sandman5000
Max Output Level: -73 dBFS
- Total Posts : 882
- Joined: 2005/05/26 02:05:56
- Location: USA
- Status: offline
Will RAID load samples, bounce or freeze faster?
If so, by a lot (say twice as fast)? Other than wasting storage space, any negatives?
thanks!
|
xackley
Max Output Level: -45.5 dBFS
- Total Posts : 2973
- Joined: 2004/01/30 09:39:49
- Location: USA
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/22 16:45:43
(permalink)
RAID 1 mirroring
on my IDE raid controller I have each mirrored drive on a seperate channel, which double the Read. It alternates between the to physical devices and delivers as much as each channel can handle.
With the 2 drives on the same channel, the Read is the same as nonRAID1, because the channel is the bottleneck.
I have not investigate SATA RAID 1 yet, as I keep putting off getting a new computer for my DAW. But the double Read speed information was included in my current cards documentation, and I tested moving the drives to different channels.
Also Windows sees the drives as one unit when using a hardware controller, there is no overhead. Configure the RAID once and forget it.
|
Sorceress Sarah
Max Output Level: -89 dBFS
- Total Posts : 81
- Joined: 2003/11/15 09:44:17
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/23 02:15:23
(permalink)
ORIGINAL: sandman5000
If so, by a lot (say twice as fast)? Other than wasting storage space, any negatives?
thanks!
The answer to your question is maybe. Mirroring does not gain any speed. A two disk stripe set will come very close to doubling throughput, but only when you stripe the disks across two channels of the disk controller. Since most of the "on-board SATA raid controllers are the Promise zero-channel, the benefits of striping are significantly reduced. Mirror for safety, stripe for speed, and don't even bother with RAID5.
|
sandman5000
Max Output Level: -73 dBFS
- Total Posts : 882
- Joined: 2005/05/26 02:05:56
- Location: USA
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/23 09:19:25
(permalink)
Thanks for the answers xackley and Sorceress Sarah.
I've been reading up on RAID and RAID 0 is the one I'm considering. I don't need to mirror because I back up all the time. I just wan't my samples to load quicker (the Reason DK 2 in particular) and also bounce to audio faster. Freeze is pretty fast already, but it could always be faster.
I didn't exactly understand the on board controller part. This is for a system I'm going to build when quad cores come out (in a month). So for maximun speed, don't use the MOBO SATA controller? Would a add on card be better (on the PCI bus?)? Would that mess with my PCI audio card (hog the bandwith?)?
|
mlockett
Max Output Level: -54.5 dBFS
- Total Posts : 2099
- Joined: 2003/11/07 17:26:14
- Location: Colorado Springs, CO
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/23 11:27:06
(permalink)
As far as if it will load samples, bounce or freeze faster, it depends if that is the bottleneck on your system. I would be surprised if freezing was more bound to cpu/ram than disc writes (dependant on how complex the track is that's being rendered).
|
sandman5000
Max Output Level: -73 dBFS
- Total Posts : 882
- Joined: 2005/05/26 02:05:56
- Location: USA
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/23 12:20:59
(permalink)
ORIGINAL: mlockett
As far as if it will load samples, bounce or freeze faster, it depends if that is the bottleneck on your system. I would be surprised if freezing was more bound to cpu/ram than disc writes (dependant on how complex the track is that's being rendered).
Well I hadn't even thought of that. How do I determine this? I'm currently using an athlon64 3700 laptop with mismatched cheap ram (1.5g) and a 7200rpm seagate. It's pretty powerful and I can do a lot with it, it just takes long to load multi velocity or bounce. So I figured I'd build my next system to really fly and be happy for a few years.
|
Sorceress Sarah
Max Output Level: -89 dBFS
- Total Posts : 81
- Joined: 2003/11/15 09:44:17
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/23 15:10:16
(permalink)
ORIGINAL: sandman5000
Thanks for the answers xackley and Sorceress Sarah.
I've been reading up on RAID and RAID 0 is the one I'm considering. I don't need to mirror because I back up all the time. I just wan't my samples to load quicker (the Reason DK 2 in particular) and also bounce to audio faster. Freeze is pretty fast already, but it could always be faster.
I didn't exactly understand the on board controller part. This is for a system I'm going to build when quad cores come out (in a month). So for maximun speed, don't use the MOBO SATA controller? Would a add on card be better (on the PCI bus?)? Would that mess with my PCI audio card (hog the bandwith?)?
