Reviews Leaderboard Database Reference Search StorageReview Discussion Reliability Survey Search About StorageReview.com Contents

Adaptec 29160 vs. Tekram DC-390U3W
  May 7, 2001 Author: Terry Baranski  
Adaptec 29160 and Tekram DC-390U3W provided by Hyper Microsystems*

* Remember, mention StorageReview.com when ordering from HyperMicro and receive FREE shipping!


Introduction

If you ask someone to name two SCSI host adapter manufacturers, his or her answer will likely include the words "Adaptec" and "Tekram" - it's safe to say that these two manufacturers are the best known amongst the online hardware-enthusiast community. If you ask someone what the significant differences between Adaptec and Tekram host adapters are, however, you'll probably get some different answers. While it's undeniable that Tekram's host adapters cost significantly less than Adaptec's, the performance issue isn't so easily agreed upon. Some claim that Tekram's host adapters perform as well as Adaptec's; others insist that Tekram cards perform worse, hence their lower price.

With that in mind, StorageReview.com's first SCSI host adapter review focuses on two equivalent models: the Adaptec ASC-29160 and the Tekram DC-390U3W. Since both cards feature similar specifications, it should be interesting to see how they perform relative to one another. Before moving on, let's go over each card's specs.


The Adaptec 29160...

The Adaptec 29160 is one of several host adapters in Adaptec's Ultra160 SCSI line. Falling in the "Entry level to mid-range servers" class, the card features the following:

  • Single SCSI channel
  • Transfer rates up to 160MB/sec
  • 64-bit PCI (32-bit compatible)
  • Internal connectors: One 68-pin for LVD SCSI, one 68-pin for Ultra Wide SCSI, one 50-pin for Ultra SCSI
  • External connectors: One 68-pin for LVD SCSI
  • Up to 15 devices
  • SpeedFlex technology ensures top performance of all connected devices, regardless of SCSI generation
  • Win2000, WinNT, Win9x/ME, DOS, Win3.1, OS/2, Netware, SCO Openserver, SCO UnixWare, and Solaris support
  • 5-year materials and workmanship warranty
In addition to the card, the retail kit comes bundled with the following items:
  • Adaptec EZ-SCSI software
  • Driver disk
  • 5-position, 68-pin LVD internal cable with terminator
  • 3-position, 68-pin Ultra Wide SCSI internal cable
  • 3-position, 50-pin internal cable
  • Complete user documentation (user's guide, installation guide, EZ-SCSI guide)
The 29160 features Adaptec's SpeedFlex technology, allowing LVD (low-voltage differential) SCSI devices to operate at maximum speed even if there are SE (single-ended) devices on the same channel. Normally, a SCSI channel can only operate in SE or LVD mode - not both. This forces LVD devices to operate in SE mode when there is one or more SE device on the channel - certainly not desirable. With SpeedFlex, however, a single LVD channel is divided into two segments: an LVD segment and an SE segment. As long as all SE devices are attached to the SE segment, LVD devices on the LVD segment can operate at their maximum speed. (Note that LVD SCSI's limit of 15 devices per channel still applies here regardless of SpeedFlex. The 29160 is therefore limited to 15 devices.) Check out the reference guide for more information on the differences between SE and LVD signaling.

Also notable is the adapter's support for 64-bit PCI. Such support is standard for Ultra160 host adapters to allow for transfer rates of 160MB/sec due to the ~100MB/sec real-world bandwidth limit of 32-bit PCI.


The Tekram DC-390U3W...

The Tekram DC-390U3W is roughly equivalent to Adaptec's 29160 with the following difference: the DC-390U3W features two separate channels (one LVD, one SE) while the 29160, as discussed, has a single channel with two segments. This yields the DC390UW two advantages: 1) a maximum (theoretical) transfer rate of 200MB/sec (160MB/sec for the LVD channel and 40MB/sec for the SE channel), and 2) support for up to 30 devices.

