Click for Mini Pro Dual Drive!
Click for Mini Pro Dual Drive!


One Click on the Banner shows support to my Sponsors

Accelerate Your Mac!  - the source for performance news and reviews
The Source for Mac Performance News and Reviews
Don't forget to check out all the other site features!

Photoshop 5.0x Performance:
Apple G3 Macs vs Older CPU Upgraded Macs vs PII/PIII PCs

PSBench's 21 Filter Test Results

Updated: 5/10/99 for PIII 500MHz results


Note: This page details Photoshop 5.0x performance. For results of Photoshop 5.5 tests (with Altivec support for G4 systems and CPU upgrades), see the applications test performance pages of many of my G4 CPU upgrade reviews. On June 20th, 20000 I posted a page with PSBench test results comparing a 1GHz Pentium III vs a G4/500 system.

Introduction:
I've gotten many mails in the past asking for Photoshop comparison tests between a Mach 5 604E based Mac and the new Apple G3 systems, and also for comparisons of Pentium II/III systems to Macs. This page lists my Photoshop 5 PSBench test results (21 filter test), using several different machines including a 9600/350 running at 400MHz,an Apple G3/300 (rev 2) running at 333/222/83, a 400MHz G3 CPU card upgraded Genesis (9500 based) and also two PCs - a Dell Pentium III 500Mhz running Windows 98 (with PIII Photoshop update) and and a dual Pentium II 450 Machine running NT 4.0.

The results show that the new G3 systems are better performers in most filter tests than the older Macs, partly due to the faster bus and RAM speed, but also possibly due to a slightly faster cache speed as well. Included in the recent updates are results with Apple's newest/fastest Mac ever - the new Blue and White G3/400 (aka Yosemite), a dual PII 450 NT machine and a Dell Pentium III 500Mhz running Windows 98 and the PIII update to Photoshop 5. As shown in the table below, the Mac is still the fastest Photoshop platform, even compared to the Intel PIII 500Mhz (Win98) and dual processor systems PII systems running NT 4.0 with a SMP kernel.

The latest update (5/10/99) includes tests run at with a Dell Pentium PIII 500MHz system (XPS-T500) running Windows 98, 256MB of RAM and with the Adobe's 5.02 Pentium III update applied (new MMXCore, Lighting and Wind filter updates).

NOTE: All tests here were run with Photoshop 5's default history setting of 20 History states and 'create first snapshot automatically enabled. As per the PSBench instructions, 1024x768, millions colors with Photoshop interpolation preferences set to 'bicubic (better)'.

How they compare:
It's interesting to note that the results of many of these systems beat the best scores shown for PSBench v5 at http://www.geocities.com/Paris/Cafe/4363/5chart.html. The 9600/400 times were slightly better than the best shown on that page for a Apple G3/300 running at the stock settings (but with 60MB allocated to Photoshop - 10MB more than I used in these tests). As I've said before, for $1600 I'm very pleased with the performance of the Mach 5 9600 especially after I overclocked it. As shown on my Mach 5 page, it has outperformed the Apple G3/300 in many real world applications tests and does so again in Photoshop vs. a stock G3/300/150/66 (but not my overclocked out 333/222/83 below).

About the Systems:
The Mach 5 is a name given to the 250-350mhz 604E CPUs that are used in 1MB inline cache CPU cards of the 8600/250, 8600/300, 9600/300 and 9600/350 Mac models. These CPU cards do not work in any other Mac models. As outlined on my Mach 5 CPU card speed settings I modified my 9600/350 CPU card to run at 400MHz. I do not suggest you modify your card. It will void the warranty and is a very delicate procedure. It has run reliably for me but others have reported problems with their cards, and the required mods are not for the inexperienced. Other Mach 5 info is available at my Mach 5 Systems page.

The Apple G3/300 is a rev 2 motherboard (RagePro) that has been modified to run at 333Mhz CPU speed, 222MHz backside cache speed and 83MHz bus speed. See the G3-ZONE for details. Again I do not suggest others overclock their machines, it will void the warranty and may not be reliable.

