Intel G3 SSD Roadmap

[scooterlibby]

my steam folder is almost that big 🙁

Hehe mine would be too if I downloaded and played all my Steam games regularly, but I usually latch on to one for awhile and have maybe that and one more game installed at a given moment. 300gb would be perfect for me.

 

[pchemist]

I think Intel is going in the wrong direction with this.

What I think a lot of people wanted was lower priced SSDs, not larger SSDs.

80GBs is plenty for a boot drive, I just need it to be fast (over 200MB/s read/write) and cheap (below $150).

I'd like to offer a different view:

(1) A large part of the consumer SSD market is for laptops, which typically have space for only one HD/SSD (you can go HD+SSD only if you get rid of the optical drive, which typically voids the warranty). Sure, you could put a boot drive in your laptop and carry around an external HD for bulk storage, but that's not a clean solution. So laptop owners like myself need a single drive large enough for everything on their laptop (which, for me, means a 240 GB drive).

(2) Mainstream consumers, even if they are desktop owners, are not going to want to use the boot drive configuration you suggest, simply because, to them, it sounds technical and messy--i.e., they're going to want a single, large SSD to replace their HD. And by around 2013, likely large capacity SSDs will be available at prices that make them relatively common. If Intel were to abandon moving towards large capacity SSDs, they would be missing out on that market. Besides, if mainstream PC owners were actually willing to implement the dual-driive configuration you suggest, they'd be doing it already (newegg has a 60 GB OCZ Vertex 2-- plenty of space for an OS + core apps-- for $126 after rebate). So in my view, what's preventing further market penetration of SSDs is not that small drives need to be even cheaper, but that the large drives -- the ones consumers really want -- aren't yet available at market-friendly prices. And the more effort Intel and others put into manufacturing and marketing these large drives, the sooner their prices will come down.

 

[frostedflakes]

Anyone have more info on "enterprise grade" MLC? Kind of surprising that they're replacing the SLC in the X25-E drives with MLC. How is it supposed to be more durable and reliable than standard MLC? Are they just binning their regular MLC chips based on quality, or is there some process change to make it better than standard 25nm MLC?

 

[capeconsultant]

Personally I think dropping real SLC chips will come back and snake bite them but good. Hope I am wrong.

 

[sub.mesa]

The enterprise MLC NAND is rated at 30k write cycles and i think that's enough. For it is not the number of write cycles per cell is important; but the total write endurance, which depends on:
- number of write cycles per cell
- efficiency of wear leveling algorithms
- total number of write cells

So you could say "this drive has a write endurance of 2 Terabytes (2000 GB)". An SSD with less write cycles but more write cells can have a higher write endurance than a smaller drive with more write cycles per cell. If you RAID several together, your write endurance will also increase. Furthermore failures on write cycles are predictable and that's good from a company perspective; so this shouldn't be a big issue for most of todays users.

Having both HDDs for mass storage and SSD as system disk would likely be used for awhile; as SSDs are only marginally better in handling large files than HDDs are but HDDs suck for any random I/O. So combining both will give you the best of both world and don't require you to spend alot of money on one expensive large SSD. SSDs also make poor backup devices; so they are not replacing HDDs just yet. Just the segment that requires IOps performance, which HDDs will *never* be able to deliver.

 

[capeconsultant]

For some reason, I have always thought of SLC as simply more reliable. Yes, more write cycles, but maybe I should use the term robust instead of reliable. To me, write cycles is one thing and reliability/robustness is another. Similar to the way spinners are more highly engineered for server environments. Not just longer lasting, but better able to stand stop/start cycles, heat ranges, etc.

I am no scientist, just an enthusiast that has always appreciated hard drives and been willing to pony up for better drives.

Funnily enough, I have never had a catastrophic HD failure. I like backups too 🙂

 

[pchemist]

Anyone have any further info. on NordicHardware's much-discussed news bit that Intel is delaying release of the G3 SSDs from Q4 2010 to February 2011 (http://www.nordichardware.com/news/86-storage/41173-intel-postpone-25nm-ssds-to-february-2011.html)? I've not seen this verified on any other site. I do recall seeing one poster (different blog, perhaps different site) say that he had written Anand about this, and Anand replied that this was different from the info. he had, but that he was trying to get confirmation or denial from Intel (of course, this is second-hand info.).