The card's full specs follow:

  • Dual SCSI channels (one LVD, one SE)
  • Up to 200MB/sec performance
  • 64-bit PCI (32-bit compatible)
  • Internal connectors: One 68-pin for LVD SCSI, one 68-pin for Ultra Wide SCSI, one 50-pin for Ultra SCSI
  • External connectors: One 68-pin for LVD SCSI
  • Up to 30 devices
  • SCO 3.2v4.2/5.0, UnixWare 2.X, DOS, Win3.1, Win95, NT 3.5X/4.X, Win2000, and Netware 3.X/4.X support
  • 5 year warranty
The following items are included along with the card:
  • Driver CD
  • 5-position, 68-pin LVD internal cable with terminator
  • 3-position, 68-pin Ultra Wide SCSI internal cable
  • 3-position, 50-pin internal cable
  • Internal-to-External 50-Pin adapter
  • Quick reference guide and user manual


Methodology...

The so-called "RAID testbed" will also be utilized for SCSI host adapter reviews. Seagate Cheetah 18XL's (18 gig versions) were used in conjunction with the adapters to assess performance.

In all 29160 tests, driver version 4.10.4000 was used along with BIOS version 2.57.2; for all DC-390U3W tests, driver version 5.01 was used along with BIOS version 4.0.

In the single-drive tests, the drive being tested was the only device present on the cable (SCSI ID 0). For multi-drive tests, SCSI ID's were used in sequential order (e.g. SCSI ID's 0 and 1 were used for the two-drive tests). All tests were run with the drive(s) on the LVD channel using the LVD cable that came with each card.

The above reference to "multi-drive tests" may be a bit confusing since this isn't a RAID review. Simply testing the single-drive performance of both cards wouldn't yield the whole performance picture. With that in mind, we've run IOMeter on up to 4 drives simultaneously to put each card's performance scalability to the test.

IOMeter's ability to simultaneously run the same test on multiple drives allows for easy assessment of scalability. Since each subtest runs simultaneously on multiple drives, the total number of outstanding IO's is the number of outstanding IO's in the subtest multiplied by the number of drives being tested - this yields up to 1024 outstanding IO's using standard SR IOMeter access patterns and four Cheetah 18XL's.


The Benchmarks...

Let's examine single-drive WinBench and IOMeter scores.

Single-drive WinBench Scores...

 RAID Testbed  Ziff Davis WinBench 99 under Windows 2000 Professional using NTFS  Details... 
Benchmark Adaptec 29160, 1 driveTekram DC-390UW, 1 drive
Business Disk WinMark 99 (KB/sec) 6250 6180
High-End Disk WinMark 99 (KB/sec) 17380 17420
AVS/Express 3.4 (KB/sec)19660 19420
FrontPage 98 (KB/sec)63120 62360
MicroStation SE (KB/sec)23860 24560
Photoshop 4.0 (KB/sec)12440 12420
Premiere 4.2 (KB/sec)12780 12800
Sound Forge 4.0 (KB/sec)14360 14640
Visual C++ (KB/sec)15240 14940
Disk/Read Transfer RateStorageReview.com
Beginning (KB/sec)35900 35900
End (KB/sec)25000 25000
Disk Access Time (ms)8.99 9.00
Disk CPU Utilization (%)2.27 2.44

 RAID Testbed Performance Graphs Details... 
Windows 2000 Professional using NTFS
Adaptec 29160, 1 drive - 6250|
Tekram DC-390UW, 1 drive - 6180|
Windows 2000 Professional using NTFS
Tekram DC-390UW, 1 drive - 17420|
Adaptec 29160, 1 drive - 17380|

Nothing stands out in the WinBench scores presented above - both cards perform pretty much identically. The 29160 holds a slight lead in the Business Diskmark test while the DC-390U3W scores slightly higher in the High-End Diskmark test. Both of these gaps fall within WinBench's margin of error, however.

Single-drive IOMeter scores...

