The Inq: Intel Launching Cheaper 34 nm SSDs in Two Weeks Time

[deputc26]

Odd that the product was announced before any reviewers/retailers had samples, I guess you gotta stop leaks somehow (though the inq got this info a week ago).

 

[Idontcare]

Originally posted by: Denithor

Originally posted by: rcpratt
Really the only new news is this TRIM bit:

TRIM Support: Not For 50nm Drives

Part of today?s announcement is the fact that Intel will enable TRIM on these 34nm drives when Windows 7 ships. Intel is planning on releasing a user downloadable firmware update that will enable TRIM support. Windows Vista and XP users will get a performance enhancement tool that presumably will just manually invoke the TRIM command. I suspect that Intel is waiting until Windows 7 to enable TRIM support is to make sure that everything is thoroughly tested. As we?ve seen with other attempts to enable TRIM, it?s a tricky thing to do.

The disappointing part of the announcement is that there?s no TRIM support for the first gen 50nm drives. As far as I can tell, this isn?t a technical limitation of the drives, but rather something Intel is choosing to enable only on the 34nm products.

...making me very happy I decided to sit out this round and for once not be one of the early adopters of new tech.

Maybe I'm just getting old or something.

Nah, I don't think it has anything to do with age. I did the iram, raid0 iram, and ramdisk based SSD gambits which all had far worse $/GB stats than the NAND-based SSD's ever had and even I sat out on all the early rounds of these flash SSD's so far.

Why? Because we have expectations, vast expectations, of the pricing freefall continuing in the flash markets as we have grown accustomed to witnessing in the thumbnail drive segment over nearly the 7yrs prior.

I didn't necessarily need raid-0 iram in my rig but I felt pretty confident gigabyte wasn't about to go through four or five iterations of performance and price improvements in the course of the two years after I bought it, so I took the plunge.

Not so with Intel and OCZ, we know they are hellbent on getting price/performance down by an order of magnitude in the next 2-3 yrs and until then I'm just fine waiting a little bit longer. That's not age talking, that's just expectations of future product enhancements preventing me from buying today's less compelling products.

 

[alcoholbob]

So it's essentially the same drive except it will have a TRIM firmware update. Just as predicted...thanks Intel

 

[Idontcare]

Originally posted by: Astrallite
So it's essentially the same drive except it will have a TRIM firmware update. Just as predicted...thanks Intel

But you cans slamz it into da ground 50% harder and itll still read yurz data just fine!

Seriously though, the higher IOPs for small files is a nice improvement...but like you are saying there is a reason these new 34nm based drives carry the exact same product name as the now older 50nm based drives.

 

[alcoholbob]

Well the lantecy numbers must have been a typo. The 50nm X25-M does not have a 4.2ms latency, that would be obscene. It's been measured by numerous publications and typically been in the .08 range, which would be 80 microseconds. 65 microseconds (the claimed 34nm performance) would be .065ms, which is a .015ms improvement (which is roughly 20%).

 

[Viper GTS]

Straight from AT's preview:

Gen 1 / Gen 2

Flash Read Latency 85 µs 65 µs
Flash Write Latency 115 µs 85 µs

Viper GTS

 

[videogames101]

Those prices are getting close to the point that I may spring for one.

 

[ilkhan]

Yeah, 80GB Intel or a 60GB vertex for the same price. Thats a hard choice.
And low enough to be worthwhile.

 

[shabby]

Originally posted by: ilkhan
Yeah, 80GB Intel or a 60GB vertex for the same price. Thats a hard choice.
And low enough to be worthwhile.

I wonder by how much ocz and others will cut prices on their ssd's, im guessing down to $2/GB at the minimum seeing that the new intel drives are $2.81/GB.

 

[Idontcare]

Originally posted by: shabby

Originally posted by: ilkhan
Yeah, 80GB Intel or a 60GB vertex for the same price. Thats a hard choice.
And low enough to be worthwhile.

I wonder by how much ocz and others will cut prices on their ssd's, im guessing down to $2/GB at the minimum seeing that the new intel drives are $2.81/GB.

It's not clear how much they can cut costs to be honest though. Intel has direct access to their NAND chips via their IM FLash joint venture with Micron, so they don't have to pay markup on the flash chips just to put them in their SSD's.

Unlike OCZ and the others who have to pay Samsung for the flash chips, and samsung doesn't sell them at cost. And contract prices for Flash have been moving higher the past month, now standing right around $2/GB after having bottomed out at almost half that price in Q4/08.

Checkout this EETimes article:

SSD demand stalls as NAND prices jump

Average pricing for 16-Gbit density multi-level cell (MLC) NAND flash rose to $4.10 in the second quarter of 2009, up from $1.80 in the fourth quarter of 2008, according to the research firm.


