Date:
Tuesday , January 13, 2009
Author:
Kyle Bennett
Google +1

Ask an Intel Solid State Drive Engineer

Intel recently offered our readers the opportunity to ask one of its engineers questions about Intel Solid State drives. You guys obliged with a more than enough questions. Today we get the answers.

continued...

Q. How do SSDs compare to traditional HDDs?

A. Quite favorably, of course smile Seriously, though, the choice of SSD vs. HDD is a trade-off like everything else. The significant disadvantages are cost and capacity. The most obvious advantage is performance, but there are other advantages that might be of interest in some applications. For example, SSDs typically consume less power to complete the same operations, which can translate into extra battery life on a laptop or energy cost savings to a data center. Mechanically, SSDs are more rugged which might make them better suited to portable applications. An SSD can be built with a lower floor cost than a HDD, which was partly responsible for the new extreme low cost "netbook" market. (Notice how the lowest cost netbooks typically use SSDs?wink SSDs can also be smaller and weigh less than HDDs, which helps keep laptops thin and light.

Let me focus on performance a bit more. Comparing throughput measurement does not tell the whole story. For example, the X25-M can read at up to 250MB/s, and an example SATA HDD might sustain 100MB/s. So based on that comparison alone, the SSD looks like it is 2.5x faster than the HDD. However, the real strength of SSDs is in random seeks. The X25-M seek time is 85 us, while a HDD may take 4 to 15 ms. Now the performance difference is more like 50x to 150x. Which is correct? The real answer depends on your application. A typical OS drive spends most of its time reading random small chunks of data (booting, loading applications, loading DLLs, etc), so the SSD will be significantly faster – closer to the 100x mark. Any application that sequentially reads large files will be more throughput-bound, and perform closer to 2.5x faster on the SSD. However, even applications that typically deal with large files (such as video editing) may see significant usability improvements beyond the 2.5x throughput increase due to reduced latency. Multitasking is also substantially improved because disk seeks are so much quicker.

There will be some applications that are not particularly well suited to an SSD. Bulk or archive storage immediately comes to mind. For files that are rarely accessed, the extra cost to store them on flash media may be wasted. Playing back video isn't going to be any slower or faster on an SSD because both will be fast enough to not cause any delay. However, an HTPC might benefit from the silence and small form factor of an SSD. The only non-obvious example I can think of is an application that relies on sequentially writing large amounts of data. The X25-M (MLC) drive only writes at 70MB/s, which might be a limiting factor. If 70MB/s is not enough, consider using the X25-E (SLC) drive, which is capable of writing at 170MB/s.

Q. Why focus on MLC flash? Isn't SLC faster?

A. Yes, SLC flash is faster but only if you compare apples to apples. It is also possible to make up most of the difference with the massive parallelism involved in an SSD. At this point, much of the SSD market is sensitive to both cost and capacity, so the extra density available from using MLC flash is a welcome advantage. I hope everyone would agree that the Intel Mainstream SSD drives are fast despite using MLC flash :)

Q. How much power do SSDs use compared to HDDs?

A. I've seen some studies that claim that SSDs use more power than HDDs, and other studies show they use less. In very general terms, a SSD and a HDD will consume similar amounts of power both during heavy operation or when sleeping. However, there are a couple of factors that give the power advantage to SSDs. Most importantly, SSDs do not have to run a spindle motor, so the active idle power use is significantly reduced. The time (and energy) taken to "spin up" or "spin down" a HDD is gone, so an SSD can be put to sleep much more quickly and much more often. Lastly, SSDs are faster so they will complete a given workload in less time, and can be put to sleep earlier to save power. All of this should contribute to some much needed extra battery life.

Q. Are there different "grades" of flash memory? Why are USB sticks so much cheaper at the same capacity?

A. Yes, there are many different qualities of flash memory components, each with a different cost. The process is similar to speed binning for CPUs. Removable media products like USB memory sticks typically use the lowest quality flash they can get away with, and if you tried to use one as a hard drive I think you would notice the difference immediately. The difference is not only speed, but also reliability. Higher quality flash chips will encounter fewer errors over its life span, which translates into a significantly reduced chance to lose data on an SSD. It is certainly possible to build cheap SSDs out of bargain flash memory, but be careful. If something looks too cheap to be true, it probably is.

Again thanks to Jonathan and Intel for sharing its resources with us.

For further discussion on this topic, please head over to this HardForum thread.