IOMeter - File Server Access Pattern
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Load = Linear
Adaptec 29160, 1 drive106.05 1.16 9.43 ms 0.84 % 126.25
Tekram DC-390UW, 1 drive105.74 1.14 9.45 ms 0.85 % 124.40
Load = Very Light
Adaptec 29160, 1 drive143.77 1.55 27.82 ms 1.27 % 113.20
Tekram DC-390UW, 1 drive143.62 1.55 27.84 ms 1.19 % 120.69
Load = Light
Adaptec 29160, 1 drive206.81 2.26 77.35 ms 1.81 % 114.26
Tekram DC-390UW, 1 drive207.74 2.26 77.01 ms 1.70 % 122.20
Load = Moderate
Adaptec 29160, 1 drive207.28 2.25 308.73 ms 1.66 % 124.87
Tekram DC-390UW, 1 drive241.53 2.62 264.89 ms 1.90 % 127.12
Load = Heavy
Adaptec 29160, 1 drive270.62 2.92 945.32 ms 2.80 % 96.65
Tekram DC-390UW, 1 drive282.68 3.06 905.01 ms 2.85 % 99.19

IOMeter - Workstation Access Pattern
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Load = Linear
Adaptec 29160, 1 drive127.41 1.00 7.85 ms 1.02 % 124.91
Tekram DC-390UW, 1 drive127.38 1.00 7.85 ms 1.00 % 127.38
Load = Very Light
Adaptec 29160, 1 drive156.96 1.23 25.48 ms 1.39 % 112.92
Tekram DC-390UW, 1 drive157.19 1.23 25.44 ms 1.25 % 125.75
Load = Light
Adaptec 29160, 1 drive218.13 1.70 73.33 ms 1.97 % 110.73
Tekram DC-390UW, 1 drive216.12 1.69 74.02 ms 1.73 % 124.92
Load = Moderate
Adaptec 29160, 1 drive223.84 1.75 285.86 ms 1.73 % 129.39
Tekram DC-390UW, 1 drive257.82 2.01 248.16 ms 2.04 % 126.38
Load = Heavy
Adaptec 29160, 1 drive284.42 2.22 899.43 ms 2.94 % 96.74
Tekram DC-390UW, 1 drive295.42 2.31 866.04 ms 3.08 % 95.92

IOMeter - Database Access Pattern
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Load = Linear
Adaptec 29160, 1 drive107.49 0.84 9.30 ms 0.99 % 108.58
Tekram DC-390UW, 1 drive107.80 0.84 9.27 ms 0.90 % 119.78
Load = Very Light
Adaptec 29160, 1 drive150.06 1.17 26.65 ms 1.27 % 118.16
Tekram DC-390UW, 1 drive150.11 1.17 26.64 ms 1.24 % 121.06
Load = Light
Adaptec 29160, 1 drive214.24 1.67 74.65 ms 1.86 % 115.18
Tekram DC-390UW, 1 drive214.80 1.68 74.47 ms 1.62 % 132.59
Load = Moderate
Adaptec 29160, 1 drive209.40 1.64 305.58 ms 1.69 % 123.91
Tekram DC-390UW, 1 drive244.04 1.91 262.19 ms 1.85 % 131.91
Load = Heavy
Adaptec 29160, 1 drive272.91 2.13 937.40 ms 2.91 % 93.78
Tekram DC-390UW, 1 drive281.66 2.20 908.36 ms 3.02 % 93.26

IOMeter - Random Write Pattern
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Load = Linear
Adaptec 29160, 1 drive170.97 1.34 5.85 ms 1.34 % 127.59
Tekram DC-390UW, 1 drive170.61 1.33 5.86 ms 1.18 % 144.58
Load = Very Light
Adaptec 29160, 1 drive180.76 1.41 22.12 ms 1.43 % 126.41
Tekram DC-390UW, 1 drive180.36 1.41 22.17 ms 1.40 % 128.83
Load = Light
Adaptec 29160, 1 drive201.81 1.58 79.27 ms 1.62 % 124.57
Tekram DC-390UW, 1 drive202.36 1.58 79.05 ms 1.54 % 131.40
Load = Moderate
Adaptec 29160, 1 drive197.11 1.54 324.62 ms 1.50 % 131.41
Tekram DC-390UW, 1 drive225.89 1.76 283.25 ms 1.70 % 132.88
Load = Heavy
Adaptec 29160, 1 drive258.85 2.02 988.47 ms 2.66 % 97.31
Tekram DC-390UW, 1 drive262.24 2.05 975.57 ms 2.74 % 95.71