"The recent increase in NAND flash pricing has benefitted memory chip makers, but also has served as a major damper on the market for SSDs used in notebooks," said Michael Yang, senior analyst for mobile and emerging memories at iSuppli, in a statement. "About 90 percent of an SSD's value consists of NAND flash memory, so with the pricing for such chips rising, consumer and corporate adoption of solid-state storage has been slowed."

Ball is totally in Intel's court here, they got the fabs and the chips and the vertical inhouse integration to create gross margins.

 

[taltamir]

Originally posted by: A5

Originally posted by: taltamir
the intel is rated at 0.065 watts idle for first gen... 0.075 watts idle and 0.15 watts load for second gen. which is unbeleiveably low... kinda sucks that they had to go to a more wasteful controller to meet the halogen free requirement by apple (first one had bromine), but thats life... we all suffer for the "green" delusions of some hippies.

You're right. We're all suffering due to the 10mW increase in idle power of an SSD...:roll:

and price increase, and having spent more time developing that modification, which could have gone into developing a faster / better controller.

 

[taltamir]

Originally posted by: Astrallite
So it's essentially the same drive except it will have a TRIM firmware update. Just as predicted...thanks Intel

no it isn't:
1. The controller is physically different, now no longer being made with bromine. (apple has a strict no Florine and Bromine rule about its hardware)
2. The NAND chips are physically different. 32nm instead of 50nm. Resulting in faster chips at half the cost to produce. Intel rates them the same, but reviews say it is 10 to 20MB/s faster on writes.
3. The firmware has been improved to be more efficient
4. IOPS DOUBLED due to the above changes... nearly tripled in 160GB drive.
5. Cost went from 320$ to 225$ for 80GB (and 410$ for 160GB)
6. Intel PROMISES to add trim to this drive via a firmware update when windows 7 comes out. Currently it does NOT have trim, just like the existing model does not have trim.

 

[alcoholbob]

Originally posted by: taltamir

Originally posted by: Astrallite
So it's essentially the same drive except it will have a TRIM firmware update. Just as predicted...thanks Intel

no it isn't:
1. The controller is physically different, now no longer being made with bromine. (apple has a strict no Florine and Bromine rule about its hardware)

You're right, I guess I won't get cancer from the new drives...

2. The NAND chips are physically different. 32nm instead of 50nm. Resulting in faster chips at half the cost to produce. Intel rates them the same, but reviews say it is 10 to 20MB/s faster on writes.

I understand the second gen Intel drives are 34nm. Did they produce a new product at 32nm? I wasn't aware of any.

3. The firmware has been improved to be more efficient

I see...

4. IOPS DOUBLED due to the above changes... nearly tripled in 160GB drive.

Who stated IOPS doubled much less tripled? Even if you combine average read and write latency the difference is 33% assuming you had an ideal workload that maximized the difference between the two drives. If you are talking about the claimed WRITE IOPS (as you love caps lock very much I will entertain you) it will be interesting what conditions they measured that in given it's disparity with write latency. One or both will be exposed with a Iometer, AS SSD, or HD Tune RA test run when the reviews come out.

5. Cost went from 320$ to 225$ for 80GB (and 410$ for 160GB)

I guess I'll make sure to remind myself that when a product price changes = new SKU

6. Intel PROMISES to add trim to this drive via a firmware update when windows 7 comes out. Currently it does NOT have trim, just like the existing model does not have trim.

Huh?

 

[sharad]

It would be interesting to see what kind of deals there are going to be on the gen 1 X25-M now.

 

[taltamir]

6. Intel PROMISES to add trim to this drive via a firmware update when windows 7 comes out. Currently it does NOT have trim, just like the existing model does not have trim.

Huh?

You said that the only thing that changed was that they added trim. I am saying that you are wrong about that, they didn't ADD trim YET. they promise to add trim in the FUTURE.
EDIT: Actually I misread that, it seems that you did mention the firmware update. Ignore that point than.

Who stated IOPS doubled much less tripled?

http://www.anandtech.com/cpuch...ts/showdoc.aspx?i=3605

 

[taltamir]

isn't it a universal complaint that it is dishonest for a company to change the innards of a device without changing its model number? I know I have complained about it plenty of times. Usually it is grounds for "bait and switch" accusations. Intel is actually changing the model number, this is very honest and good of them.

 

[sharad]

Originally posted by: taltamir
isn't it a universal complaint that it is dishonest for a company to change the innards of a device without changing its model number? I know I have complained about it plenty of times. Usually it is grounds for "bait and switch" accusations. Intel is actually changing the model number, this is very honest and good of them.

