PDA

View Full Version : Cheetah Raid for Photoshop-what am I doing wrong?



daveseeley
02-16-2005, 09:19 AM
OK...so after reading the photoshop performance guide here on MacGurus...I've been on a quest to increase photoshop 7 performance with way big files by upgrading scratch/working disk. This started when I bought into a new G5 2.5 ghz with 2.5 gigs Ram and was unimpressed with PS7 performance increase from my dual 1 gig quicksilver with sata card/drives. I cobled together a 4x 36gb 15.3 k cheetah raid with u320 cables and atto ul4s card in the pci-x fast ass slot.

My understanding was that the increase comes from save-times and when tasks exceed physical ram and go to scratch disk. I tested three different scratch disk drive setups with saving a file in ps, and then running through a long ps action with several steps on a file exceeding physical scratch ram (as all my files typically do). Each time the scratch disk was reset in PS prefs and restarted. PS shows 1.86 gb ram available. No other apps running. My boot drive is a maxtor diamond max 10 300gb with the boot partition on the outside 90gb.

4x Cheetah Raid
220mb/sec - XBench uncached write 4k blocks
288 - XBench overall disk Test Score
55 secs - Save 473 mb file to disk after opening file from another disk
21 secs - Save As over that file on the same disk
8 mins - Photoshop action = rotate canvas-flip-flatten-filters-invert-flip-fill
11 mins - Photoshop action reapeated= rotate canvas-flip-flatten-filters-invert-flip-fill


1x Cheetah
78mb/sec - XBench uncached write 4k blocks
174 - XBench overall disk Test Score
52 secs - Save 473 mb file to disk after opening file from another disk
22 secs - Save As over that file on the same disk
10:13 mins - Photoshop action = rotate canvas-flip-flatten-filters-invert-flip-fill

sata seagate 200gb
59mb/sec - XBench uncached write 4k blocks
117 - XBench overall disk Test Score
1:28 secs - Save 473 mb file to disk after opening file from another disk
25 secs - Save As over that file on the same disk
12:46 mins - Photoshop action = rotate canvas-flip-flatten-filters-invert-flip-fill

So in my real world tests: The raid had a significant speed increase over the singel Cheetah the first time I ran the action, and the single Cheetah beat out the raid on the raids second go, and neither seemed significantly better than the sata drive on the seritek card. Am I doing something wrong?

Thanks for any insights

Dave
:confused:

TZ
02-16-2005, 09:46 AM
I'll let others evaluate Photoshop and what it means ;)

PS FAQ (http://www.macgurus.com/forums/showthread.php?t=19792)
PS Benchmarks - Discussion (http://www.macgurus.com/forums/showthread.php?t=19688)


NOTE: Adobe is waffling on the Ram/Memory issues for 2.5GB-plus work stations...if 100% Memory is not working optimally, as recommended above, try reducing down to a "safer" 80%, said Adobe's Chris Cox

# Cache Levels: Adobe is recommending setting this to 4 (for now because of an OS10.3.2/PS8.0 conflict). I generally set the Cache Level to 8 on newer machines.
From http://www.gballard.net/psd/troubleshootpurgepsd.html


For Photoshop on PowerMac G5’s, we recommend a minimum of 2.5GB of RAM be installed in order to maximize Photoshop’s RAM allocation. This way, the operating system will not suffer under a "100%" or a "75%" allocation to Photoshop. Recent studies have shown that loading a G5 with 4GB of RAM provides about the maximum possible benefit. Beyond 4GB the cost/benefit ratio for Photoshop acceleration quickly deteriorates.

#10: For advanced users who have implemented a separate two or four-drive striped RAID array for Photoshop’s scratch disk, an additional speed boost is obtained by saving the image file to disk at the same volume location used for the scratch disk. This is commonly known as the "work disk/scratch disk" method.

... SoftRAID offers settings optimized for specific applications or workflows, including Photoshop.

Performance of Photoshop CS is actually about 7% faster than Photoshop 7
Photoshop Guide (http://www.macgurus.com/guides/photoshopguide.php)

Question:
Does the UL4S have the latest firmware, 3.20v2 driver, as well as 10.3.8?

Is your RAID partitioned? optmized for PS?
(You would need SoftRAID 3.1.x for this)

Do you copy the file to the RAID first before working on it? From my read of the PS Guide, that helps I think.

What you found suggests to me that one fast 15K or 10K drive (now that they are in the 89MB/sec ballpark) might suffice.