IOMeter - Sequential Write Pattern
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Load = Linear
Adaptec 29160, 1 drive114.85 28.71 8.70 ms 3.63 % 31.64
Tekram DC-390UW, 1 drive115.36 28.84 8.66 ms 2.97 % 38.84
Load = Very Light
Adaptec 29160, 1 drive114.34 28.59 34.98 ms 3.21 % 35.62
Tekram DC-390UW, 1 drive114.82 28.71 34.83 ms 3.10 % 37.04
Load = Light
Adaptec 29160, 1 drive114.90 28.73 139.24 ms 3.37 % 34.09
Tekram DC-390UW, 1 drive114.83 28.71 139.00 ms 3.19 % 36.00
Load = Moderate
Adaptec 29160, 1 drive114.66 28.66 558.15 ms 4.44 % 25.82
Tekram DC-390UW, 1 drive114.76 28.69 557.68 ms 4.29 % 26.75
Load = Heavy
Adaptec 29160, 1 drive110.91 27.73 2207.80 ms 4.46 % 24.87
Tekram DC-390UW, 1 drive112.84 28.21 1951.68 ms 4.29 % 26.30

IOMeter scores, on the other hand, are more interesting. Note that the same pattern may be seen in the File Server, Workstation, Datebase, and Random Write tests: though scores are virtually identical under linear, very light, and light load, the DC-390U3W takes the lead under moderate and heavy load. Even more interesting is that the DC-390U3W's relative advantage is much larger under moderate load compared to heavy load - 15% vs. 4%, on average.

It's impossible to say exactly why the 29160 performs so poorly under moderate load - overall, it scores slightly worse under moderate load (64 IO's outstanding) than it does under light load (16 IO's outstanding). But if we had to guess, we'd probably chalk this up to a driver issue. For whatever reason, perhaps the 29160's driver's command reordering algorithm isn't efficient with 64 IO's outstanding.

Whatever the reason, the DC-390U3W performs significantly better under moderate load, and slightly better under heavy load. This may come as a surprise to many readers given its low cost relative to the 29160.


Multi-drive IOMeter scores...

IOMeter - File Server Access Pattern - 2 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 2 drives210.98 2.29 9.48 ms 1.84 % 114.66
Tekram DC-390UW, 2 drives211.04 2.28 9.47 ms 1.60 % 131.90
Load = Very Light
Adaptec 29160, 2 drives287.34 3.12 27.84 ms 2.56 % 112.24
Tekram DC-390UW, 2 drives287.93 3.12 27.78 ms 2.39 % 120.47
Load = Light
Adaptec 29160, 2 drives415.20 4.50 77.05 ms 3.86 % 107.56
Tekram DC-390UW, 2 drives397.23 4.31 80.54 ms 3.04 % 130.67
Load = Moderate
Adaptec 29160, 2 drives415.16 4.45 308.25 ms 3.42 % 121.39
Tekram DC-390UW, 2 drives471.00 5.10 271.67 ms 3.90 % 120.77
Load = Heavy
Adaptec 29160, 2 drives485.71 5.23 1053.87 ms 4.43 % 109.64
Tekram DC-390UW, 2 drives520.02 5.64 984.23 ms 4.89 % 106.34

IOMeter - Workstation Access Pattern - 2 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 2 drives253.11 1.98 7.90 ms 2.17 % 116.64
Tekram DC-390UW, 2 drives253.17 1.98 7.90 ms 1.92 % 131.86
Load = Very Light
Adaptec 29160, 2 drives313.56 2.45 25.51 ms 2.89 % 108.50
Tekram DC-390UW, 2 drives314.51 2.46 25.43 ms 2.57 % 122.38
Load = Light
Adaptec 29160, 2 drives436.60 3.41 73.29 ms 3.93 % 111.09
Tekram DC-390UW, 2 drives420.62 3.29 76.06 ms 3.24 % 129.82
Load = Moderate
Adaptec 29160, 2 drives447.44 3.50 286.03 ms 3.67 % 121.92
Tekram DC-390UW, 2 drives493.08 3.85 259.53 ms 4.10 % 120.26
Load = Heavy
Adaptec 29160, 2 drives516.15 4.03 991.62 ms 4.70 % 109.82
Tekram DC-390UW, 2 drives541.90 4.23 944.52 ms 5.29 % 102.44

