Re: Gigabyte P35 Boards
I agree with all latest posts: drive interface isn't a limiting factor for a single drive (it may be on a raid configuration), and in fact I never stated pata drives are slower than sata counterparts (if the interface is the only difference of course: I won't compare old pata drives to newer sata ones because it's unfair). I wrote that disabling sata2 features, such as ncq, may affect performance, but I didn't explain why and this confused everyone.
IIRC Native Command Queuing is a feature which allow drives to optimize head movements in order to access to drive's sector with less revolutions, so theoretically increasing performance (heads spend a lot of milliseconds waiting to access data and read them, the real transfert rate is a lot inferior than continuous one if your files are fragmented so any improvement in access time is worthly unless we can afford some nice flash-based drives which addressing is row-column based and has a lower, constant latency); it's like scsi's tcq (but not the same thing) and requires explicit support from drive controller (drive's board), sata interface controller (chipset or expansion card) and os (drivers). NCQ should require AHCI in order to work and AHCI is available on Sata2 controllers (perhaps some sata1 too, but I can state this for sure), hence my statement.
BTW, according to Intel the ICH9 doesn't officialy support AHCI on XP (but the ICH9R does and on Vista it works on both versions!) and bgavin's Gigabyte employs this southbridge: this is a [marketing?] limitation in the Intel driver which can be resolved through a bit of driver hacking (nothing serious, only changing a line in the .ini file and adding a registry string). MSFN forum has a detailed guide about this.
Zandrax
I agree with all latest posts: drive interface isn't a limiting factor for a single drive (it may be on a raid configuration), and in fact I never stated pata drives are slower than sata counterparts (if the interface is the only difference of course: I won't compare old pata drives to newer sata ones because it's unfair). I wrote that disabling sata2 features, such as ncq, may affect performance, but I didn't explain why and this confused everyone.
IIRC Native Command Queuing is a feature which allow drives to optimize head movements in order to access to drive's sector with less revolutions, so theoretically increasing performance (heads spend a lot of milliseconds waiting to access data and read them, the real transfert rate is a lot inferior than continuous one if your files are fragmented so any improvement in access time is worthly unless we can afford some nice flash-based drives which addressing is row-column based and has a lower, constant latency); it's like scsi's tcq (but not the same thing) and requires explicit support from drive controller (drive's board), sata interface controller (chipset or expansion card) and os (drivers). NCQ should require AHCI in order to work and AHCI is available on Sata2 controllers (perhaps some sata1 too, but I can state this for sure), hence my statement.
BTW, according to Intel the ICH9 doesn't officialy support AHCI on XP (but the ICH9R does and on Vista it works on both versions!) and bgavin's Gigabyte employs this southbridge: this is a [marketing?] limitation in the Intel driver which can be resolved through a bit of driver hacking (nothing serious, only changing a line in the .ini file and adding a registry string). MSFN forum has a detailed guide about this.
Zandrax
Comment