SATA does give good performance - and Raptor or DM10 seem about even in performance - though my hat would go to the 10K Raptor personally for what you do. I think the future for most users is 4-channel SATA-II controller and drives.

kaye
02-16-2005, 10:10 AM
Dave,

I think that we need Boots to comment on your results. But I would suggest that you try his more controlled Pshop test at least a couple times. I printed out the docs so that I could follow the instructions exactly and fill out the form as I went along. I am no Pshop expert but thanks to the tests I did find out what works best for my G4s for the type of Pshop work that I do. That might give you some insight into why your results are so different the 2nd try. k

Boots
02-16-2005, 10:22 AM
Hi Dave,

Could you test your 4 x x15.3/UL4S stripe with Quickbench 2.1 Extended 100MB test? We will have a better idea from yours results to compare with other 4-drive x15.3 stripes. This way we can tell if your 4 x stripe is up to snuff.

Also, if you test with PshopTest (http://www.macgurus.com/forums/showthread.php?t=19746), I can get a much better handle on this with suggestions for better results.

Several things I see right off the bat:
More RAM will help you.
Use Photoshop CS instead of 7- it's much faster.

Edit: OK, I just figured out yer runnin a UL4 single channel- not dual. You'll get much better performance with a UL4D. I think 3 x15s saturate a single channel; I believe that is the outer limit and adding a fourth drive yields very little if any performance.

daveseeley
02-16-2005, 06:51 PM
I'm at 2.5 gigs with 100% allocated.

That's what my tests were based on. the first was a save to the scratch disk of a file that intially openned from a different disk....then a resave to the scratch disk.


... SoftRAID offers settings optimized for specific applications or workflows, including Photoshop.

So I've heard...though I don't want to drop another $130 given my preliminary results.


Performance of Photoshop CS is actually about 7% faster than Photoshop 7

I will upgrade as soon as the next version comes out...which the grape vine says is iminent.


SATA does give good performance - and Raptor or DM10 seem about even in performance - though my hat would go to the 10K Raptor personally for what you do. I think the future for most users is 4-channel SATA-II controller and drives.
Might be that I should unload the whole raid and 4 drives and go with sata.


I wanted to clean it up and added some points.

daveseeley
02-16-2005, 07:01 PM
Thanks all for the feedback!


Hi Dave,

Could you test your 4 x x15.3/UL4S stripe with Quickbench 2.1 Extended 100MB test? We will have a better idea from yours results to compare with other 4-drive x15.3 stripes. This way we can tell if your 4 x stripe is up to snuff.

Also, if you test with PshopTest (http://www.macgurus.com/forums/showthread.php?t=19746), I can get a much better handle on this with suggestions for better results.

Several things I see right off the bat:
More RAM will help you.
Use Photoshop CS instead of 7- it's much faster.

Edit: OK, I just figured out yer runnin a UL4 single channel- not dual. You'll get much better performance with a UL4D. I think 3 x15s saturate a single channel; I believe that is the outer limit and adding a fourth drive yields very little if any performance.

More RAM and photoshop ugrade I will do...but it shouldn't change the conclusions of my current quest. Where can I get quickbench...and is it free or shareware?...or does it cost?
I'll do the pshop tests soon.

The ul4s with 3x cheetahs had very similar results in xBench to the 4x...so that verifies it...though it did do better in overall score...just not write speed.

No $500 dollar ul4d for me. I'm on a budget for this experiment. the answer may well end up being liquidation and a raptor or two....in which case I'd rather not throw away dollars.

TZ
02-17-2005, 08:35 AM
Question: Does the UL4S have the latest firmware, 3.20v2 driver, as well as 10.3.8?
latest driver by 10.3.7

Here is a real problem.

OS X doesn't update the drivers needed by 3rd party controllers. You HAVE to go to www.attotech.com and get the latest drivers (3.20 v2) and ATTO Config Utility 2.80 as well. And insure what firmware is installed (System Profiler will show as would ATTO Config. If you never updated the drivers, then you have 2.04 probably - look in your Extensions folder or System Profiler even. HUGE difference.


If it truly will do that. Xbench shows real world for as somewhat lower for the 36 gb cheetahs that I have.

I would think you would get better I/O from 18GB X15.3s and save on cost seeing you don't need 143GB. One platter with two heads is usually faster than 2 platter design.

I did some tests on my Atlas 15K drives. Xbench 4K doesn't tell you anything useful.

ATTO UL4S/D Software (http://attotech.vwh.net/software/app1.html)
Drivers (http://attotech.vwh.net/software/drivers.html)
ATTO Config 2.80 (http://attotech.vwh.net/register/getfile0.php?file=xepciconf280.tgz)
UL4 Firmware (http://attotech.vwh.net/register/getfile0.php?file=xul4_320v2.tgz)

Driver version 3.20 is a maintenance release of the ExpressPCI Ultra320 driver with advanced OS X Panther support. For peak performance, ATTO recommends using this driver with your ExpressPCI UL4D/UL4S host adapter when using OS X Panther. This release, when used with the latest ATTO Configuration Tool utility version, includes support for dynamic SCSI bus rescan, allowing users to add and remove devices from the bus without a system reboot. The driver release read me file contains instructions on installing and using this driver release. You should use the latest version of the ATTO Configuration Tool and UL4D/UL4S flash file to update the adapter prior to installing the driver update. Click here to link to the download page for these utilities and flash files.

