(PeteCresswell) wrote:
> My Computer | Manage | Device Manager | (view by type) | IDE
> ATA/ATAPI controllers | Secondary IDE Channel | Properties |
> Advanced Settings | Device 0
>
> In the PC in question, this is a Western Digital Caviar Green
> drive.
>
> On this one and all the others, "Current Transfer Mode" is "Ultra
> DMA Mode 5", which I understand tb only 100 mbps - whereas 6 is
> 133.
>
> OTOH, the "System" drive is a plain old IDE drive (Western
> Digital WD3200AAJB-00J3A0) hooked into the board's IDE connector
> and it is set to UDMA 6 - or 133 mbps.
>
> Can anybody elucidate?
>
> Could it be that the Caviar Greens are inherently slower than the
> WD3200? Doesn't seem logical that what is basically a legacy
> drive would be faster than a SATA drive.... or is it?
>
> Or is there something else at work?
I've answered your question, over in the Asus group.
I can add to that answer a bit.
Historically, Intel never participated in the UDMA 133 thing.
Maxtor offered UDMA 133, but Intel steadfastly refused to update
their chipsets and continued to offer UDMA 100. Some other
chipset companies added UDMA 133 support, so if you had
a Maxtor IDE drive and say a VIA chipset, maybe it would
actually run UDMA6 at 133MB/sec.
Since Intel showed no interest in UDMA6, it should come as no
surprise if the "fake" speed offered in the interface, reports
UDMA5 instead. And it's a fake speed - once you run HDTune benchmark
and examine the "burst" field in the lower right hand corner of
the window, it'll tell you the best speed it was able to achieve
over the SATA cable. The burst test, is effectively testing the
cache RAM on the hard drive (2MB, 8MB, 16MB, 32MB cache RAM or whatever
they included). By doing a short transfer to the drive, HDTune
can time the transfer and see how long it takes, if the transfer
is stored in the hard drive cache RAM (rather than doing a longer
transfer that must be stored on the media immediately to work).
If the program gets it right, then bursting to cache RAM will
give a feeling for the "true" max transfer rate.
The very first SATA drives, used a native IDE design, and tacked
an IDE to SATA bridge chip on it. The SATA cable would be rated
for 150MB/sec, but the native IDE design would either be limited
to 100MB/sec or 133MB/sec. If you do the burst test on such an old drive,
you should be seeing 100 or 133 in the burst window. That would be
identifying an internal bottleneck inside the drive. As far as I know,
hard drives now have native SATA controllers bolted to them, so that
should no longer be an issue. But you may see it on older drives.
You'd be more likely to see it on a SATA I drive, as any SATA II
drive should be well past the "bridged" design phase.
SATA transfer rates can also be bus limited. For example, if you
were to plug a SIL3132 PCI Express card into a PCI Express x1 slot,
the slot bandwidth is 250MB/sec max. That is below the SATA II cable
rate of 300MB/sec. You'd expect the burst test, in that case, to
report a result consistent with the PCI Express x1 bus limitation.
So any time you run tests like this, you need to consider the
architecture you're using, and where the bottlenecks are.
Paul
