Posted 01/23/04 by Rob-ART Morgan, chief mad scientist
HYPOTHESIS: If I use a RAID 0 volume for the SCRATCH disk, I should experience significant speed gain over a single drive scratch when editing large documents.
TEST SCENARIO: I'm going to force Photoshop CS to use the designated scratch disk by only assigning a memory allocation of 160MB. Then I will perform standard actions to see how a 4 drive, 4 channel SATA RAID 0 volume (200+MB/s) compares to a single SATA drive (55MB/s) when each is designated primary scratch.
In addition, the test file (300MB) will be OPENed and SAVEd to the scratch volume to further test the gains of RAID vs single volume.
1. Though the RAID 0 scratch volume did OPEN and SAVE as much as 200% faster than a single scratch volume, the typical PHOTOSHOP CS ACTIONS only gained 59% in speed.
So you have to ask yourself, does the 59% gain justify the cost of the extra drives, cables, enclosures, and controller?
2. As you can see by the WHITE bars, when we increased the memory allocation from 160MB to 1800MB, the scratch volume wasn't touched. There's no substitute for lotza memory.
However, if you edit files bigger than 300MB, expect to get scratch drive hits during an editing session. Why? Because Photoshop won't let you allocate more than 2GB to itself. And you need an allocation equal to at least 6 times the size of your document to avoid major scratch hits.
3. When we duplicated these tests using Photoshop 7.01, we found it was as much as 59% more efficient in the use of the scratch disk.
However, when we increased the memory allocation to 1800MB, Photoshop CS edged out version 7.01 by 5% when performing the four actions. (OPEN and SAVE speeds were virtually identical for both.)
We plan to repeat the test with our new Ultra320 SCSI RAID set to see if it makes a better scratch volume. If we get our hands on some RAPTOR SATA drives, we'll try them as well.
Has Bare Feats helped you? How about helping Bare Feats?
RETURN to Main BARE FEATS Index Page.