Login and registration req'd for ATTO software.

Damien
02-17-2005, 10:31 AM
I have to go out for an hour or two but when I return I will run XBench over my 4 X-15 dual channel raid and see what I get

Damien
02-17-2005, 01:07 PM
Ok here is my info

I have a 4 drive, 2 channel, 18gig each X-15 RAID 0 using Apple Raid. It is on an ATTO UL4D with the 1.4.2f1 firmware and the 3.2.0 driver. (first results for the raid was using an older driver) I am NOT using Granite Cables.

Surprisingly the Raptor edged out the Single X15 18gig by just a hair



Raptor Boot drive 74gig

Results 148.75
Disk Test 148.75
Sequential 150.29
Uncached Write 190.05 79.22 MB/sec [4K blocks]
Uncached Write 175.10 71.70 MB/sec [256K blocks]
Uncached Read 154.61 24.48 MB/sec [4K blocks]
Uncached Read 109.00 44.04 MB/sec [256K blocks]
Random 147.24
Uncached Write 136.12 2.04 MB/sec [4K blocks]
Uncached Write 175.44 39.57 MB/sec [256K blocks]
Uncached Read 133.31 0.88 MB/sec [4K blocks]
Uncached Read 151.07 31.09 MB/sec [256K blocks]

Seagate Barracuda SATA 160gig

Results 106.67
Disk Test 106.67
Sequential 110.23
Uncached Write 130.67 54.47 MB/sec [4K blocks]
Uncached Write 122.07 49.99 MB/sec [256K blocks]
Uncached Read 80.89 12.80 MB/sec [4K blocks]
Uncached Read 123.77 50.01 MB/sec [256K blocks]
Random 103.34
Uncached Write 97.34 1.46 MB/sec [4K blocks]
Uncached Write 102.61 23.14 MB/sec [256K blocks]
Uncached Read 100.31 0.66 MB/sec [4K blocks]
Uncached Read 114.68 23.60 MB/sec [256K blocks]


Burly RAID <b>4x ST318452LW's using older driver, same firmware, lots of data on drive</b>

Results 263.71
Disk Test 263.71
Sequential 235.14
Uncached Write 510.74 212.90 MB/sec [4K blocks]
Uncached Write 445.77 182.54 MB/sec [256K blocks]
Uncached Read 92.66 14.67 MB/sec [4K blocks]
Uncached Read 495.57 200.23 MB/sec [256K blocks]
Random 300.18
Uncached Write 555.45 8.33 MB/sec [4K blocks]
Uncached Write 522.38 117.81 MB/sec [256K blocks]
Uncached Read 203.86 1.35 MB/sec [4K blocks]
Uncached Read 212.52 43.74 MB/sec [256K blocks]

Burly <b>4x ST318452LW's After all updates and a fresh format .. no data</b>

Results 267.14
Disk Test 267.14
Sequential 232.93
Uncached Write 452.93 188.80 MB/sec [4K blocks]
Uncached Write 472.48 193.48 MB/sec [256K blocks]
Uncached Read 92.85 14.70 MB/sec [4K blocks]
Uncached Read 481.20 194.42 MB/sec [256K blocks]
Random 313.14
Uncached Write 629.44 9.44 MB/sec [4K blocks]
Uncached Write 537.13 121.14 MB/sec [256K blocks]
Uncached Read 208.21 1.37 MB/sec [4K blocks]
Uncached Read 221.22 45.53 MB/sec [256K blocks]



Seagate ST318452LW <b>Single drive</b>

Results 148.42
Disk Test 148.42
Sequential 131.03
Uncached Write 166.83 69.54 MB/sec [4K blocks]
Uncached Write 130.89 53.60 MB/sec [256K blocks]
Uncached Read 101.37 16.05 MB/sec [4K blocks]
Uncached Read 142.26 57.48 MB/sec [256K blocks]
Random 171.13
Uncached Write 160.89 2.41 MB/sec [4K blocks]
Uncached Write 205.03 46.24 MB/sec [256K blocks]
Uncached Read 171.10 1.13 MB/sec [4K blocks]
Uncached Read 155.37 31.98 MB/sec [256K blocks]

System Info
Xbench Version 1.1.3
System Version 10.3.7 (7S215)
Physical RAM 2048 MB
Model PowerMac7,2
Processor PowerPC 970x2 @ 1.80 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1.80 GHz
Bus Frequency 900 MHz
Video Card GeForce 6800 Ultra

Boots
02-20-2005, 07:50 AM
Hi Damien,
Dave emailed me with some QB results. It looks like his 15.3k array is operating as fast as possible given that it is 4 drives on one channel of a UL4S- and not 2 or 3 drives per channel on a UL4D.

