raid 0 vs. a separate capture and render drive
 

I currently have 2 extra HD installed in my system, one if for holding media and the other is for renders. I always thought it was best to render to drives separate from the drive holding the media. Faster/smoother rendering....

But I have been reading that having a raid0 drive would give a smoother editing experience- should I convert these two drives to a raid0 and then keep the media and renders on this same raid?

thanks,
Ed

Yes, you read correctly. Raid0 gives you the better performance over a single drive.

But you're also misinterpreting the information..
You've only read a small portion of what's generally correct.

Are you using DV or uncompressed HD??
We don't know your needs or acquisition methods or bandwidth needs...

But as a rule of thumb, i'd personally keep preview files separate from media...Why thrash the discs when it's not needed??

Ed,

Maybe these two articles will be interesting:

Adobe Forums: To RAID or not to RAID, that is the...

Adobe Forums: Generic Guideline for Disk Setup

I put those two together, and came up with this:

Generally, I agree. However, if you're working with HD content on most NLEs, at least one 2-disk RAID 0 array is recommended (as a media and/or pagefile drive volume) due to the fact that most NLEs cross-convert compressed HD material to uncompressed within the editor itself for editing. But if you're going to use a RAID 0 array with two of the three disks in your system, consider adding a fourth hard drive for your previews and output. One single drive for the OS, pagefile and programs plus a RAID 0 array for everything else is workable but not recommended by me or Harm except for benchmarking purposes.

Now I have another question for you, Harm:

I have decided to break the two-disk RAID 0 array (thanks to your recommendations in the links!) in favor of three totally separate drives. Previously, I had one 1TB Samsung F3 drive as my OS/program/pagefile drive and two Seagate 7200.12 drives in a RAID 0 array for everything else. But you don't recommend a RAID 0 array if I have fewer than four or five disks. And I decided to break the array and instead install a new 2TB Western Digital Black in my main editing rig. (I could have kept the RAID 0 array and simply added the WD drive, but I don't have any spare SATA power connectors or even spare Molex connectors from my power supply.) The problem is, the 2TB Black is actually faster than the 1TB Seagate even though they both spin at 7200 rpm.

In this setup, I'm thinking of putting the 2TB Black as the media/project drive and the 1TB Seagate as the preview/media cache/export drive. I want to keep the 1TB Samsung as the system/programs drive.

My setup is simple, inexpensive, and works very smoothly for HD editing, both for native formats and Cineform DI:
SATA system drive
RAID 0 media drive
WD USB3 1TB external drive for export, project files, scratch disk, previews, etc.
I also archive System drive image to the external USB3 drive.

Randall, you disappoint me. Have you never heard of a splitter? A short cable that splits a single Molex connector into two Molex connectors? Add a simple Molex to SATA connector and you are done. Then you can simply add your WD disk to your existing setup.

I use a raid0 (two disks) for page file, media cache, local downloads, monitoring software, etc. because all of that can be easily restored/recreated or downloaded in case of disk failure. I am not fond of using raid0, or as I put it in the past, an AID0 because of the lacking redundancy for anything that is not easily recreated. For anything that you may need in the future, a parity raid is IMO the only sensible approach.

That is the reason I mention only parity raids (3/5/6 or multi-level 30/50/60) for larger arrays, possibly with a hot-spare.

Here is my problem: The nearest available Molex is too far away from the hard drives' SATA connector for an adaptor or splitter to be of much use. In the meantime, I will try to order a new cable specific to my PSU with SATA power connectors (my current PSU is partially modular).

UPDATE: I ordered the modular PSU's SATA power cable from Antec. I'm currently using only two of the five modular sockets on my current TruePower New 750W PSU.

Actually, I have only four modular sockets on my system's PSU: two 6-pin and two 8-pin (the 8-pin soclets can accept either a 6-pin or an 8-pin power connector on one end of a modular cable). And as luck would have had it, I found a spare SATA modular PSU power cable that's compatible with my PSU.

Therefore, my two-disk RAID 0 array (for my media and project files) is back in business!

However, I still ordered an extra power cable just in case.

I think separate drives or RAID0 are almost two different questions.
Although RAID0 is mostly fast enough to read and write at the same time, keeping it separated will be just as fast or faster (if the processor can keep up).

Years ago when rendering DV on an USB harddisk, I noticed a very big diference in render speeds when reading and writing was done on separated drives. USB just had far too limited bandwith to send and return the data from and to the same harddisk.
(e)SATA has much more bandwith, so any limitations will be the diskspeed and processingspeed.
I'm not an expert on harddisks, btw... so I could be wrong.

Anyway, I still render my previewfiles on a separate disk.

I agree.
I always set up such that I am reading from, and writing to different drives

Technically speaking, a two-disk RAID 0 array can transfer at almost double the rate of a single disk (assuming that the hard drives used in the two configurations are identical). But based on my own testing in PPBM5, my RAID 0 array consisting of two 1TB Seagate 7200.12 achieved a result of a relatively slow (for the drives' 500 GB per platter generation) 93 seconds in the AVI Disk test (compared to 105 seconds for a single Western Digital 2TB Caviar Black). Part of the reason might have been that the two Seagates use slightly different firmware versions (one uses CC37 and the other CC38). Another reason is that the Seagates are themselves relatively slow performers for their areal-density generation (I ran the PPBM5 AVI test with the Seagate with firmware version CC37 by itself, and it took all of 129 seconds to complete that test). The 2TB Seagate Barracuda XT, on the other hand, is a completely different animal from the 7200.12 line.

By the way, the 86-second result from my top system on the PPBM5 site was based on two Seagate 7200.12 1TB drives that had the exact same firmware version, CC38, before one of the drives failed. I am contemplating an RMA.