My Mac Pro boot OS is on a 150 Gb striped raid made from outer partitions on two 1Tb drives in bays 1 and 2. There is 95 Gb free on the boot. 8 Gb RAM.
My normal scratch is on a dedicated 150 Gb striped raid made from the outer partitions of the 1Tb drives in bays 3 and 4.
I ran the Retouche artists Photoshop speed test with the scratch on the dedicated separate scratch, and on the boot volume.
The results were:
Average time of several runs with dedicated scratch was 45.5 seconds. Average time with scratch on boot was 43.9 seconds.
Since I was expecting the dedicated scratch to be faster I was a bit surprised so I repeated the exercise on my MacBook Pro (1.83 MHz, 2 Gb RAM). Normal scratch is the boot volume which a 5400 rpm 500 Gb Samsung with 150 Gb free, no partitions. For this exercise, I connected an eSATA via an express card to provide a dedicated scratch alternative.
Average time with dedicated separate scratch was 152 seconds. Average time with scratch on boot was also 152 seconds.
All Retouche Tests were done with 40 history states and 4 cache levels, which results in about 7Gb of scratch being used. On both machines Quickbench shows the scratch as just a few percent faster than the boot.
I repeated the Mac Pro tests with the test file located on different drives, including the boot and the scratch, but there were no significant differences.
What has happened to the standard advice about dedicated scratch for Photoshop?
Any thoughts ? (other than that I have too much time on my hands!)
Important to note the buffer on those drives are the all 32MB or are some 16MB.
A drive with a 32 MB buffer is going to record data faster.
However if you are on a MacPro (Intel) which it sounds like you are, I can confirm that using your start up disk as opposed to a dedicated separate scratch will not be of any speed advantage with photoshop.
At least it does not seem that way from my own test.
I also found partitioning the drive does not seem to be necessary on the intel box?
I have a test that is fairly consistent regardless as long as you have sufficient RAM 8 GB or more a Raid O scratch and an the same amount of memory allowed.
I still find with CS 4 that using bigger tiles is helpful as wel as the Forced VM Buffer plug in.
They still seem to speed things up a bit.
My test work on my dual xeon core duo that way in 16-18 seconds ona 8 core MacPro with 2GB of RAM and with out the Raid 0 and using the startup as the scratch with no Raid configured and without the plug ins it takes about 3 minutes.
The Ram and the raid are the important things the other two help.
Try running the test while you’ve got other software running in the background, doing other things. That’s where the dedicated scratch drive is supposed to help you, because it lets other apps do their thing on the main drive, and the OS to do it’s own swapping or whatever it has to do in the course of it’s daily work, while Photoshop accesses its own drive for scratch work.
That may or may not be true but I am not certain I want to try that kind of thing.
However I have run PPRo and After Effects together with Photoshop and sometimes Illustrator and have not noticed a difference except for the fact that the Raid 0 speed things up a bit but not a great deal.
I am not so sure it is that important as it has been in the past.
What would make a big difference is if you used and external RAID 0 running off of a supported card. Since those can transfer files at a much higher rate then can the internal RAID 0.
But 64 bit might also make it less of a problem as well.
Wade – if you hit the scratch disk a lot, then 1) Get a dedicated scratch disk 2) configure a set of dedicated scratch disks as RAID0 with a hardware controller (not software)
64 bit helps, but even then you can still hit the scratch disk.
From the above and other digging around, I think the main reason my tests don’t show a benefit for a separate scratch is that I wasn’t doing any other activity during the tests. So that the boot drive was able to be a effective scratch without trying to do other things ….. I think no boot volume I/O contention is the buzzword here?
The results still surprised me. I don’t use Photoshop very often on my MBP but I think I can conclude that I won’t benefit from a dedicated scratch for it provided I don’t multitask. This obviously wouldn’t be true if the dedicated scratch was a multi-set of striped raid, but this isn’t very practical on the road.
Chris i do use a internal RAID 0 for my scratch I mostly and my MacPro works pretty well and I have no problems with Photoshop and its responsiveness and do not experience crashes.
I work with Photoshop everyday that I am not shooting. And often work with other programs at the same time.
But and if I had such a volume or decided to get really heavy into video work not just for web purposes then I would configure an external RAID 0 with sufficient space.
And if I get more into the 3D features should they develop in the right direction then that would also necessitate such a configuration. but this is not the time.
And things work very well for me. I will wait to see what Apple does.
But your advice is appreciated and I will act on it when the planets are properly aligned.
Depending on your RAM, opening a bunch of files, using a gazillion fonts on them, certain filters? who knows what else? Time Machine? Spotlight? Cron scripts running in the background at scheduled times?
Depending on your RAM, opening a bunch of files, using a gazillion fonts on them, certain filters? who knows what else? Time Machine? Spotlight? Cron scripts running in the background at scheduled times?
Mike – on our tests, we see a performance increase using a dedicated scratch disk even without multitasking other apps. But certainly, the more you have going on (iTunes, web browsing, etc.) in the background – the more you need a dedicated disk because the system disk is going to be busy.
Thanks. It seems that In a no multitasking situation the boot drive is a viable scratch (at least for a while), so whether a separate scratch drive benefits or not would depend on how fast that separate drive was compared to the boot drive. In my tests the dedicated scratch I tried was about the same as the boot drive (Quickbench) and I got the same Photoshop speed. Perhaps in your tests you were using a faster dedicated scratch?
No, we used the same drive, same connectors on the motherboard. But the OS drive was busy at the same time as the scratch drive, due to OS paging and other background stuff.
Mike you do understand why the second drive is for scratch?
the OS must write its own scratch files (they are called SWAP).
since a hard drive only has one set of read write heads, Photoshop and the OS are both trying to use that one set of read write heads at the same time. This is what slows things down.
Mike you do understand why the second drive is for scratch?
the OS must write its own scratch files (they are called SWAP).
since a hard drive only has one set of read write heads, Photoshop and the OS are both trying to use that one set of read write heads at the same time. This is what slows things down.
Oh yes, I understand that very well! The whole point of my initial post is that in the carefully done and repeated tests I did, Photoshop was NOT slowed down by having scratch on the boot volume. I do accept the points made that this wouldn’t always be the case when multitasking and background activities come into play. I was surprised by the results which fly in the face of the established wisdom.
