'Real World' Photoshop Benchmark - Speed Test

[CTYankee]

Aye, if I used PS exclusively, there is no way I would have updated, but seeing as how I use Apple FCP studio, that doesn't effect me since it's UB. I hope the results I have posted have helped, but alas I am no PS expert, just a novice that uses it on occasion.

Its very helpful...I almost want to use FCP more just so I can justify getting one! I'm 90% PS so I need the fastest PS machine I can get now...or soon. So looks like its a G5.

 

[dante@sisna.com]

10.8 vs 50...if these numbers hold true no way do I get a MacPro. THe price on the Quads is falling already (now I wish I had sprung on that quad I saw for $2100 with 4GB of RAM). Any hit I take in price is worth the time I'll save using the G5 vs my current G4 or a MP.

Yes, my Quad G5 has been a fantastic box. Best mac I've ever owned -- and like a lot of us -- I've had experience with dozens during my career.

It is very quiet and powerful. Plus RAM is cheap for it now and all apps are native. With Leopard coming out it will run even faster -- so it has a great lifespan ahead of it -- at least a year and a half.

I plan to purchase a Mac Pro next Spring with Leopard and the true 4 core processors -- at that point the Quad G5 is still a great backup machine.

DJO

 

[CTYankee]

If you start deleting that step, then it isn't testing system speed... which in my eyes is pointless.

It will show the positives of having a lot of ram and scratch disks... that is the point of it.

It was a way to look at processor speed. I agree that the upsize step does make for a better system test, but since I rarely use files that large (11x14 about as big as I go 95% of the time) deleting that step made it a more real world photographer test.

 

[CTYankee]

Yes, my Quad G5 has been a fantastic box. Best mac I've ever owned -- and like a lot of us -- I've had experience with dozens during my career.

It is very quiet and powerful. Plus RAM is cheap for it now and all apps are native. With Leopard coming out it will run even faster -- so it has a great lifespan ahead of it -- at least a year and a half.

I plan to purchase a Mac Pro next Spring with Leopard and the true 4 core processors -- at that point the Quad G5 is still a great backup machine.

DJO

One last test...could you run the 100,best radial blur test? His results were MUCH faster than my G4 so I wonder how fast the quad and dual 2.7 is.

 

[BakedBeans]

It was a way to look at processor speed. I agree that the upsize step does make for a better system test, but since I rarely use files that large (11x14 about as big as I go 95% of the time) deleting that step made it a more real world photographer test.

I disagree, and my fine art photographer that consistently prints files would have killed me if I left it out

Many or most photographers (Including me) Do that ( although it would be the last step in normality of course)

 

[dante@sisna.com]

One last test...could you run the 100,best radial blur test? His results were MUCH faster than my G4 so I wonder how fast the quad and dual 2.7 is.

Sure, running it now. Just a Radial Blur on the original test image with Amount = 100, Spin Blur, Quality = Best.

Is this the test and settings you'd like to see?

DJO

 

[CTYankee]

I disagree, and my fine art photographer that consistently prints files would have killed me if I left it out

Many or most photographers (Including me) Do that ( although it would be the last step in normality of course)

thats a 40x50" print at least. Not many people print that big. Not even the top stock houses require files that big. Is it a test that pushes all aspects of a computer, yes. Is it one that simulates the workflow of most studio photographers no. We just don't deal in such large sizes that often.

 

[CTYankee]

Sure, running it now. Just a Radial Blur on the original test image with Amount = 100, Spin Blur, Quality = Best.

Is this the test and settings you'd like to see?

DJO

yup. thanks

 

[dante@sisna.com]

Quad Core Radial Blur

One last test...could you run the 100,best radial blur test? His results were MUCH faster than my G4 so I wonder how fast the quad and dual 2.7 is.

Quad Core with Photoshop Radial Blur at Amount = 100, Method = Spin, Quality = Best. On Original Eagle Test Image.

20 History States -- Cache 4 Levels -- RAM at 90% for total of 3072 available to photoshop.

Time = 42 seconds

All Four Processors Used at 391% average usage. A real processor burner. Nice call on this test. Surprised it took this long. Any Dual 2.7 results?

DJO

 

[BakedBeans]

thats a 40x50" print at least. Not many people print that big. Not even the top stock houses require files that big. Is it a test that pushes all aspects of a computer, yes. Is it one that simulates the workflow of most studio photographers no. We just don't deal in such large sizes that often.

Its 18x30. A hell of a lot of photographers print that large. Especially Landscape, Fine art, portrait etc they often print to that size, I know many photographers in the industry, to leave it out might suit your workflow but would not represent the majority of pros (not am's on fred miranda or here)

 

[digitalassassin]

Quad Core with Photoshop Radial Blur at Amount = 100, Method = Spin, Quality = Best. On Original Eagle Test Image.

20 History States -- Cache 4 Levels -- RAM at 90% for total of 3072 available to photoshop.

Time = 42 seconds

All Four Processors Used at 391% average usage. A real processor burner. Nice call on this test. Surprised it took this long. Any Dual 2.7 results?

DJO

here's how my dual 2.7 did on same test:
20 History States -- Cache 4 Levels -- RAM at 90% for total of 2209 available to photoshop.

1:44 seconds

time for bed and dreaming of a mac pro running cs3 on leopard.......

 

[amols]

Here's what I've got on MBP 15" 2.16 Ghz, 2GB, 100GB 5400 RPM

On Tiger 10.4.7 with CS2 after restart,

5 min 18 sec,

On XP2/Boot Camp with CS2 after restart,

4 min 32 sec.

