Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

'Real World' Photoshop Benchmark - Speed Test

sneezymarble said:
EDIT: I went back and ran some more tests. The problem was that I had lowered my system voltage a few days ago while running some unrelated tests. The Photoshop test was putting my cores under a near full load and I think the system was choking a bit because the voltage was too low. I upped the voltage to where I usually keep it and now I'm getting 19s in CS4 as opposed to 21s in CS3. That's about a 10% improvement.

I think the performance gains look better on the Apple forum because most of the benchmarks are on slower systems and so the spread is higher.
Click to expand...

How on earth do you increase or decrease the voltage on a Mac Pro? Is it possible then to overclock a MacPro 1.1 system?

EDIT: Nevermind... It came up on a search...

https://forums.macrumors.com/threads/507454/
 
Here's my specs & results:

CS4 || MacPro 1.1 || X5355 2.66 8-Core || 4Gigs RAM || ZDNet's OverClocker

Performing the PhotoShop benchmark from this site: http://www.driverheaven.net/photoshop.php at two CPU speeds:
  • At 2.66 GHz
    Texturizer:............... 1.03 sec.
    CYMK CC:................. 1.27 sec.
    RGB CC:................... 1.54 sec.
    Dust & Scratches:...... 1.63 sec.
    Water Color:............. 11.32 sec.
    Texturizer 2:............. 1.21 sec.
    Stained Glass:........... 3.36 sec.
    Lighting Effects:........ 3.41 sec.
    Mosiac Tiles:............ 10.56 sec.
    Extrude:.................. 44.52 sec.
    Smart Blur:............. 62.85 sec.
    Underpainting:......... 13.92 sec.

    Total Score:..............156.62 sec.

  • At 3.06 GHz
    Total Score:..............138.81 sec.


Performing the PhotoShop benchmark from this site: http://www.retouchartists.com/pages/speedtest.html The results are:
  • At 2.66 GHz - 28.53 Seconds
  • At 2.76 GHz - 27.43 Seconds
  • At 2.86 GHz - 26.35 Seconds
  • At 2.93 GHz - 25.32 Seconds
  • [*]At 2.97 GHz - 25.53 Seconds
    [*]At 3.01 GHz - 25.08 Seconds
    [*]At 3.10 GHz - 25.53 Seconds
Me thinks PhotoShop is not very good for profiling CPUs. ;) The speed bar in ZDNet OC changed from green to yellow at 2.968 GHz (2.97 GHz) thus the colors in the list above. I guess farther up it turns red but I didn't go there! :eek:
 
Photoshop Benchmark from this Site:

27.5 seconds


PhotoShop benchmark from this site: http://www.driverheaven.net/photoshop.php

Texturizer:............... 1.3
CYMK CC:................. 1.6
RGB CC:................... 1.7
Dust & Scratches:...... 1.3
Water Color:............. 10.2
Texturizer 2:............. 1.2
Stained Glass:........... 3.2
Lighting Effects:........ 2.4
Mosiac Tiles:............ 9.1
Extrude:.................. 39.3
Smart Blur:............. 54.7
Underpainting:......... 13

Total: 139 seconds
 
Thought people might want to see what Intel's latest can do...

Intel Core i7 965 Extreme Edition, overclocked to 3.33/3.73 GHz. ("Turbo Mode", with 3.33 GHz as the full-load speed, 3.73 GHz as the single-thread speed. For this proc, the 'all cores busy' speed is, in fact, the stock speed. Only the 'only one core busy' speed is overclocked.) HyperThreading ON. 4 GB DDR3-1333 RAM (Only dual channel instead of triple channel.) 1 TB Seagate 7200.11 hard drive as the 'everything' drive; about 100 GB full. ATI Radeon 4850 512 MB.

Windows Vista SP1 64-bit, Photoshop CS4 Trial 64-bit

100% RAM Usage, 1 History State, 4 Cache Levels, OpenGL Drawing off (I saw no tangible benefit from GPU acceleration; two tests were 0.1s faster, one was 0.1s slower.) All tests run four times, with the final three averaged together. (The first run is always slower by a few percent. In my case, the final three were all identical.) For Retouch Artists, I used a stopwatch that goes to the thousandth of a second, the times were from 14.943-14.982.

Retouch Artists: 14.95 seconds.

DriverHeaven:
Texturizer: 0.5s
CMYK Colo(u)r Conversion: 0.4s
RGB Colo(u)r Conversion: 0.4s
Dust and Scratches: 0.7s
Watercolor: 7.3s
Texturizer: 0.6s
Stained Glass: 1.9s
Lighting Effects: 1.2s
Mosaic Tiles: 3.7s
Extrude: 19.3s
Smart Blur: 18.9s
Underpainting: 7.8s
Total: 62.7s

And since some have argued that this setup is an unrealistic test (100% RAM, only 1 history,) I also ran it with 20 history; 50% RAM. However, in this setup, RA was so slow I only ran one pass of it, and I'm only going to report the total for DriverHeaven. (Although DriverHeaven wasn't affected nearly as much.)