I believe his variable results and 'four drives no faster than one during-Photoshop-operations' situation is really a result of insufficient installed RAM; his OS is gettin starved...
But hopefully he can test with PshopTest and it will be easier to see exactly what is going on.

God- how I'd love to build a 6-drive Maxtor Atlas 15k II (98MB p/s each) array! :kickass: drool factor fer sure.

TZ
02-20-2005, 08:39 AM
Damien,

Two requests: could you try Speedtools extended? maybe you have? and add to RAID DB as well as post back here?

To make your Xbench more readable, cut out the redundant system info and put it in once, at the end, so it reads easier and the disk results are clearer, making the message shorter.

Someone with a G5 UL4D said that 3.20v2 make a "noticeable" difference in their results? would like to find and compare/include there in here just for base line for "optimal" performance.

Thanks!

Damien
02-20-2005, 07:35 PM
TZ, I don't have speedtools is that a free app and if so where can I get it?

The 3.2.0 driver seemed to slow my top end but increased the overall scores. I really haven't used my raid to be able to tell you the "feel" of it. It has been sitting idle for over a year due to lack of funds for the UL4D. Just did get it running again this month. :dance:

speters1
04-19-2005, 11:59 PM
I just did some tests with my 6 15k.3's and the UL4D. All of the drives have the latest firmware 007 and the card has the latest drivers and firmware. I am using Photoshop CS with 2.5g ram on a Dual 2GHZ G5. I have 85% allocated to Photoshop. My RAID is working fine. I used QB 2.1, extended test. The 100 mb read: 385mb/sec and write: 425mb/sec. The thing that I noticed is that, the time it takes to open and save a 750mb 16bit layered file, really does not change that much when I try different configurations of my RAID. With all six drives stripes accrossed two channels of the UL4D, it took 72 sec. to open. This was also the Scratch disk. When just opening up the same file from 1 15k.3, it took 73.5 sec. I made sure that everything was the same, and that I did this test right after I had just restarted. Has anyone else actually seen any real world increase in opens and saves, when comparing a RAID to a single drive?

Boots
04-20-2005, 02:18 AM
Hi Steve- welcome back! :)

Congratulations, you now have the fastest SCSI array probably anywhere!
You should submit those test numbers to the Drive & RAID Database. (http://www.macgurus.com/raid/raidmain.php). That's fantastic.

As to your comparison of opening a large Photoshop file-

I would suggest you are seriously RAM-deficient working files that large; adding 2-4 GBs more should help you quite a bit.
When CS is allocated more than 70% available RAM, she takes longer to open and longer to save files- yet filter processes tend to execute more quickly. With files that large, the Operating System wants a lot of on-board RAM for Open & Saves.
Installing more RAM will benefit the upcoming release of Photoshop CS2. It will be able to use 3GB RAM instead of 2GB, and it will be faster yet than CS, which saw a substantial performance increase over Photoshop 7.

Steve, what is the setup of your internal SATA drive(s)? That also has an impact on overall performance.

In answer to your last question, yes- I see consistently faster Open & Saves when using the Work Disk/Scratch Disk method with my tortoise-like 4 x x15.3 Cheetah striped RAID array.

If you would like to test with PshopTest (http://www.macgurus.com/forums/showthread.php?t=20218), I can make direct comparisons with you so we can scrutinize your acceleration strategy in more detail.

speters1
04-20-2005, 11:12 AM
Funny you mentioned the RAM. I just got 2 more gigs from Macguru's, so I will put that in, and report back. The test below is with the 2.5g ram.
I did download the Photoshop test. I set up the 6 drives into two partitions. The first was 68gb used for SCRATCH, then the second was used for my working files. I have my System software and all applications on the Raptor, which was installed inside the G5, no sata card. The test file was 512mb. Export clipboard on, History 60, history options checked: first snapshot & Non-Linear History, File handling: Always save, Icon, thumbnails. Adjusted Refresh Plugin: installed. Layers, channels, paths: small thumbnails. File Compatability: ask before saving layered tiff's, Enable Large Document format. Pixel Doubling: no. Async I/O: installed. Video Alpha: no. Image Cache: 8. Additional plug ins: to many to mention. Color Channels in color: no. Watermark plug-in: no. Available RAM: 85%. File Browser...: no. File Vault: off. File sharing: on. Energy Saver, Processor Setting: Highest. Hard disk sleep. off. Internet connection: none. Background apps: AppleSpell, test edit, mozilla-bin, MagicMenuHo, Activity Monitor, StuffitAVRDae, iCalAlarmSche, AppleModem, Finder, SystemUIServe, Dock, TableDriver, pbs, logingwindow,ATSServer, WindowServer.

