PlanetZ forums - scopeusers.com

Playing fast piano samples with velocity layers and sustain, I can get to hundreds of notes and that is just with one instrument. I could convert it to a stereo audio track but then I would not be able to alter the tempo track and I like to do that quite frequently. I feel also that working the hard drive a lot will reduce the life span of it.

well, sure, if you use several different large and demanding streaming libraries at the same time, that is likely to push the drive over its limit. But that is not a sensible use for the sample libs anyway: just freeze the tracks once you have them sound right, and your drive can relax again. The solution is just this one click away, really. To speed things up alternatively it would be sufficient to install different libraries that you typically use at the same time onto different physical drives. No RAID will cure your current problem, trust me.

edit: what makes you trigger so many notes with much sustain on a single instrument? I cannot think of a musical context that would really ask for such. If you really need that jumble of notes performance, install the library twice onto different drives and spread the triggers evenly among the two players...

JoeKa wrote:just freeze the tracks once you have them sound right

I am constantly changing it. This is not a good solution. I have tried this.

JoeKa wrote: I cannot think

This is your problem!

I would suggest 3 hds.
1 for the operating system and all the other programms and 2 drives in raid0 with a dedicated controller.The onboard raid controllers are taxing the cpu and are in generell not very good.with raid 0 you also double the possibility of a hd failure.1 disk craps out and the data is gone.

I always backup what is on RAID 0. I haven't had a RAID go bad because it's just for the DAW. Taxing the CPU might be a problem I guess but this is a Core 2 Quad so I will wait and see. I can get a RAID card if I have to. I think it partly depends on the mobo. It would be worth trying 3 single drives though an compare it to 1 drive with 2 in a RAID 0 for audio. I'm guessing the RAID 0 would still work better but I don't know.

It doesn't depend on the motherboard, it depends on the chipset used on the motherboard. Intel's RAID does tax cpu (he's correct) and again JoeKa was correct 2 posts above that in saying a RAID controller does NOT help in situations where more than 1 thing must be accessed at a time.

Think of it this way, enabling RAID on the controller means that even disks outside of the array (JBOD mode for them) still have to wait on the controller to finish what it's doing on the array (accessing your sample libs) before it can move onto other things. And having other data (OS/Apps/Projects) within the array makes it worse, as your controller also has to deal with the added overhead of accessing more than one part of the array. The only way around that is to add another RAID controller (or just HD controller) with its own volume(s).

What RAID does is increase throughput via stripe, and create redundancy via mirror/parity. What RAID does NOT do is increase a controller's ability to access multiple datastores across a single drive or multiple drives. So if you think your piano rompler (etc) really need more than the 80-100MB/s you'll see on average with a modern SATA2 drive, then yes RAID solves that. If you simply want the whole machine to 'do more work' with its harddrive subsystems, you need to design them to fit the workload. I outlines some RAID specific configurations above, but I agree with Fluxpod here. Add more drives to the controller in non-RAID mode (AHCI enabled, ncq enabled) and dedicate 1 per workload (1 for OS/apps/swap, 1 for your projects, 1 for sample libraries).

I might add that SAS (serial attached scsi, the modern scsi implementation) does actually quite a bit better at multitasking across multiple drives or within an array when compared to SATA. But the additional cost (of the controller and the drives, although you can substitute certain SATA drives like the Raptors) is not worth it for your workload. It doesn't make sense to throw $1500-2500 at a single RAID array when you can just design your HD subsystems better with the onboard controller. Intel's ICH9 supports up to 6 drives in AHCI mode, more than enough to spread your workload out.

On array for one application? I will have to test it out and see what works best. Your theory may be correct but the only way to find out is to test it in the real world and see. I read a RAID 0 array with 2 drives is 30% faster, more drives would be even faster as I understand it. I'm thinking even if multiple apps slows it down, it still would be better than a single drive. A single rompler accesses multiple files at the same time doesn't it? If so, that would be the same as using multiple applications.

Update: I just ordered a third drive since they are so cheap now and Cubase 5 Upgrade. The easiest thing to do will be to start doing music now with the three single drive solution since I would have to move a lot of data around to an external drive to make room for an array.

