horses for courses

In article , Floyd L. Davidson
wrote:

For example, if there are 750 images and it takes 8
seconds to complete for each, that would have added 100
minutes to the time it would take if those operations
were done interactively. With a single CPU the batch
process would cut 100 minutes out of the interactive
time (the user can do other work for that amount of
time). And with the 8 CPU machine that I run the batch
process on it reduces the total time the batch runs to
11 minutes or so.

8 seconds to convert a raw to a jpeg???

I don't think I specified outputting a JPEG image, did I?

the exported file format is not a significant factor, especially when
it's taking 8 seconds to process an image.

For a low ISO shot (little noise reduction) and a web
sized JPEG, less than 2 seconds per image.

But lets talk about a professional production workflow
intended primarily for large prints. And just for fun
lets do throw in having to use a high ISO, and thus a
high value for noise reduction.

The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

something is very wrong if it takes 50 seconds to process one image.

http://blog.phaseone.com/processing-...-gpus-capture-
one-pro-8/
The speed advantage of using OpenCL is noticeable. When using both of
the AMD D700 graphics cards in a Mac Pro (2013) computer, you can
process a full 80 megapixel RAW file in less than 1.5 seconds.
....
Performance for Mac Pro (2013), Dual D700 ATI cards, 12 core CPU:
Capture One Pro 7: 28 IQ180 files: 2.23 min * end to end
Capture One Pro 8: 28 IQ180 files: 1.34 min * end to end

here's another test, where 128 36 mp images took just over 4 minutes:
http://macperformanceguide.com/images/iMac5K/graph-CaptureOnePro.png

For real fun and games, output a PNG! That will take
over a minute per image!

you need better software.

Of course all of this is with 36MP images, and those
with a 6 MP camera won't have the same numbers.

obviously, but even with 80 mp images, it doesn't anywhere near that
long.

In article , Tony Cooper
wrote:

Sandman, of course, only agrees with himself.

except when he contradicts himself.

nospam wrote:
In article , Floyd L. Davidson
wrote:

For example, if there are 750 images and it takes 8
seconds to complete for each, that would have added 100
minutes to the time it would take if those operations
were done interactively. With a single CPU the batch
process would cut 100 minutes out of the interactive
time (the user can do other work for that amount of
time). And with the 8 CPU machine that I run the batch
process on it reduces the total time the batch runs to
11 minutes or so.

8 seconds to convert a raw to a jpeg???

I don't think I specified outputting a JPEG image, did I?

the exported file format is not a significant factor, especially when
it's taking 8 seconds to process an image.

Well it very clearly is a *significant* factor!

Roughly 1.25 times longer for a TIFF than a similar sized
JPEG. And almost 1.5 times for a PNG as a JPEG.

Not significant for one single web sized image the way
you might work, but very significant for large images to
be printed.

For a low ISO shot (little noise reduction) and a web
sized JPEG, less than 2 seconds per image.

But lets talk about a professional production workflow
intended primarily for large prints. And just for fun
lets do throw in having to use a high ISO, and thus a
high value for noise reduction.

The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

something is very wrong if it takes 50 seconds to process one image.

http://blog.phaseone.com/processing-...-gpus-capture-
one-pro-8/
The speed advantage of using OpenCL is noticeable. When using both of
the AMD D700 graphics cards in a Mac Pro (2013) computer, you can
process a full 80 megapixel RAW file in less than 1.5 seconds.
...
Performance for Mac Pro (2013), Dual D700 ATI cards, 12 core CPU:
Capture One Pro 7: 28 IQ180 files: 2.23 min * end to end
Capture One Pro 8: 28 IQ180 files: 1.34 min * end to end

You didn't notice what they are measuring. They are not
doing noise reduction on the raw sensor data, and all
they are generating is an RGB screen image.

It is essentially equivalent to what I stated took less
than 2 seconds per image with my setup. They are getting
slightly better times, but using 12 cores while I'm using 8.

here's another test, where 128 36 mp images took just over 4 minutes:
http://macperformanceguide.com/images/iMac5K/graph-CaptureOnePro.png

For real fun and games, output a PNG! That will take
over a minute per image!

you need better software.

