[DuckieHo]

http://www.anandtech.com/show/4271/i...or-use-in-ssds

Quote:

We just started testing SSDs based on IMFT 25nm NAND Flash and look at what Intel/Micron just announced? The first 8GB MLC NAND device built on a 20nm process. This is obviously an announcement of pre-production silicon, it’ll take IMFT until the second half of this year (at least) to start shipping production quality 20nm NAND.

 

[Tennobanzai]

Im guessing this is a good thing since prices will go down but the erase cycles is getting worse and worse. Is this a sign that SSD are just a short-term thing between normal HDD and something more advance?

 

[SgtSpike]

Yay! Even less life and even slower speeds, what's not to love?

 

[Seeing Red]

As predicted the drops in price after each die shrink are becoming smaller. People who are still waiting for $1 per GB will be waiting a little while longer.

Hopefully drive manufacturers tweak their firmware and controllers so we don't get the same misconceptions about the 25nm NAND being less reliable and slower.

 

[DuckieHo]

Quote:

Originally Posted by Tennobanzai Im guessing this is a good thing since prices will go down but the erase cycles is getting worse and worse. Is this a sign that SSD are just a short-term thing between normal HDD and something more advance?

However, the number of PE cycles will not effect most users.

3000 vs 10,000 will not affect home usage.

Quote:

Originally Posted by SgtSpike Yay! Even less life and even slower speeds, what's not to love?

Less life... that you would not make us of.

Slower speeds? NAND is getting faster. Current ONFI 2.2 specs for 200MT/s while ONFI 3.0 specs for 400MT/s per chip.

OCZ made the mistake of using 25nm NAND on the SandForce SF-1200/1500 controller without optimization.

 

[kppanic]

Duckie,

Do we not start seeing leakage through quantum tunneling at around 20nm?

 

[DuckieHo]

Quote:

Originally Posted by kppanic;13124292
Duckie,

Do we not start seeing leakage through quantum tunneling at around 20nm?

With the current technology, the limitation of NAND is around 20nm. However, who knows what they are researching in the labs or do with ECC.

Just a note... Samsung and Hynix have previously announced 20nm NAND but they are just not SSD quality. Changing topic.

 

[L36]

Quote:

Originally Posted by DuckieHo;13124184
However, the number of PE cycles will not effect most users.

3000 vs 10,000 will not affect home usage.

Less life... that you would not make us of.

Slower speeds? NAND is getting faster. Current ONFI 2.2 specs for 200MT/s while ONFI 3.0 specs for 400MT/s per chip.

OCZ made the mistake of using 25nm NAND on the SandForce SF-1200/1500 controller without optimization.

QFT
Even 3000 cycles for an average user is alot.

 

[Zackcy]

Hmm well be getting an SSD soon, good to hear news like this.

 

[Cheetos316]

Quote:

Originally Posted by DuckieHo;13124184
Slower speeds? NAND is getting faster. Current ONFI 2.2 specs for 200MT/s while ONFI 3.0 specs for 400MT/s per chip.

Does that exactly balance out the smaller number of channels that will be needed?

 

[DuckieHo]

Quote:

Originally Posted by Cheetos316;13127448
Does that exactly balance out the smaller number of channels that will be needed?

Not really....

...but then again, SSDs are pretty fast already. SSDs are reaching a point where users won't notice the difference between a Gen 3 vs Gen 4 model. What they and most people do notice is the price though.

 

[Ecchi-BANZAII!!!]

How would one of those SSD's benefit me that re-writes about 150-200GB per month?
Today I deleted almost 400GB just because I needed some extra space and to delete things I no longer need.
Expect it to be filled before the end of next month.
And my OS drive is reading A LOT as well since I save stuff daily to the desktop\New folder folders.

 

[MushkinNick]

Quote:

Originally Posted by Ecchi-BANZAII!!!;13127780
How would one of those SSD's benefit me that re-writes about 150-200GB per month?
Today I deleted almost 400GB just because I needed some extra space and to delete things I no longer need.
Expect it to be filled before the end of next month.
And my OS drive is reading A LOT as well since I save stuff daily to the desktop\New folder folders.

Probably last you as long as you need it to.

If that was daily usage you might see a problem.

Also take into account that the controllers feature all sorts of life-expanding stuff like write amplification, wear leveling, etc...

What others have said is basically the truth. 34nm to 25nm DOES have less PE cycles. But unless you are running a database server, you won't notice (and if you are, you should be using SLC NAND). The average user will see the NAND in their drive stop holding a charge before they use up all the PE cycles on their drive.

Speaking personally, I think the big deal with OCZ was more that they didn't inform the consumers of the change to 25nm more than anything. Eventually, everyone is going to have to go to 25nm as 34nm is pretty much getting too expensive/impossible to get. It's more about how it's done. Informing the user of the correct capacity/performance is important.

 

[SgtSpike]

Quote:

Originally Posted by DuckieHo;13124184
However, the number of PE cycles will not effect most users.

3000 vs 10,000 will not affect home usage.

Less life... that you would not make us of.

Slower speeds? NAND is getting faster. Current ONFI 2.2 specs for 200MT/s while ONFI 3.0 specs for 400MT/s per chip.