So I guess Rossetta is doing pretty good job here.

 

[spicyapple]

Amols, disk speed is diluting your results.

Try giving more memory to PS, reduce History States to 1, and skip the 300% image scaling in the action. That should more accurately test Rosetta's translation efficiency.

 

[dante@sisna.com]

here's how my dual 2.7 did on same test:
20 History States -- Cache 4 Levels -- RAM at 90% for total of 2209 available to photoshop.

1:44 seconds

time for bed and dreaming of a mac pro running cs3 on leopard.......

For sure: But 1:44 seconds for your dual 2.7 is still great. 42 seconds on the Quad G5 and only 28 seconds on the Quad Mac Pro 2.66! Wow.

Can someone else confirm that the Radial Blur Test is 28 seconds under Rosetta on the Quad Mac Pro 2.66 GHZ? I suppose Rosetta is not in play on this test and it is all processor. 28 seconds is very fast!! Wow. I guess 42 seconds is not that bad.

But still . . .

now I am dreaming of a new quad.

 

[BakedBeans]

Amols, disk speed is diluting your results.

Try giving more memory to PS, reduce History States to 1, and skip the 300% image scaling in the action. That should more accurately test Rosetta's translation efficiency.

Why take the 300% image out... testing hte processors speed and not the system speed... wont give accurate rosseta results.

 

[CTYankee]

Its 18x30. A hell of a lot of photographers print that large. Especially Landscape, Fine art, portrait etc they often print to that size, I know many photographers in the industry, to leave it out might suit your workflow but would not represent the majority of pros (not am's on fred miranda or here)

Even at those sizes (at 400+ ppi, not sure what printer thats for...most large format printers use 400 tops, 300 or even 200 most often) its not as common as you indicate. Plus it is best to NOT interpolate the photo and let the printer do it. Exceptions are when there is vector based art. I work in the industry and rarely see files this large, even on commercial jobs. To suggest most photographers work with such large files as this just isn't accurate.

Anywho. Its the data I need to see. So I asked to see it.

 

[dante@sisna.com]

1:26 Seconds

PowerMac G5 Quad. 20 History States -- 95% Ram for total of 3,072 MB used by photoshop.
Main Drive -- Maxtor Stock 250Gig
Secondary for Scratch -- Maxtor 300 gig Maxline III

******* Drives very full -- about 70% used which may account for slowdown.

Activity Monitor Launched. Most actions used reasonable amount of Quad Processors.

Photoshop CS2

File on Scratch Drive

Love my Quad G5.
DJO

New Quad Test Result on Original Test:

Time = 39 Seconds WITH Image Resize.

New Photoshop Settings: 1 History State.
4 Cache Levels
Maximum Ram used by photoshop 100% = 3072MB

Again, the History Levels made all the difference -- 39 seconds versus 1 minute 26 seconds.

Huge difference.

DJO.

 

[dante@sisna.com]

Even at those sizes (at 400+ ppi, not sure what printer thats for...most large format printers use 400 tops, 300 or even 200 most often) its not as common as you indicate. Plus it is best to NOT interpolate the photo and let the printer do it. Exceptions are when there is vector based art. I work in the industry and rarely see files this large, even on commercial jobs. To suggest most photographers work with such large files as this just isn't accurate.

Anywho. Its the data I need to see. So I asked to see it.

We have a large format 72 inch colorspan displaymaker, a 24 inch HP designjet 130 and a few 13x19 inch Epsons -- you are obviously right -- very few printers use the kind of resolution you guys are discussing.

The only time I have to process large files such as this is when an artist has a service agency scan original oil paintings on a "to-size" flatbead at 24x36 or 40x60 inches at 360 DPI. I am sure the agency charges by the scan size, but even this 1.2 gig file size is huge overkill for print or large format reproduction. It is nice to have a box that can crank through this stuff quickly. SATA Raid sure helps in this area.

DJO.

 

[amols]

Why take the 300% image out... testing hte processors speed and not the system speed... wont give accurate rosseta results.

Yeah...got 43 sec only after skipping image size test. This is too good to be true.

 

[horhey23]

5m 17s

G4 dual 1.25 - 2GB RAM - Photoshop CS1

 

[CTYankee]

Yeah...got 43 sec only after skipping image size test. This is too good to be true.

It eliminates some of the variability caused by the differences in RAM and disk speed. So its more of a processor test. The test is greatly effected by PS settings that are highly sensitive to large files. So its either standardize the PS settings or minimize their effect. I'm making a chart of all 3 'versions' (default+ history states, 1 history state, and 1 history state w/o upsizing). I'd think that a smaller interpolation would be better, or at least adjust it for more common file sizes.

 

[spicyapple]

Yeah...got 43 sec only after skipping image size test. This is too good to be true.

How does that result compare against Bootcamp at the same PS settings?

 

[amols]

How does that result compare against Bootcamp at the same PS settings?

Hate to reboot again but the result is worth the trouble,

14 sec only...

I think you are right. Image size is more of a hard drive test which has no effect under Rossetta.

 

[groovebuster]

Was curious about the performance of my old G4...

Quicksilver with a (single) 1.2GHz G4 processor upgrade, 1.5GB RAM, running PS CS2...

5 min 12 sec

Not too bad for such an old machine!

I use this puppy still every day for work (besides my Quad G5) and I am still very satisfied as long as I am not doing 'number crunching tasks' with it.

groovebuster

 

[voyagerd]

7:43

1.8 GHz Power Mac G5
2.5 GB RAM