A) Predraw Total: 35.8 Operation Total: 45.3
B) Save Total: 23.2
C) Open & Save total: 68.5
D) Test total= 113.8?
The one thing I am confused about is that in the scoring, line D, we are suppose to add the Operations total + Open & Save Total, but it seems like we are adding the Operations total twice, because the Open & Save Total already has the Operations total in it. Is this right?
Thanks,
Steve

TZ
04-20-2005, 11:35 AM
RAM prices are expected to decline further in the next two months - good news for everyone.

Steve,

Wow~ those are excellent numbers.

If it was possible, I'd want to create a new user account that loads only a minimum of processes, if possible. Or a lean system install.

Disabling 3rd party startupitems even. Some people want their "work system" to be free of any other processes, and testing would see if there is any impact.

Once you have your system optimized, then see what if any improvements CS2 and Tiger, plus the add'l RAM bring to the table and further reduce the time it takes to load, edit, save files.

I think someone said that 2.5GB RAM was the recommended configuration on G5, most people here would add "minimum" to that, and 3.5+ as optimum.

Boots
04-20-2005, 02:04 PM
Hi Steve,

Yes, the additional RAM will help for sure.

You will definitely have faster results when using the Work Disk/Scratch Disk method if you maintain the 6 x x15.3k stripe as one single volume- rather than two.

After installing the additional RAM, you will want to un-install the Adjusted Refresh Plugin- that is for single-processor machines with limited RAM.

Your additional plugins may or may not effect basic speed. The only way to find out of course, would be to test without them and with them. You can disable any one of them by re-naming the plugin- or the folder in which it resides- with a tilde in front of the name ("~").

File sharing- if active- is likely to slow your test results.

To get the PshopTest test score "Total", you add only the (A) "Operations Total" (45.3) and the (C) Open & Save Total.

The Save Total should be the sum of Test steps #4, #9, #11, and #13.

The Open & Save Total should be the sum of test step #1 and the Save Total (steps #4, #9, #11, and #13).

My guess is your step #1 was about 10 or 12 seconds? Whatever it was- add that to the Save Total (23.2) and that will the Open & Save Total. Then add that to your Operations Total....

.....78 ~ 80 seconds Test Total?

With your setup, you should see PshopTest Total numbers in the mid to low 60's.

I realize your preferences are probably set to your daily working conditions; however, you can try the "default" preferences also for the test and back off the RAM to 70%. Your score will be a bit faster with a "leaner" History settings....

But get that Adjusted Refresh plugin oughta there- that's holding you back with even the 2.5GB RAM. :(

TZ
04-20-2005, 02:18 PM
Boots,

That 2nd RAID volume is, I think, just "data storage" and isn't being used I assume, except to backup work as needed maybe. Not the safest, but would make sure maybe? that his scratch is using the fastest area of the drive, and with 6 drives, still have enough storage. (are those 36GB or 18GB Cheetah X15.3s?)

speters1
04-20-2005, 05:59 PM
In the Photoshop test, I had the first partition set up as scratch, then the second for my working files. The reason I partitioned the RAID was that I thought that it was best to use the fastest 1/3 of the drive for scratch? I am going to run some more test with the 4.5g ram, and get back. I did notice that now there is only like 1.6g available for photoshop now. I wonder if this will effect the performance? I was planning on waiting until I got CS2, before I installed this ram.

Boots
04-21-2005, 05:13 AM
If one is going to use the WD/SD method, the fastest performance is achieved by placing the (in this case the 512MB test) image file on the same volume as the "First" designated scratch disk.

To paraphrase TZ, a sensible and conservative approach to using the striped WD/SD is for a work session only- no long term storage of image files. A striped RAID of that many drives is just way too vulnerable to treat otherwise, and the method is an invitation to disaster unless the work short-term stored upon it is backed up frequently. I'm a real advocate of backing up every 15 minutes while using the WD/SD.

Steve, did you erase that first scratch volume, then drag a copy of the prepared 512MB test file on to it, then re-start the computer, then run the test?
That will give faster results. For the test, always start from a freshly re-started state; you'll get the maximum available RAM to Photoshop- which with your 2.5GB should be about 1880MB and with 4.5GB installed about 1800MB.

And what was the "File Open" time on that first test?

