Actually, that works out just about right: your 16.5MB compressed file
took 7 seconds to load on a computer that has a CPU about 2.2 times
faster than mine.
My 53MB compressed file is 3.2 times the size of yours, so if I take
the 7 seconds and multiply that by 3.2 to account for the difference in
file size, and then by 2.2 to account for your faster CPU, I come up
with 50 seconds, while I measured 61 on my computer. But no one would
claim a Celeron is as efficient with the MHz as the Athlon, so I think
we can safely say the 20% penalty I am seeing is expected.
Again, we have shown that in this case the drive speed is relatively
unimportant compared to the CPU's speed and efficiency (remember, when
the whole file was in cache so the drive wasn't even accessed it still
took my computer 55 seconds to load the file). If I'm going to do many
compressed TIFFs on this computer I would be better served with a
faster (and better) CPU than I would be with a faster drive.
And as you have amply demonstrated, even on your computer the
compression changed the load time from 4 to 7 seconds. That means even
if *all* the 4 seconds is due to drive throughput limitations, and you
went from a 7200 rpm drive to a 10,000 rpm drive (a 39% increase in
rotational speed), you would realize only a 22% increase in throughput.
Probably less, because I suspect that at least some of that 4 seconds
is due to processor activity.
However, if you increased the processor throughput by the same amount
you would get even less of a boost than by increasing the drive
throughput. Your system has pretty well-matched bottlenecks -- you
can't get a major boost dealing with compressed TIFFs by just upgrading
one thing.
--Dana
----------
From: geoff murray <geoffmurray@primus.com.au>
To: filmscanners@halftone.co.uk
Subject: Re: filmscanners: Scanning and memory limits in Windows
Date: Sunday, July 29, 2001 6:34 PM
Hi Dana,
I have just scanned an image and saved it as compressed and
non-compressed files. This particular image surprised me in that it
compressed to a remarkable degree, from 55.3mb to 16.5mb. I presume
different images will compress to different degrees depending on their
content. But, I still disagree that it should take so long to open an
image.
The uncompressed file opened in 4 seconds flat and the compressed file
opened in 6.95 seconds - I'm quite sure there is nothing wrong with
this
computer.
Geoff
----- Original Message -----
From: "Dana Trout" <dana@troutcom.com>
<filmscanners@halftone.co.uk>
Sent: Monday, July 30, 2001 8:29 AM
Subject: Re: filmscanners: Scanning and memory limits in Windows
> Now that I see you are stating load times for uncompressed files I
see
> our times are much more similar. LZW compression is very
CPU-intensive
> and there is no comparison between load times for non-compressed and
> compressed files (other than compressed files take a *lot* longer).
>
> I compress TIFFs before putting them on CDs because it's much easier
to
> manage 800 CDs than it is to manage over 1200 CDs. The reason I see a
> fairly large compression ratio is that I am saving 48-bit TIFFs, but
> the scanner uses only about 40 of them.
>
> There *is* a big hit in load and store times, though, which is the
> reason I compress *only* what's going onto the CDs. Images that are
> being played with are left uncompressed to reduce the file-handling
> times (even though the files are much larger).
>
> The upshot of this conversation is to point out (yet again) that one
> needs to look at where the bottlenecks really are. If the process is
> largely CPU-bound, as loading and storing LZW-compressed TIFFs are on
> my computer, I will realize much greater gains if I spend money to
> upgrade the CPU than if I get 10,000 rpm drives. Others who have fast
> processors and do not use LZW compression may well find their hard
> drives are the bottleneck and would be far better off upgrading them.
> It is up to each person to investigate where the bottlenecks are in
> *their own* processes and act to relieve them, instead of just
assuming
> they need a particular item (faster hard drives, more RAM, or
> whatever).
>
> As for "mucking about" with the image, lossless compression schemes
> like LZW merely encode the values in different bits than straight
> binary. It's really a code similar to Morse code -- the most popular
> values are stored in fewer bits, and the least popular take lots and
> lots of bits but because there are so few instances of them the
overall
> number of bits (and therefore bytes) needed to store the image is
> reduced. As for questions about how lossless the process is, just do
a
> file compare between the original uncompressed image and one that has
> been retrieved from a compressed TIFF. I have yet to see anything
other
> than an exact match -- if there is a difference there is something
> definitely wrong with the computer!
