. I'll also be selling off some hardware at some point to get some more funds to upgrade. And my birthday's less than a week away, so that'll hopefully bring in some funds 
.
).Naw man, RAID 5 isn't as efficient. RAID 0+1 is the way to go.5eraph wrote:I know the feeling :/ . I've had at least 6 die on me personally in 20 years. 2 WDs, 2 Maxtors, and 2 Seagate SCSIs (with 5 year warranties
If you're going to RAID and don't yet have a controller card, try RAID 5. You can chain 3 or more drives together, you still get the benefits of striping and redundancy, and you only ever lose the capacity of one physical drive on the chain to hold parity data. The only thing you'll lose compared to a RAID 10 (0+1) is the ability to access the array until a replacement drive is connected and the array is rebuilt (unless you have a hot spare, but you still have to rebuild).
RAID 5 controllers are getting pretty cheap nowadays.
What do you mean waste capacity? Assuming you get all drives the same size in a RAID 0 + 1 you won't have any extra capacity. Don't you actually lose capacity with a RAID 5 assuming you use identical drives?5eraph wrote:It's more efficient, you don't waste half your capacity, at most 33% if you have 3 drives.
Yea so you only get the space of what that one drive is worth.5eraph wrote:Not quite, RAID 1 is mirroring, to mirror 2 drives you must have a total of four, but you only have the useful storage capacity of 2. RAID 5 gets around that by using the space of one drive as parity to protect against loss on any other drive in the array. It gets more efficient with larger arrays.
Well that sounds like a bit of an annoyance right there compared to 0+1. But what I'm saying is ok with 3 80gb drives you get 160gb storage, but only 80gb of that storage is actually going to be recoverable.5eraph wrote:The space of only one drive is ever "wasted" on data used to rebuild the array (parity data). Mirroring always wastes half because an operational copy is always maintained, whereas a RAID 5 array must be rebuilt upon failure of a drive before the array's data can be accessed again.
I feel like I've been trolled. Code: Select all
http://www.acnc.com/04_01_05.htmlAny WindowsXP system is technically capable of running RAID arrays, as long as the desired amount of hard drives can be attached. It does not matter what hardware you are using. For RAID 5, merely three files need to be altered.
I found a card that seems to mention everything you listed.5eraph wrote:What I'm looking for in a RAID 5 card are these characteristics:
must be a PCI-e x1, x2, or x4 card, larger is better for throughput;
would like it to have 300 Mb/s burst SATA II hard drive support with NCQ;
must have an integrated XOR engine for accelerated parity operations;
must have at least 128MB of ECC cache, preferably expandable;
must have at least 6 SATA ports with online capacity expansion.
, and I really want to use a couple graphics cards in SLI. I might consider using just one, but it wouldn't help; the PCI-e x16 slot that would be freed reverts to x1 functionality, so it's a no-go with that motherboard.A single card will function perfectly fine in x8, there is almost zero loss of performance compared to x16. That's why the cards run in x8 SLI mode no prob. And also why AGP x8 cards are just as fast as their PCI-e x16 counterparts, because the graphics cards of today aren't pumping enough power to even make full use of AGP x8, much less PCI-E x16...5eraph wrote:I would tend to agree that DFI caters more to the enthusiast crowd, but in this case I think I may stick with the Asus board because I'm not planning on overclocking much and am having no luck finding a board with an x8 slot. Also, I like its component placement, it is entirely fanless (I may go with water-cooling later), the PCI-e x16 slots are spaced further apart to allow for better cooling of the graphics cards if I upgrade later to dual-slot cards or water cool them, and toggling the SLI mode can be done in the BIOS (not through rotating an SLI card). All things considered the DFI mobos are extremely nice and easily match or surpass Asus featurewise, but I nitpick too much.
I'm not sure if a single PCI-e GeForce card can function with only x8, I'm thinking they are made as x16 for a reason. Otherwise I'd definitely consider your suggestion of making both slots work at x8 and go with 1 graphics card and the RAID card. There's also the possibility that nVidia does something more than just convert x16+x1 (or x16+x2 in DFI's mobos) to x8+x8 by rotating the SLI card; they might bridge those connections somehow as well. I'll definitely look into it though.
The only reason I'm considering XFX is for the fact that they are factory overclocking their cards higher than anyone else at the moment. I've heard great things about BFG and have seen some of their work. Believe me, if the XFX cards don't work out I'll definitely exchange them for BFGs. Gotta love newegg.
In this approach it uses PCI slot only for power supply. It's normal SATA drive! It is detected by BIOS and need no drivers for any OS. PCI-E solution is not good for this - that kind of drive would not be detected by BIOS. We would have something like RAM-Drive with proprietary drives for every OS. May be then it’s better to add more RAM to the system and to use normal software RAM-Drive? So, Gigabyte's solution is good, only we need to wait for SATAII 300 Mb/s version and with more DIMM slots and maybe optional backup to flash on power loss.5eraph wrote:It could achieve a much higher average throughput if they integrated a dedicated custom hard drive controller, made it PCI-e based, and connected it solely to the PCI-e bus. It wouldn't be limited to 150 or 300 MB/s and the only bottlenecks I could see then would be the installed i-RAM and bus speeds.
