difference between expensive and cheap raid cards?

[sewage]

what is the difference between these RAID controllers:

http://www.newegg.com/Product/Produc...0%20119340957#

http://shop.ebay.com/i.html?_nkw=rai...card&_osacat=0


They appear to me (raid noob) to offer the same features, why the big price difference?

[Ualdayan]

The cheaper ones are similar to what is probably already integrated into your motherboard, whereas the expensive ones have (somewhat) powerful processors and even have onboard slots for 1GB of cache.

You remember the old arguments about soft modems which just provided the physical connectors you needed but your CPU still did all the work vs hardware modems that actually did the work? Or the same arguments about sound cards back in the days?

If you are just wanting to raid 2 drives together for casual use a cheap one is probably just fine. Some people however RAID together 6+ drives or such and want top performance - the higher end ones are definitely needed for those people.

[sewage]

ahhh ok I see, well....I have a raid 10 (total of 4 hdds) and its almost full. If I wanted to expand my capacity what is the best way to do this? or how to transfer the data to a new raid 10 with only one pc???

I was thinking about getting a raid card and setting up another raid 10....the problems with this is that I would need another case to hold the next 4 drives, a stronger psu or an additional psu and...how is the psu controlled?

What are some typical solutions that are available(I am not extremely interested in a NAS due to high price)?

[Pixels303]

Adaptec and LSI and AMCC (3ware) all have the ability to expand raid arrays by adding more drives. Promise, Highpoint and the others I cannot speak about as they are cheap software based devices and suffer in performance and features.

Whenever anyone upgrades a array, it is always easier to simply copy the contents of the array to a new array as most vendors suggest backing up data before expanding the array. If you run into troubles while converting the data, you chance loosing everything.

You can have two controllers in one machine, or if you have a 16 or 24 port controller, it is a snap to add more drives to your computer (If you have enough space to install them) and simply do a local copy and compare. If you need more space, you may refer to my project: http://forums.pcper.com/showthread.php?t=467077

Although there is nothing wrong with using a second PC and copying data via windows file sharing. Mind you, you will lose file security and maybe change file dates.

Final note, I would not recommend anyone using Highpoint brand. They have a non-existent customer service (Unqualified outsourced service by foreign countries) and have a good track record of selling buggy hardware. I myself have had the experience which I never plan on repeating.

[MRFS]

Be careful not to mix apples and oranges
(maybe that should read "bananas and oranges"


The old PCI bus has a bandwidth of 32 bits @ 33 MHz = 1,056 MHz;
it's considered obsolete for many reasons, one of which is that all
devices share that same bandwidth: while one PCI device is busy,
all the other PCI devices must WAIT.


There are different variants of PCI-X, but it too is considered obsolete
-- unless, of course, you have a motherboard with empty PCI-X slots;
then, it's not "obsolete".


PCI-Express is the current standard: it has 2 different generations --
PCI-E 1.0 and PCI-E 2.0. PCI-E 3.0 is currently in planning.

Under PCI-E 1.0, each "lane" is capable of 2.5 Gbps in each direction,
with 10 bits per byte (1 start bit, 8 data bits, 1 stop bit) = 250 MB/sec.
So, divide bandwidth in bits per second by 10, to get bytes per second
(as in the old serial modem protocols). Thus, x1 PCI-E lane has roughly
twice the bandwidth of the old PCI bus.

PCI-E 2.0 doubled that to 500 MB/sec in each direction.

Thus, x4 PCI-E lanes will have four times the bandwidth of each x1 Gen;
x8 lanes will have eight times the bandwidth of each x1 Gen; and
x16 lanes will have sixteen times the bandwidth of each x1 Gen.

Quite often, however, a chipset will assign fewer PCI-E lanes
than are apparent from the edge connector or from the slot size:
e.g. only x4 lanes assigned to an add-on card with an x8 edge connector
or only x4 or x8 lanes assigned to an x16 mechanical slot.

So, RTFM! (Read The Fine Manual -- not always "Fine" however


With the advent of widely available fast dual- and quad-core CPUs,
the differences between software RAID and hardware RAID have
narrowed a bit: with software RAID, an idle CPU core can do
much of the computation that a hardware RAID controller
will do with an on-board dedicated processor i.e. parity calcs.

