Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors Home PC Hardware from Made-in-China.com Search: Enter your search terms   Web www.lostcircuits.com Submit search form	.
Prices:

LOSTCIRCUITS		SHORTCUTS: The Challenge Fujitsus and Promises Assembling the RAID Building the System Test configuration HDTach vs. WB99 ATTO vs. Sandra Back to HDTach, WB99 Network File Copy Conclusions Fujitsu Drives Comments?
Internal 4-Way SATA RAID For The XPC     A Concept Study
(Review by MS, April 20, 2003)

Another benchmark that shows the same parameters using a slightly different approach of exponentially increasing transfer block size is ATTO Disk Benchmark. ATTO numbers are interesting because they show the impact of the command overhead on the overall transfer rate, depending on the sector count that is transfered. After a certain transfer block size, the effective internal media performance becomes the limiting factor and puts a cap on the I/O activity. The difference between ATTO and WB99 results mostly from the fact that ATTO averages several data points over time instead of showing each LBA. The overall read or write transfers shown are equivalent to the average of the jitters in WB99, therefore, both benchmark results are, in fact, showing identical performance.

The platters saturate at about 87MB/s sustained transfer for sequential read and some 70 MB/s for sequential writes. An interesting side dish of ATTO is that the overall transfer at any given block size divided by the sector count (1 sector = 512 Bytes) gives the number of I/Os per second. Much easier than to install and to run IOMeter, which essentially shows the same.

The last benchmark in this context is SiSoft Sandra which basically combines all relevant aspects of Winbench99 disk inspection test, ATTO and HDTach, with the exception of showing the CPU utilization which is something that is really only relevant for Windows95 users before IDE busmaster drivers were invented. SiSoft Sandra does not measure across the entire drive / platter width but stays at the OD of any given partition, which makes this benchmark more relevant for the desktop user anyway.

Once again, the numbers are the same, we have the 102 MB burst reads of HDTach and the sequential performance of WB99 and ATTO using the same averaging algorithms as ATTO. In other words, we could show 10 benchmarks that all show the same (within the variation of individual runs) instead of one that shows it all and is more reliable and robust than the others. Keep in mind that Sandra does not allow to calculate mumber of I/Os depending on transfer size, neither does it show the jitters ot WB99 nor CPU utilization. Whoever is interested in those tidbits, admittedly they are valuable for trouble shooting, will have to resolve to the individual benchmarks. For performance purposes, there is really no point in doing the yadda.. yadda ...yadda.

next page:    => HDTach revisited, WB99 Disk Winmarks =>