Norm Why wrote:
Hi,

I have a 500GB Seagate Barracuda that comes with a 5 year warrantee. I'm
having problems where various files became corrupted and don't work. I've
exchanged it once. Seagate Tools said it was slow. Maybe I'm doing something
wrong under the assumption that SATA300 and SATA600 are compatible? There
are various PCIe SATA600 adapter cards available.

I bought a PCIe USB3.0 adapter card that now works OK. The first one was
junk.

Given this info, what are the chances that a PCIe SATA600 adapter card will
make the 500GB Seagate Barracuda work flawlessly?

Thanks in advance.

I'd make sure the SATA cable isn't bent and kinked, as
that can cause errors in packets on the cable.

A Q9650 era motherboard is likely to have SATA II ports on
it, operating at the 300MB/sec rate. A SATA III drive should
be able to negotiate a SATA II rate.

You can also use the FORCE jumper to drop the rate.
Some disks have a 1x4 pin block with two jumper positions.
One position is SpreadSpectrum, the other is Force.
Install the Force jumper with power off, power up
the drive, and do a bandwidth test. Now, you'll see
150MB/sec.

Errors can occur at platter level or at cable level.

The platters are protected by ECC.

The SATA cable has some sort of CRC and as far as I know,
the protocol allows retries. SMART should have a counter for
cable errors, but just in one direction. (The other direction
should be counted by the chipset end and OS driver.) Seeing
SATA cable errors means checking the condition of the cable.

This program allows benchmarking. It also has a SMART readout
in the Health tab.

http://hdtune.com/files/hdtune_255.exe

A PCI Express to SATA card, the SATA signals stay on the card
and are fairly short. Not all SATA chips are equal - there was
a first generation Marvell that couldn't do 600 on SATA III
and transferred at around 300 or so, for unknown reasons.
And no, it wasn't negotiating 300. It was running at 600
but something was limiting transfer rate, like maybe a processor
inside the chip was limiting dataflow. Silicon Image made mistakes
like that, with some of its products, using a processor too slow
for the dataflow. But Silicon Image was implementing RAID in
hardware, so there was an excuse (it made sense at least, for
a processor to be present, but it didn't make sense that the
processor could not keep up).

This is an example of HDTune output (note - hosting site has
annoying popup adverts). You can see that C7 "Ultra DMA CRC
Error count" is zero, indicating the drive is not seeing cable
errors in packets arriving from the host. The counter is sticky,
so if the counter reads 3 on Tuesday it will read 3 on Wednesday
as well, or it might even read 4 or 40 or 400. But it can't read
a lesser number, as the counter does not reset.

https://www.turboimagehost.com/p/613...ivors.gif.html

The benchmark curve on the drive with 50,000+ hours on it,
shows almost new transfer behavior. There are no bad spots evident.
The downward spike at the beginning might be for real (an
OS has lived there), but because the spike doesn't go down close
to zero, it's not all that bad.

I'd start with HDTune, and maybe try another data cable if you
haven't tried that already. If the drive is actually slow to
transfer, that should show up in the benchmark curve.

Paul