hopefully in the future there is.

just brew it! wrote:An interesting thought: The physics that determines how quickly charges leak out of a worn cell should be reasonably well understood. X hours at elevated temperature T should be equivalent to Y days at normal operating temperatures. I'll bet you could simulate an extended data retention test by writing to a drive, then holding it at an elevated (but carefully controlled) temperature for a period of time.


In theory, yes. In practice, those similations never tell truth. There are countless amount of products that passed every "quick" test. Then after couple of years almost all units have broken down. If it's that easy, why didn't anybody (manufacturer?) test it and predict that those products will broke down after couple of years?

This can be explained many ways. Material impurities, manufacturing tolerances etc so that no two products are exactly the same. In physics we suspect that we are dealing with pure materials and ideal manufacturing process, in practice that's impossible. In the future, maybe.
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 7:37 am

Wicked Mystic wrote:In theory, yes. In practice, those similations never tell truth.

Well sure, they won't be perfect. But they are still better than not doing anything. The drive may still fail a year down the road due to dodgy capacitors or something, but at least you have gained a little confidence that the flash devices can probably do their job.
(this space intentionally left blank)
just brew it!
Administrator
Gold subscriber
 
 
Posts: 37705
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 7:44 am

just brew it! wrote:
Wicked Mystic wrote:In theory, yes. In practice, those similations never tell truth.

Well sure, they won't be perfect. But they are still better than not doing anything. The drive may still fail a year down the road due to dodgy capacitors or something, but at least you have gained a little confidence that the flash devices can probably do their job.


Better than anything, yes. Those quick tests tell if something is very much broken. Long time endurance is totally different thing and cannot be properly tested. For now.
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 9:59 am

Wicked Mystic wrote:What's max data rate? Data rate drive can transfer between SATA cable and itself? That's useless.

I'm aware. I only used that metric because it was all the WD product specs provided. (apples to apples comparison)
i5-3570K, ASRock Z77 Pro4-m, Asus GTX660 TOP, 120 GB Vertex 3 Max IOPS, 2 TB Samsung EcoGreen F4, 8GB G-Skill @1.25V, Silverstone PS07B
DPete27
Gerbil Jedi
Silver subscriber
 
 
Posts: 1656
Joined: Wed Jan 26, 2011 12:50 pm
Location: Madison, Wisconsin

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Tue Mar 04, 2014 11:47 pm

Wicked Mystic wrote:
Kougar wrote:As for Kingston's V300 they are keeping up with firmware updates. They simply changed their NAND from toggle to asynchronous so performance took a big hit. http://www.anandtech.com/show/7763/an-u ... icron-nand


Taiwan models are good, China models suck. However Taiwan models are still available. It's just hard to know what version you will get.

Flying Fox wrote:And the "performance hit", would be categorized as "crap" as any broad strokes go.


Exactly. Just have to buy Taiwan drive and no problems with that one. At least that is still MLC and not untested TLC.


Yeah, well I still consider it to be outright fraud. The new models still advertise the performance of the old models, and lets be honest the stock rating was already stretching the truth. Consumers shouldn't have to worry about which model they're getting, and the average Joe won't ever have a clue anyway unless they run AS SSD or some other program.

I would stop recommending the V300 on principle alone, except that I stopped recommending any SandForce-based drive since almost anything offers either better performance, better pricing, or just better stability. Intel's SandForce models are stable, but at their pricing there are much better performing drives to be had that are just as reliable.
Kougar
Gerbil Team Leader
 
Posts: 298
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: SSD flash: TLC vs. MLC vs. SLC

Postposted on Wed Mar 05, 2014 8:58 am

Kougar wrote:Yeah, well I still consider it to be outright fraud. The new models still advertise the performance of the old models, and lets be honest the stock rating was already stretching the truth. Consumers shouldn't have to worry about which model they're getting, and the average Joe won't ever have a clue anyway unless they run AS SSD or some other program.

I would stop recommending the V300 on principle alone, except that I stopped recommending any SandForce-based drive since almost anything offers either better performance, better pricing, or just better stability. Intel's SandForce models are stable, but at their pricing there are much better performing drives to be had that are just as reliable.


Kingston uses well known tactic. Make one good series, then degrade quality and some people make their opinion based on "good" version. Samsung did same. While 830 was good, 840 is not. Except ultra-expensive Pro-version.

I'm surprised how late this issue is receiving attention. First time I read about that issue six months ago.
Wicked Mystic
Gerbil
 
Posts: 51
Joined: Thu Oct 03, 2013 6:36 am

Next

Return to Storage

Who is online

Users browsing this forum: No registered users and 6 guests