RAID – On-Board or PCI card?

Pages: 1 2 3 4 5 6 7 8

On-Board – RAID 0

RAID 0 offered a bit more of a challenge in setup. I could not build from an existing drive, so just as with the RAID 1 test I connected two drives as masters to the Promise ports, entered the Promise BIOS and built my array. This time I had the choice of what ‘mode’ to use. You are allowed to pick one of three options when building an array: set the drive for Desktop, Server or A/V application use. These settings allow the controller to optimize performance by setting different data block sizes for the controller to use.

For this test I picked the default of desktop, since it should be optimized for the Winstone test I use. But I ran into a snag here. I use Power Quest Drive Copy to copy an exact image of my test setup from one drive to another when testing. This not only saves a lot of time, but that way I have the exact same drive and Windows setup on each drive for comparative testing. But this time it didn’t work! I Received a nice error at the end of the install and a non-usable disk array. Not only once but twice. So I manually set the drive up, at least for the first series of tests. Later, when using the Promise IDE card I was able to use Drive Copy on my first setup, but it failed on my second. That got me thinking and I then tried the onboard Promise with Drive Copy and got it to work. It seems that Drive Copy failed about 2 out of 5 tries with either Promise setup.

With the drive all setup and ready to go I booted to Windows. One thing users have reported with RAID 1 is a faster boot time, so I got out my trusty stop watch and did some testing. With the VIA or Promise setup as a single drive or RAID 1 it took about 17 seconds for the system to boot from the point the BIOS finished and the network log on screen was displayed, with RAID 0 it took about 14 seconds. Not a lot of time, but about 20% faster. However, there is one thing that really stuck out. It takes about 30 seconds for the Promise BIOS to ID and setup the drive! So compared to a system without the RAID controller the OVERALL boot time for a RAID 0 setup is 27 seconds longer when compared to a standard IDE integrated setup – nothing I would brag about.

I again ran my normal series of tests, then ran some compatibility testing with the sound, modem and network. Just like before no issues, and all worked just as it should. I also ran some applications to see if I noticed any difference in speed, which I did not except in one case. I do a lot of Power Point presentations with large picture files for Boy Scouts, and when loading or saving a very large presentation (72MB in size) it did seem to be a bit ‘crisper’ and loaded a couple of seconds faster. However, since you don’t wait a long time for it anyway it’s not really noticeable (unless you are doing back to back testing). So how is the RAID 0 performance? Take a look:

mainboard

MSI K7 Turbo R

BIOS

ver 2.7

Promise Driver

1.3

Drive

Single WD 200BB on the 686B port

2x WD 200BB – RAID 0, Desktop

Disk Blaster 128MB file size, MB/s

19.739

25.522

Winbench 99

Bus Disk KB/s

6290

11600

High Disk KB/s

19500

27300

Disk Transfer Rate

Beginning K Byte/ Sec

33300

56500

Access time – ms

11

10.1

CPU Utilization – %

0.485

0.894

Winstone 2001

Business

46.3

48.5

Content Creation

50.6

50.2

There is a significant gain with RAID 0 in raw data transfer rate, which easy to see in the tests looking at just disk transfer rates. CPU utilization stays about the same as the other Promise tests. But where you don’t see much gain is in overall system performance, as shown in the Winstone tests. Not quite 5% in the Business test and just about no change in the Content Creation. Frankly this is the opposite that I expected, since I was thinking Content Creation to be more I/O bound.

So why do we see such a large gain in disk speed, yet that gain is not reflected in the Winstone test? Simple. Winstone results are a sum of all the components and disk transfer rate is only a part of that sum. One of two things seem to be the reason, either disk transfer rate does not impact the Winstone overall score that much or the disk usage in Winstone does not benefit from the type of gain RAID 0 has to offer (I vote for #2).

How did RAID 0 rate? Just like RAID 1, at least as far as compatibility and setup, no issues at all. Some gain in system (notice I said SYSTEM, not disk) performance, depending on your disk usage.

Wait a minute, I’m getting off track here. Wasn’t the goal to look at the difference between on-board RAID and a PCI card? Sorry about that, I just can’t help pointing out what RAID has to offer and how Winstone reacts to different components. So let’s get back on track and look at the Promise FastTrak 100 TX2 PCI controller.


Pages: « Prev   1 2 3 4 5 6 7 8   Next »

Be the first to discuss this article!