Is 8GB Enough?
Seagate: Winning the Battle of the Boot Drives
What's this
The first generation of affordable consumer SSDs were fraught with buggy firmware and disappointing capacities. Performance issues abounded, and many of those who bought those low-end 32GB, 64GB, and even 80GB drives felt cheated. Now Seagate is offering hybrids with what is essentially an integrated SSD with only 8GB of capacity. Skepticism about whether this is sufficient is understandable. Does it provide enough speed benefit? Is there so little capacity that the NAND will wear out prematurely?
The issue of NAND size in hybrids ultimately leads an intriguing question: How much data does the average person use in a normal day? To answer this, Seagate turned to its employees and analyzed their use patterns over a five-day period. The findings were surprising and included the following data for the average employee
• Total data read over five days was about 19.48GB.
• Only 9.59GB of this total data access was unique. The rest were duplicate accesses.
• 2.11GB of the data represent 95% of the possible performance benefit from using flash memory.
• 0.48GB of the data represents 80% of the performance benefit.
• 0.06GB of the data represents 50% of the performance benefit.
Graphically, the data usage pattern looks like this:

Put simply, this graph shows that an SSD of just over 2GB is sufficient to accommodate 95% of a user’s potential storage acceleration. Beyond 2GB, increasing SSD capacity has almost no performance impact on user data because people just don’t use that much unique information. Equally interesting is how little caching NAND a system requires in order to realize loads of performance improvement.
“Especially with a standard consumer or corporate workload, users tend to do the same stuff over and over,” says David Burks, Seagate marketing director for hybrid storage products. “They get to work, boot up, and launch email. They load a browser, or they go into Word or Excel. Even if it’s a little custom business app they’re running, they tend to reuse the same data repeatedly, and the data they’re using tends to be moved slowly. It’s rare for someone to suddenly come up with a completely new dataset.”