Your data is about hardware, not about software.

Your software literally adds hours to processing a large
shoot compared to what I'm doing!

Of course all of this is with 36MP images, and those
with a 6 MP camera won't have the same numbers.

obviously, but even with 80 mp images, it doesn't anywhere near that
long.

--
Floyd L. Davidson http://www.apaflo.com/
Ukpeagvik (Barrow, Alaska)

(Floyd L. Davidson) wrote:

It is essentially equivalent to what I stated took less
than 2 seconds per image with my setup. They are getting
slightly better times, but using 12 cores while I'm using 8.

Correction, my measurement was not using 8 CPU's, it was
using only one (multithreading turned off).

--
Floyd L. Davidson http://www.apaflo.com/
Ukpeagvik (Barrow, Alaska)

In article , Floyd L. Davidson
wrote:

For example, if there are 750 images and it takes 8
seconds to complete for each, that would have added 100
minutes to the time it would take if those operations
were done interactively. With a single CPU the batch
process would cut 100 minutes out of the interactive
time (the user can do other work for that amount of
time). And with the 8 CPU machine that I run the batch
process on it reduces the total time the batch runs to
11 minutes or so.

8 seconds to convert a raw to a jpeg???

I don't think I specified outputting a JPEG image, did I?

the exported file format is not a significant factor, especially when
it's taking 8 seconds to process an image.

Well it very clearly is a *significant* factor!

Roughly 1.25 times longer for a TIFF than a similar sized
JPEG. And almost 1.5 times for a PNG as a JPEG.

Not significant for one single web sized image the way
you might work, but very significant for large images to
be printed.

the difference between writing a tiff or a jpeg is insignificant, which
your numbers confirm.

assume 1 second for a jpeg (which is long), then a tiff would be 1.25
seconds using your numbers. with an 8 second per image process, that
would add 1/32 of the total time, or about 3% more. it's insignificant.

For a low ISO shot (little noise reduction) and a web
sized JPEG, less than 2 seconds per image.

But lets talk about a professional production workflow
intended primarily for large prints. And just for fun
lets do throw in having to use a high ISO, and thus a
high value for noise reduction.

The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

something is very wrong if it takes 50 seconds to process one image.

http://blog.phaseone.com/processing-...-gpus-capture-
one-pro-8/
The speed advantage of using OpenCL is noticeable. When using both of
the AMD D700 graphics cards in a Mac Pro (2013) computer, you can
process a full 80 megapixel RAW file in less than 1.5 seconds.
...
Performance for Mac Pro (2013), Dual D700 ATI cards, 12 core CPU:
Capture One Pro 7: 28 IQ180 files: 2.23 min * end to end
Capture One Pro 8: 28 IQ180 files: 1.34 min * end to end

You didn't notice what they are measuring. They are not
doing noise reduction on the raw sensor data, and all
they are generating is an RGB screen image.

nope. end to end means raw to final exported image.

note that they claim processing one image is 1.5s while end to end of
28 images is about 3.3s/image. clearly there's more being done in the
latter, which at a minimum is writing out files. they certainly aren't
displaying 28 rgb images on the display.

It is essentially equivalent to what I stated took less
than 2 seconds per image with my setup. They are getting
slightly better times, but using 12 cores while I'm using 8.

how is it equivalent, if they're getting about 3s/image compared to
you're 35-50s/image? that's an order of magnitude faster.

here's another test, where 128 36 mp images took just over 4 minutes:
http://macperformanceguide.com/images/iMac5K/graph-CaptureOnePro.png

For real fun and games, output a PNG! That will take
over a minute per image!

you need better software.

Your data is about hardware, not about software.

nope. it's about software that can use a gpu to greatly accelerate
processing.

with v.7 (the previous version), there is a 3x improvement when using
the gpu:
https://phaseoneimageprofessor.files...6/benchmark.jp
g

Your software literally adds hours to processing a large
shoot compared to what I'm doing!

nope.

nospam wrote:
In article , Floyd L. Davidson
wrote:

For example, if there are 750 images and it takes 8
seconds to complete for each, that would have added 100
minutes to the time it would take if those operations
were done interactively. With a single CPU the batch
process would cut 100 minutes out of the interactive
time (the user can do other work for that amount of
time). And with the 8 CPU machine that I run the batch
process on it reduces the total time the batch runs to
11 minutes or so.