>
> Ta for now,
> --Dana
> ----------
> From: geoff murray <geoffmurray@primus.com.au>
> To: filmscanners@halftone.co.uk
> Subject: Re: filmscanners: Scanning and memory limits in Windows
> Date: Saturday, July 28, 2001 7:21 PM
>
> Hi Dana,
> I never bother to compress TIFF's, apparently there is
not
> much
> space saving anyway and I believe the less an image is mucked about
> with the
> better, even if people say that LZW compression is lossless.
>
> Geoff
>
>
> ----- Original Message -----
> From: "Dana Trout" <dana@troutcom.com>
> To: <filmscanners@halftone.co.uk>
> Sent: Sunday, July 29, 2001 5:49 AM
> Subject: Re: filmscanners: Scanning and memory limits in Windows
>
>
> > I find that little of the time spent is due to the disk drive,
which
> is
> > the reason for my comment that a 7200 rpm drive, even though it is
> 33%
> > faster than a 5400 rpm drive, will not necessarily reduce load
times
> by
> > a like percentage.
> >
> > As for my times being slow, you're right: I was quoting the
> performace
> > of my "junker" computer which is used only for scanning -- it's a
466
> > Celeron with 512MB RAM but Intel's woefully undersized L2 cache.
> > However, the times you quote make me wonder if you are loading
> > LZW-compressed TIFFs. If so, it is *definitely* time for me to
> upgrade
> > the scanner computer!
> >
> > Thanks for your comments,
> > --Dana
> > ----------
> > From: geoff murray <geoffmurray@primus.com.au>
> > To: filmscanners@halftone.co.uk
> > Subject: Re: filmscanners: Scanning and memory limits in Windows
> > Date: Saturday, July 28, 2001 6:15 AM
> >
> > Hi Dana,
> > Gee your times seem very slow. I tried loading a 56mb
> file
> > from
> > my scratch disk and it took 3.6 seconds. A 169mb file took 17
> seconds.
> > This
> > on a Win 98SE machine with a 1Ghz Athlon and 512mb of PC133 ram and
> two
> > 7200rpm hard drives. Scratch partition is not on the hard drive
which
> > has
> > PS6. 7200 rpm drives made a significant difference to overall
speed.
> >
> > Geoff Murray
> > www.geoffmurray.com
> >
> > ----- Original Message -----
> > From: "Dana Trout" <dana@troutcom.com>
> > To: <filmscanners@halftone.co.uk>
> > Sent: Saturday, July 28, 2001 7:40 AM
> > Subject: RE: filmscanners: Scanning and memory limits in Windows
> >
> >
> > > A 25% faster drive won't necessarily get you 25% faster
load/store
> > > times. PhotoShop seems to be inordinately slow in dealing with
> > > compressed TIFFs -- I got curious so I set up a cache large
enough
> to
> > > hold the whole file (53MB). The first time I loaded it into
> PhotoShop
> > > it took 61 seconds (reading from the disk). I then closed the
file
> > and
> > > reloaded it into PhotoShop (this time from the cache -- the disk
> > light
> > > never even blinked) and it took 55 seconds. And I'm reasonably
sure
> > > that a RAM cache is *much* faster than a 7200 rpm drive!
> > >
> > > BTW, Ed's VueScan takes less than 30 seconds to read the same
file.
> > > --Dana
> > > ----------
> > > From: Rob Geraghty <harper@wordweb.com>
> > > To: filmscanners@halftone.co.uk
> > > Subject: filmscanners: RE: filmscanners: Scanning and memory
limits
> > in
> > > Windows
> > > Date: Friday, July 27, 2001 12:22 AM
> > >
> > > < snip >
> > >
> > > On the other hand I'm reasonably sure the main
> > > bottleneck in my PC when dealing with large scans is the 5400RPM
> IDE
> > > drive.
> > > A 7200RPM drive would speed up loading and saving files by at
> least
> > > 25%.
> > > Two 7200rpm drives in a RAID array should be significantly
better
> > > still.
> > > Loading and saving files is the no.1 timewaster for me when
> working
> > > with
> > > film scans on my PC.
> > >
> > > Rob
> > >
> > >
> > > Rob Geraghty harper@wordweb.com
> > >
> > >
> > >
> > >
> >
>
