whats *your* ($hdparm -t) speed?

whats *your* ($hdparm -t) speed?

Post by MGatt » Wed, 19 Apr 2000 04:00:00



What is everyone here getting for $hdparm -t
speeds?  I'm especially interested in those w/ ultra DMA or SCSI
controllers cards.  I have a promise U66 w/ maxtor's latest and
greatest, and am still only getting on avg. 15 MB/sec.  (half of what
winbench reports for windoze!)
--
~MGatto~

"Tech support?!!??! We don't need no stekin tech support!?!"
Support the anti-spam movement; see <http://www.cauce.org/>

Sent via Deja.com http://www.deja.com/
Before you buy.

 
 
 

whats *your* ($hdparm -t) speed?

Post by Michael Meissne » Wed, 19 Apr 2000 04:00:00



> What is everyone here getting for $hdparm -t
> speeds?  I'm especially interested in those w/ ultra DMA or SCSI
> controllers cards.  I have a promise U66 w/ maxtor's latest and
> greatest, and am still only getting on avg. 15 MB/sec.  (half of what
> winbench reports for windoze!)

Let's see, for my Quantum Atlas 10K 9WLS (this is a 10,000rpm ultra2 drive), I
get 23.88 MB/sec.  For my 2 IBM 10K drives I get 18.93 MB/sec and 19.28 MB/sec,
with lesser amounts for the older disks.

--
Michael Meissner, Cygnus Solutions, a Red Hat company.
PMB 198, 174 Littleton Road #3, Westford, Massachusetts 01886, USA



 
 
 

whats *your* ($hdparm -t) speed?

Post by fREDDie » Thu, 20 Apr 2000 04:00:00


When I first enabled udma66 in kernel 2.2.13, I ran hdparm 4 times and got
an avg. transfer speed of 20.8mb/sec. Subsequent testing showed the speed as
high as 24.0mb/sec.


Quote:> What is everyone here getting for $hdparm -t
> speeds?  I'm especially interested in those w/ ultra DMA or SCSI
> controllers cards.  I have a promise U66 w/ maxtor's latest and
> greatest, and am still only getting on avg. 15 MB/sec.  (half of what
> winbench reports for windoze!)
> --
> ~MGatto~

> "Tech support?!!??! We don't need no stekin tech support!?!"
> Support the anti-spam movement; see <http://www.cauce.org/>

> Sent via Deja.com http://www.deja.com/
> Before you buy.

 
 
 

whats *your* ($hdparm -t) speed?

Post by Steffen Klu » Thu, 20 Apr 2000 04:00:00




>When I first enabled udma66 in kernel 2.2.13, I ran hdparm 4 times and got
>an avg. transfer speed of 20.8mb/sec. Subsequent testing showed the speed as
>high as 24.0mb/sec.

Hmm, this is what I'm getting with an Ultra-33 drive (Seagate
8GB, kernel 2.2.14). Are you sure you're running Ultra-66? Maybe
it's all but hype and no disk can actually deliver (yet) what
Ultra-66 promises? Did you try testing in single-user mode?

Cheers
Steffen.

--

Fujitsu Australia Ltd
Keywords: photography, Mozart, UNIX, Islay Malt, dark skies
--

 
 
 

whats *your* ($hdparm -t) speed?

Post by Jerry Natowi » Thu, 20 Apr 2000 04:00:00





>> What is everyone here getting for $hdparm -t
>> speeds?  I'm especially interested in those w/ ultra DMA or SCSI
>> controllers cards.  I have a promise U66 w/ maxtor's latest and
>> greatest, and am still only getting on avg. 15 MB/sec.  (half of what
>> winbench reports for windoze!)

>Let's see, for my Quantum Atlas 10K 9WLS (this is a 10,000rpm ultra2 drive), I
>get 23.88 MB/sec.  For my 2 IBM 10K drives I get 18.93 MB/sec and 19.28 MB/sec,
>with lesser amounts for the older disks.

I get 22.22 MB/sec for my 20GB IBM drive (7200 RPM), and 13.01 MB/sec for
my 13GB Western Digital (5400 RPM).  Both support UDMA66, but the transfer
rate doesn't change.
--
     Jerry Natowitz - jin at spdcc dot com
 
 
 

whats *your* ($hdparm -t) speed?

Post by Johan Kullsta » Thu, 20 Apr 2000 04:00:00





> >When I first enabled udma66 in kernel 2.2.13, I ran hdparm 4 times and got
> >an avg. transfer speed of 20.8mb/sec. Subsequent testing showed the speed as
> >high as 24.0mb/sec.

> Hmm, this is what I'm getting with an Ultra-33 drive (Seagate
> 8GB, kernel 2.2.14). Are you sure you're running Ultra-66? Maybe
> it's all but hype and no disk can actually deliver (yet) what
> Ultra-66 promises?

no hard disk can deliver data faster than about 30MB/sec.  ata-66 is
all about room for expansion.  iirc ata-66 has a mode to fire multiple
commands and wait for reply so that putting two disks on one channel
may not be so bad in the future (once hard disks, and ata controllers
and their drivers catch up).

scsi has had 40MB/sec for a long time now and recently gotten 80 and
160MB/sec extensions.  this isn't because any single drive can exceed
40MB/sec (let alone 160MB/sec), but if you've got a bank of 6 drives
you can be bus constrained.

--
johan kullstam l72t00052

 
 
 

whats *your* ($hdparm -t) speed?

Post by Martin H?yer Kristianse » Thu, 20 Apr 2000 04:00:00






> >> What is everyone here getting for $hdparm -t
> >> speeds?  I'm especially interested in those w/ ultra DMA or SCSI
> >> controllers cards.  I have a promise U66 w/ maxtor's latest and
> >> greatest, and am still only getting on avg. 15 MB/sec.  (half of what
> >> winbench reports for windoze!)