If Intel is indeed postponing release to Q1 2011, I'm not going to wait until then for their 300 GB G3, and will instead buy a 240GB OWC Mercury Extreme. Thanks!

 

[tweakboy]

That 600GB version will cost 500 dollars .. oh boy...

 

[frostedflakes]

Where did you read $500? If true, that's actually really good. Would bring SSD to under $1/GB, $0.83/GB to be exact. I really doubt the 600GB would be that cheap, though.

 

[StinkyPinky]

X25-V for my lower end laptop may be good. Give it a nice boost without hopefully breaking the bank.

 

[Mark R]

Anyone have more info on "enterprise grade" MLC? Kind of surprising that they're replacing the SLC in the X25-E drives with MLC. How is it supposed to be more durable and reliable than standard MLC? Are they just binning their regular MLC chips based on quality, or is there some process change to make it better than standard 25nm MLC?

They are basically the same dies as standard MLC but specified and tested differently:

Emlc is downclocked and has longer write timings than regular MLC. The slower write speed is most important as it allows more precise programming.

The other specifications are also changed. Regular MLC is speced for 10 years data storage time. EMLC is speced for 6 months. As data slowly corrupts with time, by requiring an EMLC controller to regularly reprogram the data, you can improve data integrity enough to allow 2x - 3x as many write cycles.

The other part of the equation is much stricter binning of the dies. Essentially, they choose the best bins for data integrity and fewest bad blocks and use those for EMLC.

 

[frostedflakes]

Very interesting, thanks for the explanation Mark. 🙂

 

[Makaveli]

Where did you read $500? If true, that's actually really good. Would bring SSD to under $1/GB, $0.83/GB to be exact. I really doubt the 600GB would be that cheap, though.

I agree with you that is way to cheap, in Dec 2009 I paid $450 for my 160GB drive.

No less than $800 for that much capacity.

 

[woolfe9999]

Here is Anand's preview of Intel's upcoming SSD release:

http://www.anandtech.com/show/3965/intels-3rd-generation-x25m-ssd-specs-revealed

Internally it’s called the Postville Refresh (the X25-M G2 carried the Postville codename), but externally it carries the same X25-M brand we’ve seen since 2008. The new drive uses 25nm IMFT Flash, which means we should get roughly twice the capacity at the same price. While Intel is sampling 25nm MLC NAND today it's unclear whether or not we'll see drives available this year. I've heard that there's still a lot of tuning that needs to be done on the 25nm process before we get to production quality NAND. The third generation drives will be available somewhere in the Q4 2010 - Q1 2011 timeframe in capacities ranging from 40GB (X25-V) all the way up to 600GB.

The bolded part is of greatest interest.

- wolf

 

[Campy]

Yes that's some very nice info wolfe, i'm really hoping the pricedrops are that big and this gives me hope of having a 240GB SSD early next year.

 

[Lee Saxon]

I would be really surprised if price/GB was actually halved. The actual memory isn't 100% of the price. There's the controller, development costs, support costs, blah blah blah...

 

[flamenko]

And don't forget the Intel premium for being the first to release. I am very curious simply to see what they are charging for the two of the same capacity, 80 and 160.

lets say they drop down to a $1.50/Gb price point with everything costed in... We will then still here from people saying that $450 is still too much for a 300Gb drive.

All in all....my bet is on a $1.80/GB pricepoint.

 

[HendrixFan]

All in all....my bet is on a $1.80/GB pricepoint.

I hope it goes lower than that. You can pick up a Vertex 2 120GB drive for $210AR, so right around $1.80. There was a hot deal on that same Vertex 2 drive a week or two back for $180AR which put it at $1.50/GB.

Since the stats on this G3 drive from Intel seem to point at roughly the same performance as the Vertex 2, I really hope we see something pushing more towards $1.50/GB.

 

[LokutusofBorg]

Since the stats on this G3 drive from Intel seem to point at roughly the same performance as the Vertex 2, I really hope we see something pushing more towards $1.50/GB.

For that very reason, they're going to *have* to use price as their marketing push, unless they pull out some performance improvements before release.

