HD Tune: QUANTUM FIREBALL CR8 test results

Firmware version : A5U.1200
Serial number : 128909171916
Capacity : 7.9 GB (~8.5 GB)
Buffer size : 418 KB
SMART : Yes
Standard : ATA/ATAPI-4
Supported mode : UDMA Mode 4 (Ultra ATA/66)
Current mode : UDMA Mode 4 (Ultra ATA/66)

Transfer Rate Minimum : 9.1 MB/sec
Transfer Rate Maximum : 14.6 MB/sec
Transfer Rate Average : 12.7 MB/sec
Access Time : 14.3 ms
Burst Rate : 51.9 MB/sec
CPU Usage : 2.1%

HD Tune: WDC WD1200JB-00DUA3 test results

Firmware version : 75.13B75
Serial number : WD-WMACM2128806
Capacity : 111.8 GB (~120.0 GB)
Buffer size : 8192 KB
SMART : Yes
Standard : ATA/ATAPI-6
Supported mode : UDMA Mode 5 (Ultra ATA/100)
Current mode : UDMA Mode 5 (Ultra ATA/100)

Transfer Rate Minimum : 26.4 MB/sec
Transfer Rate Maximum : 47.8 MB/sec
Transfer Rate Average : 40.5 MB/sec
Access Time : 13.4 ms
Burst Rate : 59.0 MB/sec
CPU Usage : 7.0%

HD Tune: FUJITSU MPE3084AE test results
Supported mode : UDMA Mode 4 (Ultra ATA/66)
Current mode : UDMA Mode 5 (Ultra ATA/100)
Burst Rate : 2.3 MB/sec
CPU Usage : 99.0%
Машина: Cel 400, MB: PC-Chips (Sis 620), 64 RAM

What's a good hard drive benchmark?


The Official StorageReview.com Take

In ascending order, from worst to best:

1. SiSoft? Sandra 200x

Strengths: None

Weaknesses: Inconsistencies between test runs; unrealistic results for its access time and transfer rates; no facilities for application-level testing.


2. ATTO Powertools

Strengths: Measures STR using several different block sizes.

Weaknesses: STR is NOT representative of real-world performance. The program is susceptable to a Windows XP issue involving SCSI that many if not most applications are not.


3. HDTach 2.xx

Strengths: Measures read and write STR. Presents a reasonable measurement of access time.

Weaknesses: STR is NOT representative of real-world performance. The STR measurements are taken from sample sectors staggered across the drive, often rendering results for RAID configurations useless. This is because this method often means a particular sample reading doesn't cross stripe boundaries, therefore producing results equal or similar to single-drive results.


4. IOMeter

Strengths: Highly-configurable synthetic tests that can reasonably simulate multi-user performance.

Weaknesses: Does not take locality effects into account and thus is totally useless for assessing single-user performance.


5. WinBench? 99

Strengths: Exhaustive, repeatable, and accurate read STR measurement. Reasonably accurate random access time measurement. Disk WinMarks? are one of the few tools that actually attempt to replicate application-level access patterns.

Weaknesses: No provision for testing write STR speeds. STR is NOT representative of real-world performance. Disk WinMarks? were recorded using dated operating systems and applications and "double cache" (weigh the effects of OS-level caching in both the initial record and subsequent playback stages). More information on WB99's deficiencies may be found here (cache).


6. StorageReview Desktop DriveMark 2002

Strengths: Exhaustive, repeatable, and accurate application-level performance measurements using a contemporary OS and application suite. Avoids the key WB99 Disk WinMark? shortfall of double-caching.

Weanesses: Not publicly available; results published only at SR.


Summary: There is no publicly available benchmark that can accurately simulate single-user performance. As a result, individuals are limited to using accurate low-level benchmarks as a diagnostic rather than a measure of speed. For this purpose, WinBench? 99's Disk Inspection suite remains best. If a given sample's WB99-measured access time and STR matches the results published by SR, then it is safe to assume that the drive's application-level performance also matches those found in the SR review.

For multi-user performance, IOMeter remains a viable choice. Remember, however, that it can not assess single-user performance. Otherwise commendable sites such as xbitlabs.com unfortunately perpetuate this flawed use of IOMeter.




Additional comments by reader skrossklass:


Windows
Probably the best freely available hard drive benchmark is German c't magazine's h2benchw, which they use themselves when reviewing hard drives. It measures the interface and physical transfer rates, access time, includes some application patterns and can do data integrity testing. Text and Postscript log (tranfer rate graph, if created is found in the latter). Requires an empty unpartitioned disk (write access) for accurate application benchmark results. Works on device level. Command line tool for Windows NT, 2000, XP, Server 2003, etc. German, English and Dutch interfaces.
A graphical benchmark suite that used to be quite popular in days gone by and is in terms of hard drive performance still useful for measuring the read access time and to create nice STR graphs is ETestingLabs' WinBench 99 (cache). The drive benchmarks operate on partition level. Test results can be saved to its database, the transfer rate graph can be exported in Windows bitmap (BMP) format. Virtually any 32-bit Windows starting from 95 and NT. English interface.
StorageReview uses a suite of utilites known as the Intel Performance Evaluation and Analysis Kit: Storage Performance Toolbox, (IPEAK SPT) for short, to evaluate typical single-user drive performance. It was an $800 commercial program that has unfortunately been discontinued. For server-related performance measurement, SR uses IOMeter (cache). Note that while IOMeter is capable of measuring performance under multi-user loads, it is not capable of assessing single-user performance.