Benchmark Photoshop Speed Test

[blackviolet]

there isn't any point in getting one, now. CS3 will not only support intel Macs natively, it will also be properly optimised as a multi-threaded windows app. then when the intel 'PowerMac' replacements come out, look at both platforms. the Mac will let you run both OS universes.

[PacAce]

Originally Posted by blackviolet

there isn't any point in getting one, now. CS3 will not only support intel Macs natively, it will also be properly optimised as a multi-threaded windows app. then when the intel 'PowerMac' replacements come out, look at both platforms. the Mac will let you run both OS universes.

I don't get it. I thought CS3 is supposed to be "Universal", meaning it will run natively on both Intel-based and PowerPC-based machines. The only difference between the two would be the compiler used to generate the binary executable files.

[blackviolet]

the universal binary will run on both intel and PPC macs - not on windows boxes. i wouldn't recommend buying a G5 now - sure the PPC version will be great, but as a hardware platform, you would be limiting yourself. getting a Conroe Intel based Mac when they come out makes good sense as you have the ability to run osx, xp, vista, linux, etc.

the current CS2 windows code is not nearly as optimised as it should be to reap the benefit of todays dual core processors. merely making an application multi-threaded doesn't make it optimised. once CS3 is out, hopefully the app will run close to the same on intel based mac and windows boxes on similar hardware. that's what i meant above.

[PacAce]

Originally Posted by blackviolet

the universal binary will run on both intel and PPC macs - not on windows boxes. i wouldn't recommend buying a G5 now - sure the PPC version will be great, but as a hardware platform, you would be limiting yourself. getting a Conroe Intel based Mac when they come out makes good sense as you have the ability to run osx, xp, vista, linux, etc.

the current CS2 windows code is not nearly as optimised as it should be to reap the benefit of todays dual core processors. merely making an application multi-threaded doesn't make it optimised. once CS3 is out, hopefully the app will run close to the same on intel based mac and windows boxes on similar hardware. that's what i meant above.

OK, thanks for the clarification. That makes more sense than when I read (or should i say "misread") your previous post.

[blackviolet]

it was largely my fault staying up to watch the world cup and then posting early in the morning - i haven't had much sleep over the last week or so...

[gbrandon]

Okay, I stumbled across this forum by chance after reading some posts in the 2cpu.com forums, but heres my results. Also, please note I know as much about photoshop as the average person knows about quantum physics, so please be patient. I used this test solely for fact that my gf uses photoshop and as I just built this computer for video editing, I figured she could use it for her illustrator and photoshop needs. This test seemed to be a good benchmark for tweaking my system.

Specs: Homebuilt
Supermicro H8DC8 Motherboard
2 -285 Opteron processors (dual core for a total of 4 cores)
8 gigs ram.
1 Maxtor SCSI U320 150 gig drive (O/S and applications)
1 Raid 0 array with 2 WD Raptor 1500's. (300 gig partition) used for scratch disk
Windows XP X64 bit edition, with Photoshop CS2 version 9.0.1

When I first built the system, there was a single raptor for the scratch disk. I ran the test and got 1:45. I then played with the settings in photoshop and changed scratch disk from 1)c 2)d to 1)D 2)C and also moved the memory used to 75% (i think it was around 2 gigs ram to use at 75%) I re-ran the test and got 1:15. Smoking!!

I then read some photoshop forums and it said to check out the efficency. I noticed mine was around 85% but it sometimes dropped to 20%. I also noticed when watching performance under windows task manager that it was using alot of page file, but all 4 processors rarely got above 25% usage (some spikes to 100 for very very short periods of time). I figured ram was the bottleneck. I then bought another western digital drive and created a raid 0 array and re-ran the test. I got :57 seconds. Damn, thats impressive. I wanted a comparison, so I ran the test on a old athlon fx-57 with 2 gig ram and 1 raptor 75 gig and got 2:45.

I then bought 4 more gigs of ram and ran the test and suprisingly, my test results went UP to 1:15. Hmm...I also noticed on the photoshop setting for memory that max ram available to use was still 2786. ?? I thought I read on the adobe forum that photoshop could use 4 gigs of ram. I wonder whats goin on? Anyway, I read this forum and took the advice of some previous posters. I went to preferences, and under the general tab I swicthed history states from 10 to 1, and then closed all the little windows on the right hand side of the main window of photoshop except the test script window. (i hope you understand that last sentence) I then re-ran the test. And guess what....

:34 seconds. WOW!!. I feel thats pretty impressive. But I still want to know why photoshop is limited out for memory at 2786. Anyone have any suggestions?

Thanks..

Geoff

[CyberDyneSystems]

O-kay, I just had a dismal showing for the system I have, which is very similar to the Dual Opteron listed directly above this post...
(Iv'e got the slower 2.2GHz Opteron 275s and my disk subsystem is NOTHING like Geoff's awesome set up! )

I have a few problems.

1.Wrong version of PSCS?:
I'm running PSCS or PS8 not CS2.. action runs, but I can;'t find any indication what version this test was intended for.