> >Let's see, for my Quantum Atlas 10K 9WLS (this is a 10,000rpm ultra2 drive), I
> >get 23.88 MB/sec.  For my 2 IBM 10K drives I get 18.93 MB/sec and 19.28 MB/sec,
> >with lesser amounts for the older disks.

> I get 22.22 MB/sec for my 20GB IBM drive (7200 RPM), and 13.01 MB/sec for
> my 13GB Western Digital (5400 RPM).  Both support UDMA66, but the transfer
> rate doesn't change.

hdparm -t measures how fast data can be read of the disk, and not how
fast cached data can be transferred.

The benefit of UDMA 66 in current systems is very modest, because
current controllers are limitid to one UDMA 66 device per controller.
The point of higher bandwidth is that the controller would be able to
saturate more devices, but because of the one device restriction on UDMA
66, performance are dominated by the slower device-interfaces.

SCSI beats the *out of IDE if disks > 2 (maybe even 1)

cheers
Martin

 
 
 

whats *your* ($hdparm -t) speed?

Post by fREDDie » Thu, 20 Apr 2000 04:00:00


Of course I'm sure it's an ultra-66 drive. Enabling it involves a kernel
hack in 2.2.13. And keep in mind that consistently reaching the max burst
speed of 66 is by most accounts not realistic, no matter what Promise
"promises". But it works fast enough for me, I'm satisfied.




> >When I first enabled udma66 in kernel 2.2.13, I ran hdparm 4 times and
got
> >an avg. transfer speed of 20.8mb/sec. Subsequent testing showed the speed
as
> >high as 24.0mb/sec.

> Hmm, this is what I'm getting with an Ultra-33 drive (Seagate
> 8GB, kernel 2.2.14). Are you sure you're running Ultra-66? Maybe
> it's all but hype and no disk can actually deliver (yet) what
> Ultra-66 promises? Did you try testing in single-user mode?

> Cheers
> Steffen.

> --

> Fujitsu Australia Ltd
> Keywords: photography, Mozart, UNIX, Islay Malt, dark skies
> --

 
 
 

whats *your* ($hdparm -t) speed?

Post by pe.. » Thu, 20 Apr 2000 04:00:00



Quote:> The benefit of UDMA 66 in current systems is very modest, because
> current controllers are limitid to one UDMA 66 device per controller.

Does this mean one *active* UDMA-66 device or one UDMA-66 device
period?  Are you allowed one UDMA-66 and one UDMA-33 device?
I haven't bothered to take the plunge to UDMA-66 quite yet...

Regardless of these restrictions, I would *never* put more than one
device on an IDE controller if speed were an issue, especially since
IDE controllers are so cheap (even the UDMA-66s nowdays)

Quote:> The point of higher bandwidth is that the controller would be able to
> saturate more devices, but because of the one device restriction on UDMA
> 66, performance are dominated by the slower device-interfaces.

Not entirely true.  7200 and 10K RPM drives will show better
throughput on UDMA-66 vs. UDMA-33 controllers in a single drive per
controller situation.  (or so I have read in several places.)

Quote:> SCSI beats the *out of IDE if disks > 2 (maybe even 1)

A Ferrari will beat the *out of a Hyundai, so what?  There is no
comparison here.

Dollar for dollar, IDE will buy you much more performance for your
money in all single user situations and most multi-user situations, up
to the performance limitations of IDE.

If I needed 40 gigs of storage for an enterprise database, you had
better believe I'd build it with four 9G U2W SCSI drives on two
controllers.  Even w/o hardware RAID, that would be US$1000-$1500.  If
I was just storing MP3s, I could get away fine with 1 5400 RPM 40G
drive.

On my desktop machine (dual P2-333, 66mhz bus) I can tell absolutely
no difference between a 7200RPM UDMA-33 drive and a 7200 RPM UW SCSI
drive.  YMMV.

-p.

 
 
 

whats *your* ($hdparm -t) speed?

Post by Doc Shiple » Fri, 21 Apr 2000 04:00:00



> What is everyone here getting for $hdparm -t
> speeds?  I'm especially interested in those w/ ultra DMA or SCSI

<BIG grin>
 Don't try it on a full-blown SuSE box at 12:00:30 AM

Would you believe on a 9G lvd drive behind a Tekram DC390F UW - 2.84
MB/sec

--
 Doc Shipley
 Network Support
 TARL Labs, UT
 Austin, Texas

 
 
 

whats *your* ($hdparm -t) speed?

Post by CaT » Tue, 25 Apr 2000 04:00:00






>>> What is everyone here getting for $hdparm -t
>>> speeds?  I'm especially interested in those w/ ultra DMA or SCSI
>>> controllers cards.  I have a promise U66 w/ maxtor's latest and
>>> greatest, and am still only getting on avg. 15 MB/sec.  (half of what
>>> winbench reports for windoze!)

>>Let's see, for my Quantum Atlas 10K 9WLS (this is a 10,000rpm ultra2 drive), I
>>get 23.88 MB/sec.  For my 2 IBM 10K drives I get 18.93 MB/sec and 19.28 MB/sec,
>>with lesser amounts for the older disks.

> I get 22.22 MB/sec for my 20GB IBM drive (7200 RPM), and 13.01 MB/sec for

I bet we have the same HD. I get this figure aswell with my 20Gig IBM. :)

My other IBM only does 12.61MB/sec. :/

--

        'He had position, but I was determined to score.'
                -- Worf, DS9, Season 5: 'Let He Who Is Without Sin...'