[radix]

Originally Posted by Maestro

radix,

Did not want to try and respond to each of your points in quotes, however it is obvious you are familiar with the technology, I am too. I Manage mass storage technology for a large company and I use to working as an engineer in a very well known computer company doing drive qualifications as one of my jobs.

No SSD today is using NAND flash with 10K write and erase cycles those were based on 56nm technology which were taken off line last years by all the manufactures Toshiba being the last. Today they are all shipping 34nm technology and it has 3K cycles, and they are moving to 2Xnm technology next year it will have 1.5K.

You mention the eMLC as having 30K, which it true but the way they are achieving this is by slowing the device down, there is more to it than that but that is the basic approach which mean the it will not preform as well.

Another important point with SSD, they should not be spec on write and erase cycles which they had avoided talking about but will share if asked, they should be spec on total GB transfer and still maintain specifications. I been asking for this for years and they are now specing the SSD this way and you would be surprised. Basically, you can transfer so much data and then the device starts to drop in performance or it is wore out can not be use, not the case with HDDs time and POH is the limiting factor on HDD but not how many bits you can write.

Imagine if every car on the road stopped working at 100K miles not to be fix or usable, that is what you get with SSD.

Also the links you list are Intel propaganda and I will it at that.

Your correct HDD have blocks that fail, and the HDD moves data around and spare the blocks out and some times you get an unrecoverable error, just the fact of life with HDDs, but HDD do not wear out like SSD do. You can write a block on an HDD millions of times and it will not degrade just because of the bits being flipped.

What I mean by erasing data is the SSD as part of their wear leveling algorithms move data around to all the cells and every time it changes the data or flips the bit in a cell it is a write and erase cycle. Also as the OS creates and deletes scratch and temporary file this adds to this problem.

Now on performance, I personally have not see any data from any manufacture demonstrating the performance of their SDD over time, as it is worn, as it is filled to capacity. I have asked for it from Intel, Toshiba, SanDisk, Samsung, and Micron, oh by the way Intel uses Micron Flash and none of them can backup their claims other than how it performs new out of the box.

NAND Flash is dying as an industry, unless they can solve the data reliability problem, it will continue to useful in products like MP3s but any device that has what is called data turnover can not use it beyond the current technologies.

nTo all SSD are the same, there are only 4 companies who make Flash and 50 companies making SSD, some have a clue and most have no idea what they are doing. Personally, buying SSD from companies like WD, Seagate or Toshiba is good since they understand the entire storage industry and understand data integrity issue due to their 30 years of HDD experience. Everyone else is just trying to sell you a product.

I am not saying SSD are bad, I am saying it depends on the application and that most consumers have no clue about it. They see and hear all the great things about SSDs, but much of it is based on SLC technologies, drives that use a combination of SLC and MLC technologies or also incorporate SDRAM. Yes you can get 64GB SDD in the $150 price range but it does not have the performance you think, you are going to be spending more like $500 to $1000 for high performing SSD that can keep up with a server class drive you showed.

I personally rather use a HDD with a back up than spend 3X on a SSD and then buy a HDD to back up still.

Even at 3000 write cycles, it's still not that big an issue, here's how the math works for a 160GB drive:

lark ~ % bc -l
bc 1.06
Copyright 1991-1994, 1997, 1998, 2000 Free Software Foundation, Inc.
This is free software with ABSOLUTELY NO WARRANTY.
For details type `warranty'.
160 * 3000 / 1.5
320000.00000000000000000000
320000 / 40
8000.00000000000000000000
8000.00000000000000000000 / 365
21.91780821917808219178

The forumla was gleaned from this article:

http://techon.nikkeibp.co.jp/article...090528/170920/

AFAICT, the 1.5 number is probably the product of coefficients for both wear leveling and write amplification, if I had to guess. Nevertheless, that's 22 years for a drive with a write endurance of 3k per cell for a user that writes 40GBs of data per day.

http://techreport.com/articles.x/15433

If you don't want to crunch through the math, Intel estimates that the 80GB X25-M will last for five years with "much greater than" 100GB of write-erase per day. That's a relatively long time for much more data than most folks are likely to write or erase on a daily basis.

The drive in question has a 55nm process size however. Still, most users don't write half that much, and this still doesn't even account for over provisioning. Most companies that use flash use SLC drives, because they can afford to. MLC, even at 25nm is probably fine for consumers.