We just received our Thunderbolt 2 test kit, including a Pegasus2 R4 and four 1TB Samsung 840 EVO SSDs. We’re working on a full review that looks at several configurations and usage scenarios, but we wanted to share some early data on sequential performance using the SSDs in RAID 0.
Using the Mac Performance Guide’s DiskTester application, we ran the Sequential Suite via both first-generation Thunderbolt and Thunderbolt 2. Our first-gen Thunderbolt testing device was a 2011 27-inch iMac and our Thunderbolt 2 device was a 2013 6-core Mac Pro. In both cases, the Pegasus2 was connected directly to the Mac, with no other devices or displays on the chain.
Because our four SSDs should be able to easily saturate the first-gen Thunderbolt’s bandwidth, we expect a decent performance increase. We won’t get the theoretical maximum bandwidth of 2,560 MB/s, but we’re looking for a noticeable bump that will make the expensive upgrade to Thunderbolt 2 worth it.
In terms of writes, Thunderbolt 2 enjoys a noticeable advantage across the range of transfer sizes. With transfer sizes at 1 megabyte and up, users can expect about a 40 to 60 percent increase in sequential performance compared to the same setup under the first-generation Thunderbolt.
Reads, while still better in every instance, don’t reveal as great of a performance improvement. Small transfer sizes are 7 to 23 percent faster with Thunderbolt 2, while large transfer sizes get about a 40 percent improvement at most.
With peak write speeds of 1318 MB/s and peak reads of 1303 MB/s, the Pegasus2 with an SSD RAID 0 configuration does indeed offer strong performance. But we have to admit we’re slightly disappointed, and we’ll be looking further at any bottlenecks in the device’s RAID controller and configuration. Speeds greater than 1 gigabyte per second have long been available via PCIe and internal RAID solutions, so we hope we’ll be able to find a way to squeeze out a bit more speed from this otherwise compelling setup.
Stay tuned for a more thorough review of the Pegasus2, including benchmarks using traditional hard drives and other standard RAID setups. If you have any ideas or suggestions for areas you’d like us to investigate, be sure to drop us an email.