Do you have a second internal SATA drive installed and if so, what is it?

speters1
04-21-2005, 11:54 AM
With 4.5g installed I now only have 1619mb available, not 1800. I have only actually logged a couple of hours with this new ram installed, and using 100% allocation, but it appears that I get the beach ball more frequently now.
I have two internal drives. The drive that my system and all apps are on is a Western Digital Raptor. The second drive is what ever came inside my G5. I usually use this for backup.
When I ran the Photoshop test, I put the 512mg file on the second partition. I will try as you suggested an put it on the first partition. I am always sure to make sure that I do my test after a fresh restart.

TZ
04-21-2005, 12:12 PM
Even with 4GB of installed RAM, it is possible to create this condition "Starving" the Operating System, with a 512MB test file and a higher than "70%" memory setting. However, enough installed RAM can increase the overall speed of Photoshop operations at settings higher than "70%".

With larger amounts of installed RAM, settings between "70%" and "100%" can provide faster overall results, though slower File Save times may occur. Use PshopTest to find the optimal setting for your particular system.

http://www.macgurus.com/forums/showpost.php?p=83304&postcount=16

Boots
04-21-2005, 12:13 PM
I would say some running background process is using a large amount of RAM- if that is what you have available to Photoshop immediately after a restart; it really should be in the neighborhood of 1770~1800MB.

Steve, you might want to try testing at a 70% allocation first- then gradually work your way up to a higher RAM allocation. 100% is not always the fastest and can lead to the beachball effect in PS.
In fact, I found with 4GB RAM installed a 70% allocation was generally pretty optimal- especially for smaller image files.
Sometimes 100% is faster, but there is usually a tradeoff between faster filter executions and slower file saves.

speters1
04-21-2005, 03:02 PM
I thought that I read that this is normal. The more ram you put in these machines the less it has available for photoshop. This makes no sense to me, but I swear I read this somewhere. I did not change anything when I put in the ram. The only apps running are the same ones I listed before I put in the ram. I will try and lower the percentage.

TZ
04-22-2005, 06:10 AM
A number of people and man-hours went into testing it, and to have a way to test what does work, and how to get there. Print out and go through the PShop test, Guide, FAQ.

I recommend PShop Troubleshooting (http://www.gballard.net/psd/troubleshootpurgepsd.html) for 2nd opinion.

What is running is also anything that was installed. Try to create a test system that is lean and mean as it will run better.

Background apps:

AppleSpell,
test edit, mozilla-bin,
MagicMenuHo,
Activity Monitor,
StuffitAVRDae,
iCalAlarmSche,
AppleModem, Finder,
SystemUIServe,
Dock,
TableDriver,
pbs,
logingwindow,
ATSServer,
WindowServer.

Boots
04-22-2005, 07:12 AM
I thought that I read that this is normal. The more ram you put in these machines the less it has available for photoshop. This makes no sense to me, but I swear I read this somewhere. I did not change anything when I put in the ram. The only apps running are the same ones I listed before I put in the ram. I will try and lower the percentage.

Yes, it is normal- and you did read it somewhere. But it stays within a certain range:

With 2-2.5GB RAM installed:

Immediately after a restart- and without opening any other applications- then opening Photoshop, the "Available RAM: " will be roughly 1840-1880MB. The exact figure will depend on the quantity and type of any running background processes. If there are also applications which have been opened automatically by a script or added to System Preferences>Accounts>Startup Items, these will further decrease the actual figure of "Available RAM: "

With 3+GB RAM installed:

Immediately after a restart- and without opening any other applications- then opening Photoshop, the "Available RAM: " will be roughly 1700-1800MB. The exact figure will depend on the quantity and type of any running background processes. If there are also applications which have been opened automatically by a script or added to System Preferences>Accounts>Startup Items, these will further decrease the actual figure of "Available RAM: "

If you look at Activity Monitor (/Applications/Utilities/Activity Monitor), it will show you the memory being used by those running processes under the column "Real memory". That should help.


The reason I partitioned the RAID was that I thought that it was best to use the fastest 1/3 of the drive for scratch? I am going to run some more test with the 4.5g ram, and get back.

Yes, generally speaking. And if the volume is empty to begin with, you are using the first 1/3 of the volume across all 6 drives; the scratch file is not going to be spread out all over the width of the drives- it'll land on the outer parts first then work it's way inward as it gets larger from additions during a work session.

If the array is used only for the scratch disk or for the WDSD method- and not for storage ( :weeskull: ), "short-stroking" it by partitioning it doesn't really gain you anything. I tested this with my 4-drive array of 15.3k Cheetahs and the results were the same for both configurations. These SCSI drives have such terrifically consistent performance from the outside to the inside of the platters, but with ATA or SATA drives, it's a different story and "short-stroking" may indeed improve scratch disk performance.