The Yosemite G3/400 is a overclocked G3/350/DVD/6GB IDE drive base system overclocked to 400MHz.

Also included are scores from a Genesis (9500 motherboard based) system with XLR8 400MHz CPU upgrade (reviewed here).

The PC systems included both a Dell Pentium III 500MHz XPS-T500 running Windows 98 and an NT 4.0 system I built from scratch - a dual Pentium II 450MHz (ASUS P2B-DS) with onboard SCSI, 256MB Ram, Diamond FireGL 4000 graphics card, Plextor UltraPlex CD and Yamaha 4416 CDR running NT 4.0 with service pack 4 and MP kernel. The components used were top quality, from the ASUS motherboard (used by many Linux builders), Cheetah 10,000 RPM Ultra-Wide SCSI hard drive, Plextor UltraPlex CDrom, 256MB PC100 Ram, PC Power & Cooling 300W power supply, etc.. This is not a slapped together clone, I've been building, servicing and support PC/Win9x/NT machines for over 10 years.I selected the finest components possible - this machine is faster than most 'pre-built' NT systems I've seen. It's odd the more complex filters took so long on this system, actually longer than some people report on a single PII 400 system. Could the Diamond FireGL4000 drivers be a factor? I used the latest version at Diamond's site, but the card is not a very fast 2D performer, however these tests mostly stress CPU/FPU performance I believe. I didn't run the tests under Service Pack 3, only with service pack 4. I will rerun test tests again when I can, perhaps with a different video card (RivaTNT) just to see if that has any affect. I expected the NT machine to win, especially after the initial filter times were much faster than any other system I'd tested, and general applications performance of the machine is very responsive.

The PowerForce 533/267 scores were obtained from a PowerForce G3 466/311/1MB ZIF Upgrade installed in an Apple G3/300 (rev 2) (see my CPU Cards page for a full review). This particular ZIF and system combination ran reliably at 533MHz CPU speeds with a 2:1 cache ratio. Other samples and/or system combinations may not be reliable at these same speeds.

Configurations:
The first test (far right) of the 9600/400 had 192MB of 60ns FPM RAM, OS 8.1, no Libmoto or Speed Doubler used. The Graphics card was a ATI RagePro 3D card with 8MB of SGRAM, using ATI driver release v3.1. The 9600/400 had a 400MHz 604E with 1MB of inline cache running at 100MHz and a system bus speed of 50MHz. Subsequent tests of XLR8's 400MHz G3 Upgrade (see my Review)
used the same basic 9600 but with 320MB of RAM and a Radius Thunder 3D card running drivers/bios v1.07 (as noted in the table).

The Genesis system tested with the XLR8 G3/400 had 512MB of 60ns FPM RAM, OS 8.1, no Libmoto or Speed Doubler used. The Graphics card was a IXMicro Ultimate Rez card with 8MB of VRAM, using driver release v1.02. The XLR8 G4/400 had a 400MHz G3 with 1MB of backside cache running at 200MHz and a system bus speed of 50MHz. The boot disk was a wide Ultra-SCSI Seagate Cheetah connected to a ATTO PCI SCSI card (see my SCSI and Graphics card review list for info on these products).

The initial test of the Apple G3/300 was a rev 2 system with 160MB of 10ns SDRAM, OS 8.1, no Libmoto or Speed Doubler used. The onboard ATI RagePro graphics chip was used with 6MB of SGRAM. The G3 was running at 333MHz G3/750 CPU, 1MB of backside cache running at 222MHz and a system bus speed of 83MHz. Subsequent tests used the same system with a PowerForce G3/ZIF card with 400MHz CPU speed, 1MB of backside cache running at 266MHz and a 66MHz system bus speed as noted in the table below. Subsequent tests

The Blue and White (Yosemite) G3 tests were run on a 350/DVD/6GB IDE system overclocked to 400MHz. 192MB Of PC100 SDRAM was installed. The new G3 models run a 100MHz bus speed and 66MHz Rage128 16MB Video card. This system was running OS 8.5.1 (shipping version).

