|
Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home |
. | . |
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ |
LOSTCIRCUITS |
||
| Fujitsu MHT2060AH SA, HighPoint RocketRAID 1640 and Level5 RAID All in an XPC | ||
|
(Review by MS, April 23, 2004) | ||
|
Fujitsu MHT2060AH ATA-100 |
RAID Level5 Performance
For those not familiar with the principles of RAID Level5 I highly recommend our As the HDD spins" series. Briefly, all data are striped into blocks and the parity value of all data is written to a separate drive on a rotating basis, thus avoiding the overproportional stress on the "parity drive" that is a feature of RAID Levels 3 and 4. Also, whereas RAID Level5 can be rather fast in read applications and makes best use of the HDD capacity available, it is burdened with low write performance since every write consists of several steps from updating the data over modifying the metadata and finally writing the parity values to the respective drive. In addition, while a Pentium4 should have enough power to handle the parity calculations, it is not made for this kind of operation which reflects in the CPU usage as we'll show below.
Sequential Performance
For grins, this is what HDTach 2.61 generates and 2.70 is not any prettier. The "low" performance results from the HDD internal performance being faster than what the interface can handle which results in a buffer overflow of the cache. As a consequence, the heads cannot write data from the platters into the cache --- but meantime, the disc spins on and the drive has to wait one full rotational latency until the requested LBAs come under the head again. The benchmark was run with a PS/2 mouse as pointing device.
HDTach / USB Mouse
One interesting effect we noticed was that using a USB mouse changed the trace characteristics of HDTach. Overall the traces became even more noisy than they were before. This is not an artifact since we could replicate the findings using WinBench99.
WinBench99 Drive Inspection Test
With the original 2.0 BIOS and drivers, WB99 only generated a pattern similar to that we saw in HDTach (the HDTach results were all run after flashing to the 2.03 BIOS and loading the 2.03 drivers - for simplicity reasons and better resolution, a 6 GB partition was used here). Upgrading to the latest drivers and BIOS noticeably improved the situation, however, note the cyclic depressions in the sequential reads when a USB mouse was used (center - 12 GB partition at OD). Changing to PS/2 mouse reduced the noise levels considerably (right).
Interestingly, the overall sequential performance is in the order of what one would expect from bundling three drives together in a RAID0 configuration, namely approximately 90 MB/sec sustained transfer at the drive's OD.
next page: => ATTO, SiSoft Sandra =>
If you enjoyed reading this article and found it useful, please consider making a small donation to LostCircuits.