RAID-0 Shootout with ATA-100 and ATA-133 - Page 3
06-Feb-02 - Article by Chainbolt

Results, Conclusions, and Recommendations:
The following table shows the results for all test and all settings.

All Results

Configuration

Single Drive

RAID 0 with 2 Drives

RAID 0 with 4 Drives

Drive

WD

Maxtor

WD

Maxtor

Maxtor

WD

Maxtor

Maxtor

ATA/UDMA

100/5

133/6

100/5

133/6

133/6

100/5

133/5

133/6

Interface

IDE

HPT 374
PCI

TX/4
PCI

HPT 372
On-board

Dynamic
Disk

TX/4
PCI

HPT 374
PCI

Dynamic
Disk

HDTach 2.61

Disk Access Time in milliseconds

13.7

11.6

14.0

11.7

N/a

13.9

10.3

N/a

Read Spead average in MB/sec

24.3

35.1

30..3

50.7

N/a

41.2

42.4

N/a

Write Speed average in MB/sec

20.0

20.4

26.,5

27.6

N/a

31.6

35.0

N/a

CPU utilization in %

3.2

7.8

12.8

24.8

N/a

18.9

25.3

N/a

Winbench 99 (2.0): Disk Inspection Test

Disk Access Time in milliseconds

9.36

8.82

10.9

8.43

12.5

10.1

8.3

14.4

Transfer Rate Average in MB/sec

33.7

41.8

44.4

57.7

62.0

54.6

68.8

63.9

CPU Utilization in %

3.84

5.99

13.8

25.9

27.5

42.1

68.1

65.1

Winbench 99 (2.0): High End Disk Mark

Weighted Suite

21,500

24,900

34,600

48,400

35,400

43,300

52,900

45,400

AVS Express

17,500

31,700

39,200

48,100

42,100

54,300

56,300

52,500

Front Page 98

170,000

148,000

193,000

308,000

195,000

292,000

301,000

254,000

Microstation SE

30,000

48,300

38,500

68,000

51,700

63,500

56,800

54,000

Photoshop 4.0

10,400

11,900

18,300

14,800

14,900

14,000

18,400

16,000

Adobe Premier 4.2

19,600

25,800

28,000

56,800

37,100

32,600

62,000

47,300

Sound Forge 4.0

32,600

29,900

54,400

147,000

49,000

76,000

171,100

73,800

Visual C++ 5.0

24,200

21,700

36,900

80,200

38,300

56,200

81,500

62,400

ZD Content Creation 2001

Score

76.9

81.5

87.0

96.8

85.1

96.2

99.8

89.0

We observed huge differences between single test runs with Winbench99 as shown in the screenshot below with the results of the 10-test run with 4 Maxtor D740X on the HPT 374 RAID controller in RAID 0. The subtests of the High-end Disk Mark suit are running only for a rather short time. This is obviously resulting in the observed fluctuations. We also regard the differences between RAID 0 x 2 and RAID 0 x 4 with the High-end Disk Mark as rather exaggerated, probably due to the same reason. As explained already, the focus of Content Creation 2001 is on audio and video editing. The applications used by Content Creation like Adobe Premier, Adobe Photoshop and Sound Forge are working with much larger files and also running longer than in Winbench 99, and we therefore regard the results with Content Creation as more valid for the evaluation of RAID 0 than Winbench 99.


Results of High-end Disk Mark test repetition - Raid 0
with 4 Maxtor D740X on HPT 374, note fluctuation

Summarized Comparison ATA 100 versus ATA 133

 

 

ATA 100
WD400BB

ATA 133
Maxtor D740X

Difference

Transfer Rate (Winbench 99)

Single disk

33.7 MB/sec

41.8 MB/sec

24.0%

Access Time (Winbench 99)

