Strange Intel Mobo Behavior

Tim Daneliuk tundra at tundraware.com
Mon Jun 4 06:23:22 UTC 2007


r17fbsd at xxiii.com wrote:
> At 02:08 PM 6/3/2007, Tim Daneliuk wrote:
>> 3) Both the MOBO and drive are SATA-300 rated, but 6.2 insists that
>>    the drive is running at SATA-150.  I have verified that the drive
>>    has no jumper forcing it into this mode.
> 
> Don't know about the other issues, but I have a Dell with similar Intel 
> components, and it did the same SATA thing when I put a new drive in it 
> yesterday:
> 
> atapci1: <Intel ICH7 SATA300 controller> port 
> 0xfe00-0xfe07,0xfe10-0xfe13,0xfe20-0xfe27,0xfe30-0xfe33,0xfea0-0xfeaf 
> irq 20 at device 31.2 on pci0
> ad4: 76293MB <Maxtor 6L080M0 BACE1G10> at ata2-master SATA150
> ad6: 305245MB <Seagate ST3320620AS 3.AAK> at ata3-master SATA150
> 
> The ad6 drive is supposed to do SATA-300, but realistically, other 
> bottlenecks dictate it's not going to get anywhere near the '150 speed, 

Could you comment a bit more on why you think this is so.  I would
think that with modern processors and buses, a machine with light load
ought to be able to drive SATA-300, but I've never actuall tested for
it myself.

> so I'm not terribly worried about it.  cp'ing a 4GB file to /dev/null 
> yielded 57MB/sec.
> 
>   -RW

I get around 50MB/sec or so with about 2G file, so we're in the same
ballpark.  In round numbers, this is 1/3 the theoretical throughput
of a SATA-150 or 1/6 that of SATA-300.  Now, I *am* curious on what
the bottlenecks are.  50MB/sec isn't a whole lot different that what
I'd expect out of a modern PATA drive.  So, noting the better cabling
and the wide availability of on-board RAID, it sure looks to me like there
is no compelling argument to be made for SATA in non-RAIDed environments.
I'm guessing the drives are the same ones as their PATA counterparts, just
with different interface electronics, so we're not going to see SCSI-like
reliability and/or performance under load.

I can understand some overhead due to system dispatching and multitasking,
but in a lightly loaded machine (as mine was when I did the test) with
2G of memory and dual 3.2G processors, it seems very strange that the
drive should run at 1/3 or less the stated interface speed.  What am I
missing here, I wonder...

----------------------------------------------------------------------------
Tim Daneliuk     tundra at tundraware.com
PGP Key:         http://www.tundraware.com/PGP/



More information about the freebsd-questions mailing list