Richard Parker's blog

OCZ Colossus Performance in AHCI vs IDE mode on the Intel DP55KG Desktop Board

Posted in Technology by Richard on March 28, 2010

For those of you who don’t know, I’m currently building a new high specification PC to handle the day-to-day rigours of software development. Like a lot of developers, I’ve decided to embrace the new SSD drive technology due to the massive performance increase to be had over traditional ‘plate drives’.  

I’ll provide more details about the PC’s specification soon, but I wanted to share some information about the various ways in which SSDs can be configured and how they affect performance. There are plenty of articles out there regarding SSD design, from a technical point of view, but I haven’t found any that have compared AHCI to IDE so that I could figure out which might give my Colossus the best chance of performing well. So, I decided to run my own basic test.  

Motherboard  

For this test, I am using the Intel DP55KG motherboard running BIOS version KGIBX10J.86A (17th Feb 2010). The DP55KG has 8 on-board SATA ports, six of which are powered via the board’s PCH, and two via an integrated Marvell 88E6145 chip (check out this post for an excellent review of the DP55KG).  

The SSD  

The drive under scrutiny here will be the OCZ Colossus 120GB. On paper, according to OCZ, the drive supports read and write speeds up to 260MB/sec, with sustained write around 140MB/s. This was the best drive I could find of this capacity in the price bracket.  

The 3.5" 'Colossus' SSD by OCZ

The test  

I wanted to know, “Which performs better, AHCI or IDE?” followed by “will IDE mode on the Marvell chip out perform IDE mode on the Intel PCH controller?”. To find out, I will use Passmark Performancetest 7.0 from Passmark. It provides a simple, standardised means of testing my SSD’s.  

  • To test the Intel PCH:
    • Configure the controller in IDE  mode
    • Install Windows 7 Ultimate
    • Install the latest Intel Chipset drivers
    • Restart
    • Install Passmark PerformanceTest 7.0
    • Run tests
    • Repeat these steps, but next time configure the controller in AHCI mode.
  • To test the Marvell Chip:
    • Connect the OCZ Colossus to port 0 on the board
    • Verify controller is in IDE mode (AHCI is not supported by the Marvell chip)
    • Install Windows 7 Ultimate
    • Install the latest Intel Chipset drivers
    • Restart
    • Install Passmark PerformanceTest 7.0
    • Run tests

Results  

Here are the raw test results. Note that the figures quoted are in MB/s.: 

Test Name Intel PCH (AHCI) Marvell (IDE) Intel PCH (IDE)
Disk – Sequential Read 109.9 94.3 111.3
Disk – Sequential Write 131.0 46.1 131.8
Disk – Random Seek + RW 41.8 37.2 41.4
Disk Mark 1022.7 642.1 1029.0
PassMark Rating 2372.6 1489.7 2387.3

Here’s the corresponding bar chart (click to enlarge):  

Intel DP55KG PCH IDE vs AHCI, vs Marvell IDE

 

 
Please note: in the results above, “This computer” refers to Intel PCH in IDE mode.
 
Conclusion
  
Based on the information above, for three identical Windows 7 installations on the same test PC, the result is fairly obvious: the Marvell controller won’t get you very good disk performance. Use ports 1 to 6 on the Intel DP55KG instead. IDE mode seems to outperform AHCI, if only marginally.
 
Having read various articles on the topic, I know it is unusual for anyone to actually experience the quoted performance speeds of the manufacturers. In this instance, though, I’m curious to learn if there’s anything else I can do to get closer to those, as in my tests – for sequential reads, for instnace – I’m well over 150MB/s slower than the manufacturer’s quoted speed (albeit still way faster than most 7,200RPM SATA “Plate” drives).
 
I’ll continue reading to see what else I can do but for now, at least, I’m happy with the added performance benefit and satisfied that I’ve chosen the best mode and chipset for my system.

Possibly related posts: (automatically generated)

Tagged with: , , , , ,

leave a comment

Leave a Reply