External H.D. For Scratch Disk?

Bob Williams wrote in news:fzg7f.90871$lq6.57620
@fed1read01:

PS likes to have its Scratch Disk on a separate physical drive.
1. Can one use a USB 2 External H.D. for this purpose?
Or must it be an Internal EIDE Drive?
Is USB 2 fast enough to make a separate scratch drive practical.
2. If the main drive is a SERIAL ATA (which handles data differently
than EIDE), is it OK to just put the Scratch Disk on a separate partition of the main drive?
Bob Williams

usb 2 external is still to slow

puting it on a seporate partition has little or no advantage

—
Art, like morality, consists in drawing the line somewhere.

Bob Williams writes …

PS likes to have its Scratch Disk on a separate physical drive.
1. Can one use a USB 2 External H.D. for this purpose?
Or must it be an Internal EIDE Drive?

You can use an external USB or Firewire (1394) disk for scratch.

Is USB 2 fast enough to make a separate scratch drive practical.

No. I tested this once with a longish action on a large file … with scratch on a separate internal HD time was 321 sec, with scratch on the C drive 411 sec (28% slower), with scratch on a 1394 external HD 522 sec (63% slower), with scratch on a USB 2 drive it took 745 sec (132% slower).

Bill

In article <lKF7f.96531$ says…

Bill Hilton wrote:

Bob Williams writes …

PS likes to have its Scratch Disk on a separate physical drive.
1. Can one use a USB 2 External H.D. for this purpose?
Or must it be an Internal EIDE Drive?

You can use an external USB or Firewire (1394) disk for scratch.

Is USB 2 fast enough to make a separate scratch drive practical.

No. I tested this once with a longish action on a large file … with scratch on a separate internal HD time was 321 sec, with scratch on the C drive 411 sec (28% slower), with scratch on a 1394 external HD 522 sec (63% slower), with scratch on a USB 2 drive it took 745 sec (132% slower).

Bill

Thanks Bill and "Kingdom"
It was very interesting to see quantitative results for using various drives for Scratch space.
I was impressed to see that using the C drive for scratch was as good as it was. Adobe makes such a big point of using a separate physical drive from C, that I thought the difference in performance would be at least 100% or more. With huge hard drives so inexpensive nowadays, it seems hardly worth the hassle of installing a separate drive for scratch. Did you set up a separate partition on your main drive or just use the unpartitioned C drive? Did you give PS any special instruction as to where your scratch disk was or did you just let Windows allocate space as needed?
Bob Williams

An aside re: partitions. In the "old days," read pre-CS, PS could only handle up to ~ 4GB per Scratch Disk. Partitioning allowed up to ~ 16GB in 4GB parcels. With CS – CS2, you can still allocate 4x Scratch Disks, however, the size usable is now approximately infinity (oh, there is a physical limit, but I doubt that any of us will ever have it available).

The only other reason for partitioning would be to simply have the ability to "clean up" should there be any TMP files ever left lying around. As I find PS seldom crashes (holds breath, as the PS-gods snarl and hurl lightening bolts at his computer), the liklihood of needing to do this is greatly diminished.

Now, do not misunderstand Bill’s data. His tests were for "external Firewire/ USB" HDDs for Scratch. The best situation would be to have one physical HDD for OS/programs, and another/others PHYSICAL "internal" (directly attached to SATA/SCSI/etc.) HDDs for Scratch Disk(s).

Also, thanks Bill for that data,
Hunt

In article wrote:

In article <lKF7f.96531$ says…

Bill Hilton wrote:

Bob Williams writes …

PS likes to have its Scratch Disk on a separate physical drive.
1. Can one use a USB 2 External H.D. for this purpose?
Or must it be an Internal EIDE Drive?

You can use an external USB or Firewire (1394) disk for scratch.

Is USB 2 fast enough to make a separate scratch drive practical.

No. I tested this once with a longish action on a large file … with scratch on a separate internal HD time was 321 sec, with scratch on the C drive 411 sec (28% slower), with scratch on a 1394 external HD 522 sec (63% slower), with scratch on a USB 2 drive it took 745 sec (132% slower).

Bill

Thanks Bill and "Kingdom"
It was very interesting to see quantitative results for using various drives for Scratch space.
I was impressed to see that using the C drive for scratch was as good as it was. Adobe makes such a big point of using a separate physical drive from C, that I thought the difference in performance would be at least 100% or more. With huge hard drives so inexpensive nowadays, it seems hardly worth the hassle of installing a separate drive for scratch. Did you set up a separate partition on your main drive or just use the unpartitioned C drive? Did you give PS any special instruction as to where your scratch disk was or did you just let Windows allocate space as needed?
Bob Williams

An aside re: partitions. In the "old days," read pre-CS, PS could only handle up to ~ 4GB per Scratch Disk. Partitioning allowed up to ~ 16GB in 4GB parcels. With CS – CS2, you can still allocate 4x Scratch Disks, however, the size usable is now approximately infinity (oh, there is a physical limit, but I doubt that any of us will ever have it available).