Retouch Artists: 3 min, 22 sec.
DriverHeaven: 63.7s

Tomorrow (or maybe the next day,) I'll see if I can do a comparison of less capacity, higher-speed RAM vs. more capacity, slower-speed RAM. (Specfically, Three-channel 3 GB of 1600 MHz vs. One-channel 8 GB of 1066 MHz.) This run was on my home machine; any future testing will be on a machine at work. (And if I'm lucky, I'll get to slap it on a RAID of much faster drives.)

Edit: Okay, benchmarked a near-identical system; only once with 3 GB of 1600 MHz, and once with 8 GB of 1066 MHz. Very little difference between the two configs. Dropping to 2 GB of RAM was dismal, though. Both tests appear to be slightly RAM speed bound once you're past the 2 GB mark. The Retouch Artists test had a difference of about 7%. 14.1s for the faster/less capacity vs. 15.2s for the slower/more capacity. DriverHeaven was 61s vs. 71.5s, or about 17%. This system was running at the same CPU speeds; but with a RAID-0 of SSDs as the only storage. (So the SSDs also appear to have had nearly no effect as well.) And an nVidia GTX 280 instead of the Radeon I have at home.
 
Intel Core i7 speeds

ehurtley,
thanks for posting those impressive tests. have you done a comparison between 32-bit and 64-bit windows versions? I'd be curious how much of a speed up there is in CS4 between the 2 OSes.
paul
 
OSXconvert said:
ehurtley,
thanks for posting those impressive tests. have you done a comparison between 32-bit and 64-bit windows versions? I'd be curious how much of a speed up there is in CS4 between the 2 OSes.
paul
Click to expand...

Alrighty, here you go:

Code: 
[b]64-bit 32-bit Diff%[/b]
0.6    0.7    14.3%
0.5    0.7    28.6%
0.5    0.5    ----
0.8    0.8    ----
8.1    8.4    3.6%
0.6    0.7    14.3%
2.1    2.1    ----
1.2    4.1    [i]70.7%[/i] [b]<- Yowzas![/b]
4.1    4.8    14.6%
20.4   23.9   14.6%
19.9   19.8   -0.5%
8.2    8.5    3.5%
[i]67 s   75 s   10.7% (Total)[/i]

Yes, that's a 70% reduction in time for the"Lighting Effects" step. I verified it four times in each mode. The seemingly large percentages in Texturizer and CMYK conversion are because the times involved are so small. The margins from multiple runs of each had each test varying by 0.1; so in some cases, they flip-flopped. (Sorry, I forgot to do an average of multiple runs.)

Overall, it appears that 64-bit mode gains you about 10%.

Edit: And while I'm being statistical..... I went ahead and messed with the settings in the BIOS. I did one 64-bit run with half the cores disabled and HyperThreading off, then again with only one core enabled, but HyperThreading on. (All my previous tests were with all four cores on, HyperThreading on; so the processor appears to the OS as 8 cores. These two new tests both appear to the OS as two cores, but in different ways. Didn't go to one core, no HT because Vista itself becomes slow that way.) I saw no tangible, measurable difference in times for the two real cores test. Times varied by 0.1-0.3s from what I reported above, but my testing conditions weren't rigorously scientific (not a 'clean' OS install, my tests above weren't done immediately after a reboot, whereas these new ones were, etc,) so I can chalk those up to margin of error.

The one-core-plus-HT test, though; showed that Photoshop really likes two real cores on a few of the operations. HT doesn't make it happy. Total time was 69.3. The two Color Conversions, Dust and Scratches, and Lighting Effects each took over twice as long; but as they are all short operations, didn't have a huge impact. And, oddly, a few of the longer tests took less time. (Which could be because this was done immediately after a reboot, I'll call the faster results due to margin of error.)

Maybe if I have a few spare hours at work in a few weeks, I'll do a properly scientific analysis with many variables. (32-bit/64-bit, XP/Vista, amount of RAM, speed of HD, etc, etc, etc...) If I had the time or energy to install Boot Camp on my Mac, I'd do a comparo of OS X vs. Windows. (I don't have a 64-bit Mac, so I can't do a 32-Mac vs. 64-Win comparison, sorry.)
 
Hi, my HP 950C keeps complaining about the color cartridge. The cartridge is the correct type but the printer keeps saying the cartridge does not presence or it is the wrong type. I have tried cleaning the contact as HP web site suggested but no luck. I am looking a way to bypass the cartridge checking. How does HP inkjet printer detect whether the cartridge is the correct type or not?
 
MacPro 2.8 Dual, 10GB Ram, OS 10.5.4 , 1 scratch disk (partition on some internal WD 7k drive).

Retouch Artists test, Photoshop CS4, 100% Ram, Cache 4.

