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:

In Serial ATA, the situation is much more streamlined. SATA drives use a point to point connection without being interrupted from the controller. Therefore. as soon as a SATA drive receives a data request, the drive itself, being a bus master, can set up the DMA channel itself and start transferring data to the HBA. This streamlined protocol is referred to as First Party DMA or short FPDMA.

From the above, especially when comparing it against the handshaking explained on the last page, it should be clear that the SATA specifications will dramatically reduce the command overhead, the number of FIS required alone is only half and, moreover, the streamlined protocol will dramatically improve transfers especially of smaller files as they are encountered in standard gaming and office applications.

SATA "Out Of Order Data Delivery" To Reduce Rotational Latencies

The main focus of SATA from a protocol standpoint is the reduction of command overhead and the concomitant reduction of latencies. However, keep in mind that latencies in HDDs are mainly mechanical latencies with rotational latencies being the main issue to worry about. However, even there, SATA offers an improvement, which is only possible because of the implementation of FPDMA.

That is, it is written into the SATA specifications that the drive does not need to begin data transfers at the starting LBA, rather, it can start at any LBA within the requested range and add the data from the "earlier LBAs" after getting to the "missed" starting point during the ongoing rotation. This "Out Of Order Data Delivery" can dramatically reduce the rotational latency since the highest rotational latency incurred for the entire transfer is one full rotation. Without "Out Of Order Data Delivery", the rotational latency is one full rotation plus the initial rotational latency.

Almost a "worst case scenario", the head has missed the first requested LBA (1) in a series of 10 LBAs as part of one coherent data transfer. In any standard scheme, the drive would need to wait one full rotational latency before the head can start reading the data. In SATA, however, courtesy of the "Out Of Order Data Delivery" as part of the SATA specs, the drive can start reading and transferring the data to the HBA at any given LBA within the requested senquence. In the case illustrated, this would cut the rotational latencies for the entire transfer almost in half.

Out Of Order Data Delivery is also an excellent example for another benchmark criterion becoming rather obsolete since the relevance of the so called Random Access Time of a SATA drive is being is being reduced to be of almost purely academic interest. In other words, why would anyone care about the Random Access Time, when in most cases, the data transfers can be initiated much earlier and the total service time becomes almost exclusively a function of the rotational speed.

This time will be identical for all drives running at the same spindle speed. The only other factor that matters is the seek latency, which, in about 45% of all real life cases will be a track-to-track latency and that one is in the order of 0.7 to 0.85msec for all drives currently available - therefore, differences are almost non-existent. As it stands, there are no good and/or free benchmarks to evaluate these parameters, at least not at the moment -- if they were, however, they would all show the same results for all drives.

next page:    => Stay Tuned For Command Queuing =>