The new SSD Thread (Benchmarks, Brands, News and Advice)

Yep, basicly more Indilinx drives on the market.

 

As I said, it's just my opinion and I won't argue, but must say that files are one that can be fragmented and defragmented, not disk (HDD or SSD).

 

I said so. I said explicitly I only defragmented files, not the disks as a whole. And that I used defraggler for this, as it can do that very simple.

But from numbers, my disk was hell and at some point, even the most tiny added latencies and additional work for the system may sum up.

 

Sure that's a possibility. It's typically because you have services starting at boot time (like antivirus) that need network access to update themselves before they will allow booting to continue. Wifi routers tend to be slow in terms of authenticating clients, so that can also add some delay. Solution - switch to Linux / FreeBSD, OSes that don't need to screw around with BS like that.

 

No, not entirely useless. Even if physical fragmentation is meaningless, the problem is that it still has an impact at the logical layer. The ATA command set "streams" by saying "read X blocks starting at address Y" - but you can only stream this way if the data is logically contiguous. If you have a thousand blocks all fragmented across a thousand different logical block addresses, then you need to issue 1000 commands to read them, instead of just one command with a length of 1000. And if your drive doesn't support command queueing, or the queue is too small, then the command overhead can easily make this orders of magnitude slower than just issuing a single ATA command.

 

Does defragging on an ssd specify which blocks to write to? Wouldn't that be counteracting on the write wear leveling? I would think that the write wear leveling would decide where to write to and defragging wouldn't really have much effect.

 

Well played my friend! Well played! That is a good point!

 

You've completely missed the point. Defragging certainly has no effect at the physical level because of wear-leveling. But fragmentation still has a performance impact due to logical block address concerns.

 

They've turned up in Germany already

http://www.computerbase.de/preisvergleich/a444063.html

Also Norway, from what i read on xtremesystems...

In other news, I just got an email from Dell saying my order has been delayed another week. . Maybe I should cancel it and see what the market looks like by then.

 

Darn, I don't see any specifications...

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16820233087

$285 is a great price.

 

Exactly. Logical file fragmentation results in the issuance of multiple I/O requests by the filesystem to read the same file since each fragment requires a separate I/O request. This easily impacts overall drive response, more so for magnetic HDDs and less for SSDs.

HDDs have higher latencies relating to seek/rotation/settle-time so the effect is more pronounced when the multiple I/O requests end up as separate seeks for each 'fragment' of the file at the physical block level. SSDs on the other hand do not suffer as much for sequential read/writes or random reads, but suffer greatly for random writes- too many I/O requests and the time required for erase+write for entire physical blocks (for each request) slows things down.

So SSDs do benefit if the I/O is sequential rather than random (writes). A properly defragged filesystem with the free space consolidated, improves the chances of a fresh write being sequential rather than random. Note: it's not a guarantee but an increase in the probability. If the free space is consolidated, then the probability of physical blocks, corresponding to those logical blocks, being 'empty' is higher, and this speeds up the write.

 

yup, that's what i said. and that's why i hope to one day be able to drop the logical layer and just "feed the ssd directly". database CRUD style. so the ssd can do the wearleveling on itself, and is not "blind" like it is right now (espencially without TRIM).

that way, there would be 0 write amplification, finally 0 fragmentation, and a very simple interface.

one day, maybe..


anyways, looking at the intel numbers from that price-comparison:

732,7761 US$ for the current 160GB SSD
536,5220 US$ for the next 160GB SSD

so the new one costs around 73% of the old one.

in europe, prices are always higher.. if we reproject that onto the current ebay prices, that would mean the 160GB version could be available soon for 365$ (and the 80gb for 229$)

that would be *yadda yadda*

 

A ~25% price drop is nice, that should definitely put some pressure on the other vendors. But it's still over the $2/GB mark, and I expect the Indilinx will stay above as well. So I guess I won't cancel my Dell order just yet.

 

no need to cancel anyways yep

lets wait for the 320gb price..

if the price growth is linear with size, then the 320gb one would be at 637$ hehe. that would be a good price

btw, i don't believe that myself but it would rock

 

What I know about Intel 34nm SSD "Postville"

Consumer PC/Mobile Segment

-Halogen Free
-32MB wear levelling buffer
-90MB/s sequential write
-AES 128-bit Encryption
-Advanced NCQ Features: Enhanced performance through status aggregation
-Advanced Smart Support: Improved drive statistics to monitor drive life

Workstation/Server

-Power Safe Write Cache
-All other features similar to Consumer/Mobile Segment(except sequential speeds of course)

For 320GB pricing, I assume it'll be slightly less than 2x the price of the 160GB one as the current generation pricing is.