IOMeter - Database Access Pattern - 2 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 2 drives213.23 1.67 9.38 ms 1.93 % 110.48
Tekram DC-390UW, 2 drives213.96 1.67 9.34 ms 1.80 % 118.87
Load = Very Light
Adaptec 29160, 2 drives299.50 2.34 26.71 ms 2.61 % 114.75
Tekram DC-390UW, 2 drives300.48 2.35 26.62 ms 2.41 % 124.68
Load = Light
Adaptec 29160, 2 drives429.25 3.35 74.52 ms 3.90 % 110.06
Tekram DC-390UW, 2 drives414.59 3.24 77.17 ms 3.13 % 132.46
Load = Moderate
Adaptec 29160, 2 drives419.21 3.28 305.27 ms 3.64 % 115.17
Tekram DC-390UW, 2 drives477.51 3.73 268.00 ms 4.53 % 105.41
Load = Heavy
Adaptec 29160, 2 drives490.89 3.84 1042.73 ms 5.14 % 95.50
Tekram DC-390UW, 2 drives521.87 4.08 980.73 ms 5.53 % 94.37

IOMeter - Random Write Pattern - 2 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 2 drives338.13 2.64 5.91 ms 3.01 % 112.34
Tekram DC-390UW, 2 drives339.53 2.65 5.89 ms 2.79 % 121.70
Load = Very Light
Adaptec 29160, 2 drives357.42 2.79 22.37 ms 3.34 % 107.01
Tekram DC-390UW, 2 drives358.06 2.80 22.33 ms 3.03 % 118.17
Load = Light
Adaptec 29160, 2 drives398.55 3.11 80.26 ms 3.65 % 109.19
Tekram DC-390UW, 2 drives393.32 3.07 81.34 ms 3.13 % 125.66
Load = Moderate
Adaptec 29160, 2 drives389.42 3.04 328.62 ms 3.22 % 120.94
Tekram DC-390UW, 2 drives443.95 3.47 288.22 ms 3.95 % 112.39
Load = Heavy
Adaptec 29160, 2 drives461.45 3.61 1109.11 ms 4.54 % 101.64
Tekram DC-390UW, 2 drives486.32 3.80 1052.54 ms 4.93 % 98.65

IOMeter - Sequential Write Pattern - 2 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 2 drives236.34 59.08 8.46 ms 7.19 % 32.87
Tekram DC-390UW, 2 drives236.15 59.04 8.46 ms 6.77 % 34.88
Load = Very Light
Adaptec 29160, 2 drives235.45 58.86 33.97 ms 7.16 % 32.88
Tekram DC-390UW, 2 drives235.28 58.82 33.99 ms 6.57 % 35.81
Load = Light
Adaptec 29160, 2 drives236.38 59.10 135.36 ms 7.20 % 32.83
Tekram DC-390UW, 2 drives234.19 58.55 136.62 ms 6.74 % 34.75
Load = Moderate
Adaptec 29160, 2 drives233.45 58.36 538.69 ms 7.75 % 30.12
Tekram DC-390UW, 2 drives230.42 57.61 542.55 ms 8.08 % 28.52
Load = Heavy
Adaptec 29160, 2 drives231.44 57.86 2193.51 ms 7.89 % 29.33
Tekram DC-390UW, 2 drives230.09 57.52 2177.68 ms 8.07 % 28.51



IOMeter - File Server Access Pattern - 3 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 3 drives315.00 3.42 9.52 ms 2.56 % 123.05
Tekram DC-390UW, 3 drives315.13 3.42 9.52 ms 2.48 % 127.07
Load = Very Light
Adaptec 29160, 3 drives427.12 4.59 28.09 ms 3.77 % 113.29
Tekram DC-390UW, 3 drives428.33 4.64 28.01 ms 3.49 % 122.73
Load = Light
Adaptec 29160, 3 drives617.15 6.67 77.75 ms 5.51 % 112.01
Tekram DC-390UW, 3 drives535.75 5.78 89.58 ms 4.35 % 123.16
Load = Moderate
Adaptec 29160, 3 drives617.31 6.68 310.94 ms 5.20 % 118.71
Tekram DC-390UW, 3 drives689.65 7.44 278.31 ms 6.14 % 112.32
Load = Heavy
Adaptec 29160, 3 drives670.82 7.28 1144.73 ms 6.35 % 105.64
Tekram DC-390UW, 3 drives722.89 7.84 1062.17 ms 6.91 % 104.62