The Pentium III 500MHz was a Dell XPS-T500 system running Windows 98 and had the Photoshop 5.02 Pentium III update applied (new MMX Core, Lighting and Wind filer updates). Photoshop was set to use 50% of available RAM (256MB installed) so no swap file would be used. An AGP Voodoo3 3000 (166MHz core clock) video card was used set to 32bit color and using the latest standard drivers at the time of the test (not the Quake3 specific drivers that have issues with some non-Q3 apps). Some filters showed dramatic improvements from the SSE (Katmai/MMX2 instruction features of the PIII), but overall I was disappointed in the score. I can only suspect that others reporting better performance do not have Photoshop's interpolation set to Bicubic as stated in the PSBench instructions perhaps or are not timing the test accurately. The system felt very responsive, especially on restoring the test image. The PIII's total time for the 21 filter test trailed most faster G3 Macs however.

The dual Pentium II 450MHz had NT 4 (SP4) set for maximum foreground application performance. Photoshop was set to use 50% of available RAM (256MB installed) so that it would not use the PS swap file for the tests (this is in the basic readme of the PSBench tests, but some people seem to have missed that). NT 4 itself has a 400MB Swap file on a dedicated 1GB partition (the first partition, the fastest) on a U2 SCSI Atlas III drive. However as anyone that bothered to read the PSBench test info knows, the idea is to allocate enough ram to Photoshop so that it does not use its swapfile at all, which is separate from the OS's swapfile.

The standard NT system level tasks were running (nothing out of the ordinary from a standard configuration except for Adaptec's DirectCD). No screensaver or anti-virus software was active. Virtual memory was active (as is normal in NT) but no Photoshop swap file use was detected (as is desired for this test). Photoshop was run from a UW SCSI disk connected to the onboard Adaptec 2940U2W chipset's UW SCSI connector. Photoshop 5 is listed as supporting symmetric multiprocessing under Windows NT, and the PC was running the MP kernel. Adobe's Photoshop 5 Features page has it listed near the bottom - here is Adobe's quote:

" Support for symmetric multiprocessing under Windows NT, and under Mac OS systems that support Apple's multiprocessor API"

The 533MHz G3 CPU scores were obtained from a PowerForce G3 466/311/1MB ZIF Upgrade installed in a revision 2 Apple G3/300 base system, running OS 8.1, with 160MB of RAM. It is the identical base system, OS and extensions set that was used for the other Apple (beige) G3 tests.

All systems were running Photoshop 5.02 (latest version as of the test date).

The PSBench scores below are the times in seconds that each machine took to perform each of the 21 filter operations on a 10MB image. Photoshop was allocated 50MB of RAM on each machine.

The table on this page is getting too wide to add more systems, but check my CPU Cards page for results of these tests with faster 3rd party ZIF CPU upgrades (such as the PF 466Mhz ZIF that ran 550MHz in the Blue and White G3 and 533Mhz in the Beige G3. As shown on the Review, its PSBench time was 117.4 seconds running 533MHz in the Beige G3.)


PS5Bench Test

Apple G3 w/PF466 ZIF
533MHz CPU Speed
66MHz Bus
160MB SDRAM
1MB L2 at 267MHz
RagePro 6MB
OS 8.1 VM OFF
Dell PIII 500MHz
100MHz Bus
256MB SDRAM
512K L2 at 250MHz
Voodoo3 3000 16MB
Win98
PShop5 PIII Update
Dual PII 450MHz
100MHz Bus
256MB SDRAM
512K L2 at 225MHz
FireGL4K 31MB
NT 4 SP4
MP Kernel
Cheetah 10K RPM UW HD
B&W G3/400
100MHz Bus
192MB SDRAM
1MB L2 at 200MHz
Rage128 16MB
OS 8.5.1 VM OFF
Apple G3 w/PF400 ZIF
66MHz Bus
160MB SDRAM
1MB L2 at 266MHz
RagePro 6MB
OS 8.1 VM OFF
Genesis w/XLR8 G3/400
400MHz G3
50MHz Bus
512MB RAM
1MB Backside cache at 200MHz
Ultimate Rez
OS 8.1 VM OFF
9600 w/XLR8 G3/400
400MHz G3
50MHz Bus
320MB RAM
1MB Backside cache at 200MHz
Radius Thunder 3D
OS 8.1 VM OFF
9600/400
(OC stock 350MHz card)
400MHz 604E
50MHz Bus
192MB RAM
1MB L2 at 100MHz
ATI RagePro 3D v3.1
OS 8.1 VM OFF
Apple G3/333
(OC 300 CPU)
83MHz Bus
160MB SDRAM
1MB L2 at 222MHz
RagePro 6MB
OS 8.1 VM OFF

