Click for SMOKING FAST SATA 6G SSDs!
Click for SMOKING FAST SATA 6G SSDs!


A Click on this Banner shows your site support to my Sponsors

Accelerate Your Mac!  - the source for performance news and reviews
The Source  for Mac Performance News and Reviews
Don't forget to check out all the other site features!
Performance Tests: PCI Ultra2 SCSI Controllers
Published: 2/24/99 (updated 3/6/99)
Which Ultra2 SCSI Card Performs Best?
How does Ultra2 SCSI Compare to the new UltraATA Drives?

Note: This article was written in Feb. 1999. Since models, prices and performace change frequently, for the latest articles and reviews on faster controllers and drives, see the Storage Articles page.


Introduction:

This page will attempt to illustrate the performance differences between the three most popular Mac Ultra2 SCSI controllers (Adaptec 2940U2W, ATTO U2 ExpressPCI Pro, Initio Miles U2) in both real world and pure benchmarks. In these tests the only variable was the SCSI Card, as the exact same system, PCI slot, hard drive, U2 SCSI cable and terminator were used with all 3 cards. Different drives, drivers or systems may produce different results but the point illustrated here is how the cards compared to each other under identical conditions and hardware.

I think you will also find the performance comparisons to the low-cost UltraATA (IDE) drives like the Maxtor DiamondMax Plus very enlightening. The Maxtor drive is the same one used in the (beige) G3 IDE upgrade and B&W G3 Slave drive articles. Performance of the DiamondMax Plus was shown to be excellent last year in the original G3 IDE article, but for the first time I'll be able to illustrate how it compares to Ultra2 SCSI.

My objectives in this review are to answer the following questions:

  1. What is the best Ultra2 SCSI card for the Mac?
  2. How does performance compare to a fast UltraATA Drive?
  3. Are rumors of 2940U2W cards not running in Ultra2 mode true?
  4. What is the best value or consumers? For Demanding Users?

Update: one other question in my mind was if the 2940U2W 'SpeedFlex' design (see below) really preserved U2 speeds (40Mhz) when a Single-Ended (SE) SCSI drive was attached along with a U2 drive. Rob Huigsloot wrote that his tests with the ATTO Tools benchmark showed a EZ230 external drive did not affect the U2 disk's rate. My later tests with a ZIP drive attached to the SE port showed that U2 disk performance was not affected.


Test System Configuration:

For these tests I used a (beige) Apple G3 system with 400MHz XLR8 ZIF (400/200/66) upgrade installed, 160MB of RAM, Virtual Memory Off, running OS 8.1. No 3rd party extensions were installed (i.e. No SpeedDoubler, etc.). All cards were tested in the center PCI slot with no other PCI cards installed. The same Ultra2 SCSI cable and terminator was used for all tests. The disk was an Ultra2 SCSI Quantum Atlas III 18GB drive using Apple's drive setup 1.6 driver. Tests were done on a 4GB partition (first partition) since MacGurus will be sending a 4.5GB Seagate U2 Cheetah for the second part of these tests, which will be done in my B&W G3/400.

SCSI Cards Tested:

  • Adaptec 2940U2W (firmware v1.02f3)
    (the latest beta firmware is only for B&W G3s). The only Ultra 2 SCSI card that includes narrow and standard Ultra-Wide SCSI connectors with a unique SpeedFlex design which promises no U2 performance degradation when running non-U2 drives. The narrow/UW connector feature makes this an attractive option for IDE based B&W G3 owners. No LED indicators of Ultra2 mode. Kit includes a complete 4 cables (internal to external narrow scsi, and internal narrow/UW/U2 w/terminator).
    Model Number: 2940U2W Mac kit (incl. cable set)
    Price: Originally $449 at Adaptec's Store [Card Image]

  • ATTO Ultra2 ExpressPCI Pro (latest firmware v1.3.3)
    Only Ultra2 Connectors onboard, attaching non-U2 devices will reduce card clock speed. Uses sub-miniature 68pin external connector. LEDs show card status and U2 mode. My package included PC and Mac drivers/firmware. ATTO has a history of solid performing cards on both Platforms.
    Model No: EPCI-UL2S-000
    Price: Originally $299.98 (street) [Card Image]