Windows supports its own software RAID, but you cannot
boot from such a software RAID.

Another feature of good add-on RAID controllers is the
management software which is bundled with that hardware,
e.g. a Windows GUI that permits RAID array creation,
and device monitoring. For example, a good RAID controller
can send an email message when it detects a fault condition
in any of the component drives: that type of feature can be
very handy for larger server farms.

Intel's ICHx I/O controller hubs come with their IMSM software --
an acronym for Intel Matrix Storage Manager.

Also, the size of the on-board cache will make a noticeable
difference in performance: the larger the hardware cache
the better the overall performance, in general. These caches
usually come in sizes ranging from 128MB and 256MB at lower end
to 512MB, 1GB and 2GB at the higher end.

Similarly, the size of the cache integrated in each HDD
can also make a big difference, particularly as the
number of component drives increases. Thus, a RAID 0
with 4 x HDDs each with 32MB cache will produce
an "effective" HDD cache of 128MB (4 x 32MB).

Lastly, the very latest RAID controllers now support
SATA/6G and SAS/6G interfaces, e.g. Intel's RS2BL040
and RS2BL080 hardware RAID controllers, built by LSI.

Some motherboard manufacturers are offering a
compromise add-on controller that uses "PCI-E x1 Gen2" logic:
x1 Gen2 means 500 MB/second, however, NOT 600 MB/second.

The "6G" standard means 6 Gbps / 10 = 600 MB/second (full bandwidth).


p.s. We've been using several different Highpoint models,
and they have all worked flawlessly.


MRFS

[sewage]

I thought all cards were hardware raid, how can you tell if the card is using software raid?

So you can have two controllers on one mb? During boot, do you get some kind of choice to setup arrays using different cards or does the second card have to be a software raid?

What if the hardware raid controller fails? Is it possible to limp the array on to a totally different controller to recover the data?

[Pixels303]

If you do benchmarks on RAID, you will find that with RAID 5, windows has a theoretical limitation which cannot be exceeded regardless of the speed of processor used. I used a single core 1.5ghz AMD Socket A cpu and a quad core 2.5Ghz AMD AM2+ cpu, and comparing benchmarks, they were within 20% of each other performance wise while using software RAID5, the same drives and different on-board controllers.

Hardware controllers have at least 10 times the benefit in speed over software raid and do not suffer data loss with BIOS upgrades and changes to software unlike windows RAID.

Some controllers (Highpoint mentioned) have much less efficiency than true hardware controllers. I don't care about how much they have improved since I last used them, but I find their advertising deceptive as I have been ripped off, and reading other peoples reviews regarding Highpoint have been generally poor. Highpoint may work flawlessly for you, but when you run into an issue with upgrading or fixing something, they have very poor customer service. If it works flawlessly, you will find the LSI, ADAPTEC and 3ware brands far speedier and loaded with additional useful features. I always found that using drives as single drives were always faster than RAID1 or RAID5, unless using a True-Hardware based controller. If your data is important, do not use software RAID or Highpoint. Spend the money and buy LSI, 3ware or Adaptec.

I have had up to 5 controllers on a motherboard once. If you have a good BIOS, it can select which devices to boot from with a choice of controllers. Most often, you create only one boot device, and the BIOS tries to load from all devices with only one available boot drive - then your ok. With older PCI controllers, being shared sometimes has BIOS limitations as to how many controllers are usable. Manufacturers sometimes state that only one controller is usable on any given system. If you use PCI-E, the likelyhood of using more than 1 is slim as they are almost always PCI-E4X cards or bigger, and most mainboards are equipped with only 2 PCI-E 16X and the rest 1X slots. Being that most people have a 16X video card, that leaves you with only one controller.

If the controller fails, no big deal, just plug in a new one (SAME TYPE) and your up and running again. Some software (Rstudio) has the ability to recover data from arrays without using hardware controllers (Direct to SATA ports on mainboard). Data recovered is always questionable with moving an array to a different controller. With Rstudio, you may run out of memory when recovering more than 2 Terrabytes of files. However you need available storage equivalent to what is being recovered, and that is not easily come by with HUGE arrays as the cost of storage gets big with more than eight drive systems.

I answered a few questions I hope.

[sewage]

ya I think this is plenty of info to chew on for a bit, thanks guys!