Braincell, I've used scsi RAID since 1995, I assure you what I'm telling you is correct. One array per workload is the ideal, the rest of the info you may have heard sounds like potential heresay relayed from info originally off of computer enthusiast sites that portray single benchmarks as indicators for things they are not. Consumer RAID controllers (including Intel onboard) do NOT have the ability to read & write to more than one array per controller simultaniously, though they do allow you to define more than one array per controller as it's not really even a true array (it's a software array handled entirely by the cpu and no different than what you can get with ZFS and other filesystems that support software RAID natively).

RAID increases throughput up to the limits of the bandwidth reasonably available (not theoretical limits) by allowing data to be read/written to more than one drive interface at a time (NOT via multiple drive controllers though). Increasing access times (seeking etc) is increased by faster spindle speeds (RPM), increased aerial density (packing the data tighter on the platters) and improved drive mechanisms, not by RAID. Handling multiple reads/write requests at once is addressed by having more than 1 controller in play, though NCQ in theory was supposed to improve things some. With the sort of streaming data you're dealing with it can actually be a hindrance, oddly enough. NCQ only 'steers' the drive head near data that's in the request/write queue on the drive itself, which is only as large as the cache for that drive, and can actually cause stalling in streaming large amounts of data for reasons I won't go into in this paragraph.

SCSI/SAS controllers are better than SATA in regards to the last point (multiple reads & writes). The controllers there are smart enough that when paired with on-drive SCSI/SAS drive logic you can actually read & write to multiple devices at once, as the controller will pass the request off to the drive and move onto another in the meantime. I actually LOVE the 'feel' of a machine running u320 SCSI & 15k rpm drives and almost went SAS for my new Xeon rig. It would be nice would be to have that FEEL again (it does feel smoother even in this SATA2 era) of nothing ever 'hanging' while a drive was being accessed, but the cost is just prohibitive for the minor *actual* performance increase I'd see over SATA--in terms of throughput for a given workload. And that throughput is only higher because of the increased spindle speed again (SCSI/SAS drives actually have a lower aerial density to decrease the error rate, which is one of the reasons that they have higher spindle speeds--seek times being the other).

So I can have faster speed on one instrument or slower speed but more instruments.

Let's just assume you are correct; that using three independent drives is the best option. I would still like to test it in the real world. There are a lot of variables such as chipset, the speed of the software, OS and how they interact with the other hardware. All of these things are not exactly the same as in your configuration. Maybe I will go for the speed of the RAID 0 and freeze tracks. I am curious to see how much I can do with it before it causes audible problems. Perhaps I will push it to the limit and then try without a RAID if I can't do what I want to. My piano takes several minutes to load; from what I have read, a RAID 0 array would really make this faster.

Yes a RAID 0 would increase preload performance. As for 'real world tests', I have 5 drives in this system right now (my main box),

One drive is for my graphics & web work (clientelle) and only has a partition towards the end of the disk for storing some extra audio data (not performance related). Win7 also lives on this drive atm in a middle partition, used for playing around & testing but nothing serious. We can ignore this drive for our conversation here I think.

Another drive is actually dedicated to OSX, with the OS/apps/etc on the first partition and some project data on the second, this drive is guid partitioned and not available outside of OSX. I use OSX primarily for Logic but we'll ignore that drive as well since it's OSX and focus on the Windows audio aspects now...

The 3 drives that are left are setup exactly as we've stated.

Primary drive is OS/apps/swap, 3 partitions: First partition is Xp32 for music & general stuff. Second partition is Vista64 for when I need to do large graphics projects (3d renders mostly, some video compositions). Third partition is again data storage (I tend to always use the last 1/3 of my drives for data storage that won't see a lot of frequent use, as this keeps the main use of the drives in the higher performing regions of the platters).

Secondary drive is 2 partitions. Samples & Romplers on the first partition, music storage on the 2nd partition (only the last 1/3 of the drive).

Third drive is 3 partitions. First partition is audio projects (recording & bounced clips), second partition is actually the same thing but in HFS+ for OSX, and the third partition is used for OSX data storage (again HFS+ formatted).

When doing projects, I don't use romplers the size of what you use but I do have some relatively large romplers and tons of samples. I can get several vsti based romplers, quite a few other vsti's and on average 20-30 tracks of audio going (on the separate drive from my romplers/samples) and I generally only see drive usage listed in Ableton & Cubase around the 30% mark. It's not the same workload as you exactly, but works fine here and I suspect would for most.