  • Initio Miles U2 (using latest firmware v1.01)
    (Update 1.02 will be used in the B&W G3 tests) Only Ultra2 Connectors onboard, attaching non-U2 devices will reduce card clock speed. Most affordable of the Mac Ultra2 SCSI cards. Uses standard 68pin external connector. LEDs show card status and U2 mode. Package includes U2 cable/terminator and optionally, SoftRaid 2.1.5 (the fastest RAID driver I've seen). I noted that my Miles U2 card had a faster oscillator than the other cards (appx. 44MHz vs the standard 40MHz U2 oscillator). Initio said that some early cards used these due to availability of the 40Mhz parts. This faster clock (10%) is almost certainly responsible for the faster performance of this particular card.
    Model No: INI-A100U2W MAC
    Price: Originally $199.95 (w/Softraid) [Card Image]

Drives Tested:

Quantum Atlas III 18GB Ultra2 (7200RPM, 1MB cache). Complete drive specs are here. This drive is targeted at server/hi-reliability applications. According to NewMedia's magazine tests last last year, this was near the bottom of the list on performance among 7200RPM Wide SCSI drives. Spec'd at 1.6 inches high, it will not fit on the B&W G3's drive tray. Tests results shown here were with Apple's drive setup v1.6, but I'm sure with a better driver performance would improve. (Macgurus.com sent An Ultra2 Seagate Cheetah (10,000RPM) drive after the tests were completed.)

A preliminary test with the U2 Cheetah is shown below in the MacBench 5 tests results graph. The Cheetah used SoftRaid 2.1.5's disk drive. Note scores were near 2000, demonstrating that the drive, not the interface is almost always the limiting factor in a single drive setup (non-RAID).


HFS vs HFS+:

Tests were done in both HFS and HFS+ formats to see if there was any performance penalty (increased overhead from larger volume bitmap, especially on larger file I/O). Surprisingly HFS+ was actually a better performer in most tests. In Windows 9x, FAT32 (a similar scheme to HFS+) shows generally a 10% performance hit over FAT16 in most cases. From what I saw in benchmarks and finder file copies the space saving (smaller minimum allocation blocks) HFS+ format showed no negative performance impact on the Mac in standard file I/O. Video capture is one area where I still have my doubts however.

Update 3/6/99: To see graphs of the ATTO Tools tests of R/W performance for each card, click on the desired card in the table below.

Links to ATTO Test Graphs (HFS/HFS+)
2940U2W
ATTO U2
Miles U2

Compare 0-8MB Rates at a Glance - Click on the small graph to the right to open this tall chart in a separate window. Graph includes peak/sustained rates as well as performance across file size. If you'd prefer a horizontal format graph - click here.


Finder Test Results:

For tests of actual Finder file I/O performance I copied a 124.6MB System folder (1,287 files) from the stock 6GB beige G3 IDE drive (not optimized) to the Ultra2 drive and recorded the total time from start to finish. To remove the IDE drive from the equation, I recorded the time to duplicate that folder on the U2 SCSI drive. Tests were run 3 times and the average time in seconds are shown below (best scores in red):

OS 8.1 Finder File Copy Results
(Copy folder w/1,287 files - 124.6MB) total)
SCSI Card
Copy Files
Duplicate Files
Adaptec 2940U2W
(HFS)
104.06 sec
59.84 sec
Adaptec 2940U2W
(HFS+)
101.62 sec
48.25 sec
ATTO U2
(HFS)
105.5 sec
59.66 sec
ATTO U2
(HFS+)
102.19 sec
48.22 sec
Miles U2
(HFS)
100.53 sec
59.03 sec
Miles U2
(HFS+)
100.21 sec
47.97 sec

Scores were very close but the Miles U2 was faster by a small margin. Note that HFS+ was the best performing format (a trend that would continue).

Peak/Sustained Rate Tests:

The ATTO Tools utility has a benchmarking function, which I personally like for the adjustability, large selection of test file sizes, graphing and for the fact it does not use the system disk cache (although it has an option to do so). Bypassing the system disk cache gives a better indication of the card and drive performance, eliminating the effect of a large cache setting in the memory control panel - which inflates scores in other benchmarks like MacBench.

ATTO Tools v2.1.1 Benchmark Results
PR = Peak Read, SR = Sustained Read, PW = Peak Write, SW = Sustained Write.
All rates are in MegaBytes/Second.