The AT article says model number remains the same. Or did you mean to say that Intel is honest because they are not changing the model number but making the disk faster and cheaper? I see plenty of companies doing that. The difference then becomes the release year or the product generation - first gen, second gen, etc.

 

[rcpratt]

It's still the X25-M, but the part number has G2 at the end now. It will be perfectly simple to distinguish between the two.

 

[biostud]

The price seems to be ~33% lower in the shops for the 80gb compared to the old one (EU shops)

 

[Idontcare]

Originally posted by: taltamir
isn't it a universal complaint that it is dishonest for a company to change the innards of a device without changing its model number? I know I have complained about it plenty of times. Usually it is grounds for "bait and switch" accusations. Intel is actually changing the model number, this is very honest and good of them.

It is, sure the model number changed but product label has not and they means there will be a non-zero number of customers who buy one thinking they are getting second gen stuff but they get home and end up with first gen stuff.

However really this is no different then steppings. Everyone lusts after getting the latest and best CPU stepping (D0 anyone?) but the actual SKU/part number (i7 920) remains unchanged leaving customers to do their own legwork to ensure they don't get an outdated stepping at that point.

It is definitely unhelpful product obfuscation, and it only serves to benefit Intel and their distributors clear out inventory of the older generation drives and cpus.

 

[Denithor]

Originally posted by: Viper GTS
Straight from AT's preview:

Gen 1 / Gen 2

Flash Read Latency 85 µs 65 µs
Flash Write Latency 115 µs 85 µs

Is there any kind of diminishing returns associated with this improvement? IE at some point does a reduction in latency no longer result in perceived "faster" performance?

 

[Idontcare]

Originally posted by: Denithor

Originally posted by: Viper GTS
Straight from AT's preview:

Gen 1 / Gen 2

Flash Read Latency 85 µs 65 µs
Flash Write Latency 115 µs 85 µs

Is there any kind of diminishing returns associated with this improvement? IE at some point does a reduction in latency no longer result in perceived "faster" performance?

Yes. I would be quite surprised if anyone will be able to tell from system response and human experience whether a computer has gen1 versus gen2 X-25M under the hood.

For your compute bound workloads that involve small files (the kind of stuff I work with) then it will make a difference in the amount of work my rig gets done hour after hour, but my perception of "fast" is probably saturated once we get beyond Vertex levels of 4k IOPs TBH.

To bring home the point, my ramdisk has read/write latency around 100ns worst case (0.1µs) and the 4k random R/W is around 600MB/s and it doesn't feel any snappier to me than my iram drives which have about 1/10 the random 4K read/write capability.

At this point I think the consumer level SSD is pretty much perfected with the X-25M drives, needs Trim of course and we won't turn down higher bandwidth if its free to us, but what we all really want to see is $/GB come below 1 and head on towards 0.1.

 

[imported_halcyon]

Correct me if I'm wrong:

1. Gen1 will not get TRIM via firmware upgrade (is this official from Intel?)
2. Gen2 is 2-2.5x faster in 4K random write IOPS than Gen1 (see anandtech)
3. Gen1 due to lack of TRIM can slow down at least 60% (in IOPS / write) after being in use (rec PC Perspective's test on this) compared to a fresh drive
4. Combining the above, in real-life use, Gen 2 can be 5x - 6x as fast as G1 in random write IOPS 4K

If all of the above are true, it certainly looks like Gen2 makes a lot of sense purely from the speed/firmware support perspective (TRIM) and not just price.

 

[taltamir]

Originally posted by: Idontcare

Originally posted by: taltamir
isn't it a universal complaint that it is dishonest for a company to change the innards of a device without changing its model number? I know I have complained about it plenty of times. Usually it is grounds for "bait and switch" accusations. Intel is actually changing the model number, this is very honest and good of them.

It is, sure the model number changed but product label has not and they means there will be a non-zero number of customers who buy one thinking they are getting second gen stuff but they get home and end up with first gen stuff.

However really this is no different then steppings. Everyone lusts after getting the latest and best CPU stepping (D0 anyone?) but the actual SKU/part number (i7 920) remains unchanged leaving customers to do their own legwork to ensure they don't get an outdated stepping at that point.

It is definitely unhelpful product obfuscation, and it only serves to benefit Intel and their distributors clear out inventory of the older generation drives and cpus.

only as far as people who do not CARE enough to even do the legwork. I stopped caring about things for people when they don't care themselves.
Today you have to make a willful to choice to be ignorant about such things. Even the most ignorant and tech unsaavy has heard of google.

 

[jatwell]

So, can you buy them anywhere yet?? 🙂