Is BIOS raid considered software raid?

[ThePiston]

If you use the onboard raid instead of a PCI card, is that considered software raid?

Is there a performance difference between an onboard mobo raid and a PCI card?

 

[VirtualLarry]

If you're paying less than $200 for the RAID card, then most likely it is a software RAID implementation.

 

[ThePiston]

so a mobo raid has the HW for raid but all of the work is done by the CPU (?)

and a nice raid card (over $200) has an onboard CPU that does all of the work, correct?

 

[IntelEnthusiast]

I like to remember back in the day of the US Robotic 56K hardware modems. There were all these $25 modems and then there was the $100 US Robotic modems. The difference was that the US Robotic modems have their own processor to do the work while all of the others (software modems) would use the CPU for their computational work.

If the board has RAID on any of our desktop boards and chipsets it is going to be software based.

 

[ThePiston]

ok, cause I know you can do raid in Windows too so I wasn't sure about the difference between that kind of truly software raid and an onboard mobo raid

 

[ThePiston]

I thought I'd post this from Tom's. They are talking about software vs hardware raid:

"Why Software is Better"

"We asked Mike Mihalik if a system with a dedicated, hardware-based RAID controller would be more appropriate for a test such as this. He said no. These days, processors like the Core i7 in our MacBook Pro are so fast that they keep resource utilization on RAID 0 computation down to less than 2% on average—a negligible amount that has essentially no impact on other applications. Going with a discrete controller wouldn’t improve the performance we'll be reporting. Also, hardware RAID controllers tend to use proprietary formatting routines that dictate where data gets written on each drive in each enclosure. With OS-based RAID, such as that provided by Apple’s Disk Utility, users can pull out drives, switch their positions around, and the system will reassemble the array in seconds. This can’t be done in conventional hardware-based RAID configurations."

http://www.tomshardware.com/picturestory/623-thunderbolt-performance-5big.html

go to picture 6 of 18

 

[greenhawk]

I thought I'd post this from Tom's. They are talking about software vs hardware raid:

"Why Software is Better"

the range of products that offer raid it too large to just sum up features on the software vs hardware level.

The part about drives moving not being supported is a feature of good hardware raid controllers. If the person was talking about a cheap pci raid card, then you will have that problem of moving drives breaks arrays. It is also possible to see this sort of feature on motherboard/bois raids.

Only a OS based software raid gets around it at the cheap end.

As to cpu load, for raid 0 and raid 1, the software vs hardware raid is pointless as there is no real cpu processing needed, just data shifting/moving. If fact, there was a point when cpus started getting better where some older hardware raid cards would be slower than software raid for raid 0 and raid 1. the reason being the interface between the drives and the cpu.

for all other raid levels, hardware raid is still generally faster in write speeds (where performance between software and hardware is most easily seen).

 

[tweakboy]

Either way the fastest RAID is ,,, kind built into the mobo. gl

 

[Mark R]

so a mobo raid has the HW for raid but all of the work is done by the CPU (?)

and a nice raid card (over $200) has an onboard CPU that does all of the work, correct?

That is correct. However, RAID 0, RAID 1 and RAID 10 are trivially simple algorithms, and the CPU usage is barely measurable even on something like an atom. On something like a multi-core ivy bridge, the CPU usage is utterly negligible (like less than 2% of 1 core).

Good quality RAID cards do have on-board CPUs. These are more useful when you are using a more complex RAID configuration - like RAID5 and RAID6. These specialist cards use CPUs which are specifically tuned to be very fast at the specific RAID5 and RAID6 algorithms.

However, when compared with a modern general purpose CPU, these RAID CPUs are very weak and can become the bottleneck if you are running RAID6 with 8 or more fast hard drives. In contrast, repeating the same test with software raid on a sandy bridge CPU, had the drives maxed out with 10% CPU usage of 1 core. In practice, this is a very contrived scenario, and the bottleneck will only be relevant in the very unusual situation of long sequential writes.

Where hardware RAID cards win (and which cannot be replicated in software) is in the use of battery (or flash) backed cache RAM. This cache has a magical effect on RAID5/6 performance (for random writes) and drastically improves data integrity. Even with a weak on-board RAID processor, the benefits of the cache typically make it a more desirable solution to software for datacenter use.