Originally posted by: VirtualLarry
Originally posted by: Redviffer
By going from the IDE to SATA you take information off the PCI bus, which is limited to 133MB/sec, the performance increase isn't really the drive, but in how you are accessing/transferring the information.
Doesn't really make much of a differance with only 1 drive, but if you raid then you will see the benefit.
Uhm, aside from the few very recent chipsets, with integrated SATA, most instead have their PATA controllers not on the PCI bus (since PATA has been "integrated" into chipsets for a long time), but the SATA is implemented using an add-on chipset, which is bridged to the PCI bus.
So for any of those motherboards, you would see an overall performance DECREASE, if you moved your drives from PATA to SATA.
Also, VaG's argument that somehow, because the highest-performing SATA HD, outperforms the highest-performing PATA HD, that the natural conclusion must be therefore that SATA outperforms PATA, is false.
Couldn't I use the same erronous logic, to state that by moving from an ATA (parallel OR serial), to Ultra320 SCSI, that I could shave nearly 3-4 ms off of the access time? Certainly, the performance increase MUST be due to the interface chage, and not due to the mechanics of the HDs themselves, of which the 15K RPM drives are currently only available in U320 SCSI interfaces, and not IDE.
For all of the benchmarks that I've read so far, between otherwise identical models of SATA and PATA HDs, their performance is basically identical, and sometimes the SATA version is actually slightly slower, due to two factors, one being that the mobo SATA controller chip is likely to be bridged to the PCI bus, instead of on some faster proprietary inter-chipset bus, and another due to early SATA model HDs themselves, using a PATA-to-SATA converter chip on the HD, basically.
It is true, that unless PATA evolves another speed revision to the interface (UDMA mode greater than 6), then the theoretical maximum for the SATA-I interface (150MB/s), is slightly faster than the PATA interface (UDMA mode 6 == ATA133 == 133MB/s). However, "native command-queueing", is only an optional feature in SATA-I, and will be manditory in SATA-II, as I understand it. That one singular feature, is what could potentially differentiate HD performance in a multi-drive array. I've some experience with TCQ on PATA HDs, IBM GXPs since the 14GXP support it, as do most of the Promise controller cards/drivers. Let me tell you, the performance difference on a loaded desktop system was very noticable. I wish that all ATA HDs, parallel OR serial, would support this feature. It allows IDE to approach the performance efficiency of SCSI, which still IMHO can't be touched by other alternative high-bandwidth storage interfaces for local storage.
I don't plan on moving to SATA for another 2 years, at least. I DO plan on taking advantage of as many deals as I can, on cheap PATA storage, due to the erroneous perception that PATA is somehow "slow", or "obsolete".
(And for all of the performance advantages of a single WD Raptor HD, I can easily match those same performance levels, with a (cheaper) PATA RAID array, and have far MORE storage space to boot. This is in terms of a home-user scenario, not a low-cost enterprise-level storage server/subsystem, in which the hot-pluggability advantage of SATA does come into play.)
Edit: And yes, I AM a nerd.