IOMeter - Workstation Access Pattern - 3 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 3 drives376.47 2.94 7.96 ms 3.26 % 115.48
Tekram DC-390UW, 3 drives378.39 2.96 7.92 ms 2.96 % 127.83
Load = Very Light
Adaptec 29160, 3 drives466.51 3.64 25.72 ms 4.13 % 112.96
Tekram DC-390UW, 3 drives468.84 3.66 25.59 ms 3.72 % 126.03
Load = Light
Adaptec 29160, 3 drives648.46 5.07 74.00 ms 5.67 % 114.37
Tekram DC-390UW, 3 drives578.70 4.52 82.93 ms 4.77 % 121.32
Load = Moderate
Adaptec 29160, 3 drives665.49 5.20 288.45 ms 5.44 % 122.33
Tekram DC-390UW, 3 drives724.40 5.66 264.99 ms 6.40 % 113.19
Load = Heavy
Adaptec 29160, 3 drives715.97 5.59 1072.46 ms 7.07 % 101.27
Tekram DC-390UW, 3 drives754.71 5.90 1017.31 ms 7.74 % 97.51

IOMeter - Database Access Pattern - 3 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 3 drives317.22 2.48 9.45 ms 2.99 % 106.09
Tekram DC-390UW, 3 drives318.73 2.49 9.41 ms 2.76 % 115.48
Load = Very Light
Adaptec 29160, 3 drives443.95 3.47 27.02 ms 4.36 % 101.82
Tekram DC-390UW, 3 drives446.05 3.48 26.89 ms 3.95 % 112.92
Load = Light
Adaptec 29160, 3 drives638.29 4.99 75.18 ms 6.36 % 100.36
Tekram DC-390UW, 3 drives557.30 4.35 86.11 ms 4.85 % 114.91
Load = Moderate
Adaptec 29160, 3 drives621.13 4.85 309.04 ms 5.40 % 115.02
Tekram DC-390UW, 3 drives700.38 5.47 274.07 ms 6.71 % 104.38
Load = Heavy
Adaptec 29160, 3 drives679.80 5.31 1129.54 ms 6.82 % 99.68
Tekram DC-390UW, 3 drives732.15 5.72 1048.65 ms 8.01 % 91.40

IOMeter - Random Write Pattern - 3 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 3 drives502.30 3.92 5.97 ms 4.56 % 110.15
Tekram DC-390UW, 3 drives503.38 3.93 5.96 ms 4.10 % 122.78
Load = Very Light
Adaptec 29160, 3 drives530.44 4.14 22.62 ms 4.53 % 117.09
Tekram DC-390UW, 3 drives532.08 4.16 22.55 ms 4.55 % 116.94
Load = Light
Adaptec 29160, 3 drives592.97 4.63 80.92 ms 5.68 % 104.40
Tekram DC-390UW, 3 drives565.16 4.42 84.92 ms 4.94 % 114.40
Load = Moderate
Adaptec 29160, 3 drives577.93 4.52 332.15 ms 5.03 % 114.90
Tekram DC-390UW, 3 drives656.31 5.13 292.47 ms 6.36 % 103.19
Load = Heavy
Adaptec 29160, 3 drives639.34 4.99 1200.93 ms 6.09 % 104.98
Tekram DC-390UW, 3 drives684.34 5.35 1121.89 ms 7.00 % 97.76