Single disk

9.36 ms

8.82 ms

5.7%

CPU Utilization

Single disk

3.84%

5.99%

55.9%

Winbench99 High End Diskmark

Single disk

21.5

24.9

15.8%

RAID 0 with 2 disk

34,600

48,400

22.2%

RAID 0 with 4 disk

40,300

52,900

31.3%

Content Creation 2001

Single disk

76.9

81.5

6.0%

RAID 0 with 2 disk

87.0

96.8

11.3%

RAID 0 with 4 disk

96.2

99.8

3.7%

Raid 0 with 2 drives versus 4 drives:
The massive performance advantage of RAID 0 with 2 disks over a single disk, be it ATA 100 or ATA 33, is clearly visible in all scores. Depending on the application it’s ranging from 25% in Content Creation to 50% in the Winbench 99 “High-End Disk Mark”. This huge performance gain is in line with the results of a similar comparison with the same benchmarks at Tom’s Hardware Guide. A difference between RAID 0 with 2 disk and RAID 0 with 4 is also visible, but only ranging between 4% and 10% depending on the hard-disk. This difference is probably too small for most users to rectify the much higher cost for 4 harddisks in case somebody is considering a 4-channel RAID controller only because of the assumed speed advantage. As explained in the introduction, there are of course some other good reasons to go for a 4-channel controller. Interesting is the strong showing of RAID 0 by Dynamic Disk. It’s slower than RAID 0 with a hardware controller, but not very much. It’s the cheapest way to get RAID 0, because it’s purely software based (only in Windows 2000 and XP Professional). The drawback is that the OS cannot be installed within the RAID 0 array. That means that one advantage of RAID 0 the faster loading of the OS is not present. We also found that RAID 0 with Dynamic Disk seems to provide a more “even” STR diagram than RAID 0 with a hardware controller. The following screenshots are STRs taken with the Winbench 99 “Disk Inspection Test” for RAID 0 arrays with 4 Maxtor D740X disk on the HPT 374 and with Dynamic Disk.


Winbench 99 Transfer Speed Diagram RAID 0
with 4 Maxtor disks on HPT 374


Winbench 99 Transfer Speed Diagram RAID 0
with 4 Maxtor disks with Dynamic Disk Windows XP

ATA 133 Maxtor D740X versus ATA 100 Western Digital WB400BB:
The direct comparison between a single ATA 133 D740X Maxtor and a single ATA 100 WD400BB shows that the Maxtor disk is in almost all performance categories ahead of the Western Digitals. Transfer rates, access time, and most important the scores in applications are all clearly better with ATA 133. We achieved a transfer rate of around 41MB/sec with a single ATA 133 Maxtor disk, which is 24% better than the tested ATA 100 Western Digital disks. This is a very good result and in line with the findings of techchannel.de. Even when running the Maxtors on an ATA 100 interface we found them to be slightly faster than the WD400BB (results not shown). The only point where the ATA 100 WD400BB could score was the significantly lower CPU utilization. The same trend is visible in RAID 0. The tested ATA 133 Maxtor disks were running applications up to 31% faster in Winbench 99 than their Western Digital counter parts. But again, we regard the 11.3% difference in Content Creation 2001 more credible. That’s not enough to be noticeable, when running applications. But faster is faster, and given the competitive pricing of the Maxtor D740X series, these ATA 133 are a good choice for somebody who is considering to upgrade his ATA 66 system. On the other hand, it’s obvious that the speed gain is not enough to warrant a change from ATA 100 drives. Although we tested the non fluid-bearing version, they were running more silent than the Western Digitals and also cooler. We also liked the activity LED the Maxtors are sporting. Pricing, performance and features make the Maxtor D740X a good buy for anybody with an ATA 133 compatible board.


Transfer rate with HDTach Western Digital WD400BB


Transfer rate with Hdtach Maxtor D740X

ATA 133 Rocket 404 versus ATA 100 TX/4:
The ATA 133 Rocket 404 is easy to install, fast, issue free, and comes with a flexible RAID management utility which offers some interesting features like the capability of arranging and deleting RAID arrays from within Windows, e-mail failure notification and a warning device. It can handle 8 disks and is 30% cheaper than the competing 4-channel controller from Promise. We do recommend the Rocket 404 for any user who is looking for a 4-channel RAID controller. As already mentioned RAID 0 with 4 disks is faster than RAID 0 with 2 drives, regardless ATA 100 or ATA 133. But the gain of, for example, 10% in Content Creation is rather small and does not justify the additional cost for 2 additional disks. The additional data safety of RAID 0+1 or the storage aspect with JBOD are probably better reasons for a 4-channel controller than the relatively small gain in speed. The TX/4 left us with mixed feelings. The point is not that this ATA 100 controller is slower than the ATA 133 Rocket 404. That was to be expected, as we were comparing ATA 100 with ATA 133. What counts more is that the TX/4 is apparently incompatible with the KT266 chipset, has immature drivers, is only working with 4 devices, has a mediocre RAID management utility, and lacks support by the manufacturer. We don’t recommend the TX/4 for anybody running a KT266 or KT266A board. And we doubt that the TX/4, being 30% more expensive, is any longer competitive against the faster, better equipped, and in our experience issue-free Rocket 404.

Final Words:
Maxtor is marketing their ATA 133 drives as the fastest IDE solution until serial ATA will replace IDE. Motherboards with serial ATA support will probably not be available for another year in quantity and even after their introduction it will take many years until boards with IDE ports have completely vanished. Until then ATA 133 disks are the fastest IDE solution and the Highpoint 372 and 374 Raid controllers are a good match from them.

Is your PC in the PC Database yet?
Other Recent Reviews:

DIY Parallel-Cable
Remote LCD!

TwinMOS PC2700 DDR

MSI K7N420 Pro
nForce Motherboard!