Intel claims SSD superiority, tries to dispel myths

San Francisco (CA) – Our love/hate relationship with hard drives may be coming to an end … that is if Intel has its way. For more than a year, Intel has been touting their upcoming solid state drives and at the recent Intel Developer Forum in San Francisco, Intel basically trumpeted the end of the spinning magnetic platter era. Lower power consumption, faster speeds and longer life were the reasons people would be buying SSDs in the coming months and Intel engineers gave plenty of numbers to back up their claims. The engineers also claimed that not all solid state drives are created equal and, you guessed it, Intel SSDs were at the top of the dog pile.

At IDF we sat in on an interesting talk given about SSDs in extreme gaming applications. Chris Saleski, an Initiatives Manger at Intel’s Storage Technologies Group, and Jack Weast, Architect with Intel’s Consumer PC Group, showed slide after slide of performance numbers of the upcoming X18 and X25 mainstream SSDs. These drives will come in 1.8-inch and 2.5-inch form factors (as the model numbers imply) and will start with 80 GB capacities that will scale up to 160 GB in the fourth quarter.

Compared to what Intel called “competitor A” and “competitor B”, the X18/25 drives were four to five times faster in some benchmarks. Saleski and Weast touted the drive’s transfer rate which was up to 250 megabytes a second for sequential reads and 75 MB/sec for writes of 4KB and above. All drives aren’t created equal and Weast added, “the manufacturer of the SSD makes a huge difference.”

At another talk give by Intel Fellow Al Fazio, Intel touted SSDs longevity and tried to dispel the myth that SSDs wear out quickly. Fazio’s slides showed that the X-18/25 SSDs have a mean time before failure (MTBF) rating of 1.2 million hours, which is on par with modern server hard drives. In addition, he claimed that the drives can withstand a workload of 100 GB worth of writes a day for five years.

But is this enough? Intel’s spokesperson for the storage group, Deb Paquin, told us that most people only write a few gigabytes worth of data each day, so hitting 100 GB/day would be difficult. “During onsite tests with our own employees, we found that most people used about 2GB to 3 GB a day, and the highest power user we had was much less than 20GB,” Paquin said.

So if you extrapolate Intel’s data, then at 2-3 GB a day an SSD should outlast our frail human bodies, but what if the drive was under heavy load, like that of a server or perhaps even a rogue or poorly written program?

Intel’s claim of 100 GB/day for five years totals approximately 182500 GB worth of writes or 182.5 terabytes. Let’s make a big assumption that we can max out the 70 MB/sec max transfer rate until the SSD dies, this gives us 19 minutes to write the entire 80 GB disk and 30 days to hit 182.5 terabytes. Now I don’t know about you, but that doesn’t look too shabby considering most people would notice something bad was happening after an hour or two, let alone 30 days.

Paquin tells us that engineers in the Intel labs have actually completely worn out an SSD, but unlike regular hard drives that fail catastrophically (gotta love that thunk or high-pitched whine of a head crash), you probably won’t know if your SSD is actually failing. Flash memory has a limited number of write and erase cycles, but modern SSDs perform wear leveling which spreads out the writes throughout the disk. Essentially the drive wears uniformly and without the user knowing any better. If some blocks are getting close to being worn out, the drive controller rewrites the data to spare blocks and then remaps the logical address to the new location, so basically your operating system doesn’t know any better. In fact, modern SSDs contain extra capacity just for this purpose.

Paquin adds that a drive that has completely worn out (meaning no more spare blocks) will become read-only, so you shouldn’t lose any data. But she cautioned that SSDs really shouldn’t be used for long-term storage. “No flash SSD should be thought of as an archival storage, but it is important that the drive remain reliable over the serviceable life, which our SATA drive does by the means just described,” she said.

But ok, so if SSDs are so much better than hard drives, why aren’t our computers full of them by now? Cost is definitely one factor as a decent 64 GB SSD will set you back several hundred dollars these days, but beyond cost the slow adoption of SSDs can be summed up in one word ‘fear’.

Most computer users have lived with hard drives for decades and I personally remember my wonderfully huge Commodore 64 hard drive that I bought in my high school days. And while we know a hard drive can crash at any moment, we’re comfortable in the fact that we’ll definitely know that something is wrong and we’re ok with that compared to the “silent” death of an SSD.

But on the enterprise level, Paquin argues that SSD drive death doesn’t have to be silent and that SMART technology (Self-Monitoring Analysis and Reporting Technology) will test for and diagnose drive problems before they become a problem. SMART actively monitors drive blocks and warns the user of faulty sectors, read/write errors and temperature variations. Proponents of the technology say this could help operating systems predict when an SSD is about to fail, but according to some people the technology doesn’t always work as advertised.

John Christopher, a Senior Data Recovery Engineer at DriveSavers - company that has made a name for itself in recovering hard drive data – told us that DriveSavers has been performing an increasing number of data recoveries on flash devices with most of the problems being data corruption and accidental deletion and formatting. Christopher said SMART didn’t work as advertised in the hard drive arena and we can infer that it might not do such a great job in SSDs as well.

“S.M.A.R.T. technology showed a lot of promise when it was introduced but hasn't really panned out as a method of providing an early warning system to identify potential hard drive failure,” said Christopher.

Christopher agrees that some type of early warning system should be implemented for SSDs and indeed all types of storage media. “Carmakers build systems into vehicles they manufacture that monitor oil levels, brake wear and so forth, so it would make sense to have something similar to future storage devices that we rely on daily to hold critical data,” he told us, but he adds that so far it’s “probably too early to predict how reliable these devices will be in the long term.”

And that pretty much sums it up… it’s still too early and most people are taking a wait and see approach to SSDs. Logically we know they’re probably better, but emotionally the story is different.