The performance differences between the Savvio drives at 10,000 RPM and their 15,000 RPM brothers are substantial. The Savvio 15k.1 shows huge performance advantages in all low-level benchmarks such as the access-time measurements or the data-transfer rates. However, these benefits come at the expense of smaller maximum storage capacities (73 GB as opposed to 146 GB with 10,000 RPM) and higher costs. The 15k drives did particularly well in our I/O benchmarks Compare Prices on SAS 15,000 RPM Drives.
If you intend to run several of these babies in a large RAID array we recommend that you select the RAID controller carefully, as it might easily bottleneck your array. We used an ICP 5085BL controller, which is based on a fast Intel IOP333 processor running at 800 MHz, but it still maxes out at 300 MB/s. This equals 75 MB/s per hard drive when four drives are used. The benchmark results show that a single Savvio 10k.2 can deliver more than 90 MB/s.
We also found that the ICP controller offered excellent I/O performance, as we achieved 20% more I/O operations per second with the faster Savvio 15k.1 drives. The performance doesn't even stagnate with large command queue depths of 64 or more outstanding commands. Even with a degraded RAID 5 array where one drive fails, the 15k.1 will still perform as fast as a fully-functional RAID 5 array with 10,000 RPM drives. This difference is tremendous for servers that have to process large storage workloads.
If you simply want to increase data transfer speeds, opting for 15,000 RPM drives is not worth the cost. You will have to live with half of the maximum capacity, too. Using RAID controllers that are optimized for throughput and simply adding one more drive to the array will give you a better result. However, there might be another possible application for drives such as the Savvio 15k.1: as a system hard drive inside a PC enthusiast's desktop, since the Savvio 15k.1 is faster than any SATA drive. But you should be willing to pay a premium for the hard drive-and for the SAS controller.