Comments

Rotate 90

0.7

0.3

0.4

0.7

1.1

1.0

1.0

1.5

1.1

Odd that the 533Mhz Upgrade was not faster on this test. First runs were slower than 2nd/3rd runs on most Macs.

Rotate 9

2.7

2.8

2.1

3.2

4.2

4.4

5.1

4.8

4.4

 

Rotate .9

2.5

2.6

1.9

3.0

3.1

3.6

3.8

3.6

3.5

 

Gaussian Blur 1

1.2

0.8

1.2

1.3

1.4

1.8

1.8

1.7

1.5

 

Gaussian Blur 3.7

3.2

2.2

1.5

3.6

3.7

5.1

4.9

5.1

4.2

 

Gaussian Blur 85

4.3

4.7

3.0

4.8

5.6

7.8

7.2

8.4

6.2

 

Unsharp 50/1/0

1.5

0.9

1.3

1.9

1.8

2.3

2.3

2.2

2.0

 

Unsharp 50/3/7/0

3.6

2.3

1.5

4.0

4.2

5.7

5.4

5.9

4.7

 

Unsharp 50/10/5

3.6

2.6

1.7

4.0

4.2

5.8

5.6

6.1

4.7

 

Despeckle

1.8

3.2

1.9

2.4

2.4

2.6

2.9

2.9

2.8

 

RGB-CMYK

3.8

6.0

6.5

5.1

5.0

5.2

5.6

6.0

5.9

 

Reduce Size 60%

1.0

1.1

0.7

1.3

1.3

1.5

1.7

1.7

1.5

 

Lens Flare

3.6

4.7

5.2

4.4

4.5

5.4

5.9

6.6

5.3

 

Color Halftone

4.2

7.6

13.0

4.1

5.4

6.3

6.8

8.3

6.4

 

NTSC Colors

3.0

5.4

4.3

3.9

3.8

4.2

4.6

6.0

4.7

 

Accented Edges

8.1

13.8

13.1

10.8

10.7

11.1

12.6

13.2

12.8

 

Pointillize

12.4

16.9

18.6

15.9

16.4

17.6

19.7

19.8

19.6

 

Water Color

17.0

29.6

26.7

22.6

22.5

23.3

26.0

27.3

26.9

 

Polar Coordinates

4.6

10.7

12.8

5.6

6.1

7.3

7.9

9.1

7.2

 

Radial Blur

27.7

45.0

48.9

36.4

35.7

37.6

38.2

40.0

42.4

 

Lighting Effects

6.9

2.5
(PIII Opt.)

10.7

9.6

9.6

10.3

11.3

11.8

11.1

 