Red = Best U2 SCSI Rate,   Blue = Case where IDE drive is faster
Card
Driver
& Format

32KB
Test

128KB
Test

512KB
Test

2MB
Test

8MB
Test

Adaptec 2940U2W
Bios 1.02f3
Apple DS 1.6 HFS
PR 19.11
SR 17.57
PW 6.28
SW 5.44
PR 30.05
SR 26.35
PW 20.67
SW 14.90
PR 43.96
SR 40.81
PW 27.07
SW 11.27
PR 47.36
SR 13.87
PW 16.58
SW 11.65
PR 32.08
SR 12.82
PW 17.42
SW 11.89
Adaptec 2940U2W
Bios 1.02f3
Apple DS 1.6 HFS+
PR 19.55
SR 18.16
PW 6.34
SW 5.57
PR 30.33
SR 28.66
PW 15.74
SW 8.78
PR 45.53
SR 40.11
PW 26.07
SW 10.33
PR 43.78
SR 13.79
PW 12.54
SW 11.36
PR 30.16
SR 12.85
PW 19.44
SW 11.87
Atto U2
Bios 1.3.3
Apple DS 1.6 HFS
PR 22.21
SR 20.49
PW 6.15
SW 5.36
PR 33.11
SR 29.25
PW 20.07
SW 14.50
PR 50.90
SR 46.75
PW 17.76
SW 11.62
PR 53.54
SR 14.09
PW 16.15
SW 11.60
PR 35.36
SR 12.84
PW 17.18
SW 11.90
Atto U2
Bios 1.3.3
Apple DS 1.6 HFS+
PR 22.27
SR 20.27
PW 6.10
SW 5.36
PR 33.25
SR 31.55
PW 16.05
SW 8.19
PR 53.72
SR 45.70
PW 24.08
SW 10.29
PR 50.53
SR 13.91
PW 15.62
SW 11.35
PR 33.78
SR 12.89
PW 18.13
SW 11.86
Miles U2
Bios 1.01
Apple DS 1.6 HFS
PR 20.43
SR 18.85
PW 6.31
SW 5.56
PR 35.01
SR 31.00
PW 14.43
SW 8.33
PR 50.93
SR 45.58
PW 16.47
SW 10.78
PR 50.33
SR 13.85
PW 19.91
SW 11.84
PR 38.18
SR 12.88
PW 18.68
SW 12.02
Miles U2
Bios 1.01
Apple DS 1.6 HFS+
PR 21.40
SR 20.06
PW 10.12
SW 5.81
PR 32.49
SR 29.01
PW 18.27
SW 12.46
PR 54.91
SR 46.65
PW 16.55
SW 11.16
PR 53.82
SR 13.88
PW 15.17
SW 11.64
PR 36.47
SR 12.84
PW 18.11
SW 12.00
Maxtor IDE
(10GB 7200rpm)
Apple DS 1.6
PR 13.56
SR 13.17
PW 13.44
SW 12.64
PR 15.19
SR 8.77
PW 15.01
SW 11.65
PR 15.04
SR 9.92
PW 15.30
SW 12.95
PR 15.10
SR 12.59
PW 14.81
SW 13.73
PR 13.76
SR 13.59
PW 14.23
SW 13.93
(Maxtor IDE Scores shown for Ref, taken from previous tests in the same system)
0-32KB
Test
0-128KB
Test
0-512KB
Test
0-2MB
Test
0-8MB
Test

Note: This test did not use the system disk cache to better indicate the actual drive and interface performance.

[Update: There are currently far faster U2 SCSI drives and IDE drives that the samples here. And IDE drives and interfaces are far faster than the ATA/3 (16.6MB/sec max) shown above. For results of more recent tests (including ATA/66 & Ultra 160 SCSI), see the Storage Topics page. Included is a review of the 40GB IDE Maxtor drive that sustained nearly 30MB/sec read and write rates in a B&W G3. The IDE IBM 75GXP drives can sustain nearly 35MB/sec read and write rates.]

Atto Tools Tests - Best Overall Ultra2 SCSI Card: - Initio Miles U2

Results were very close on many tests and there was no runaway winner but the Miles2 was consistently at or near the top, often finishing 2nd by less than the margin of error (repeatability). In the end, price and features may be the deciding factor. Note the IDE drive does well in sustained rates, but the interface and drive does not have the peak potential of Ultra2 SCSI. In general consumer use, I suspect the IDE drive's file copy/write performance does not show this limitation. I'll know more tonight when I run finder tests on the Maxtor 10GB IDE drive in the beige G3.

Until then, for a look at the Maxtor and stock G3 6GB IDE drive performance (300MHz CPU speed) - see my (beige) G3 IDE Upgrade article.