My personal choice would be an LSI Logic 2 channel U320 SCSI card and a pair of 73GB 15000 RPM Seagate Cheetahs, one on each channel, stripe across them using the striping capabilities of the hardware. Don't do "software raid" from the Windows Logical Volume Manager. This would be in leiu of a Fibrechannel SAN. Others here may offer differing opinions. (And probably will.)
|
jacktheexcynic
Max Output Level: -44.5 dBFS
- Total Posts : 3069
- Joined: 2004/07/07 11:47:11
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/24 21:05:39
(permalink)
for everyone's benefit, the 'R' in RAID does not stand for "really fast", it stands for "redundant".
in mirror mode your write times will suffer since you have to push twice the data. you will get better read times but it will never be twice as fast. yes, in theory, no in reality since even though you can read from two places the data eventually has to be ordered correctly and processed in one channel.
in striped mode you gain some write performance (again theoretically) since you can write to two drives at once. the reality is you probably won't notice that much. read performance can actually be worse here since you have to wait for both drives to get things in gear (and give you the data in the right order). sata 3.0g channels may have changed things but sata I RAIDed drives sometimes had worse real-world read performance compared with the same drives in non-RAID mode. there's a tom's hardware or anandtech.com article from probably a year ago or more on that so things may have changed.
if you want fast drives get the western digital 150gb raptors (10k rpm). beyond that, 15k SCSI 320 drives are probably somewhat faster but you'll pay a huge premium over the raptors (which are expensive to begin with). striping could help here but you better back those suckers up. if either drive fails you lose them both (which is why RAID 0 is an oxymoron - redundant it is certainly not).
and as sarah mentioned you can go with fibrechannel SAN if money is no object. however the cheapest remedy is to realize that all hard drives are slow compared to cpus and memory and bouncing 30 tracks will take at least a minute or so.
|
sandman5000
Max Output Level: -73 dBFS
- Total Posts : 882
- Joined: 2005/05/26 02:05:56
- Location: USA
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/25 12:13:59
(permalink)
ORIGINAL: jacktheexcynic
for everyone's benefit, the 'R' in RAID does not stand for "really fast", it stands for "redundant".
in mirror mode your write times will suffer since you have to push twice the data. you will get better read times but it will never be twice as fast. yes, in theory, no in reality since even though you can read from two places the data eventually has to be ordered correctly and processed in one channel.
in striped mode you gain some write performance (again theoretically) since you can write to two drives at once. the reality is you probably won't notice that much. read performance can actually be worse here since you have to wait for both drives to get things in gear (and give you the data in the right order). sata 3.0g channels may have changed things but sata I RAIDed drives sometimes had worse real-world read performance compared with the same drives in non-RAID mode. there's a tom's hardware or anandtech.com article from probably a year ago or more on that so things may have changed.
if you want fast drives get the western digital 150gb raptors (10k rpm). beyond that, 15k SCSI 320 drives are probably somewhat faster but you'll pay a huge premium over the raptors (which are expensive to begin with). striping could help here but you better back those suckers up. if either drive fails you lose them both (which is why RAID 0 is an oxymoron - redundant it is certainly not).
and as sarah mentioned you can go with fibrechannel SAN if money is no object. however the cheapest remedy is to realize that all hard drives are slow compared to cpus and memory and bouncing 30 tracks will take at least a minute or so.
Well thanks everyone for the explanations. So...there isn't really much to be gained by RAID0? I already have a 10krpm raptor on another DAW, but didn't notice it being faster than the 7200rpm.. I'd never even heard of the fiberchannel, but it's way too pricey. Is there anything I can do to speed up the loading of samples or bouncing down audio?
|
jacktheexcynic
Max Output Level: -44.5 dBFS
- Total Posts : 3069
- Joined: 2004/07/07 11:47:11
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/25 22:10:09
(permalink)
ORIGINAL: sandman5000
So...there isn't really much to be gained by RAID0? I already have a 10krpm raptor on another DAW, but didn't notice it being faster than the 7200rpm..
not as far as SATA is concerned. if you've got a recently-made big (say 300gb+) 7200rpm SATA drive it will probably be just as fast or faster as a second-gen raptor (10k 74gb). the new SATA II raptors (150gb) will spank pretty much anything but you won't notice unless you do a timed transfer of a large file. for bouncing tracks you probably won't notice much - samples may load somewhat faster.
Is there anything I can do to speed up the loading of samples or bouncing down audio?
faster disks... =) faster cpu for bouncing, more memory for sample loading...
|
tomek
Max Output Level: -77 dBFS
- Total Posts : 684
- Joined: 2004/03/21 18:43:22
- Location: Vancouver B.C.
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/25 23:59:48
(permalink)
I would not bother with RAID0 for the following reasons:
1) The new drives (vertical bit technology) are super fast.
Get the 750GB Seagate you will not be disappointed with the speed
2) Raid0 increases the chance of data loss, and complexity.
You not only double you data loss chance by adding another drive,
but it increases even more, because the RAID itself can screw up.
Make music not computers.
3) You want a DAW to be as quiet as possible. Fewer drives = less drive and fan noise
4) RAID0 may / may not increase your performance.
I understand it's good for increased sustain rate, not access time.
If you stream sample with gigastudio, go for it.
If not, it may not help you access "find" files faster (depends on your setup_
That's allot of work, risk, and money to take a chance.
Non-scsi two drive RAID0 is just not worth the effort for the most part.
Why are you so set on Intel Quad core?
It's the 1st incarnation, and is going to suck compared to the "true quad cores"
I'm waiting for it too, but do we have any proof that Sonar will see a worth while performance gain?
Most existing applications do not.. Quad core might be a waist of cash. We'll see I guess..
Tomek.