PS5Bench Index

117.4

165.7

177.0

148.6

158.6

169.9

180.3

192.0

178.9

Total Time

System

Apple G3 w/PF466 ZIF
533MHz CPU Speed
66MHz Bus
160MB SDRAM
1MB L2 at 267MHz
RagePro 6MB
OS 8.1 VM OFF
Dell PIII 500MHz
100MHz Bus
256MB SDRAM
512K L2 at 250MHz
Voodoo3 3000 16MB
Win98
PShop PIII Update
Dual PII 450MHz
100MHz Bus
256MB SDRAM
512K L2 at 225MHz
FireGL4K 31MB
NT 4 SP4
MP Kernel
Cheetah 10K RPM UW HD
B&W G3/400
100MHz Bus
192MB SDRAM
1MB L2 at 200MHz
Rage128 16MB
OS 8.5.1 VM OFF
Apple G3 w/PF 400 ZIF
66MHz Bus
160MB SDRAM
1MB L2 at 266MHz
RagePro 6MB
OS 8.1 VM OFF
Genesis w/XLR8 G3/400
400MHz G3
50MHz Bus
512MB RAM
1MB Backside cache at 200MHz
Ultimate Rez
OS 8.1 VM OFF
9600 w/XLR8 G3/400MHz
400MHz G3
50MHz Bus
320MB RAM
1MB Backside cache at 200MHz
Radius Thunder 3D
OS 8.1 VM OFF
9600/400
(OC 350 card)
400MHz 604E
50MHz Bus
192MB RAM
1MB L2 at 100MHz
ATI RagePro 3D v3.1
OS 8.1 VM OFF
Apple G3/333
(OC 300 CPU)
83MHz Bus
160MB SDRAM
1MB L2 at 222MHz
RagePro 6MB
OS 8.1 VM OFF

 

Note: A stock Apple G3/300 scored (at best) a 196.6 (300/150/66, 128MB Sdram, 60MB allocated to Photoshop) as shown at: http://www.geocities.com/Paris/Cafe/4363/5chart.html




Summary:
I have to say I was surprised that the Pentium III 500Mhz did not score better. Granted the Adobe Pentium III update only contained a few improvements but did have a new MMX Core. There was a dramatic improvement in the lighting test from the new optimize filter, but overall scores were still disappointing. Since the Dell XPS-T 500 was one of the highest rated (if not the highest) Pentium III 500 system at this time (per Maximum PC magazine and other reviews), using this system hopefully will remove the questions about my homebrew dual PII machine not being 'optimized' or having some flaw (and note the dual PII still outperformed the single PIII on some filters).

In case anyone is not aware of it, Adobe says Photoshop 5 supports symmetric multiprocessing under Windows NT, and the dual PII 450 PC was running the NT MP kernel. The results show a fast Mac is still the best overall Photoshop machine (at least in the PSBench wide array of filter tests), as the PC's total time for the 21 tests exceeded that of the faster Macs listed in the table (it placed 4th of the 7 systems tested, besting a G3/333/222 system by less than 2%). The PC was the best performer at most of the shorter duration filters, but on the more time intensive ones it really lost ground to the faster G3 Macs.

What is not shown in these scores is file IO times, which was normally much faster on the PC (reloading the image for instance). The PSBench tests do not take this into account however.

As of this article date (Early 1999), it appears the new Blue and White G3's (aka Yosemite) are the fastest overall Photoshop machines Apple has made, but consider if the cost and lack of legacy interfaces (no serial ports, no scsi standard, no floppy) and compatibility issues (most are being addressed now via firmware and driver updates). The B&W G3 is the best Mac from Apple to date in the areas of price/performance, ease of servicing and value.

The original purpose of this page was to compare the Mach 5 (350MHz 604ev Mac, overclocked to 400Mhz) to the G3/300 class of machines. Considering the Mach 5 (at 400/100/50) was running a 33MHz slower memory bus (Beige G3/333/222 with bus speed set to 83MHz) and 122MHz slower cache bus, the scores were very close. Despite the low PCI/memory performance of the Mach 5's Kansas motherboard, it performed better than I expected. In fact as mentioned above, the 9600/400 outperformed a stock Apple G3/300 at these same tests, so if you need 6 PCI slots, the 9600 series of machines are still very attractive.


Back to XLR8YOURMAC.COM
Your Source for the best in CPU/SCSI/VIDEO card reviews, daily news, and more!

Copyright © Mike, 1998-1999.
No part of this site's content is to be reproduced in any form without permission.
All brand or product names mentioned here are properties of their respective companies.
Disclaimer: Users must read and are bound by the Site Terms & Conditions of Use.