The Benchmarks...
Note: Our RAID testing methodology was disclosed in Battle Of The Titans: The Promise SuperTrak100 vs. 3ware's Escalade 6400.
For the following benchmarks, the "write cache" setting of both cards was enabled. The MegaRAID's DMA setting was enabled, and tests were done using a stripe size of 64k for both cards. The FastTrak was tested with driver version 1.30 (Build 33) and BIOS version 1.30 (Build 12), while the MegaRAID was tested with driver version 2.40 and BIOS version 2.410.
WinBench results...
Let's examine base, single drive scores for each card to establish a baseline against which RAID arrays may be compared.
|
|
| Benchmark |
Single-Drive Base Scores |
| Abit's SL6 on-board ATA | Promise Ultra66 | MegaRAID IDE100 | FastTrak 100 |
|
| Business Disk WinMark 99 (KB/sec) |
6484 |
5820 |
6304 |
6128 |
| High-End Disk WinMark 99 (KB/sec) |
16220 |
14460 |
15740 |
15400 |
| AVS/Express 3.4 (KB/sec) |
13600 |
13160 |
16260 |
15320 |
| FrontPage 98 (KB/sec) |
65860 |
50980 |
61980 |
62520 |
| MicroStation SE (KB/sec) |
21480 |
18320 |
20880 |
20520 |
| Photoshop 4.0 (KB/sec) |
8534 |
8668 |
8618 |
8598 |
| Premiere 4.2 (KB/sec) |
14760 |
12440 |
14320 |
13860 |
| Sound Forge 4.0 (KB/sec) |
18540 |
16040 |
15620 |
15200 |
| Visual C++ (KB/sec) |
17000 |
14200 |
14600 |
14460 |
|
| Disk/Read Transfer Rate | StorageReview.com |
|---|
| Beginning (KB/sec) |
29800 |
29733 |
29800 |
29667 |
| End (KB/sec) |
17500 |
17500 |
17500 |
17433 |
|
| Disk Access Time (ms) |
15.2 |
15.3 |
15.2 |
15.1 |
|
| Disk CPU Utilization (%) |
2.9 |
3.0 |
4.8 |
3.0 |
| |
[an error occurred while processing this directive]
The only thing that stands out about the above scores is the MegaRAID's significantly higher CPU utilization... upwards of 2% higher than that of the FastTrak. This wouldn't likely be noticeable to users, but it's interesting nonetheless.
WinBench RAID 0 scores...
|
|
| Benchmark |
2 Drives, RAID 0 |
|
3 Drives, RAID 0 |
|
4 Drives, RAID 0 |
| MegaRAID IDE100 |
FastTrak 100 |
MegaRAID
IDE100 |
FastTrak 100 |
MegaRAID IDE100 |
FastTrak 100 |
|
| Business Disk WinMark 99 (KB/sec) |
6176 |
6124 |
|
6236 |
6180 |
|
6340 |
6356 |
| High-End Disk WinMark 99 (KB/sec) |
18100 |
9424 |
19340 |
10700 |
19200 |
8652 |
| AVS/Express 3.4 (KB/sec) |
16020 |
17280 |
16840 |
16740 |
15042 |
17420 |
| FrontPage 98 (KB/sec) |
62100 |
61380 |
17960 |
62180 |
61540 |
61880 |
| MicroStation SE (KB/sec) |
19060 |
19000 |
14920 |
18480 |
17420 |
17380 |
| Photoshop 4.0 (KB/sec) |
11780 |
4502 |
14540 |
5096 |
15880 |
3886 |
| Premiere 4.2 (KB/sec) |
16880 |
5858 |
18300 |
7346 |
17620 |
5716 |
| Sound Forge 4.0 (KB/sec) |
18760 |
8350 |
19740 |
9396 |
19340 |
6866 |
| Visual C++ (KB/sec) |
17300 |
10780 |
17680 |
12100 |
17740 |
10508 |
|
| Disk/Read Transfer Rate | StorageReview.com |
|---|
| Beginning (KB/sec) |
59467 |
58700 |
|
75567 |
80600 |
|
76567 |
80667 |
| End (KB/sec) |
34933 |
34667 |
52200 |
51667 |
69767 |
68467 |
|
| Disk Access Time (ms) |
15.0 |
15.1 |
|
15.2 |
15.4 |
|
14.9 |
14.8 |
|
| Disk CPU Utilization (%) |
4.7 |
3.0 |
|
4.8 |
3.1 |
|
90.4 |
88.7 |
| |
- AMI MegaRAID IDE100, RAID 0, 2 Drive STR Graph
- Promise FastTrak 100, RAID 0, 2 Drive STR Graph
- AMI MegaRAID IDE100, RAID 0, 3 Drive STR Graph
- Promise FastTrak 100, RAID 0, 3 Drive STR Graph
- AMI MegaRAID IDE100, RAID 0, 4 Drive STR Graph
- Promise FastTrak 100, RAID 0, 4 Drive STR Graph
Several notables... First, each card's Business Disk Winmark performance in RAID 0 remains virtually unchanged from its single-drive performance. Does this imply no performance improvement in RAID 0 with both of these cards? Unlikely, so we'll avoid drawing such a conclusion until we go over additional results.