8 seconds to convert a raw to a jpeg???

I don't think I specified outputting a JPEG image, did I?

the exported file format is not a significant factor, especially when
it's taking 8 seconds to process an image.

Well it very clearly is a *significant* factor!

Roughly 1.25 times longer for a TIFF than a similar sized
JPEG. And almost 1.5 times for a PNG as a JPEG.

Not significant for one single web sized image the way
you might work, but very significant for large images to
be printed.

the difference between writing a tiff or a jpeg is insignificant, which
your numbers confirm.

assume 1 second for a jpeg (which is long), then a tiff would be 1.25
seconds using your numbers. with an 8 second per image process, that
would add 1/32 of the total time, or about 3% more. it's insignificant.

Pay attention to detail. It isn't just the write time,
it's the time to generate the formatted data set that is
to be written. It goes up progressively with the size
of the image, and amounts to several seconds for full
sized image. That amounts to hours of time for a large
shoot. And that is significant if your software
requires that you sit there twiddling your thumbs while
the disk spins.

For a low ISO shot (little noise reduction) and a web
sized JPEG, less than 2 seconds per image.

But lets talk about a professional production workflow
intended primarily for large prints. And just for fun
lets do throw in having to use a high ISO, and thus a
high value for noise reduction.

The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

something is very wrong if it takes 50 seconds to process one image.

http://blog.phaseone.com/processing-...-gpus-capture-
one-pro-8/
The speed advantage of using OpenCL is noticeable. When using both of
the AMD D700 graphics cards in a Mac Pro (2013) computer, you can
process a full 80 megapixel RAW file in less than 1.5 seconds.
...
Performance for Mac Pro (2013), Dual D700 ATI cards, 12 core CPU:
Capture One Pro 7: 28 IQ180 files: 2.23 min * end to end
Capture One Pro 8: 28 IQ180 files: 1.34 min * end to end

You didn't notice what they are measuring. They are not
doing noise reduction on the raw sensor data, and all
they are generating is an RGB screen image.

nope. end to end means raw to final exported image.

Read the article you cited. Stop mixing Apples and Prunes.

note that they claim processing one image is 1.5s while end to end of
28 images is about 3.3s/image. clearly there's more being done in the
latter, which at a minimum is writing out files. they certainly aren't
displaying 28 rgb images on the display.

It is essentially equivalent to what I stated took less
than 2 seconds per image with my setup. They are getting
slightly better times, but using 12 cores while I'm using 8.

how is it equivalent, if they're getting about 3s/image compared to
you're 35-50s/image? that's an order of magnitude faster.

I'm doing the same thing in a measured 1.758 seconds per image.

here's another test, where 128 36 mp images took just over 4 minutes:
http://macperformanceguide.com/images/iMac5K/graph-CaptureOnePro.png

For real fun and games, output a PNG! That will take
over a minute per image!

you need better software.

Your data is about hardware, not about software.

nope. it's about software that can use a gpu to greatly accelerate
processing.

Read the article you cited. Mixing Apples and Prunes
is a waste of time.

with v.7 (the previous version), there is a 3x improvement when using
the gpu:
https://phaseoneimageprofessor.files...6/benchmark.jp
g

Your software literally adds hours to processing a large
shoot compared to what I'm doing!

nope.

Try it. So far as I can tell, because nobody has
suggested a way to do it, there is no way using
Lightroom that images can be batch processed with unique
configurations per image. Batching only files that use
the same configuration doesn't help at all.

Your repeated "nope" is just confirmation of ignorance.

--
Floyd L. Davidson http://www.apaflo.com/
Ukpeagvik (Barrow, Alaska)

On 5/19/15 PDT 1:54 PM, Tony Cooper wrote:
On Tue, 19 May 2015 10:33:02 -0700, John McWilliams
wrote:

On 5/19/15 PDT 5:51 AM, Floyd L. Davidson wrote:
nospam wrote:
In article , Floyd L. Davidson
wrote:
For example, if there are 750 images and it takes 8
seconds to complete for each, that would have added 100
minutes to the time it would take if those operations
were done interactively. With a single CPU the batch
process would cut 100 minutes out of the interactive
time (the user can do other work for that amount of
time). And with the 8 CPU machine that I run the batch
process on it reduces the total time the batch runs to
11 minutes or so.