OCZ made the mistake of using 25nm NAND on the SandForce SF-1200/1500 controller without optimization.

What if those same optimizations that were made for the 25nm drives were made for the 34nm drives? Would the 34nm drive still outpace the 25nm drive?

If so, then we're still getting an inferior product moving from 34nm to 25nm.
Quote:

Originally Posted by MushkinNick;13127996
Probably last you as long as you need it to.

If that was daily usage you might see a problem.

Also take into account that the controllers feature all sorts of life-expanding stuff like write amplification, wear leveling, etc...

What others have said is basically the truth. 34nm to 25nm DOES have less PE cycles. But unless you are running a database server, you won't notice (and if you are, you should be using SLC NAND). The average user will see the NAND in their drive stop holding a charge before they use up all the PE cycles on their drive.

Speaking personally, I think the big deal with OCZ was more that they didn't inform the consumers of the change to 25nm more than anything. Eventually, everyone is going to have to go to 25nm as 34nm is pretty much getting too expensive/impossible to get. It's more about how it's done. Informing the user of the correct capacity/performance is important.

What if you run a server, but cannot afford an SLC drive? Then 34nm NAND is awesome. And it's being taken away...

Oh well, there's always the used drives market if I end up needing another one down the road.

BTW, what do you expect the actual price savings will be in going from 34nm to 25nm?

EDIT: Oh, and I'm loving my Callisto drives, they're working great for the server!

 

[MushkinNick]

Quote:

Originally Posted by SgtSpike;13128357
What if you run a server, but cannot afford an SLC drive? Then 34nm NAND is awesome. And it's being taken away...

Oh well, there's always the used drives market if I end up needing another one down the road.

BTW, what do you expect the actual price savings will be in going from 34nm to 25nm?

EDIT: Oh, and I'm loving my Callisto drives, they're working great for the server!

Well, with all the provisioning and wear leveling, it may still work just fine for a Server. Just depends on whether that server is doing a ton of writing. Home server? MLC is probably fine.

Enterprise level Database server? Eh...

I wish we could stick to 34nm, but eventually it'll all be gone as everyone that build it moves to 25nm. We don't build the NAND, just the drive. That said, we'll use the best 25nm we can, and obviously make the consumer aware of any changes to density/performance/etc...

I can't speak for the price savings right now. I can't remember all the paperwork I signed when I started here, but I think some of it gives them the right to sell my organs if I release NDA info. Once they become available for sale, I'll let you know.

Glad you are enjoying the Callistos

 

[MIenEsEFOoL]

Quote:

Originally Posted by MushkinNick;13128572
Well, with all the provisioning and wear leveling, it may still work just fine for a Server. Just depends on whether that server is doing a ton of writing. Home server? MLC is probably fine.

Enterprise level Database server? Eh...

I wish we could stick to 34nm, but eventually it'll all be gone as everyone that build it moves to 25nm. We don't build the NAND, just the drive. That said, we'll use the best 25nm we can, and obviously make the consumer aware of any changes to density/performance/etc...

I can't speak for the price savings right now. I can't remember all the paperwork I signed when I started here, but I think some of it gives them the right to sell my organs if I release NDA info. Once they become available for sale, I'll let you know.

Glad you are enjoying the Callistos

Haha, hope they compensate you enough for risking your organs.
Great thread BTW, very interesting and informative read as I read in class.

 

[SgtSpike]

Quote:

Originally Posted by MushkinNick Well, with all the provisioning and wear leveling, it may still work just fine for a Server. Just depends on whether that server is doing a ton of writing. Home server? MLC is probably fine. Enterprise level Database server? Eh... I wish we could stick to 34nm, but eventually it'll all be gone as everyone that build it moves to 25nm. We don't build the NAND, just the drive. That said, we'll use the best 25nm we can, and obviously make the consumer aware of any changes to density/performance/etc... I can't speak for the price savings right now. I can't remember all the paperwork I signed when I started here, but I think some of it gives them the right to sell my organs if I release NDA info. Once they become available for sale, I'll let you know. Glad you are enjoying the Callistos

Yeah, hard saying. It says that my Vertex has gone through 8000 write/erase cycles, but then shows 100% SSD life left in the SMART data, so I'm not sure what to believe anymore. I guess I'll just use them 'til they die and do the analysis then.

Also, I'm glad you'll be letting me know when your organs are up for sale! I could really use another kidney...

 

[Krusher33]

Yay for technology!

 

[Hephasteus]

I still don't get the packaging. Why do they have to keep putting them in aluminum boxes and using weird interfaces.

http://www.fusionio.com/products/

 

[MushkinNick]

Quote:

Originally Posted by SgtSpike;13131441
Yeah, hard saying. It says that my Vertex has gone through 8000 write/erase cycles, but then shows 100% SSD life left in the SMART data, so I'm not sure what to believe anymore. I guess I'll just use them 'til they die and do the analysis then.

Also, I'm glad you'll be letting me know when your organs are up for sale! I could really use another kidney...

The P/E rating is per-cell. So it may be that it's only gone through 8,000.

Hmmm, I could use some parts for my motorcycle.... and it's not like I don't already have two kidneys...