On the test bench this week is Intel’s newest SATA 3Gbps SSD offering, the 320 series 160GB SSD. This is designed to replace the vaunted X25-M SSD. Featuring fast read and write speeds, and vastly improved 4K IOPs specifications, this drive on paper at least, lives up to being a high end SSD designed with SATA 3Gbps in mind. Built using 25nm NAND flash memory the new drive is inherently less expensive and comes in larger capacities than previous SSDs. Why would Intel make a SATA 3Gbps SSD now, when SATA 6Gbps drives are already pushing the maximum bandwidth capable on enthusiast systems today? To really understand the answer to that question, we should take a quick look at Intel as a company.
Intel started out in 1968 as the Integrated Electronics Company, and invented the x86 architecture in 1978 a 32bit processor still in use in almost every home user machine today. Intel is at or near the top in sales of CPUs, Chipsets, Graphics Processors, Network Interface Chips and much, much more. When Intel decided to produce Solid State Drives, they went at it with open arms, producing some of the highest regarded SSDs made.
So why SATA 3Gbps? Intel’s Answer, “Most computers users today are still limited to SATA 3GBps controllers.” As the top producer of CPUs, GPUs and motherboard chipsets, Intel is in an excellent place to know what computers are using today.
This points out that this drive is meant as an upgrade to existing technology. Various sizes are available ranging from 40GB to 600GB in density. Intel SSD 320 prices, based on 1,000-unit quantities, are as follows: 40GB at $89; 80GB at $159; 120GB at $209; 160GB at $289; 300GB at $529 and 600GB at $1,069. PC users that did not want to spend an extra $400+ dollars a year ago on an SSD that was most likely too small as an upgrade when they purchased directly from the manufacturer, can now do it themselves.
Intel even provides tools to help the end-user to make the move to a Solid State drive. Providing both an SSDTOOLBOX, and a data migration tool, which is based on Acronis True Image Home Edition which will only work with an Intel drive installed in your machine.
The Intel 320 series SSD is a third generation SSD produced from Intel. Built using new higher density 25nm NAND flash memory, using the SATA 3Gbps interface it is designed as a more affordable, higher performance and more reliable version of the previous generations X25-M SSD. The 320 series is targeted at all consumer and business options from mobile to server deployments.
As a review sample the product shipped in a large cardboard box, with enough insulation around to ensure, the smaller cardboard box did not get damaged in anyway; Those pictures are not necessary, as they are not retail packaging. The drive itself, is a 7mm design, much thinner than most SSDs of any data density. To make the drive more suitable to all users, they added a small black ring to the top of the drive. This works as a spacer, to increase the height to 9.5mm. For laptop users, this is a “standard” size that is most common and simplifies installation in many hard drive caddies.
Note: The bottom plate is slightly tarnished aluminum, though this should not be indicative of retail sample quality.
Each plate internally, has a thin, clear plastic film to keep the drive from making contact with the aluminum frame. The thinner top portion of the drive is embossed with a arcing line from the back of the drive to one side near the front. This limits flex of the top plate and allows them to use a lighter material.
Opening the drive up, from the top down instead of the bottom up like other drives that use a 9.5mm casing, also designed in, we see the bottom of the PCB. Which holds two NAND flash chips. These 2 chips are 25nm designed in a joint venture with Micron Technologies. The IC number 29F64G08ACME1, is only for the two chips located on the back (top) of the PCB. These are both 25nm 64 Gigabit chips. These chips provide up to 8GB (10%) of extra space for redundancy per chip. This has been seen before on SandForce controllers that usually use 7% over-provisioning cache.
The two NAND devices are 64G in model, which translates to 8 GB single die. These chips are used both for NAND redundancy, to cover for failed locations on the stroage NAND, and additional memory space.
Flipping the board over we can quickly notice the remaining 10 NAND chips. Each of the 10 chips is labeled 29F16B08CCME1. Like the redundancy devices on the back of the PCB, these chips are designed in a joint venture with Micron. These storage chips are 8 GB Dual die NAND memory, for a grand total of 160 GB. Each NAND is based on an 8K page size.
The largest single chip on the PCB is the storage controller itself. PC29AS21BA0 was used in G2 designs of Intel SSDs, and has proven to a be a reliable controller for over 2 years. Changes have been made, such as real-time AES-128 bit encryption and the afore mentioned redundancy feature.
The other 2 main devices embedded are a Hynix Mobile SDRAM controller, and Kemet KO-CAP capacitors. H55S5162BFR-60M identifies the Mobile SDRAM cache, as a 1.8v 512 Mb chip and runs at 166 MHz. The chip is used for caching processes to the controller and does not store the data itself.
The six Kemet Tantalum KO-Caps (T520) are each rated to 6.3V and 470 microfarads +/-10%. In a parallel circuit, total capacitance is the sum of the parts which in this case leads to a .00282 farad rating. The use of heavy duty tantalum polymer caps in parallel over a single larger cap is an earmark towards durability and reliability. A parallel circuit will still maintain some charge in the event that one or more components fail, allowing cached data to be written to the NAND chips. Note: The event of one of the caps failing before a NAND cell is highly unlikely due to the quality of the caps used.
Intel documents that the 320 series drives will have better performance with larger data transfers (particularly 8GB data sizes or 100 minutes in length). This is due to a command called PAGE READ CACHE MODE. By checking registered pages into a buffer, the next page can be moved into place without waiting for the first page to be sent to storage. While using slower individual cycles, the speed kicks in when multiple requests are handled because of the increased data rate, despite the latency loss. It should be pointed out, that performance in writes improves as much as 33% over single threaded requests. This is not the same as increased queue depth so standard testing does not equal the field like on other drives.
In January of this year the IOMeter project at sourceforge has a new release candidate, and will be monitored as a possible benchmark in the near future when it goes live. At this time there is some speculation on the inconsistency of compressible data rates in random, pseudo random and repetitive data, which would in no way affect the Intel drive, but would affect other controllers.
As a compromise CDM 3 x64 was run at standard settings and again with the largest data test size available to test the benefits of the new command structure for paged data.