Micron unveils its first 1TB SSD -- for under $600

[Pardus]

Article

At about 60 cents per gigabyte, the M500 is priced well below average SSD pricing

 

[sub.mesa]

Old news dude? ;-)

The M500 is potentially a very good SSD. But it is not yet available, and might still take some months before you can actually buy them. Firmware issues may determine how long this is going to take. Let's hope they take their time and deliver a good product instead of rushed job.

 

[VirtualLarry]

The big question I have, which wasn't answered by that article was - at that price per GB, is it using TLC NAND? I would think that it would have to be, unless the price per GB of 128Gbit NAND is that much lower.

 

[sub.mesa]

It is using newer 20nm MLC NAND. The Crucial M4 and those generation SSDs use 25nm MLC NAND.

What I want to know is whether they have kept the 8KiB NAND page size which was introduced in the 25nm generation NAND. There is a tendency to make the pages bigger which makes it easier (lesser power/calculations) to achieve sequential performance. But I prefer these pages to be kept at a minimum. Your filesystem assumes you are using 4KiB sector alignment by default. Not many people change this setting manually to 8KiB for the 25nm NAND generation SSDs.

 

[Kippa]

I think Plextor have announced a 768gb SSD and that is TLC based. Personally I'd wait a while see how the TLC drives pan out and see how the last in the long term before committing to buying an TLC drive.

 

[sub.mesa]

If you care about reliability there is only one thing to look at: does it have power-safe capacitor protection, or not. Currently only the Intel 320 and Crucial M500 support this kind of protection in the consumer SSD market. Given that the M500 will be much cheaper than the 320, this could really translate to a popular SSD that is inherently reliable.

All other consumer-grade SSDs are inherently unreliable. Want to know how many times you have rolled in the casino of SSD death?! Well take a look yourself, scan your SSD for its SMART data and look at the raw value of Unexpected Power-Loss Count. It's not zero huh?! You know what that means? Your valuable data was at risk from the jaws of death, but the jaw was already feasting on another prey at the time. You may not be that lucky next time.

 

[Coup27]

If you care about reliability there is only one thing to look at: does it have power-safe capacitor protection, or not. Currently only the Intel 320 and Crucial M500 support this kind of protection in the consumer SSD market. Given that the M500 will be much cheaper than the 320, this could really translate to a popular SSD that is inherently reliable.

All other consumer-grade SSDs are inherently unreliable. Want to know how many times you have rolled in the casino of SSD death?! Well take a look yourself, scan your SSD for its SMART data and look at the raw value of Unexpected Power-Loss Count. It's not zero huh?! You know what that means? Your valuable data was at risk from the jaws of death, but the jaw was already feasting on another prey at the time. You may not be that lucky next time.

You can buy really small UPS's now rated enough for a computer for around $100 if you're that concerned. It will still work out a lot cheaper to buy a consumer 1TB SSD with a small UPS than an enterprise SSD with capacitors.

 

[sub.mesa]

A UPS will not prevent you from unexpected power-loss. This is a misconception. From the perspective of the storage device - the SSD - unexpected power-loss includes all power loss that was not preceded by an ATA STANDBY IMMEDIATE command.

So with UPS you would be protected against real power failures. But actually few of the Unexpected Power-Loss comes from real power failures - most come from software deficiencies. A crashed OS, power saving gone bad, people who disconnect SATA thinking you can simply do that. There have also been bugs in Windows which cause the power to be triggered too early and the ATA STANDBY IMMEDIATE command never had a chance to be actually processed by the host. In other words: a UPS alone is not enough.

See also: http://www.intel.com/content/www/us...-series-power-loss-data-protection-brief.html

 

[Ao1]

^ I agree. Below are the unsafe shutdown counts for the SSD’s I have in my system at the moment. (My 840 Pro does not show the unsafe shutdown counts)


All those events are ATA STANDBY IMMEDIATE rather than power loss. (OK I’ve lost power once that I can recall in the last 2 years)


I haven’t lost data or had a failed SSD due to these events, but I suspect that is pure luck. For sure my next SSD purchase will be based on a SSD that has power caps as SSD performance is already beyond my needs.


X25-M G2 – 517 unsafe shutdown counts (4,550 power on hours)
Plextor M5Pro – 345 unsafe shutdown counts (978 power on hours)
Plextor M5S – 110 unsafe shutdown counts (159 power on hours)

 

[Hellhammer]

What I want to know is whether they have kept the 8KiB NAND page size which was introduced in the 25nm generation NAND. There is a tendency to make the pages bigger which makes it easier (lesser power/calculations) to achieve sequential performance. But I prefer these pages to be kept at a minimum. Your filesystem assumes you are using 4KiB sector alignment by default. Not many people change this setting manually to 8KiB for the 25nm NAND generation SSDs.

20nm 128Gb/die NAND has a page size of 16KiB, 64Gb die still has 8KiB. Increase in page/block sizes reduces the die size as you'll need less internal connectors (I can't remember the fancy word Micron uses for them) per GB, which reduces production costs (i.e. cheaper SSDs for us). Most controllers also have a limitation of how many pages/blocks they can deal with (which is why we need 128Gb die to get more 1TB SSDs), so keeping the page size the same while doubling the capacity per die wouldn't really help.

 

[GlacierFreeze]

I haven’t lost data or had a failed SSD due to these events, but I suspect that is pure luck. ...


X25-M G2 – 517 unsafe shutdown counts (4,550 power on hours)
Plextor M5Pro – 345 unsafe shutdown counts (978 power on hours)
Plextor M5S – 110 unsafe shutdown counts (159 power on hours)

Dunno. Not sure if I'd consider no data loss with a whopping 974 unsafe shutdowns as being lucky. More likely that just about all of those unsafe shutdowns that registered weren't the kind that lead to data loss. Most could be false positives for all we know. Heck, compare the unsafe shutdown count to power on hours. Huge differences. M5S has nearly 1 unsafe shutdown per power on hour. Smells fishy to me.

 

[Ao1]

Just reporting what the SMART data states. There are a number of reasons for an unsafe shutdown….
Intel states:

CO Shutdown Count
This attribute reports the cumulative number of unsafe (unclean) shutdown events over the life of the device. An unsafe shutdown occurs whenever the device is powered off without STANDBY IMMEDIATE being the last command. Use the Raw value for this attribute.

Someone posted some good info here:
http://communities.intel.com/message/132477#132477

 

[Emulex]

you do realize that most drives are in idle most 90% of the time. the crucial m4 is is idle mode except when doing random writes pretty much.

 

[sub.mesa]

Many SSDs perform background garbage collection - also known as aggressive GC. Crucial M4 and Intel 320 are exceptions; these use foreground GC. Foreground means the SSD only is doing something when you send commands to it. Background means it performs autonomous actions.

So when it is 'idle' it may actually be very busy reading and writing. A power-loss at this time could seriously damage your mapping tables. You can check this theory with a power consumption monitor, preferably DC on the SSD alone.