Ivy Bridge Testing – Part 2
Once again the improved bus speed should prove beneficial to small file performance and none queued Disk performance.
AS SSD
| Raid Controller | LSI MegaRAID 9271-8i |
| Drives | Intel 330 60GB SSD x 5 |
| Raid Level | Raid 0 |
| Stripe Size | 16k |
Compared to the five drive array on the sandy bridge we see a dramatic increase in sequential read performance, 1172MB/s in fact. Write performance sees a marginal increase at 62MB/s over the sandy bridge. small increases are seen in the 4K read and write performance. Read performance increases by 1MB/s, and write performance increases by 3MB/s. The 4K-64thrd shows a similar result. Read performance increases by 2.5MB/s.
Crystal Disk
Crystal disk Follows AS SSD more closely this time. Having used an identical Stripe size. Sequential read performance increases by 1039MB/s. Write performance falls back a little due to diminishing returns from the stripe size. Only by 4MB/s though compared to the sandy bridge results. The 512K Read results increase by 1092MB/s, and read results decrease 13MB/s/. The 4k read results increase 4MB/s, and write results increase 7MB/s over the sandy bridge results. The 4K QD32 read results increase by 4MB/s, which is identical to the 4k results. Write results however stay the same.
Summary of Ivy Bridge AS SSD & Crystal Disk Mark Testing
The difference between the PCIe 3.0 and 2.0 bus once again make themselves clear. The more traditional behavior of SSD reads outperforming that of writes clearly becomes visible in these benchmarks. The fact that data is split to multiple drives on read also shows clearly as well In that the performance scales mostly inline with the number of drives added to the chain.
If we where testing this on a Sandy Bridge E system. We would probably see an increase in performance beyond what we are seeing on the Ivy bridge system. The added I/O performance and bus quality would likely yield numbers much closer to the theoretical maximum for PCIe 3.0 throughput.
