We will start with simple throughput benchmarks: reading or writing as fast as it can in a sequential order. These are easy tasks for hard drives and will show us the maximum sustained throughput the drive is able to offer. This is what is called STR: Sequential Transfer Rate.


HDTune
First, I was interested if I could reproduce the same figures with HDTune as claimed by users on the net. After benchmarking, I incorporated both the figures from the AAE and AAK disks in the same graph for easy comparison. The results are pretty self-explanatory:



Interpretation: blue is for the AAE disk, red is for the AAK disk. Notice the horizontal trend of the AAK disk, which usually indicates there is a cap on the performance imposed by either the bus interface or the controller. Since we're using the same nForce controller here featuring an embedded bus, there should be no cap at all. Yet there it is, the AAK disk holds a flat line until about 65% of the rated capacity, where it features the same throughput level as the AAE disk.


Unix file system
I also was very interested in the filesystem throughput. I chose the Unix File System for this task and measured the sustained throughput over a period of 50 seconds.

Sequential Read (MB/s)
AAE	Read	78.8
AAK	Read	64.1
Sequential Write (MB/s)
AAE	Write	67.7
AAK	Write	64.5
Benchmark interpretation: higher is better

Interpretation: the AAE gains a solid advantage when reading. The write scores are pretty low for a reason: when writing to a UFS filesystem, it will have 2 read request every second. This causes the disk to have to seek to those sectors in between, limiting the write throughput. This limits the difference in write scores.


Blocksize crunch with DD
DD is a small program used in Linux and UNIX that can transfer data from point A to point B, using various parameters. It makes an excellent tool for measuring the throughput with various blocksizes. This is under-the-hood performance. The results are pretty shocking:

DD	Read			Write
Blocksize	AAE	AAK	%	AAE	AAK	%
1024K	81.14	69.78	16,3%	81.38	78.79	3,3%
128K	81.22	69.76	16,4%	81.51	79.10	3,0%
64K	81.22	68.89	17,9%	81.51	78.23	4,2%
32K	81.22	53.75	51,1%	81.46	70.95	14,8%
16K	75.07	37.36	100,9%	68.01	59.53	14,2%
8K	53.27	21.58	146,8%	50.38	45.44	10,9%
4K	34.52	11.94	189,1%	31.58	29.66	6,5%
2K	17.91	6.29	187,5%	16.97	16.25	4,4%
1K	9.60	3.20	200,0%	8.55	8.31	2,9%
512	5.11	1.66	207,8%	4.35	4.21	3,3%
Benchmark interpretation: higher is better
Interpretation warning: results are synthetic in nature and may not tally with real-world performance

Interpretation: wow! These are some incredible differences. The AAE drive simply blasts away the AAK drive in the lower blocksizes when reading. Writes seem less affected though, with a maximum difference of 14,8% in favor of the AAE drive. In fact, the AAE drive wins all the scores. But I would never have imagined the difference to be so huge, an advantage of over 200% means the AAE disk is triple as fast as the AAK disk using the same block size.

This in itself does not mean anything though - it is a synthetic benchmark. But it probably is a good source of the performance differences we will see in the next two pages.