RME fireFace 400, Q6600 w/ all the bells and whistles, Acess Virus KB, Mics, soundcraft compact4, Orange AD30HTC into Orange 1x12, DIY FX Pedals, '73 Guild S-100, '97 Guild Starfire II
|
Clydewinder
Max Output Level: -72 dBFS
- Total Posts : 941
- Joined: 2005/02/28 22:34:40
- Location: Milwaukee, WI USA
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/26 17:51:34
(permalink)
Quad core might be a waist of cash. We'll see I guess.. i wish i had a waist of cash...
|
tomek
Max Output Level: -77 dBFS
- Total Posts : 684
- Joined: 2004/03/21 18:43:22
- Location: Vancouver B.C.
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/26 22:29:32
(permalink)
ya me too!
I JUST bought a x2 3800 and crapy mobo (ASRock 939DUAL-VSTA) haha
RME fireFace 400, Q6600 w/ all the bells and whistles, Acess Virus KB, Mics, soundcraft compact4, Orange AD30HTC into Orange 1x12, DIY FX Pedals, '73 Guild S-100, '97 Guild Starfire II
|
sandman5000
Max Output Level: -73 dBFS
- Total Posts : 882
- Joined: 2005/05/26 02:05:56
- Location: USA
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/26 23:20:21
(permalink)
ORIGINAL: tomek
ya me too!
I JUST bought a x2 3800 and crapy mobo (ASRock 939DUAL-VSTA) haha
Yeah, that is the feeling I'm trying to avoid. I figure with a quad core (!), I'll be set for a while (a few weeks at least  ). I may even try to build a dual quad core (Octacore!), if I can afford it (who needs food?). It might be worth it just to go around telling people "I have an OCTACORE!". But you make a good point about waiting a few months longer for some compatibility testing and the cheaper quads (and maybe AMD will have something by then). Now I'm thinking maybe faster RAM would load my samples faster? I ordered my laptop with 256mb and then put in a gig of cheap RAM. It hasn't been an issue at all in almost two years. I'm just getting tired of waiting for my samples to load. I guess I could do everything with my current system and freezing, but....when the creativity strikes, having to wait kind of kills the flow.
|
Sorceress Sarah
Max Output Level: -89 dBFS
- Total Posts : 81
- Joined: 2003/11/15 09:44:17
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/27 03:08:03
(permalink)
ORIGINAL: jacktheexcynic
for everyone's benefit, the 'R' in RAID does not stand for "really fast", it stands for "redundant".
in mirror mode your write times will suffer since you have to push twice the data. you will get better read times but it will never be twice as fast. yes, in theory, no in reality since even though you can read from two places the data eventually has to be ordered correctly and processed in one channel.
in striped mode you gain some write performance (again theoretically) since you can write to two drives at once. the reality is you probably won't notice that much. read performance can actually be worse here since you have to wait for both drives to get things in gear (and give you the data in the right order). sata 3.0g channels may have changed things but sata I RAIDed drives sometimes had worse real-world read performance compared with the same drives in non-RAID mode. there's a tom's hardware or anandtech.com article from probably a year ago or more on that so things may have changed.
if you want fast drives get the western digital 150gb raptors (10k rpm). beyond that, 15k SCSI 320 drives are probably somewhat faster but you'll pay a huge premium over the raptors (which are expensive to begin with). striping could help here but you better back those suckers up. if either drive fails you lose them both (which is why RAID 0 is an oxymoron - redundant it is certainly not).
and as sarah mentioned you can go with fibrechannel SAN if money is no object. however the cheapest remedy is to realize that all hard drives are slow compared to cpus and memory and bouncing 30 tracks will take at least a minute or so.
Old used Fibrechannel gear can be had dirt cheap on eBay. Compaq RA4100 arrays run around $200 bucks, Then put disks in it. But it's faster then hell. . . The snag is they only work with one kind of FC controller. . . The CPQFC card, which is getting scarce and is expensive. Another $200. And we still have not bought disks. Fortunately this array uses U160s, and even with 10kRPM disks, when you stripe across 12 disks, you get speed. But. That's expensive. The ebst bang for the buck in raw speed in RAID is, as I said above, a two-channel U320 controller with hardware striping, like the LSI Logic u320 controllers (used extensively in IBM servers), a pair of 15kRPM 73GB Seagate Cheetahs with one on each channel, stripe across both disks. You'll feel it. What makes this work is the two pipes to the bus. When you do it this way, you really can write or read both disks simultaneously.
|
jacktheexcynic
Max Output Level: -44.5 dBFS
- Total Posts : 3069
- Joined: 2004/07/07 11:47:11
- Status: offline
RE: Will RAID load samples, bounce or freeze faster?
2006/10/27 19:55:51
(permalink)
ORIGINAL: Sorceress Sarah
What makes this work is the two pipes to the bus. When you do it this way, you really can write or read both disks simultaneously.
yeah that's the issue with most consumer-grade stuff - most of it is single channel and for scsi if they have dual channel the other channel is usually for external devices. you've got to get into server-class stuff for two internal channels. and then you have to be careful - some of the dell powervault stuff is crap. maybe when i'm famous i'll get dual channel raid... 
|