IOMeter - Sequential Write Pattern - 3 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 3 drives299.37 74.84 10.02 ms 7.33 % 40.84
Tekram DC-390UW, 3 drives300.08 75.02 9.99 ms 9.06 % 33.12
Load = Very Light
Adaptec 29160, 3 drives303.15 75.79 39.57 ms 8.16 % 37.15
Tekram DC-390UW, 3 drives313.43 78.36 38.28 ms 9.49 % 33.03
Load = Light
Adaptec 29160, 3 drives302.79 75.70 158.51 ms 8.51 % 35.58
Tekram DC-390UW, 3 drives312.96 78.24 153.37 ms 10.13 % 30.89
Load = Moderate
Adaptec 29160, 3 drives297.90 74.47 643.51 ms 8.26 % 36.07
Tekram DC-390UW, 3 drives311.57 77.89 587.42 ms 11.37 % 27.40
Load = Heavy
Adaptec 29160, 3 drives300.16 75.04 2556.07 ms 8.31 % 36.12
Tekram DC-390UW, 3 drives311.05 77.76 2424.89 ms 11.51 % 27.02



IOMeter - File Server Access Pattern - 4 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 4 drives421.10 4.54 9.50 ms 3.62 % 116.33
Tekram DC-390UW, 4 drives420.50 4.52 9.51 ms 3.34 % 125.90
Load = Very Light
Adaptec 29160, 4 drives569.32 6.15 28.10 ms 5.07 % 112.29
Tekram DC-390UW, 4 drives571.06 6.18 28.01 ms 4.79 % 119.22
Load = Light
Adaptec 29160, 4 drives825.26 8.94 77.53 ms 7.43 % 111.07
Tekram DC-390UW, 4 drives683.18 7.35 93.66 ms 5.61 % 121.78
Load = Moderate
Adaptec 29160, 4 drives824.46 8.91 310.44 ms 7.32 % 112.63
Tekram DC-390UW, 4 drives900.25 9.76 284.31 ms 8.59 % 104.80
Load = Heavy
Adaptec 29160, 4 drives868.02 9.41 1179.46 ms 7.93 % 109.46
Tekram DC-390UW, 4 drives901.46 9.76 1135.77 ms 8.59 % 104.94

IOMeter - Workstation Access Pattern - 4 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 4 drives504.97 3.95 7.92 ms 4.25 % 118.82
Tekram DC-390UW, 4 drives504.43 3.94 7.93 ms 4.08 % 123.63
Load = Very Light
Adaptec 29160, 4 drives623.49 4.87 25.65 ms 5.50 % 113.36
Tekram DC-390UW, 4 drives625.31 4.89 25.58 ms 5.19 % 120.48
Load = Light
Adaptec 29160, 4 drives867.11 6.77 73.80 ms 7.92 % 109.48
Tekram DC-390UW, 4 drives741.50 5.79 86.29 ms 6.23 % 119.02
Load = Moderate
Adaptec 29160, 4 drives887.80 6.94 288.33 ms 8.50 % 104.45
Tekram DC-390UW, 4 drives954.02 7.45 268.28 ms 9.80 % 97.35
Load = Heavy
Adaptec 29160, 4 drives931.71 7.28 1098.91 ms 9.49 % 98.18
Tekram DC-390UW, 4 drives954.61 7.46 1072.48 ms 10.09 % 94.61

IOMeter - Database Access Pattern - 4 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 4 drives425.92 3.33 9.39 ms 3.96 % 107.56
Tekram DC-390UW, 4 drives426.12 3.33 9.38 ms 3.68 % 115.79
Load = Very Light
Adaptec 29160, 4 drives594.87 4.65 26.89 ms 5.77 % 103.10
Tekram DC-390UW, 4 drives596.20 4.66 26.83 ms 5.26 % 113.35
Load = Light
Adaptec 29160, 4 drives853.36 6.67 74.98 ms 8.41 % 101.47
Tekram DC-390UW, 4 drives701.36 5.48 91.23 ms 6.17 % 113.67
Load = Moderate
Adaptec 29160, 4 drives831.50 6.50 307.84 ms 8.01 % 103.81
Tekram DC-390UW, 4 drives914.48 7.14 279.89 ms 9.25 % 98.86
Load = Heavy
Adaptec 29160, 4 drives876.75 6.85 1167.87 ms 8.88 % 98.73
Tekram DC-390UW, 4 drives915.86 7.16 1117.89 ms 9.81 % 93.36