Something I should mention here too, for the samples/rompler drive I keep the partition that content is on defragged on a regular basis. The 'project' partition that I do my actual audio work on I do NOT defragment, as multitrack audio will tend to record to the disk with the data interleaved in a way that aids playback performance. Attempting to make each file contiguous with a defrag app will actually increase the work the drive has to do. When I finish up projects or put them aside, I move them to a data storage partition somewhere or back them up & remove them. This keeps that partition below 60% usage and performing well...

Taking Tiger Mountain by Strategy... so to say with Brian Eno
thanks a million for sharing your insights about disk systems, Valis

cheers, Tom
(this just had to be said)

1/3 price cut for intel ssd

http://www.pcmag.com/article2/0,2817,2340525,00.asp

becoming tempting!

I am going to wait one more year.

Seagate, WD and everyone better watch themselves. once they get off the stupid SATA bus (for memory chips ..no)
i can see Intel, OCZ, Corsair, and so on taking over the entire storage business.
however since its "just chips" there might be fierce competition with many players like in the memory business.
you just need a factory that can make circuit boards and populate chips on them, rather than expensive fabrication of mechanical parts to extremely tight tolerance and cleanliness
(its kind of sad to see yet another nice mechanical thing at such a high standard go by the wayside)
even if the hard drive makers stay competitive, they will be "just another" manufacturer making boards.

maybe hi fi freaks will find some reason that hard drives "sound better" than memory chips and seagate can sell them to the tube heads

It seems to me that spinning magnetic platters will remain in use for some time to come for data storage, with SSD's replacing them in the next several years for runtime use (OS, apps, our audio tasks etc). SSD's priced at the CONSUMER level have a LONG way to go before they can compete on a dollar/GB (soon to be TB) basis, and the event horizon to Moore's law is still a theoretical barrier. SSD's still also need to make improvements in wear-levelling, or actually not taking wear at all (instead of simply writing around it). As they stand now within a few years the usable area on a drive is reduced substantially, even from reads.

The latest crop have started to address speed and access time issues (some of the older drives take as long as one second to start to move data!)

Braincell, if you want something that will improve performance for your romplers, I heartily recommend buying a single WD Velociraptor 300GB model.

Neutron wrote:1/3 price cut for intel ssd

http://www.pcmag.com/article2/0,2817,2340525,00.asp

becoming tempting!

Also covered here with additional info (and not as verbose):
http://www.dailytech.com/article.aspx?newsid=14188

Neutron wrote:...you just need a factory that can make circuit boards and populate chips on them, rather than expensive fabrication of mechanical parts to extremely tight tolerance and cleanliness ...

eh... chips are made in factories of extremely tight tolerances and cleanliness....

cheers, Tom

astroman wrote:

Neutron wrote:...you just need a factory that can make circuit boards and populate chips on them, rather than expensive fabrication of mechanical parts to extremely tight tolerance and cleanliness ...

eh... chips are made in factories of extremely tight tolerances and cleanliness....

cheers, Tom

with no aluminum molding, milling, turning, stamping, welding, coil winding, making shaped magnets, assembly and all those other expensive things not needed for chips.

I fixed all problems with the piano rompler by raising the amount of RAM it uses to the maximum of 1 gigabyte for the two microphone positions I am using in the largest preset. I am still going to try it with a RAID though since the 3rd drive I ordered is arriving tomorrow. The single drive for the OS and programs and the RAID for all samples. I was considering the Hackintosh until I read that they strongly recommend IDE drives over SATA which is what I have. Valis uses a Mac. I see a potential difference between the Mac versus a PC. Apple just started using the Intel chip. Who knows what problems they might be having. Astro is right Apple isn't what it used to be.

yes, they've changed...
(a Billie Holiday tune comes to memory...)
but to their defense it must be stated that after all it was customer demand that sealed the change
Apple was faced with the paradoxon of decreasing sales while customer satisfaction increased
the average duty cycle of an Apple was at least 3 times longer than the PC those days
At the same time it was easy to spot that people bought the M$ crap like mad, though Apple had a technically superior and easier to maintain small office solution.
Effectively resulting in much lower overall cost outweighing the higher entry price of the machine several times.

But People are People - and they don't mind facts
when it comes to sales decision they are hordes driven by tribal instinct and such stuff - but definetely not by their calculator or spredsheet, let alone reasoning
How would you decide facing those facts ?
Let the company go down with your head high - or give the market what it demands ?
Apple did the latter, but at least the machines are still visually appealing and easy to operate...

cheers, Tom