Photoshop CS3 Performance vs. CS2

[mac57]

I downloaded the CS3 beta a few days ago and promised to report on its relative performance vs. CS2, when run on a PowerMac G5. As you can see from my signature block, I have a fairly recent PowerMac G5 with 2.3 GHz dual core, with 2.5 GB of RAM.

Subjectively, CS3 feels a LOT snappier. My favorite improvement is a little one. Brush resizing on the fly is now fast on crisp. On CS2, even on my machine, there is a noticable lag between the keypress and the result on screen. On CS3, it is instantaneous.

For those who don't know what I am referring to, you can dynamically resize the brush for most tools by using the "[" key to decrease the brush size and the "]" key to increase it. I use this ALL the time. On CS2, you press the key and wait for a second or so before the brush size changes. On CS3, it is instantaneous. This is a small thing, but for a base operation like this which gets used all the time, it is a great improvement.

The objective numbers don't backup my subjective feel, but here they go. I have measured only the truly obvious, since I don't have test suites and test equipment. I measured the cold and warm startup times for Photoshop CS2 and Photoshop CS3, and the same thing for their respective versions of Bridge. Cold startup time is the number of seconds it takes from click on the dock until you have the full application up and running, the first time you run it after a restart. Warm restart time is the same thing, but after you have already run it at least once before (and it is thus cached at least to some extent in memory).

Here are the results:

Photoshop CS2:
- Cold start: 11 seconds
- Warm start: 4 seconds

Bridge CS2:
- Cold start: 12 seconds
- Warm start: 5 seconds

Photoshop CS3:
- Cold start: 13 seconds (worse)
- Warm start: 4 seconds (same)

Bridge CS3:
- Cold start: 11 seconds (better)
- Warm start: 3 seconds (better)

So, a mixed bag. Some start times are better, some are worse. This notwithstanding, overall, the application just FEELS snappier. I am going to have to use it a lot more to decide why this is. I have to admit to being heavily influenced by the snappiness of the brush resizing.

By the way, the new interface for CS3 has some nice refinements, like the ability to make your tool bar a single column vs. two columns. Like the cursor resizing time, it is a small thing, but nice.

As I work with CS3 more, I may post more results, or perhaps evaluation of some of the new tools - the new selection tools look VERY interesting.

 

[CrossBones]

I wonder how those results would vary using the Intel machines?

 

[deus_ex_machina]

it's supposed to fly

 

[PunkRockTuba]

it's supposed to fly

It does


A thing I just noticed in photoshop cs3 is you can control how much ram it uses.

 

[D3v1L80Y]

You can do the same thing in CS2 with the memory as well.

 

[PunkRockTuba]

You can do the same thing in CS2 with the memory as well.

Oh really, I never knew that.

 

[CrossBones]

I didn't know that either.

 

[D3v1L80Y]

Yep, its nothing new. :black:

 

[CrossBones]

Yep, its nothing new. :black:

Same in PS7...I have both PS7 and CS2...same stuff.

 

[Alexis]

Here are the times on my 20" Intel Core2Duo 2.1Ghz iMac:

PS CS2 cold 24 secs
PS CS3 cold 12 secs

PS CS2 warm 11 secs
PS CS3 warm 3 secs

 

[dtownley1]

I downloaded the CS3 beta a few days ago and promised to report on its relative performance vs. CS2, when run on a PowerMac G5. As you can see from my signature block, I have a fairly recent PowerMac G5 with 2.3 GHz dual core, with 2.5 GB of RAM.

Subjectively, CS3 feels a LOT snappier. My favorite improvement is a little one. Brush resizing on the fly is now fast on crisp. On CS2, even on my machine, there is a noticable lag between the keypress and the result on screen. On CS3, it is instantaneous.

For those who don't know what I am referring to, you can dynamically resize the brush for most tools by using the "[" key to decrease the brush size and the "]" key to increase it. I use this ALL the time. On CS2, you press the key and wait for a second or so before the brush size changes. On CS3, it is instantaneous. This is a small thing, but for a base operation like this which gets used all the time, it is a great improvement.

The objective numbers don't backup my subjective feel, but here they go. I have measured only the truly obvious, since I don't have test suites and test equipment. I measured the cold and warm startup times for Photoshop CS2 and Photoshop CS3, and the same thing for their respective versions of Bridge. Cold startup time is the number of seconds it takes from click on the dock until you have the full application up and running, the first time you run it after a restart. Warm restart time is the same thing, but after you have already run it at least once before (and it is thus cached at least to some extent in memory).

Here are the results:

Photoshop CS2:
- Cold start: 11 seconds
- Warm start: 4 seconds

Bridge CS2:
- Cold start: 12 seconds
- Warm start: 5 seconds

Photoshop CS3:
- Cold start: 13 seconds (worse)
- Warm start: 4 seconds (same)

Bridge CS3:
- Cold start: 11 seconds (better)
- Warm start: 3 seconds (better)

So, a mixed bag. Some start times are better, some are worse. This notwithstanding, overall, the application just FEELS snappier. I am going to have to use it a lot more to decide why this is. I have to admit to being heavily influenced by the snappiness of the brush resizing.

By the way, the new interface for CS3 has some nice refinements, like the ability to make your tool bar a single column vs. two columns. Like the cursor resizing time, it is a small thing, but nice.

