Computer Graphics with Photoshop - <PING> Bill Hilton RE:Scratch Disk Tests

Author

<PING> Bill Hilton RE:Scratch Disk Tests

Hunt

2005-11-29, 10:14 pm

Bill,

I ran a test on my laptop with different Scratch Disk configs. All of my
timings were very close. Take a look at my article and tell me where you and I
may have differed on the results.

Look forward to hearing from you on this one, as I am puzzled.

Hunt

Bill Hilton

2005-11-30, 3:15 am

>Hunt writes ...
>
>Bill,
>I ran a test on my laptop with different Scratch Disk configs. All of my
>timings were very close. Take a look at my article and tell me where you and
>I may have differed on the results.
>
>Look forward to hearing from you on this one, as I am puzzled

Hi Hunt,

First off we do live in the same city (figure the odds), I'm down here
in the south end (poor side of town) about 45 minutes from you ...
small world.

Second, thanks for running all these tests, I know it takes a lot of
time. Harvey will probably thank you as well :)

OK, your data is in the other post but I'll answer you here ...

>I ran a test on my laptop with different Scratch Disk configs. All of my
>timings were very close

I saw USB and FW slowing down my system quite a bit on my DESKTOP but
my tests on my laptop agree with what you saw. I posted the results in
the "any luck testing usb on laptop" thread but basically for 35 mm and
645 scans it was slightly faster with USB as scratch and for a 6x7 cm
scan slightly slower. As I mentioned this was surprising based on the
desktop tests using the same action and same files and I re-ran
everything to double check.

So we are pretty much in agreement here even though your laptop is much
faster than mine (actually faster than my 2 year old desktop :).

>Take a look at my article and tell me where you and
>I may have differed on the results

Even though we got similar results for the laptops there were some big
differences in how we did the tests so I'll go thru those ...

You described your test file as "671.3MB, 16bit/Channel color, 15
Layers, 5 Adjustment Layers, 2 Paths, 2 Alpha Channels, 7084px x 2853"
.... what I tested was a flat tiff of various sizes, up to about 10,800
x 8,800 pixels for the 6x7 scan file. I think it might throw the test
times off if you have a lot of layers since the steps will only run on
the one layer you have selected and that layer might not have much
data, but I'm not sure. At any rate I'd run the action on a flat tiff.

> Purge All was run before each test

When I change the scratch disk settings in Prefs there's a note
"scratch disks will remain in use until you quit Photoshop" (I'm using
CS) ... I *think* that might mean it keeps the previous settings until
you log off and back on (not sure of this) so each time I change this
Pref I close Photoshop and re-launch it ... if I'm right about the
changes not showing up until you "quit Photoshop" then maybe you were
changing the settings but they weren't actually implemented ... in
other words all the tests were done with the same hardware scratch
disk. But I'm not certain that's how it works ... at any rate closing
and re-opening ensured I had the same starting point each time.

Also you said something about scratch size of 1.1 GB ... not sure if
that's your starting scratch requirement or the final one (I check the
"scr" number in the bottom left, which is set to show "scratch sizes"
.... if this is your final number after the action ran then I think you
never get out of RAM so the hard drives aren't accessed ... this might
happen if the layer the operations are run on has little data ... can
you verify this? On my 6x7 cm test I was eating up to 2.2 GB of
scratch disk (I only had 323 MB of free RAM) for example.

At any rate this is all conjecture ... if you want to run one more test
(I know, I know) then flatten the file and save it off, set your
scratch RAM much lower in Prefs, say to 512 MB, and close Photoshop
each time you change the disk location setting and see whether or not
you see a time difference between say C and USB (which should be the
two extremes). If not then we declare victory and quit I guess :)

Bill

Bill Hilton

2005-11-30, 6:14 pm

>>Hunt writes
[color=darkred]
> I replied ...
>
>When I change the scratch disk settings in Prefs there's a note
>"scratch disks will remain in use until you quit Photoshop" (I'm using
>CS) ... I *think* that might mean it keeps the previous settings until
>you log off and back on (not sure of this)

I checked this earlier today and it works like I assumed, that is, if
you change the scratch disk in Prefs but don't quit Photoshop it still
uses the same disk. You have to quit and re-open Photoshop to switch
to the new scratch disk. Maybe I misunderstood what you wrote but if
you just did Purge after changing this Pref setting then that would
explain why your times are all pretty much the same, it's using the
same disk each time?

Bill

Hunt

2005-11-30, 6:14 pm

In article <1133331953.216602.210020@g47g2000cwa.googlegroups.com>, bhilton
665@aol.com says...
>
>
>Hi Hunt,
>
>First off we do live in the same city (figure the odds), I'm down here
>in the south end (poor side of town) about 45 minutes from you ...
>small world.