The only other reason for partitioning would be to simply have the ability to "clean up" should there be any TMP files ever left lying around. As I find PS seldom crashes (holds breath, as the PS-gods snarl and hurl lightening bolts at his computer), the liklihood of needing to do this is greatly diminished.
Now, do not misunderstand Bill’s data. His tests were for "external Firewire/ USB" HDDs for Scratch. The best situation would be to have one physical HDD for OS/programs, and another/others PHYSICAL "internal" (directly attached to SATA/SCSI/etc.) HDDs for Scratch Disk(s).

Also, thanks Bill for that data,
Hunt

IME USB2 uses lots more CPU cycles than FW. This makes USB fine for CD burning, backup, etc, but when you are using CS with a swap didk your goal is to overlap computing with IO and you want the IO to use as few CPU cycles as possible. If you’ve got a hot sh*t dual CPU system you might not notice.

At PhotoExpo I asked one of the guys demonstrating Aperature on a G4 laptop about using a FW device as a TMP disk to get more parallel IO.

He said that it would be a net win but warned that no-name FW backpacks vary wildly in the amount of CPU cycles they need based on which controller chip is used. He obviously said that an Apple FW box would be wonderful.

To change the topic, one reason to dedicate a small partition to something like PS work disk is to minimize the fragmentation a bit.

—
a d y k e s @ p a n i x . c o m

Don’t blame me. I voted for Gore.

Bill Hilton wrote:

Bob Williams writes …

PS likes to have its Scratch Disk on a separate physical drive.
1. Can one use a USB 2 External H.D. for this purpose?
Or must it be an Internal EIDE Drive?

You can use an external USB or Firewire (1394) disk for scratch.

Is USB 2 fast enough to make a separate scratch drive practical.

No. I tested this once with a longish action on a large file … with scratch on a separate internal HD time was 321 sec, with scratch on the C drive 411 sec (28% slower), with scratch on a 1394 external HD 522 sec (63% slower), with scratch on a USB 2 drive it took 745 sec (132% slower).

Bill

Thanks Bill and "Kingdom"
It was very interesting to see quantitative results for using various drives for Scratch space.
I was impressed to see that using the C drive for scratch was as good as it was. Adobe makes such a big point of using a separate physical drive from C, that I thought the difference in performance would be at least 100% or more. With huge hard drives so inexpensive nowadays, it seems hardly worth the hassle of installing a separate drive for scratch. Did you set up a separate partition on your main drive or just use the unpartitioned C drive? Did you give PS any special instruction as to where your scratch disk was or did you just let Windows allocate space as needed?
Bob Williams

Hunt wrote:

In article <lKF7f.96531$ says…

Bill Hilton wrote:

Bob Williams writes …

PS likes to have its Scratch Disk on a separate physical drive.
1. Can one use a USB 2 External H.D. for this purpose?
Or must it be an Internal EIDE Drive?

You can use an external USB or Firewire (1394) disk for scratch.

Is USB 2 fast enough to make a separate scratch drive practical.

No. I tested this once with a longish action on a large file … with scratch on a separate internal HD time was 321 sec, with scratch on the C drive 411 sec (28% slower), with scratch on a 1394 external HD 522 sec (63% slower), with scratch on a USB 2 drive it took 745 sec (132% slower).

Bill

Thanks Bill and "Kingdom"
It was very interesting to see quantitative results for using various drives for Scratch space.
I was impressed to see that using the C drive for scratch was as good as it was. Adobe makes such a big point of using a separate physical drive from C, that I thought the difference in performance would be at least 100% or more. With huge hard drives so inexpensive nowadays, it seems hardly worth the hassle of installing a separate drive for scratch. Did you set up a separate partition on your main drive or just use the unpartitioned C drive? Did you give PS any special instruction as to where your scratch disk was or did you just let Windows allocate space as needed?
Bob Williams

An aside re: partitions. In the "old days," read pre-CS, PS could only handle up to ~ 4GB per Scratch Disk. Partitioning allowed up to ~ 16GB in 4GB parcels. With CS – CS2, you can still allocate 4x Scratch Disks, however, the size usable is now approximately infinity (oh, there is a physical limit, but I doubt that any of us will ever have it available).

The only other reason for partitioning would be to simply have the ability to "clean up" should there be any TMP files ever left lying around. As I find PS seldom crashes (holds breath, as the PS-gods snarl and hurl lightening bolts at his computer), the liklihood of needing to do this is greatly diminished.
Now, do not misunderstand Bill’s data. His tests were for "external Firewire/ USB" HDDs for Scratch. The best situation would be to have one physical HDD for OS/programs, and another/others PHYSICAL "internal" (directly attached to SATA/SCSI/etc.) HDDs for Scratch Disk(s).

Also, thanks Bill for that data,
Hunt

Yes! I understood that Bill’s experiment involved using "External USB and Firewire drives" but in one experiment he just used his plain old C drive for scratch space and it was pretty fast.
In fact, his "longish action on a large file" was only 28% slower than when he used a separate physical drive. That would hardly be noticeable in "ordinary" operations.
Bob Williams

"Al Dykes" wrote

[…] when you are using CS with a swap didk your
goal is to overlap computing with IO and you want the IO to use as few CPU cycles as possible.

You want overlap seeks on the hard drives. That happens best with drives on the bus.