SCSI 15000 versus 2x10000 SATA on RAID 0?

[gazzer82]

Hello

I am looking at building a new Catalyst system with a Mac Pro and am trying to work out whether it is better to go for the traditional SCSI system we normally use or to take advantage of the SATAII interface in the mac pro and use two 10000rpm 150GB SATAII drives in a Raid0 configuration to get 300GB 20000rpm drive access for a considerably reduced cost.

Has anyone tried this/got any experience with it??

Cheers

Gareth

[Gian]

Hello,... We have some MacPros, and we use the inside interface from the Macs with 3 x 250 Gb SATA II RAID 0. We are pretty happy with them.
I do not know your experiences with the SCSI systems, but we here did not had very luckly, so I have no comparasion for that.
What I can say is we can achive to play with NO frame loss 30fps in our case, up to at least....

1 mov with 2400 x 800 PhotoJpeg 75%
or
1 mov with 1024 x 768 PhotoJpeg Uncompressed

2 mov with 1024 x 768 PhotoJpeg 75%

4 DV movs from the original Cat Library.

I am conservative in that, and of course depends a lot of the movie, and other factors. I do not know what you can achive with the SCSI, but for us here is doing the job, and is cheapper and easier to work than the SCSIs.

[gazzer82]

Cheers Gian thats really good to know.

Anyone else here driven a system like that any harder? Be interested to know how far it could be pushed before it started dropping frames?

Cheers

Gareth

[tharding]

It is all about access time not spindle speed and besides 2 x 10000 rpm does not equal 20000 rpm total. Scsi still has fast seek times which is why playback can be better. The disk is able to randomly access the data faster. Remember that unlike a non linear editing system the Mac has no way of knowing what you want it to do next with Catalyst. So it is really depends on how fast the system can get to any piece of data with no notice.

A RAIDs real advantage is its ability to capture or play out a wide bandwidth of data in one hit. You need the balance of both with Catalyst or any media server. A single high speed SCSI drive tends to have that balance.

If you look back through the forum you will see a lot of posts about this including some speed tests that Richard did some time ago.

Personally I use an Xserve RAID which uses some really heavy handed caching in its onboard RAM to give really consistent playback of many layers. Prior to that I used Raptors set up in 4 drive raids.

Obviously it is all about what you can afford and what level of performance you need.

Have a bit of a read on the Barefeats.com and Macgurus.com websites where they explain both technologies fairly well.

Cheers

Toby

[samsc]

Cheers Gian thats really good to know.

Anyone else here driven a system like that any harder? Be interested to know how far it could be pushed before it started dropping frames?

Cheers

Gareth

sata drives tend to work horribly when you need lots of layers- doesnt matter what raid you use. raids dont help.

a single scsi 15000rpm will work so much better for most normal uses.

you will be dissapointed with how your 'raid' works.

[samsc]

Hello,... We have some MacPros, and we use the inside interface from the Macs with 3 x 250 Gb SATA II RAID 0. We are pretty happy with them.
I do not know your experiences with the SCSI systems, but we here did not had very luckly, so I have no comparasion for that.
What I can say is we can achive to play with NO frame loss 30fps in our case, up to at least....

1 mov with 2400 x 800 PhotoJpeg 75%
or
1 mov with 1024 x 768 PhotoJpeg Uncompressed

2 mov with 1024 x 768 PhotoJpeg 75%

4 DV movs from the original Cat Library.

I am conservative in that, and of course depends a lot of the movie, and other factors. I do not know what you can achive with the SCSI, but for us here is doing the job, and is cheapper and easier to work than the SCSIs.


ah but you see gian - you can only play back 4 dv movies from your raid- that is pretty typical.
and not very good. you would do a lot better with a scsi drive-

from a scsi drive you can do 6- 8 layers. because the access time is much faster- for most applications this is what people need.

also dont use 75% photojpeg you need to use 50-60% - believe me - the performance of photojpeg above 70% drops by 30% per layer - for no increase in anything. dont do it.
keep it below 75% at all times.


richard

[samsc]

Hello,... We have some MacPros, and we use the inside interface from the Macs with 3 x 250 Gb SATA II RAID 0. We are pretty happy with them.

1 mov with 2400 x 800 PhotoJpeg 75%
or
1 mov with 1024 x 768 PhotoJpeg Uncompressed

2 mov with 1024 x 768 PhotoJpeg 75%

4 DV movs from the original Cat Library.


gian - please can you do a little test-

remove raid - just work with 1 drive - and give us the result of your movie tests

[gazzer82]

Hello

Just thought i would fill you in on my findings.

I have tried a setup using a Mac Pro, Dual Quad 3.0Ghz, 4GB RAM, with an ATI Radeon X1900XT, outputting the video at 1024x768 on DVI to a second display.

I have then setup two video drives,

1 - 150GB 15000rpm SCSI Drive
2 - 2x150GB 10000rpm SATA Drive, In Raid 1 (Mirrored)

I am using the latest release of Catalyst, and all the current updates from Apple.

All test conducted with DV footage, motion JPEG to follow :)

Results:
SCSI
DV at 720x576 - 5 Layers till frames dropped
DV at 1024x768 - 4 Layers till frames dropped

SATA
DV at 720x576 - 5 Layers till frames dropped
DV at 1024x768 - 4 Layers till frames dropped

This is with no processing and playing full screen, not that it matters as teh CPU never went above 20%.