2. PSCS Limited to 2GB RAM?:
This seems to mean that even with PSCS set to use 100% memory, it will only address a little under half? (1.8GB as opposed to the full 4GB available? )

3. CPU's not being used:
On THIS test, PSCS does not seem to want to use all the CPU power available? When I actually use PSCS for editing, I see my CPUs redline all the time, particularly using the same Gaussian blur filter that this speed test uses and seems to bog down the most. In my real world Guassian tests, I fly through them with CPU utilization bumping 100%, .. here it seems to stop dead on this same filter, with CPU barely bumping 14% ??
For the duration of this test the CPU's never seemed to go up above perhaps 65% usage.. and more often was way below 50%
(**EDIT after setting to 1 history state this changes a lot, CPUS are relined often enough, until that last blur filter)

I'm assuming this is hard drive IO related, (**even after the history state revelation I still think this) and indeed, I have yet to install any separate scratch disks for this system, that coupled with the fact that PSCS is ignoring half the installed RAM seems to mean a much more disk intensive action.
I ran it a second time with the history states at 1, and even so, that least Gaussian blur filter, CPUS were down at 0% - 10% for a full 15-20 seconds, memory dropped from 1.8GB down to under 500MB, and the system just seems to hang there, (disk IO was not even flashing, and drives were idle)

Score as of now: 1minute 48 seconds.

So I'm actually already approaching Leo's awesome PowerMac G5, and currently I'm doing this with only a single physical hard drive array (two 400GB drives in RAID1 mirrored)

I think if I could get PSCS to use more ram and get away from the disk IO, and thus let the action take full advantage of the currently Idle CPU time, I would be in much better shape,

[StewartR]

CDS, I'm sure I speak for all POTN members when i say I'm disappointed with that. Surely you can do better?

Originally Posted by CyberDyneSystems

[StealthLude]

hahahahaha

58 seconds =)

Home Brew 2.8 Dual Xeon Server Cluster (x 2 servers) with 1.5 TB Raid 5 Array on a 3Ware Raid Card with 128 MB Drive Casche, 4 Gigs Ram. Using Windows Server OS

PS Scratch Disk are 2 x Raptor 36 gig drives on Raid 0

[StealthLude]

Nice Server Room in the post above, I wish mine looked like that.

I got 2 x Homebrew servers in SuperMicro rack mount cases. Using a Dell 1/2 sized rack with APC power backup and Watchguard VPN Firewall.

Sigh, I dont have lots of L glass b/c i decided to buy hard drives instead. =)

[pturton]

Half hour to download the test files by dialup but repeatable
56 seconds
to run the test.

CS2 9.02
AMD Athlon 64x2 Dual Core Processor 4600+ 2.41 GHz
3 GB Ram
Asus M2N-E Motherboard
3 WD SATA 250 Gig drives
NVIDIA GeForce 7300 GS video card
20 inch and 19 inch monitors both @ 1600x1200

[CyberDyneSystems]

56 seconds.. nice.

It was more sensible to have the system builder install only the system volume, so I have now placed my order @ Newegg for three more hard drives.

Two 500GB drives for a 1/2 TB mirrored array, for the photos, and a Raptor fro the scratch disk. Didn't wan't to dedicate two whole drives in RAIDzero and the power overhead required just for a scratch disk that would be waaaay too large for what it needs to do, so no RAID ZERO for me. This the Raptor as the fastest single disk solution available without a SCSI controller.

I'm thinking once I have this in place, my disk IO issues will be over

[pturton]

Originally Posted by CyberDyneSystems

56 seconds.. nice.

It was more sensible to have the system builder install only the system volume, so I have now placed my order @ Newegg for three more hard drives.

Two 500GB drives for a 1/2 TB mirrored array, for the photos, and a Raptor fro the scratch disk. Didn't wan't to dedicate two whole drives in RAIDzero and the power overhead required just for a scratch disk that would be waaaay too large for what it needs to do, so no RAID ZERO for me. This the Raptor as the fastest single disk solution available without a SCSI controller.

I'm thinking once I have this in place, my disk IO issues will be over

One thing that may be helping my system is the fact that I have only my Internet access, DVD/CDR burning support and photo software installed plus a ffew minor utilities.

I am not using RAID, only for the occasional bug. Going from 2 GB RAM to 3 GB made a noticeable difference. my RAM is only running at 633 MHz, BTW - the 800 MHz RAM was more than I can afford. After using a PIII 700 laptop for 5+ years with WIn Me, this system is a real pleasure accept for the dialup Internet access. Hopefully, someday, I can purchase another video card that doesn't steal 384 MB RAM from my system board.

Your system sounds like it could be a winner once you add your new drives and move up to CS2 or CS3 when available.

Have you tried the /3GB switch in the Boot.ini? Until I added this switch, PS only saw 1.6 GB but now it sees 2.6 GB.

Now I'd better spend the time and learn how to use PS properly.

[CyberDyneSystems]

I have not tried that... some day when I get home from work I will have to give it a go!

[pturton]

Originally Posted by CyberDyneSystems

I have not tried that... some day when I get home from work I will have to give it a go!

Just in case you do not have the syntax handy:


[boot loader]
timeout=12
default=multi(0)disk(0)rdisk(0)partition(1)\WINDOW S

[operating systems]
multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Micro soft Windows XP" /noexecute=optin /fastdetect /userva=3000 /3GB /noguiboot