 

70 to 90... not that much of a jump. I was hoping for at least 100, if not closer to the X25-E's 170... the X25-E should go to 250 easily.

 

Price is lower than 1rst gen drives: $276.01
from:
http://www.nextwarehouse.com/item/?846017_INTEL_SSDSA2MH080G2C1

 

I'll bet retailers in the USA are trying to unload their current stock at the higher prices before listing these newer, faster, cheaper drives and having to eat whatever leftover stock there is...

 

So is it wise to order now intel postville, or I should wait?

 

Give it a week to see what Intel is going to actually bring out (and what the retailers are going to charge us) and then order. You might can get a really good deal on Intel's last gen SSD around the same time.

 

either a 160gb or a 320gb.......decisions decisions. LOL

 

So far prices in Europe for intel postville:
80 Gb ~200 euro
160 Gb ~ 385 euro

I hope it will not get higher.

 

$261 for 80GB from a University computer store in the USA:

http://www.uarkcomstore.com/cart/look.php?item_id=5915313

I'm assuming that the 2 prices found so far ($276 / $261) are with virtually no discounting. I think we can reasonably plan on $250 as the street price and perhaps as low as $225 in the very near future.

 

and then we can wait for ebay in the very near future, too can't wait for i-hope next week

 

Sweet. Sounds good. I was hoping for more of a price drop than that. Like, an affordable 160GB, but I can settle with an 80GB as long as I move all my movies to an external drive.

 

*hugs my ikea home server, loving non-local-storage*



yep, my media center + my music pc most likely will have both a 80gb intel ssd in the end.

what do i do with my mtron 64gb, then? they're not 2.5" btw..

oh right, my dad wants one, i guess..

a friend of mine would like the other one, i think

we'll see......

 

In Greece, 1st gen 80GB is 286 and you can preorder the equal capacity postville for 239. So it's hardly 16-17% cheaper.

 

Welcome mate

Greek market is a "strange" one.

 

Thank's. I waited for the new ones for my fujitsu crap-top whose disk died but I got tired and bought the smallest 7200 WD Black instead and I think I'll wait till the autumn.

 

Intel 80GB Gen 2 - $246 USD

http://www.fadfusion.com/selection.php?product_item_number=10068600248

Intel 160GB Gen 2 - $486 USD

http://www.fadfusion.com/selection.php?product_item_number=10068600249

edit: I posted this in my other thread but since this is "THE" SSD thread, I decided to include it here - some info on Intel drives:

http://www.intel.com/cd/channel/reseller/asmo-na/eng/products/nand/feature/index.htm

 

interesting.

 

Well, drat.

 

Never heard of Fad Fusion.

Looks a little sketchy... Anyone here order from them before?

 

The MSRP on fadfusion is listed as $265. Good to know.

So some discounted estimates:
10% off = $239
15% off = $225
20% off = $212

I'd guess that places like newegg and amazon will start out with a street price of 10% off. Looking good! This makes postville cheaper $/GB than the Indilinx-based drives. I wonder how OCZ, SuperTalent, Transcend, etc will respond?

 

engadget has reported it.....how long more before it becomes a reality

 

Indilinx will have to drop their prices to be competitive or else everyone will go for Intel or Samsung - which is good for us A couple months ago the Intel X25-M was mid $200s, hopefully that means we'll see this gen dip to mid $100s! I might be dreaming now, but hopefully it doesn't remain a dream for long...

 

How much do you think the cost of the Indlinx controller is compared to the Intel? I'm curious to know what they charge for the controllers and if that is the big difference. I was under the assumption that the NAND flash memory made up the largest cost, and that the cost between the Indilinx controllers and the jmicron's are less than $10. What about the cache also, how much does that factor into the cost of Indlinx SSD's?

 

The main cost is the NAND memory. Intel is using a newer, lower fab process. Whoever uses Indilinx controllers will have to sell for less profit or switch their NAND manufacturer to really lower their cost of production. I think most companies use Samsung for their NAND memory.