Cheers

Gareth

[samsc]

All test conducted with DV footage, motion JPEG to follow :)

Results:
SCSI
DV at 720x576 - 5 Layers till frames dropped
DV at 1024x768 - 4 Layers till frames dropped




gareth-

dv doesnt work at 1024x768 - you can create the movies but it doesnt work- all quicktime does is resize a 720x576 movie. DV is both fixed size and fixed bit rate.

---

did you install the latest version of the atto scsi drivers - the ones on any cd didnt work last time i tried?

http://www.attotech.com/software/drivers.html

its shown as version 4.1 on the UL5D driver page.
you will need to login and create an account - if you havent.

if you didnt you need to as the default driver has very poor performance.

[gazzer82]

Hi Richard

Yes i do have the latest drivers from atto installed, and just to add some more detail, the SCSi drive is a Fujitsu MAX3073NP, and the SATA drives are the Western Digital Raptors WD1500ADFD.

I can't believe i never new that about DV, doh, i am just about to encode the same video footage as Motion JPEG at 60% and see how they compare.

Cheers

Gareth

[samsc]

Hi Richard

Yes i do have the latest drivers from atto installed, and just to add some more detail, the SCSi drive is a Fujitsu MAX3073NP, and the SATA drives are the Western Digital Raptors WD1500ADFD.

I can't believe i never new that about DV, doh, i am just about to encode the same video footage as Motion JPEG at 60% and see how they compare.

Cheers

Gareth

there are many different dv settings too.

you need to check which dv codec you used.

you need to email me your pdf profile so i can check your files.

[gazzer82]

Hmm,

I have just run a test with motion jpeg and the results are very odd, here are the complete results.

SATA
Motion JPEG 60% - 1280x720 = 2 Layers before frames dropped
Motion JPEG 60% - 720x576 = 4 Layers before frames dropped
DV 100% - 720x576 = 8 Layers

SCSI
Motion JPEG 60% - 1280x720 = 1 Layers before frames dropped
Motion JPEG 60% - 720x576 = 2 Layers before frames dropped
DV 100% - 720x576 = 7 Layers

Very very odd.

PDF profile on it's way asap.

Gareth

[Gian]

Sorry did not get back here for long time,... I am touring at the moment, so not so much time. I did tryied some of our files on a single diskd before.

Richard I will make some better tests soon I get back,.. but from what I recall, without the RAID set up, I can achive on a single SATA drive, sometimes 3, sometimes 2 original DV files from the library. Depends a lot on how much information you have on the movie.

Regarding big files (like 2400 x 600) or HD 1920 x 1080, with compressions in beetween 60 and 80% ( I got upper that 70% gets bad, but some movies, speccially with black background and white lettering, gets some strange white points arround the letters.) I never got those movies playing at 30 fps.
They always still arround 27, 28, sometimes less. The RAID gets them to play at 30 fps.

Funny, I had a movie 2400 x 600 ( black and white on the begining, and then gets some colors), on the BW he got the 30 fps on a single SATA drive, when reach the colored part, drops to 24, 25 fps. Very funny....

I will make some more tests next week soon I get back. Cheers to all....

[OlliR]

After having a talk with Oli Metcalfe this summer when he was with Muse (and his Catalysts) at Rock am Ring I tried the 3 x 500G SATA Raid Level 0 (Striped).
Works fine for me. Access time seems to be just a little slower than the 15K SCSI times.

I ran 2 full HD 1920 x 1080 Photo-JPEG 75% movies at 30 fps.

Starting another HD playback led to dropping frames on two layers.

But I could run another 720x576 Photo-JPEG 75% movie at 30 fps.

I will do some more testing with SD Movies on more layers.

System setup: Mac Pro 8xCore, 4 GB, 1x500GB System, 3x500GB Content

[samsc]

this all depends on what you want to do-

----

i test hundreds of codecs- i have very clear graphs- showing performance on many different systems.
im working on presenting them a little clearer- in talking to people the detail was too technical.

[samsc]

this all depends on what you want to do-

----

i test hundreds of codecs- i have very clear graphs- showing performance on many different systems.
im working on presenting them a little clearer- in talking to people the detail was too technical.

[samsc]

Hmm,
Very very odd.

PDF profile on it's way asap.

Gareth

yes your results are odd. and not what you got the day before.

do you have a terminator on your scsi disc?

[samsc]

i test all sorts of systems-
i have just spent several weeks looking at photojpeg performance-

the graphs here are using an esperance 2GB RAM disk- which give me the absolute maximum possible performance- because 'disc' speed is almost infinite.
you cant get any faster than this - right now - whatever RAID you use.

these tests are for 4 layers - on 8core intel and quad g5.
and for 2 layers - with 2 layers its possible to do almost anything-

this is a graph of 84 photojpeg files - in 5% steps from 35% to 100% - and from size 640x480 to 1920x1080.

these graphs show the raw cpu time necessary to do these things.

the axis are codec and time in milliseconds.

the clumping of points around 40ms is when testing has broken down and tests are not valid.

intel 3.0ghz is faster than quad g5 at most data points - but not all.

[gazzer82]

Cheers for sharing that, really interesting to see the performance hit going from 95% to 100%.

In answer to your question, yes the drive is terminated. My results may have been slightly different as i re-ran things to double check the results and it seems to depend a small amount on the exact clip selelcted as to how may layers will play back.

Cheers

Gareth

[samsc]

Cheers for sharing that, really interesting to see the performance hit going from 95% to 100%.
Gareth

100% is uncompressed.

there is also a hit from 70% to 75%

---

inconcistent results means that something is wrong with your setup.

you have inconisistent monitor refresh rates too.

neither of your results are trustworthy - yet.

your scsi value is too low.
---

richard