MacBench 5.0 Tests:

The combination of the Atlas III drive and the Apple driver provided disappointing MacBench disk scores, even with a 4MB Disk cache setting. As shown in my Ultimate Mac Project Genesis MacBench disk scores, using SoftRaid 2's disk driver on a single Seagate Cheetah (10,000RPM) Ultra-Wide (not Ultra2) drive and UW (not U2) controller produced Disk scores that were nearly twice as high (1604). The Atlas III 18GB drive is not in the same performance class as the Cheetah. An Ultra2 4.5GB Seagate Cheetah drive just arrived as this was being posted (courtesy of MacGurus). I quickly installed SoftRaid 2.1.5's driver on the disk and ran the MacBench disk tests to illustrate the difference A better drive/driver combination makes. The results are obvious in the graph below:

Macbench 5.0 tests

(Maxtor 10GB IDE DiamondMax Plus drive scores restored in the graph above, they were inadvertantly omitted when the Cheetah scores were added.)

Based on the above and my own observations, here are the answers to the questions posed in the introduction:

  1. What is the best Ultra2 SCSI card for the Mac?
    Based on performance, price and the bundled SoftRaid driver, the Initio Miles U2 card is my pick. B&W G3 owners that have no SCSI card and need to use legacy (ZIP, Jaz, CDR or other legacy drives) and still want the option of Ultra2 should consider the 2940U2W. Althought more expensive and not quite as fast (with the current firmware at least), the narrow and UW SCSI connectors and Speedflex design is attractive and could save a PCI slot. My later tests showed that Speedflex does seem to work as advertised.

  2. How does performance compare to a fast UltraATA Drive?
    Sustained rates on the Maxtor DiamondMax Plus 7200RPM drive were actually better than the Atlas III (not one of the faster U2 drives however). Peak rates were up to 3 times faster with Ultra2 SCSI, but initial tests show that finder copies for instance are comparable between the two. Bottom line is IDE is the better value for all but the most demanding users, or those that need high performance RAID.

  3. Are rumors of 2940U2W cards not running in Ultra2 mode true?
    Based on the >40MB/sec peaks shown in the ATTO Tools tests, I'd say these rumors are false. Tests with a SCSI ZIP drive attached to the SE port showed that the Speedflex design works as promised (preventing non-U2 devices from affecting U2 drive performance).

  4. What is the best value or consumers? For Demanding Users? For the average user or consumer, IDE is the better value, with a 10-17GB drive costing less than any of these Ultra2 Controllers. ProMax's $99 dual IDE controller is also tempting, adding the capability to have 4 internal drives even in many older Macs. However if you need or dream of 40MB/sec+ RAID and/or support for a large number of drives, then IDE may leave you wanting. Ultra2 SCSI has the bandwidth to handle your current and future needs.

Summary:

The Initio Miles 2 was the best performer overall and had the lowest price tag. But the real winner here may be the low-cost IDE drives (and the consumer). Make no mistake - if you want high performance RAID or more than two devices per channel, then SCSI is the way to go. The performance potential and flexibility for very demanding users may justify the additional cost of SCSI over IDE.

SCSI's command queuing and higher bandwidth interface is the preferrred choice in a true multitasking environment. When high performance RAID is needed, the limitations of IDE can be a serious constraint. IDE RAID exists, but is far more limited than SCSI as far as number of drives supported and IO rates. Advertisements I've seen listed 24MB/sec as the ceiling, far below SCSI RAID performance with even low cost drives.

The good news is that even SCSI fans can be thankful for the proliferation of IDE drives, as the competitive market has helped dramatically lower costs of SCSI disks. I've seen prices drop hundreds of dollars in the last six months alone (and ATA/66 is just around the corner). Everyone wins as competition drives prices down and performance up.

In the final analysis, the average consumer or business user is well served by the new low-cost, high performance IDE drives. Those that push the limit will find Ultra2 SCSI, with the right drives and drivers at least, still the preferred choice.

Watch the main site daily news at www.xlr8yourmac.com for notice of an update covering tests in the B&W G3.



Related Links:


Back to WWW.XLR8YOURMAC.COM

Copyright © XLR8YOURMAC.com, 1999.
Thanks to my better half for tolerating all the time I spend on this site.

No part of this sites content or images are to be reproduced or distributed in any form without written permission. Please ask before using quotes or images.

All brand or product names mentioned here are properties of their respective companies.
Legal: Users of the web site must read and are bound by the terms and conditions of use.