As I work with CS3 more, I may post more results, or perhaps evaluation of some of the new tools - the new selection tools look VERY interesting.

some interesting results there

have you had a chance to compare certain heavy duty tasks?
Some ideas that come to mind;
•transforming/scaling large docs or layers
•certain filters - blur/liquify etc.

I'm wondering if there's much or any difference when it comes to something like this.

 

[DrQuincy]

Here are the times on my 20" Intel Core2Duo 2.1Ghz iMac:

PS CS2 cold 24 secs
PS CS3 cold 12 secs

PS CS2 warm 11 secs
PS CS3 warm 3 secs

Looking good! Are you able to do a benchmark on say applying a filter?

 

[CrossBones]

Here are the times on my 20" Intel Core2Duo 2.1Ghz iMac:

PS CS2 cold 24 secs
PS CS3 cold 12 secs

PS CS2 warm 11 secs
PS CS3 warm 3 secs

I cannot wait!

 

[mac57]

Is there anyone out there with a 2.66 GHz or even better, a 3.0 GHz Mac Pro who could report similar timings for it?

 

[Alexis]

Some benchmarks with a 300dpi 34 x 26cm image. First figure is CS2, second is CS3

Smart blur, high quality
30s / 19s

Extrude 30x30
20s / 11s

Plastic wrap full detail
18s / 3s

Enlarge to 500%
61s / 36s

 

[mac57]

Alexis, what type of Mac were these figures taken from?

 

[Ice Cream Man]

What really rubs me the wrong way about that, is that if they can make a universal Binary that works that much better on PPC, what was their problem that they could make the PPC code just as snappy.

Discrimination, that's what it is. They purposely made crap code on PPC, to make the performance suck. Now that it's associated with Intel, they're singing the praises now. It's not like they couldn't have done it. They were just either lazy, or discriminating.

Added

Here are the times on my 20" Intel Core2Duo 2.1Ghz iMac:

 

[MacHeadCase]

I downloaded the CS3 beta a few days ago and promised to report on its relative performance vs. CS2, when run on a PowerMac G5. As you can see from my signature block, I have a fairly recent PowerMac G5 with 2.3 GHz dual core, with 2.5 GB of RAM.

Subjectively, CS3 feels a LOT snappier. My favorite improvement is a little one. Brush resizing on the fly is now fast on crisp. On CS2, even on my machine, there is a noticable lag between the keypress and the result on screen. On CS3, it is instantaneous.

For those who don't know what I am referring to, you can dynamically resize the brush for most tools by using the "[" key to decrease the brush size and the "]" key to increase it. I use this ALL the time. On CS2, you press the key and wait for a second or so before the brush size changes. On CS3, it is instantaneous. This is a small thing, but for a base operation like this which gets used all the time, it is a great improvement.

The objective numbers don't backup my subjective feel, but here they go. I have measured only the truly obvious, since I don't have test suites and test equipment. I measured the cold and warm startup times for Photoshop CS2 and Photoshop CS3, and the same thing for their respective versions of Bridge. Cold startup time is the number of seconds it takes from click on the dock until you have the full application up and running, the first time you run it after a restart. Warm restart time is the same thing, but after you have already run it at least once before (and it is thus cached at least to some extent in memory).

Here are the results:

Photoshop CS2:
- Cold start: 11 seconds
- Warm start: 4 seconds

Bridge CS2:
- Cold start: 12 seconds
- Warm start: 5 seconds

Photoshop CS3:
- Cold start: 13 seconds (worse)
- Warm start: 4 seconds (same)

Bridge CS3:
- Cold start: 11 seconds (better)
- Warm start: 3 seconds (better)

So, a mixed bag. Some start times are better, some are worse. This notwithstanding, overall, the application just FEELS snappier. I am going to have to use it a lot more to decide why this is. I have to admit to being heavily influenced by the snappiness of the brush resizing.

By the way, the new interface for CS3 has some nice refinements, like the ability to make your tool bar a single column vs. two columns. Like the cursor resizing time, it is a small thing, but nice.

As I work with CS3 more, I may post more results, or perhaps evaluation of some of the new tools - the new selection tools look VERY interesting.

I truly appreciate you taking the time to post your findings, mac57! You and I both have a PowerPC Mac and most reviews are only looking at CS3 from the Intel-based Mac user base perspective.

Thanks again!

 

[mac57]

Ice Cream Man, the numbers would tend to suggest that the PowerPC code is actually very good indeed. The numbers I posted are for a PowerMac G5 at 2.3 GHz. The other numbers that have been posted were for a 2.1 GHz iMac. As you can see, they are pretty close, implying that the Intel Native stuff and the PPC Native stuff perform at about the same rate.

The really bad numbers that were posted for CS2 vs. CS3 compared CS2 on an Intel Mac (hence PPC code via Rosetta) and CS3 on the same Mac (Universal binary, hence running native). So, the CS2 PPC code didn't stink there - it was just being interpreted. On its native platform, you get the numbers I posted, which are quite respectable.

 

[mac57]

MacHeadCase, you're welcome! Now, I am REALLY looking forward to the released product, and seeing what CS3 on a 3 GHz Mac Pro performs like vs. my 2.3 GHz PowerMac G5. I am hoping for really good results (approximately 50% better?). We shall see.