8 seconds to convert a raw to a jpeg???

I don't think I specified outputting a JPEG image, did I?

For a low ISO shot (little noise reduction) and a web
sized JPEG, less than 2 seconds per image.

But lets talk about a professional production workflow
intended primarily for large prints. And just for fun
lets do throw in having to use a high ISO, and thus a
high value for noise reduction.

The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

For real fun and games, output a PNG! That will take
over a minute per image!

Of course all of this is with 36MP images, and those
with a 6 MP camera won't have the same numbers.

I agree with this post.

What's going on today? I just agreed with nospam on something, and
you are agreeing with Floyd on something. Red letter day.

Sandman, of course, only agrees with himself.

Do you need to take pot shots at him so frequently?

You do know, don't you, that that's a rhetorical question?

In article , Floyd L. Davidson
wrote:

Roughly 1.25 times longer for a TIFF than a similar sized
JPEG. And almost 1.5 times for a PNG as a JPEG.

Not significant for one single web sized image the way
you might work, but very significant for large images to
be printed.

the difference between writing a tiff or a jpeg is insignificant, which
your numbers confirm.

assume 1 second for a jpeg (which is long), then a tiff would be 1.25
seconds using your numbers. with an 8 second per image process, that
would add 1/32 of the total time, or about 3% more. it's insignificant.

Pay attention to detail. It isn't just the write time,
it's the time to generate the formatted data set that is
to be written. It goes up progressively with the size
of the image, and amounts to several seconds for full
sized image. That amounts to hours of time for a large
shoot. And that is significant if your software
requires that you sit there twiddling your thumbs while
the disk spins.

nobody said anything about twiddling thumbs.

For a low ISO shot (little noise reduction) and a web
sized JPEG, less than 2 seconds per image.

But lets talk about a professional production workflow
intended primarily for large prints. And just for fun
lets do throw in having to use a high ISO, and thus a
high value for noise reduction.

The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

something is very wrong if it takes 50 seconds to process one image.

http://blog.phaseone.com/processing-...-gpus-capture-
one-pro-8/
The speed advantage of using OpenCL is noticeable. When using both of
the AMD D700 graphics cards in a Mac Pro (2013) computer, you can
process a full 80 megapixel RAW file in less than 1.5 seconds.
...
Performance for Mac Pro (2013), Dual D700 ATI cards, 12 core CPU:
Capture One Pro 7: 28 IQ180 files: 2.23 min * end to end
Capture One Pro 8: 28 IQ180 files: 1.34 min * end to end

You didn't notice what they are measuring. They are not
doing noise reduction on the raw sensor data, and all
they are generating is an RGB screen image.

nope. end to end means raw to final exported image.

Read the article you cited. Stop mixing Apples and Prunes.

i did read it. did you?

nowhere does it say it's only an rgb image.

it also states:
In Capture One Pro 8, we have carefully optimized the processing
pipeline to become as parallel as possible, resulting in the fastest
processing of any Capture One version * and any current current RAW
processor.

note that they claim processing one image is 1.5s while end to end of
28 images is about 3.3s/image. clearly there's more being done in the
latter, which at a minimum is writing out files. they certainly aren't
displaying 28 rgb images on the display.

It is essentially equivalent to what I stated took less
than 2 seconds per image with my setup. They are getting
slightly better times, but using 12 cores while I'm using 8.

how is it equivalent, if they're getting about 3s/image compared to
you're 35-50s/image? that's an order of magnitude faster.

I'm doing the same thing in a measured 1.758 seconds per image.

you wrote:
The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

now you're claiming 1.8s/image? both can't be true.

here's another test, where 128 36 mp images took just over 4 minutes:
http://macperformanceguide.com/images/iMac5K/graph-CaptureOnePro.png

For real fun and games, output a PNG! That will take
over a minute per image!

you need better software.