The High-End Disk Winmarks, however, are another story. The MegaRAID's scores do indeed scale somewhat in RAID 0... and we do believe they are representative of performance. The FastTrak's High-End scores, however, are far too low to be believable, and are almost certainly the result of the same WinBench Disk Winmark issue we encountered in the Escalade 6400 vs. SuperTrak 100 article. This prevents us from comparing the two cards here.
When it comes to sustained transfer rates, both cards scale quite nicely when going from a single drive to 2 and 3-drive RAID 0 arrays. Neither card, however, yields an STR improvement when going from 3 to 4 drives - the FastTrak limits transfers to around 80MB/sec, while the MegaRAID limits around 76MB/sec.
As far as CPU utilization goes, the FastTrak consistently scores 1.50-1.75% better than the MegaRAID in this area. As with the Escalade 6400 vs. SuperTrak 100 review, WinBench reports ridiculously high CPU utilization in 4-drive RAID 0 arrays. As mentioned in that review, we believe this to be a flaw in WinBench itself.
WinBench RAID 1 scores...
|
|
| Benchmark |
2 Drives, RAID 1 |
| MegaRAID IDE100 |
FastTrak 100 |
|
| Business Disk WinMark 99 (KB/sec) |
6304 |
4450 |
| High-End Disk WinMark 99 (KB/sec) |
15400 |
2920 |
| AVS/Express 3.4 (KB/sec) |
15900 |
15340 |
| FrontPage 98 (KB/sec) |
60860 |
49020 |
| MicroStation SE (KB/sec) |
21100 |
19840 |
| Photoshop 4.0 (KB/sec) |
8626 |
1056 |
| Premiere 4.2 (KB/sec) |
13380 |
1614 |
| Sound Forge 4.0 (KB/sec) |
14900 |
2370 |
| Visual C++ (KB/sec) |
14360 |
3734 |
|
| Disk/Read Transfer Rate | StorageReview.com |
|---|
| Beginning (KB/sec) |
29800 |
29567 |
| End (KB/sec) |
17500 |
17400 |
|
| Disk Access Time (ms) |
15.2 |
13.6 |
|
| Disk CPU Utilization (%) |
4.7 |
3.1 |
| |
In RAID 1, the MegaRAID exhibits little improvement over a single drive. This indicates that the adapter doesn't balance requests between the two drives in the array. Given the MegaRAID's cost-sensitive nature, this should be no surprise.
Unfortunately, the FastTrak's High-End Disk Winmark scores continue to be blatantly unrepresentative, and are therefore useless for comparison purposes. The FastTrak's Business Disk Winmark scores, however, while at first glance looking suspicious, may indeed be accurate - as we'll see later, IOMeter backs them up to an extent.
Also of interest is the Fastrak's significantly lower RAID 1 average access time. This would be expected if the FastTrak balanced loads between the 2 array drives; however, we see no other indications that any balancing is occurs - the FastTrak's Business Diskmark, STR, and (as we'll see) IOMeter scores in RAID 1 don't concur.
WinBench RAID 01/10 scores...
|
|
| Benchmark |
4 Drives, MegaRAID IDE100 in RAID 10, FastTrak 100 in RAID 01 |
| MegaRAID IDE100 |
FastTrak 100 |
|
| Business Disk WinMark 99 (KB/sec) |
6020 |
4302 |
| High-End Disk WinMark 99 (KB/sec) |
17420 |
3080 |
| AVS/Express 3.4 (KB/sec) |
15320 |
14820 |
| FrontPage 98 (KB/sec) |
60080 |
50180 |
| MicroStation SE (KB/sec) |
18340 |
17600 |
| Photoshop 4.0 (KB/sec) |
11460 |
1136 |
| Premiere 4.2 (KB/sec) |
16160 |
1696 |
| Sound Forge 4.0 (KB/sec) |
17780 |
2506 |
| Visual C++ (KB/sec) |
16680 |
3894 |
|
| Disk/Read Transfer Rate | StorageReview.com |
|---|
| Beginning (KB/sec) |
59067 |
59267 |
| End (KB/sec) |
34933 |
34800 |
|
| Disk Access Time (ms) |
15.1 |
13.3 |
|
| Disk CPU Utilization (%) |
4.7 |
3.1 |
| |
Although the MegaRAID's RAID 10 business Disk Winmark score comes in lower than that of a single drive, its High-End Diskmark score is indeed higher, likely due to doubling of STR. WinBench also reports a slightly lower access time relative to a single drive; this may be within the test's margin of error.
As far as the FastTrak goes, the Diskmark scores are obviously not representative of performance. They are simply abysmal, despite a high STR and significantly lower reported access time than that of a single drive.
IOMeter Scores...
