PDA

View Full Version : SCSI and GS performance



Simon Ravn
06-27-2002, 07:42 AM
As a little followup to my \"160 voices\" topic, I would like to find out if SCSI is a sure-fire solution - if SCSI drives with 5ms. seektimes will make GS do 160 unique voices without any problems. Anyone who tried this? I can see SCSI drives have become quite a bit cheaper than last time I took a look, so I might be interested in this solution. Also, are there any 10.000 RPM SCSI drives which are about as quiet as IDE Seagate Baracuda IV?

Mark_Knecht
06-27-2002, 07:50 AM
Simon,
I\'m sure you already know about this, but just in case you don\'t, this site is great for comparing drives:

http://www.storagereview.com/ (\"http://www.storagereview.com/\")

Take care,
Mark

eggman
06-28-2002, 12:05 PM
James,

There is a better way of doing so...

If you wish to use two IDE drives to double the performance, I would recommend putting them on two seperate IDE channels (a cheap Ultra IDE card would provide you these).
Then, create a spanned volume (Win2000 and XP only).

For example, if you have two 60gb IDE drives, you can create one volume (\"D:\" for example), which will hold both drives, and its perfomance will be twice better than a single drive (Windows will split the reading / writing equally between these two drives).

Cheap and friggin\' fast!

gabriels
06-28-2002, 01:58 PM
How would this idea of using 2 drives improve performance. What would determine which drive GS goes to for any one note. If a all samples for an instrument happen to be on one drive, wouldn\'t GS just go to that drive for all it\'s voices trying to play that instrument?
Gabriel

Mark_Knecht
06-28-2002, 04:06 PM
True - the dual drive idea only works if the required notes are actually split among the drives. However, this is managable, I think. I can put my piano on drive 1 and my strings on drive two, for instance, and since I know the note density of the two MIDI tracks, it will work.

But you\'re right, if I try to do 160 note on only the piano, or only the strings, then it fails.


Originally posted by gabriels:
How would this idea of using 2 drives improve performance. What would determine which drive GS goes to for any one note. If a all samples for an instrument happen to be on one drive, wouldn\'t GS just go to that drive for all it\'s voices trying to play that instrument?
Gabriel<font size=\"2\" face=\"Verdana, Arial\">

gabriels
06-28-2002, 06:06 PM
Well, this makes sense after all. I\'d like to hear the results if you test your theory. I might go the same route.

I wonder how long it will be until we can put enough memory in our computers so we can dispense with the process of streaming off the drives. I wouldn\'t think we have too long to wait, and hopefully Tascam will take advantage of that to create software that doesn\'t have some of the drive induced weaknesses.
Gabriel

Simon Ravn
06-28-2002, 06:26 PM
I already have my GIGS spread across two different drives on two different channels though.

JamesGrote
06-28-2002, 11:48 PM
Or would a better idea be to simply get two big IDE drives and split gig files between them. This would idealistically have each drive only handling half the polyphony, hopefully reducing the affects of seek time, AND still be much cheaper than one SCSI drive.

Just an idea...

eggman
06-29-2002, 06:30 AM
No no no, there is no need to manually split the files between drives, the feature I was talking about is called \"Striped Volume Set\", when two (or more) drives are formatted in Win2k/Xp to a Striped Volume Set, The OS will split the files equally between those drives.

What happens is, for example, if you copy Piano.GIG into drive D:, The OS starts writing in 64kb parts. First it writes 64kb to one drive, then 64kb to the other and so on, till the end of the file.

This way, when you read the sample from the hardrives, you get twice the speed since it has 2 ide channels to read from. (NEVER put them on the same channel, it will be a bottle-neck).

Simon Ravn
06-29-2002, 10:19 AM
Eggman, ok that\'s called RAID.. images/icons/smile.gif

gabriels
06-29-2002, 12:12 PM
What makes me nervous about raid or \"software raid\" in a live situation, is that you double your vulnerability to drive failure. A drive may only go down once in a blue moon, but now you have a chance of a drive going down in half a blue moon.

I have the same fears each time I increase the number of synths or effects devices i use, especially if they become an integral part of my sound.
Gabriel

eggman
06-29-2002, 02:35 PM
Well... if you are blessed with a high budget, you can always get a hardware RAID 5, this way you get fault tolerance, if one of your hard-drives fail, you can keep on playing and take your time to get it fixed.

eggman
06-29-2002, 11:00 PM
Well, Software RAID is not really RAID... images/icons/grin.gif

Roman Beilharz
07-01-2002, 02:55 AM
Simon,

I know the Barracuda ATA4 is a very high rated, very fast HD, but I\'ve read about some strange slow-down issues when e.g. used in a stripe-set. As far as I can remember, problems may occur, when two of them are getting used due to their caching-strategy. Just to make sure: did you test any other drive? Maxtor 5T030H3 or IBM\'s fast drives have performed well for me. Using Win2000 Pro I am able to continously stream 160 voices without any glitch having my samples balanced on the two Maxtor\'s.

All the best

Roman Beilharz

Simon Ravn
07-01-2002, 06:45 AM
Roman, I don\'t use a RAID so............ And yes I have IBM and just got a new WD too. I dont believe this is a transfer rate issue. The WD\'s are a tad faster than the Seagates, but I don\'t expect any miracles.

Simon Ravn
07-01-2002, 10:14 AM
Originally posted by Roman Beilharz:
Simon,
Using Win2000 Pro I am able to continously stream 160 voices without any glitch having my samples balanced on the two Maxtor\'s.

<font size=\"2\" face=\"Verdana, Arial\">I don\'t know how to interpret that statement. If you can stream 160 UNIQUE voices continously, then you have a winner system. But I never heard of anyone being able to do that. 160 voices means you use sustained samples on all channels and you use different notes (samples) all the time. If you can do that with 160 voices, then it\'s impressive. I can get 160 voices no problem with many of my performances, including shorter samples, non-chromatically sampled patches and sample repetition.