 

The last couple of days are going to be the hardest part. Well, that and waiting for whenever the drives will actually start shipping.

So do we think there will be reliability improvements with the 2nd gen models?

 

Usually the controllers were the majority of the costs werent they? The nand didn't catch up in the latest round of SSD's until the MLC flash went up in price (SLC unfortunately still has very expensive memory )

Samsung makes good NAND, but if Intel's 32nm tech pays off they might start supplying to ALOT of the current SSD makers. Thats going to give us nice lower prices all around... and make Intel a ton of money

Plus the people already putting out 256GB drives now can offer decently priced 512GB drives to the market... and that rocks. Then SSD's can be used as storage as well as OS/APP drives, that makes me happy

There weren't really any reliability in the 1st gen Intel's. They had the crippling slow down, but Intel fixed that almost instantly after it was discovered (with in a week if i remember correctly)

All the 2nd gen are giving us is better capacity, slightly better speeds (assuming they are using the same controllers) and better price .

 

That's what I thought about the cost.
I also read about certain controllers only using certain NAND, is that a true limitation or is it possible to mix up Intel NAND + Indilinx contrller?

 

Correct me if I'm wrong, but I believe Indilinx controllers can only use Samsung NAND. And of course, the Samsung controllers can only use Samsung NAND. Current JMicron controllers can only use Samsung NAND also. So if I understand correctly, these MFRs don't have the option of just using a lower-cost NAND - they are stuck with Samsung as this is a limitation of the controllers. However, JMicron's upcoming controller (JMF612) is supposed to be able to use *any* MFR's NAND - which will bring prices down considerably as all sorts of 3rd-tier SSD manufacturers can enter the market with products that use 3rd-tier NAND.

Also, I believe the cost of the controllers are about $10 for JMicron and $15 for Indilinx. I don't know about the Samsung controller but I suspect it is about the same. Since both JMicron-based and Indilinx-based drives use Samsung NAND, the difference in cost is mostly due to markup. In other words, Indilinx-based drives (like the Vertex) should be able to drop their price down to JMicron levels if they want to. They will have to drop their price a lot in response to postville. At around $250 for 80GB, it will take a lot of coaxing to get me to consider another SSD.

 

160g $472:
http://www.compsource.com/pn/SSDSA2MH160G2C1/Intel_211/

80g $239:
http://www.compsource.com/pn/SSDSA2MH080G2C1/Intel_211/

So where are the 320giggers!!!???

 

Lowest price listed now of $239? I wonder what the big retailers like Amazon and Newegg will price it at...

First benchmarks for postville are up:

http://translate.google.com/transla...hives/004532.html&sl=ja&tl=en&history_state0=

Unfortunately the benchmarks are using a pre-production firmware and are not impressive. I am confident that subsequent benchmarks for retail versions will show an improvement from Intel's G1 drives...

EDIT:

WOW...just WOW...

I just found a link that states (translated from japanese):

 

Those speeds are TERRIBLE compared to the G1!

 

OK, so that Japanese blog is claiming that:

4kb random write 50nm (gen1): 3,300 IOPS
4kb random write 34nm (gen2): 6,600 IOPS (80GB model)
4kb random write 34nm (gen2): 8,600 IOPS (160GB model)

In brief, postville doubles random write speeds. Note that the above specs are with a very high queue depth of 32 and an 8GB section of the disk (as per Intel's spec sheet). Such a high queue depth makes it very difficult to compare this specification with other drives because they usually test at a queue depth of only ONE to inflate the benchmark results.

So to compare, Anandtech has published (most of) the following IOPS results using IOMeter (8GB, Queue depth = 3, 3 minutes run time):

OCZ Solid (JMicron): 6
OCZ Apex (2 x JMicron): 8
Kingston V-Series (JMicron/Toshiba): 20
5400rpm HDD: ~50
7200rpm HDD: ~80
OCZ Vertex (Indilinx): 1,700
Intel G1 80GB: 6,000
Intel G2 80GB (estimated*): 12,000

*To compare with Anandtech's previous results, we can estimate that Intel's G2 80GB (postville) will be twice as fast (based on the japanese blog post).

 

Exactly. I'm sure we will see Intel's 80GB for <$200 soon, thus cutting into Indilinx based drives profits further...

 

you know what? it's sexy!!

 

I'll agree with you there. I've always liked the minimalistic look