IOMeter - Random Write Pattern - 4 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 4 drives675.39 5.28 5.92 ms 6.17 % 109.46
Tekram DC-390UW, 4 drives674.68 5.27 5.92 ms 5.92 % 113.97
Load = Very Light
Adaptec 29160, 4 drives714.39 5.58 22.39 ms 6.80 % 105.06
Tekram DC-390UW, 4 drives713.50 5.57 22.42 ms 6.05 % 117.93
Load = Light
Adaptec 29160, 4 drives798.59 6.24 80.10 ms 7.66 % 104.25
Tekram DC-390UW, 4 drives744.49 5.82 85.95 ms 6.62 % 112.46
Load = Moderate
Adaptec 29160, 4 drives778.14 6.08 328.94 ms 7.25 % 107.33
Tekram DC-390UW, 4 drives874.24 6.83 292.71 ms 8.63 % 101.30
Load = Heavy
Adaptec 29160, 4 drives816.36 6.38 1254.07 ms 7.97 % 102.43
Tekram DC-390UW, 4 drives873.64 6.83 1171.88 ms 8.81 % 99.16

IOMeter - Sequential Write Pattern - 4 Drives
IOMeter TestsIO/secMB/secResponse TimeCPU Util.IO/CPU%
Adaptec 29160, 4 drives298.09 74.52 13.41 ms 7.85 % 37.97
Tekram DC-390UW, 4 drives297.30 74.33 13.45 ms 9.21 % 32.28
Load = Very Light
Adaptec 29160, 4 drives300.95 75.24 53.16 ms 8.50 % 35.41
Tekram DC-390UW, 4 drives318.84 79.71 50.18 ms 10.29 % 30.99
Load = Light
Adaptec 29160, 4 drives305.93 74.48 209.18 ms 8.84 % 34.61
Tekram DC-390UW, 4 drives318.74 79.69 200.74 ms 10.80 % 29.51
Load = Moderate
Adaptec 29160, 4 drives311.18 77.80 822.43 ms 8.52 % 36.52
Tekram DC-390UW, 4 drives318.17 79.54 782.09 ms 11.73 % 27.12
Load = Heavy
Adaptec 29160, 4 drives315.08 78.77 3235.51 ms 8.56 % 36.81
Tekram DC-390UW, 4 drives318.19 79.55 3203.59 ms 12.03 % 26.45

Interestingly, the same pattern emerges when comparing scores between the two cards utilizing multiple drives. Note that scores for all tests are virtually the same under linear and very light loads in the File Server, Workstation, Database, and Random Write tests. With 16 IO's outstanding, however, the story changes somewhat - this time, the 29160 boasts a significant (up to 20% in the 4-drive tests) lead over the DC-390U3W. Under moderate and heavy loads, however, the DC-390U3W takes the lead (up to 10% under moderate load; up to around 8% under heavy load) just as it did in the single-drive tests.

The reason for the surprising reversal when transitioning from light to moderate loads is something that we also saw in the single-drive tests: the 29160 performs no better with 64 IO's outstanding than it does with 16. Again, this phenomenon was consistent throughout all of the single and multi-drive tests.

Both adapters turn in similar sequential write performances with the DC-390U3W perhaps taking a very slight lead in the three and four-drive tests.


Conclusion...

Given the 29160's tendency to retail for about $100 more than the DC-390U3W ($300 vs $200, on average), we're somewhat disappointed with the Adaptec's performance. We expected the 29160 to perform significantly better than the DC-390U3W to justify its higher cost - what we found, however, was somewhat worse performance. In IOMeter, both cards score virtually the same in all tests under linear and very light loads. And although the 29160 takes a significant lead under light load in the multi-drive tests, it falls behind the DC-390U3W under moderate and heavy load in almost every one of our tests. Both cards turn in similar WinBench performance; however, IOMeter better represents the applications that these cards are targeted towards (high-end workstations and entry-level servers).

With that in mind, we have a hard time recommending the 29160 over the DC-390U3W. The only reason to pay more for the 29160 would be to have a card with the Adaptec name on it. To how many readers does this apply?


HOME | ARTICLES | LEADERBOARD | PERFORMANCE DATABASE | REFERENCE GUIDE
COMMUNITY | RELIABILITY SURVEY | SUPPORT SR! | ABOUT SR |

Copyright © 1998-2005 StorageReview.com, Inc. All rights reserved.
Write: Webmaster