Your data is about hardware, not about software.

nope. it's about software that can use a gpu to greatly accelerate
processing.

Read the article you cited. Mixing Apples and Prunes
is a waste of time.

nobody is mixing anything.

with v.7 (the previous version), there is a 3x improvement when using
the gpu:
https://phaseoneimageprofessor.files...6/benchmark.jp
g

Your software literally adds hours to processing a large
shoot compared to what I'm doing!

nope.

Try it. So far as I can tell, because nobody has
suggested a way to do it, there is no way using
Lightroom that images can be batch processed with unique
configurations per image. Batching only files that use
the same configuration doesn't help at all.

that's exactly what lightroom excels at.

Your repeated "nope" is just confirmation of ignorance.

nope.

nospam wrote:
In article , Floyd L. Davidson
wrote:
shoot. And that is significant if your software
requires that you sit there twiddling your thumbs while
the disk spins.

nobody said anything about twiddling thumbs.

Yes you have!

....
You didn't notice what they are measuring. They are not
doing noise reduction on the raw sensor data, and all
they are generating is an RGB screen image.

nope. end to end means raw to final exported image.

Read the article you cited. Stop mixing Apples and Prunes.

i did read it. did you?

nowhere does it say it's only an rgb image.

And nowhere does it say anything about writing a file,
but it does talk about the video card used, and parallel
processing of GDU's.

it also states:
In Capture One Pro 8, we have carefully optimized the processing
pipeline to become as parallel as possible, resulting in the fastest
processing of any Capture One version * and any current current RAW
processor.

Which of course supports my point, and makes your
discussion absurd.

how is it equivalent, if they're getting about 3s/image compared to
you're 35-50s/image? that's an order of magnitude faster.

I'm doing the same thing in a measured 1.758 seconds per image.

you wrote:
The batch time to make each conversion and output a TIFF
image will run more like 35 to 50 seconds each. Higher
times for an 8 bit TIFF than a 16 bit, and higher time
for more noise reduction.

now you're claiming 1.8s/image? both can't be true.

Pay attention to the details! They *never* suggest they
are timing the production of a TIFF file including noise
reduction. What they are timing is putting an image
through the video card onto the monitor.

As I noted previously, that takes less than 2 seconds
for the hardware and software that I use, except I'm
timing writing it to a JPEG file rather than to the
screen. (The screen probably takes less time, but I
view everything remotely via ethernet, and the box with
the monitors on it is not particularly fast either.)

here's another test, where 128 36 mp images took just over 4 minutes:
http://macperformanceguide.com/images/iMac5K/graph-CaptureOnePro.png

For real fun and games, output a PNG! That will take
over a minute per image!

you need better software.

Your data is about hardware, not about software.

nope. it's about software that can use a gpu to greatly accelerate
processing.

Read the article you cited. Mixing Apples and Prunes
is a waste of time.

nobody is mixing anything.

Get off it nospam. You can't keep anything straight.

I can't tell if it is due to confusion or if you are
just making an effort at contortion. Either way, the
effect is the same: your discussion is worthless.

with v.7 (the previous version), there is a 3x improvement when using
the gpu:
https://phaseoneimageprofessor.files...6/benchmark.jp
g

Your software literally adds hours to processing a large
shoot compared to what I'm doing!

nope.

Try it. So far as I can tell, because nobody has
suggested a way to do it, there is no way using
Lightroom that images can be batch processed with unique
configurations per image. Batching only files that use
the same configuration doesn't help at all.

that's exactly what lightroom excels at.

Really? Why hasn't anyone described how to do it?
Batching only files that use the same configuration is
one thing. Generating, without excess tedium, a separate
configuration per image and then doing the conversions as
one massive batch process is very different.

What has been done so far just takes a single configuration,
and adds images files to a list of those processed with that
one configuration.

I have no doubt that it can be done, the question is only
one of whether the interactive part of the process is
smooth and efficient, or time consuming. It seems not, as
no one is interested in using that...

Your repeated "nope" is just confirmation of ignorance.

nope.

More of that ignorance nospam. You are famous, or
infamous, for it.

--
Floyd L. Davidson http://www.apaflo.com/
Ukpeagvik (Barrow, Alaska)