History 20 : 65 sec.

History 10 : 43 sec.

History 1 : 20 sec.




FWIW, below are some former results for my old G5 + PS CS/CS2 -->


Horst said:
DP 2.0 Ghz, Rev. B, 4.5 GB Ram, internal SATA scratch disk (half full), Radeon X800XT .

Photoshop CS2 / 9.01, Cache: 2


History 15, Ram 70% : 1.58

History 1, Ram 85% : 1.17

History 1, Ram 85%, 'Bigger Tiles' Plugin enabled : 1.06

History set to more realistic 10, Ram 85%, 'Bigger Tiles' Plugin enabled : 1.40
Click to expand...



Horst said:
My results with Photoshop CS / 8.01, Cache levels: 2
CS2 results see above.


History 1, Ram 95% : 1.10

History 10, Ram 95% : 1.40

History 20, Ram 95% : 1.53

I upped the Ram allocation a little for CS, as it isn't able to use more than 2GB, as opposed to the 3GB CS2 is capable of using.
Click to expand...
 
64 bit vs. 32 bit vista

ehurtley,
Thanks for running those tests. So, about 10% speedup with Vista 64 vs 32 in CS4. Not bad. I suspect OSX running on your hardware would be around 25-30% slower than Vista 64. Leopard and CS4 are just not that well-optimized for the multiple cores and hyperthreading. When Snow Leopard comes out next year, I bet the gap will narrow a lot, but we'll have to wait and see.
Paul
 
Just curious. Any of you folks equipped the Intel X-25M SSD on any of your setups yet? I wonder how much improvement does the Intel SSD add.
 
Jiten said:
Just curious. Any of you folks equipped the Intel X-25M SSD on any of your setups yet? I wonder how much improvement does the Intel SSD add.
Click to expand...

None.

I tested on a Western Digital "GreenPower" 1 TB drive as the sole drive on one system, and on a pair of X-25M SSDs in a RAID-0 on a second (otherwise identical) system. The difference in scores between the two systems was so small as to fall within a reasonable margin of error.

The SSDs improve many other things; but this is not one of them.

Next week or the week after I should have a chance to do many runs in many different configurations.

edit: here you go:

SSD scores:

Retouch Artists: 14.1 and 15.2 seconds (two different runs; sorry, I don't recall what the difference between the runs was. I know it was the same hardware both times, with a two-SSD RAID. Maybe it was 64-bit vs. 32-bit OS?)

DriverHeaven:
Code: 
0.7  0.7
0.5  0.6
0.5  0.6
0.8  0.9
8.2  8.3
0.7  0.8
2.2  2.3
1.3  1.3
4.1  4.1
21.2 21.8
20.8 21.3
8.7  8.8

Again, I don't recall the exact differences. Maybe I just ran two separate passes. (It was a couple weeks ago now, I just found my results in a text file that all I did was type in the results, no descriptive text.)
 
i did the original test on my Intel (core2duo) 2.13ghz with 2GB RAM and an eVGA 8800GTA 320mb, running XPSP2 with CS3

i was lazy so i didnt restart, or change any settings, and i had mIRC, avast scanner, opera, daemon tools and steam running in the background

3m 14s.
 
Vista 64 SP1, CS3 32bit = 16.50 seconds
Vista 64 SP1, CS4 64bit = 15.02 seconds

It appears CS4 64bit is roughly 10% faster in the RA artist test. Will do the driverheaven test later, as well as CS4 OS X on the same machine.

Though I might have a Core i7 @ 4Ghz machine up and running soon. That should be fast :D
 
I did the test on CS3 and CS4 on my MBP C2D 2,33 ghz 2 gb ram, 160 gb 5400 rpm:

CS3: 54 secs!

CS4, with GPU accelleration on: 54 secs!!

SO much for speed enhancement of CS4 - I dont get it!!!
 
apfhex said:
On my Mac Pro 2.66GHz with 2GB RAM. I ran the test in Photoshop CS3 Beta first, quit, let my system "recover" (the RAM usage made it pretty sluggish), then ran it in Photoshop CS2.

Retouch Artists Speed Test results:
CS3 - 1:02
CS2 - 2:17
Click to expand...
Maybe no one cares about an old Mac Pro but for my own interest I wanted to revisit this test with my upgraded system. What's changed: using Photoshop CS4, 6GB RAM (though Photoshop can only use 3GB), using non-boot drive (WD Caviar 7200RPM 8MB cache) as primary cache disk and boot drive (WD Caviar 7200RPM 16MB cache) as secondary, and of course 8800GT. Not sure how much CS4's OpenGL drawing features make a difference in the test. Oh, and running Leopard, but I don't know how much that affects it either.

I didn't restart sooo the result might not be optimal, but I quit all open apps and set Photoshop's preferences according to the test's conditions.

Results:
CS4 - 0:36
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back