Boots
04-22-2005, 08:19 AM
Looking at my system, and comparing with Steve's running background processes....

His background apps:

AppleSpell (this opens when you open TextEdit) uses about 1.5MB
test edit (I think you must mean TextEdit) uses about 6MB
mozilla-bin (Mozilla the browser correct?) uses about 32MB
MagicMenuHo
Activity Monitor uses about 10MB
StuffitAVRDae
iCalAlarmSche uses almost 7MB
AppleModem
Finder uses more memory the more Finder windows are open
SystemUIServe uses more memory the more application are open
Dock
TableDriver uses about 2.5MB
pbs
loginwindow
ATSServer
WindowServer uses more memory the more windows are open

So there is 60MB + god-knows-what for stuffit on steroids.

Steve, in other words when you did the first PshopTest test, it looks like you had a number of major applications running in the background....

speters1
04-22-2005, 10:34 PM
So I ran some more test. This time with the 4.5g installed. Everything else was the same . I did try different drive configs and ram allocation.
100% allocation, 6 drives, one partition, scratch and save to the same raid:
Test total= 69
85% (1689gb)allocation, 6 drives, one partition, scratch and save to the same raid:
Test total= 64.8
75% allocation, 6 drives, one partition, scratch and save to the same raid:
Test total= 65.1
85% allocation, 4 drives for scratch, 2 for saves:
Test total= 65.2

I realize I could get faster results with less history states, etc. I ran the test the way I need to have my computer for my working conditions. I am happy with the 4 drive, 2 drive scenerio. This is a bit safer, with out that big a loss in performance.

I did run the test again with my 750mb layered 16bit file. It was a lot faster with the new ram, but the difference between using the 6drive scratch/work and the 4 drive scratch, 2 work, was 202.2sec for the 6 drives and 201.0 for the 4/2 configuration. For this test I opened, flattened, increased size by 300%, convert to 8bit, etc, etc.
As I said, I am happy with my results and just wanted to share them with everyone.

Boots
04-23-2005, 04:10 AM
So I ran some more test. .....As I said, I am happy with my results and just wanted to share them with everyone.

That's great Steve- I'm glad to see your numbers in the "expected" range. Now we know your system is running more or less OK with no big obvious problems. :)

But for our benefit, could you post the test results details? If you could- for each test run, please list:

a) the "Available RAM" in actual Megabytes
b) the "Maximum Used by Photoshop" in actual Megabytes and the percentage
c) the "Operations Total"
d) the "Save Total"
e) the "Open & Save Total"

Also, could you verify or correct the previous post about your running background processes and apps? Was that Mozilla the browser and TextEdit running?

Thanks Steve, this would help us a lot. :)

TZ
04-23-2005, 08:12 AM
One interesting thing is there is increased seek latency as you increase the number of drives that have to be accessed in a RAID, which is what you might be seeing.

Taking these results, and building a table of other system configurations, might help the next person, and us as, to what works the best for someone and with a ballpark for how much to prioritize any upgrade.

speters1
04-23-2005, 05:18 PM
Background Apps:
AppleModem*
Activity Monitor
MagicMenuHo*
Finder
SystemUIServe*
Dock
pbs*
TableDriver
loginwindow
ATSServer*
WindowServer*
StuffitAVRDae
iCalAlarmSche
*= are these apps needed? I did not ad them.

Dual 2Ghz G5, 4.5gb ram, OS 10.3.8 , ATTO UL4D, Softraid 3.1.1, External RAID: 6 36gb Cheetah 15k.3's, OS and Applications on WD Raptor.

The test file was 512mb. Export clipboard on, History 60, history options checked: first snapshot & Non-Linear History, File handling: Always save, Icon, thumbnails. Adjusted Refresh Plugin: installed. Layers, channels, paths: small thumbnails. File Compatability: ask before saving layered tiff's, Enable Large Document format. Pixel Doubling: no. Async I/O: installed. Video Alpha: no. Image Cache: 8. Additional plug ins: to many to mention. Color Channels in color: no. Watermark plug-in: no.

Test 1: 100% allocation (1689gb ram), Scratch and Work on all 6 Cheetahs
Operation Total=42.3
Save Total=18
Open & Save Total= 26.7
Test Total=69

Test 2: 85% allocation (1436gb ram), Scratch and Work on all 6 Cheetahs
Operation Total=38.1
Save Total=17.9
Open & Save Total= 26.7
Test Total=64.8

Test 3: 75% allocation (1267gb ram), Scratch and Work on all 6 Cheetahs
Operation Total=38.7
Save Total=17.5
Open & Save Total= 26.4
Test Total=65.1

Test 4: 85% allocation (1436gb ram), Scratch on 4x Cheetahs, Work on 2xCheetahs
Operation Total=38.4
Save Total=18.1
Open & Save Total= 26.8
Test Total=65.2

You guys can delete the older threads if you want, since alot of this is repeated.

