Navigate: 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 Search Prices:

Conclusion

After several months of testing and going through different driver and BIOS versions, and in finally ending up throwing away the bulk of all benchmark results because a slight change in some binaries made them as obsolete as the conclusions and interpretations we derived from the results, this is finally coming to an end. There are numerous holes left in the benchmarks but one needs to draw a line somewhere and we are not going to post some 20 pages of irrelevant or redundant statistics and benchmarks. we have a number of take-home messages, though, and let's go through them one by one.

Level5 RAID For The Home User

The increasing requirements of reliability and storage space make RAID Level5 a very appealing solution, at least on paper. In reality, there are a number of restrictions. first of all, it appears as if Software Level5, that is without any dedicated XOR processor does not really cut it. This is not aimed against HighPoint, on the contrary, we have to applaud their almost heroic effort on the design and implementation of Level5 on a sub $100,- RAID controller card. It is just that apparently a standard desktop processor is way out of its league when it comes to XOR processing, otherwise, how would a 3.2 GHz Pentium4 completely bog down under any sustained write load. Of course, this opens up an entire new field of benchmarks revolving around XOR processing, using the write performance of a soft Level5 RAID configuration as a CPU benchmark.

RocketRAID 1640

The RocketRAID 1640 has a few other interesting anomalies, for example the very low burst transfer rates we saw in HDTach are in stark contrast with most of the single drive and RAID benchmarks we saw in any of the other benchmarks. Whether this is simply another bug of HDTach or whether the other benchmarks are misreporting is something we cannot answer with the means at our hands.

What was interesting also was the increase in performance with a new BIOS version and drivers that, unfortunately, had the undesirable side effect of intermittently stalling the USB mouse functionality among others. What that means in most cases is that the performance of one single device is boosted by setting the "Current Privilege Level" (CPL) in the driver code to a low numerical value (0 or 1 for kernel or superuser CPL, respectively) which allows the driver and the device to maintain priority over all other processes running on a system level. In a case like that, e.g. mouse input would be deferred until any fairness algorithm kicks in and allows other devices to arbitrate for system resources. Clearly, this might improve performance of the device in question but it would happen at the expense of some system performance. Interestingly, however, the negative effects were not as pronoounced with HyperThreading enabled. In any event, the RocketRAID has some very interesting capabilities and appears worth its price as long as it is used in the correct environment with the correct tasks and in a suitable configuration. One example for a Level5 configuration would be the use as a webserver, where the write performance is of subordinate importance and space, read performance and redundancy matter most.

Rebuilding the Level5 Array

We have no benchmarks to show how long this takes, the reason is very simple, it depends on the content and the degree of filling up of the drive how long it takes to rebuild an array in case one drive is lost. We formatted one drive on a different system and put it back into the Level5 array that did not contain any data and the drive was immediately reintegrated into the array with full functionality. In other words, we could make up some benchmarks here but what would they really show?

Fujitsu MHT2060AH-SA

In many ways this little HDD has been the hilight of the review. In all honesty, the performance of these drives is almost in the same league with any but the most recent ATA100 drives, only the size and weight are only a fraction. The interface is mad fast, the drives support SATA 2.0 specs including NCQ and the main thing that sets them apart from the higher rpm drives is the rotational latency. Rotational latencies will, however, lose their importance with the introduction of out of order delivery, one of the features enabled by Native Command Queuing. All in all, the Fujitsu MHT2060AH SA is the first hard disk drive we are giving an Editor's award for overall performance and feature set.

next page:    => More =>