 

[Emulex]

kind of like AMD versus INTEL - in the business world it doesn't matter wtf AMD sells; folks are buying intel like no tomorrow. Why? you won't get fired for buying Intel..

even if intel isn't the best situation (see all the netburst machines still in offices?)

 

[Makaveli]

Here is Anand's preview of Intel's upcoming SSD release:

http://www.anandtech.com/show/3965/intels-3rd-generation-x25m-ssd-specs-revealed



The bolded part is of greatest interest.

- wolf

That statement doesn't mean much.

It mearly says they should be the same price Anand himself doesn't know what the actual pricing will be yet.

For all you know it will be the same thing as the Radeon 5000 series which stayed well above MSRP.

What they say and what actually happends once retailers get the products are two different things.

 

[Voo]

It mearly says they should be the same price Anand himself doesn't know what the actual pricing will be yet.

For all you know it will be the same thing as the Radeon 5000 series which stayed well above MSRP.

What they say and what actually happends once retailers get the products are two different things.

Actually it sounds like Anand is basing that claim on the technical facts (i.e. 25 vs 34nm flash), so the question is how much of the savings Intel want to give on to their consumers.

If the controller is really only a revised Postville, they should better drop the prices accordingly because the performance increase probably won't be that great. But considering Intels large ressources a "just as good or marginally better than a SF controller" would be a big letdown in my book :/
But that's all speculation at this point.. can't wait for benchmarks (including the rumoured 25nm flash chips from SF)

 

[Makaveli]

I'm looking forward to seeing those benchmarks!

I'm hoping for another Firmware release for the current G2 drives when the 120GB G2 drive get released in nov or dec.

 

[flamenko]

I am curious why all are not mentioning the obvious which is going to be an even shorter write endurance period for the drives. It is going to mean that the present 80Gb X25 will have a longer life span than the new X25 with 25nm NAND. I am very curious on how much less but consider this...

When we went from slc to 2 bit mlc we reduced our write endurance from 100000 to 10000 writes per cell which means it was only 10% of the original. Now we are going from 2 to 3 bit which is going to reduce it even more.

Now, for those that know a bit about the industry, this will be overcome with the larger 300Gb and 600Gb drives because larger capacity means less total writes to the entire SSD. Its actually a simple calculation that means that larger capacities will last longer but....

What about the 80 and 160Gb SSD in 25nm NAND being released? I am very curious what their total lifespan prediction will be in comparison to the present. This would also bring a thought about something else not mentioned which is the size of the over provisioning that Intel will have in these drives.

Just a thought.

 

[frostedflakes]

I am curious why all are not mentioning the obvious which is going to be an even shorter write endurance period for the drives. It is going to mean that the present 80Gb X25 will have a longer life span than the new X25 with 25nm NAND. I am very curious on how much less but consider this...

When we went from slc to 2 bit mlc we reduced our write endurance from 100000 to 10000 writes per cell which means it was only 10% of the original. Now we are going from 2 to 3 bit which is going to reduce it even more.

Now, for those that know a bit about the industry, this will be overcome with the larger 300Gb and 600Gb drives because larger capacity means less total writes to the entire SSD. Its actually a simple calculation that means that larger capacities will last longer but....

What about the 80 and 160Gb SSD in 25nm NAND being released? I am very curious what their total lifespan prediction will be in comparison to the present. This would also bring a thought about something else not mentioned which is the size of the over provisioning that Intel will have in these drives.

Just a thought.

Is there even going to be an 80GB G3? It sounds like the price points are going to stay about the same, and capacity is just going to increase (for example, 80GB will be dropped and 160GB will replace it at about the same price, 300GB will replace 160GB, etc.). If so, the larger capacities will help, because that means more flash to spread wear over. Everything else being equal, a 160GB drive should have about twice the write endurance of an 80GB drive. So depending on how much you lose in terms of write endurance from the 34nm->25nm shrink, the new 160GB SSDs based off 25nm MLC could have higher write endurance than the 80GB G2.

Also IMFT's flash is still 2 bit per cell right? I thought the 3 bit per cell flash was meant more for USB flash drives and stuff like that, not high performance applications like SSD.