Extremely "small world!" AND, it seems to be getting smaller all of the time.
>
>Second, thanks for running all these tests, I know it takes a lot of
>time. Harvey will probably thank you as well :)
>
>OK, your data is in the other post but I'll answer you here ...
>
>
>I saw USB and FW slowing down my system quite a bit on my DESKTOP but
>my tests on my laptop agree with what you saw. I posted the results in
>the "any luck testing usb on laptop" thread but basically for 35 mm and
>645 scans it was slightly faster with USB as scratch and for a 6x7 cm
>scan slightly slower. As I mentioned this was surprising based on the
>desktop tests using the same action and same files and I re-ran
>everything to double check.
>
>So we are pretty much in agreement here even though your laptop is much
>faster than mine (actually faster than my 2 year old desktop :).
>
>
>Even though we got similar results for the laptops there were some big
>differences in how we did the tests so I'll go thru those ...
>
>You described your test file as "671.3MB, 16bit/Channel color, 15
>Layers, 5 Adjustment Layers, 2 Paths, 2 Alpha Channels, 7084px x 2853"
>... what I tested was a flat tiff of various sizes, up to about 10,800
>x 8,800 pixels for the 6x7 scan file. I think it might throw the test
>times off if you have a lot of layers since the steps will only run on
>the one layer you have selected and that layer might not have much
>data, but I'm not sure. At any rate I'd run the action on a flat tiff.

You could be correct with this thought, and I will re-run with a flattened
image. I did choose a "full image" layer {a sky, in my case, that was full
image, the lower section being contained, visually, by another layer(s)}.
Also, I forgot to point out that both drives were formatted with NTFS (I have
another Maxtor which is FAT32, so it can be transported between a WinME
machine).
>
>
>When I change the scratch disk settings in Prefs there's a note
>"scratch disks will remain in use until you quit Photoshop" (I'm using
>CS) ... I *think* that might mean it keeps the previous settings until
>you log off and back on (not sure of this) so each time I change this
>Pref I close Photoshop and re-launch it ... if I'm right about the
>changes not showing up until you "quit Photoshop" then maybe you were
>changing the settings but they weren't actually implemented ... in
>other words all the tests were done with the same hardware scratch
>disk. But I'm not certain that's how it works ... at any rate closing
>and re-opening ensured I had the same starting point each time.

For my tests, I did re-start CS after each Scratch Disk selection was made,
just so that the new Prefs would be in effect.
>
>Also you said something about scratch size of 1.1 GB ... not sure if
>that's your starting scratch requirement or the final one (I check the
>"scr" number in the bottom left, which is set to show "scratch sizes"
>... if this is your final number after the action ran then I think you
>never get out of RAM so the hard drives aren't accessed ... this might
>happen if the layer the operations are run on has little data ... can
>you verify this? On my 6x7 cm test I was eating up to 2.2 GB of
>scratch disk (I only had 323 MB of free RAM) for example.

Scratch Size was from the PS window (lower left) with image data. I did
consider that much of what I was doing might well be in RAM, and think that
taking the "flattened" pano, maybe with the resolution run up, just for the
test, might stress the system a bit more. In my curiosity, I did get a fair
amount of HDD (blinking light) activity during the Action, but maybe a much
larger file will tax it more. I'm sure that the amount of RAM will taint the
test, as we are looking at HDD I/O connections here.
>
>At any rate this is all conjecture ... if you want to run one more test
>(I know, I know) then flatten the file and save it off, set your
>scratch RAM much lower in Prefs, say to 512 MB, and close Photoshop
>each time you change the disk location setting and see whether or not
>you see a time difference between say C and USB (which should be the
>two extremes). If not then we declare victory and quit I guess :)
>
>Bill

Oh don't quit!!! Your tests have been very useful in the debate on how one
should configure their systems for Scratch Disk usage. In a perfect world,
we'd be able to just stick a bunch of RAIDs in, just for PS. On a workstation,
this is certainly an option, with the cost of HDD storage space nowadays, but
on a laptop, one runs out of real estate quickly. When the paying project is
done, I'll take your advice and re-run the tests, maybe throwing in a few more
steps to the aciton with a much larger file.

Thanks for the thoughts,
Hunt

Hunt

2005-11-30, 6:14 pm

In article <1133363398.558261.95680@o13g2000cwo.googlegroups.com>, bhilton665
@aol.com says...
>
>
>
>I checked this earlier today and it works like I assumed, that is, if
>you change the scratch disk in Prefs but don't quit Photoshop it still
>uses the same disk. You have to quit and re-open Photoshop to switch
>to the new scratch disk. Maybe I misunderstood what you wrote but if
>you just did Purge after changing this Pref setting then that would
>explain why your times are all pretty much the same, it's using the
>same disk each time?
>
>Bill

In my tests, CS2 was re-started after each change. In all tests, after the
setup of the Action and the first test, I really just checked that Purge was
greyed out. In my very first test, I had not done this, and the times varied
from 3:01.64' (without Purge, and C:\ Scratch Disk), to 2:01.65' & 2:18.25'
after Purge>Purge All. I wanted to make sure that all later tests had nothing
in memory, but in all cases, there was nothing to Purge. Note that in my first
run, there was a difference of ~1:00' with a full History, etc. in memory.

Thanks for your input, and I will follow up with phase II of the tests.

Hunt

PS, just watching the "gas gauge" bar during the various steps in the Action,
it seemed that some of those ran much faster with different Scratch Disk
sizes, but the totals, seemed to be about the same - maybe boredom, just
watching gas gauges?