Boots
04-24-2005, 09:43 AM
Thanks very much Steve.

I notice your list of running background apps now does not include those items from your previous list.

So please correct me if I am wrong: with 4.5GB installed RAM- after you restart the computer, allowing about a minute wait after the desktop fully loads...
...then opening Photoshop CS....(and nothing else)
...then immediately checking Photoshop menu>Preferences>Memory & Image Cache,

You see:

"Available RAM: 1689MB"?

The other items- StuffitAVRDae, MagicMenuHo- are most likely components of Stuffit Deluxe and may be able to be disabled if you don't need them- but I doubt they use much RAM. iCalAlarm Scheduler can be disabled if you don't need it from iCal preferences.

But what is really interesting to me is that if this is true, then you are right- (and I'm wrong :o ) that above 4GB installed RAM, "Available RAM:" for Photoshop decreases even further, and perhaps doesn't level off in the way that I thought.

Could you also check the reading for "Available RAM" with the SCSI array turned off? (shut down computer, turn off power to Burly, restart computer. To reverse, shut down computer, power on Burly, wait 30 seconds, restart computer)

-------

Now on to your test results and a comparison with mine:

Your tests #2 and 3# are the fastest. Notice when the memory allocation is decreased, file open and files saves are faster but processor operations are slower. The reverse is true when memory allocation is increased. So there is a bit of tradeoff and logic might suggest-
If you save frequently, notch the memory allocation back a bit...
If you save only occaisionally, ratchet it up a bit...

My best score G5 Dual 2GHz, 4GB RAM, WDSD method on 4x15.3k SoftRAID stripe with Raptor startup, Available RAM: 1802MB, Memory allocation 85%, 1532MB:

Operation Total=39.7
Save Total=18.8
Open & Save Total= 26.8
Test Total=66.5

Your best score G5 Dual 2GHz, 4.5GB RAM, WDSD method on 6x15.3k SoftRAID stripe with Raptor startup, Available RAM: 1689MB (?), Memory allocation 85%, 1436MB:

Operation Total=38.1
Save Total=17.9
Open & Save Total= 26.7
Test Total=64.8

Steve, how are you formatting the SoftRAID volumes? I found the 128k stripe size was faster than the 64k stripe size. If so, you may be able to get a bit more out of the array that way.
I would also un-install the Adjusted Refresh plugin if you haven't already- I'm pretty sure that would increase your speed even more. :xphat:

speters1
04-24-2005, 09:21 PM
the 1689 was with only the apps I listed the last time running, and it was a fresh restart with nothing else open or running. I will turn off my RAID and see if that makes a difference.
I am also using the 128k in softraid, and I did take out the refresh plugin.
I left my history state at 60 and left all of my plugins, in, I have alot. I could probably get a little better time if I dropped down the History states, but I like being able to go back that far.

Boots
04-25-2005, 05:38 AM
the 1689 was with only the apps I listed the last time running, and it was a fresh restart with nothing else open or running. I will turn off my RAID and see if that makes a difference.
I am also using the 128k in softraid, and I did take out the refresh plugin.
I left my history state at 60 and left all of my plugins, in, I have alot. I could probably get a little better time if I dropped down the History states, but I like being able to go back that far.

Thanks Steve- that is indeed an eye-opener, the lower-than expected "Available RAM".
I think this could be a result of your many plugins and not the usual circumstance.....
I have not tested with more that 4GB installed RAM but I have seen other user reports indicating "Available RAM:" within the 1700-1800MB range on G5s with 4.5 and even 5GB RAM installed.

How many additional plugins do you have installed?

Do you use them all? If not, you can disable any one of them without un-installing by putting the tilde (~) character in front of the actual plugin or the folder containing the plugin. The plugins are usually located within one of the folders in /Applications/Photoshop CS/Plug-Ins.
You can verify which plugins are active by checking the list: Photoshop menu>About Plugin>.

Like you, I use a lot of History states and the same settings for History Options. Wonderful part of Photoshop for sure!
I once tested a 500 state History setting to see the effect on scratch disk size. I did a gazillion individual brushstrokes etc. It was a 32MB image file and by the time I had recorded 500 states, the scratch file had ballooned to 40 Gigabytes!
You can set History to 1000 actually. :eek:

speters1
04-25-2005, 07:12 PM
I have around 18 plugins installed. I will run some more test in about a week or two. I have several jobs that I need to finish up.
You guys can check out my work at www.stevepetersdigital.com

Boots
04-26-2005, 08:05 AM
Thanks Steve. I checked out your site. Nice work! :)