> The new 2.1 kernels do the Triton busmastering in a completely different
> way which should work with more or less every Triton chip out there,
> including the ones on the 430VX/430TX and all the 440 boards.  I'm not
> entirely sure how much improvement they get on UDMA type drives.  In any
> case, I would predict 2.2 (the stable series incorporating all the 2.1
> improvements) will be out in the next couple of months, so my guess is
> that RedHat 5.2 will include this kernel version. :)

here's hoping....

> In any case, the difference in performance is not too great, unless your
> applications are very disk intensive.  If they are, then any server or
> multiuser system should be using SCSI anyway.  UDMA drives are still IDE
> drives, and they still impose excessive loads on the CPU.  They just don't
> impose as much as they used to. :)

well, hdparm -tT says there isn't much difference between PIO and DMA
modes for EIDE in terms of thruput, but in the real world DMA mode is
a huge gain. the difference for my EIDE drives on my old triton 1
system is night and day. with DMA off any kind of disk activity
noticeably affects interactive performance. with DMA on the drives can
grind away indefinitely and i don't know the difference other than the
noise. i also have SCSI drives on an NCR810 PCI controller and some
simple benchmarks i've done (cp of large files, dd to and from the
drive) show that if anything the SCSI drives use a little more CPU
than the EIDE drives. where SCSI rocks is when i do transfers between
drives. SCSI to SCSI can be almost twice as fast as EIDE to EIDE even
though the disks in question have very similar thruputs (or in one
case EIDE and SCSI models of the same drive). 

> >     Got the Fujitsu purely because it was cheaper.  Have had good luck with
> > them at work, but like anything else, you get what you pay for.  Think I'll
> > stick with this one.
> 
> I've heard some horror stories about the Fujitsus.  Of course I have a few
> of them myself and none of them ever so much as blinked.  I guess it all
> depends on what "batch" your drives come out of.  In any case, Fujitsu
> seems to be pretty good about RMA's on their defective drives...

yeah, the ones that went bad here were replaced very promptly. western
digitals have been dropping like flies around here so i wouldn't touch
them with a 10' pole. we have a lot of seagates and i can't think of
any that have died other than the 'cudas that succumbed to
insufficient cooling. i like IBM drives a lot, but they're harder to
find and tend to cost a little more. i kinda worry sometimes that
reliability is taking a back seat to all of the price slashing on HDs.

tim

-- 
  Time is like fingers        A    |  Tim Pickering ([EMAIL PROTECTED])
Gesturing for me to stop -- Quake  |  UA Steward Observatory
   Why is the sun up?       Haiku  |  Tucson, AZ  85721     520-621-6523


-- 
  PLEASE read the Red Hat FAQ, Tips, Errata and the MAILING LIST ARCHIVES!
http://www.redhat.com/RedHat-FAQ /RedHat-Errata /RedHat-Tips /mailing-lists
         To unsubscribe: mail [EMAIL PROTECTED] with 
                       "unsubscribe" as the Subject.

Reply via email to