Motherboards today have hardware level raid?

Tengis

Supreme [H]ardness
2FA
Joined
Jun 11, 2003
Messages
6,578
Are there any nice intel motherboards with built in hardware raid? Is all of the raid nowadays still software? Havent kept up with a lot of this stuff lately... :eek:
 
There are no consumer boards with built-in hardware RAID. It's possible that some server boards have it built in, but I wouldn't know about that with any certainty.
 
Yea consumer boards just have software raid. If you just want to run a raid 1 you should be fine. If you want a raid 5 or something get a hardware card.
 
i may be wrong but hardware raid is when there is a raid controller built onto something....and from what i have seen a great many boards today have built on hardware level raid......

Software raid is when you raid drives in windows or linux or what ever else OS.....
 
HW raid uses its own processor to calculate party, etc.

SW level RAID offloads these calculations to the CPU - however another advantage to HW raid is that if the rest of the system goes down you can physically move the drives and controller to another system and your data should (hopefully) be intact. This advantage would seem to be negated by HW level raid controller on the motherboard, but in a server environment that shouldn't matter as identical systems should be around.

SW raid can be setup either within the OS or through the bios, but I believe they are both SW level raid.
 
i may be wrong but hardware raid is when there is a raid controller built onto something....and from what i have seen a great many boards today have built on hardware level raid......

Software raid is when you raid drives in windows or linux or what ever else OS.....

Hardware RAID uses actual RAID controllers, whereas consumer boards with RAID rely on software usually incorporated into the BIOS (or in the cases of motherboards with extra drive controllers, in the firmware of those controllers). Although they do have drive controller chips, those chips don't actually have built-in RAID controllers. Read more here: http://en.wikipedia.org/wiki/Redundant_array_of_independent_disks#Firmware.2Fdriver-based_RAID
 
... which is why I dont know why mobo makers even bother putting it on the boards anymore. Half the time the 'on board' RAID is buggy or doesnt even work, and if you are a 'power user', chances are you arent going to use it anyways, opting for an actual card.

besides, with SSD's gaining speed/market, RAID's days are numbered. In some ways, maybe already gone (since if you had the money, and wanted the speed, why wouldnt you just get a SSD?). So why mobo makers still add RAID onto the boards, I dont know.
 
... which is why I dont know why mobo makers even bother putting it on the boards anymore. Half the time the 'on board' RAID is buggy or doesnt even work, and if you are a 'power user', chances are you arent going to use it anyways, opting for an actual card.

besides, with SSD's gaining speed/market, RAID's days are numbered. In some ways, maybe already gone (since if you had the money, and wanted the speed, why wouldnt you just get a SSD?). So why mobo makers still add RAID onto the boards, I dont know.

Intel RAID is still plenty good for the vast majority of people. There's no need for mobo manufacturers to spend a lot of money putting proper RAID controllers on their boards when a cheap drive controller chip with some firmware will be perfectly sufficient.
 
My Gigabyte GA EX58 Extreme has built in hardware Raid.
There are 4 extra Sata ports to be used as 'ordinary' HDD ports, or enable Raid and boot up, configure the drives and that's it.
Raid can also be used with Raid drivers on the other 6 Sata ports, so 10 ports in all......just about covers everything eh ? :D
 
My Gigabyte GA EX58 Extreme has built in hardware Raid.
There are 4 extra Sata ports to be used as 'ordinary' HDD ports, or enable Raid and boot up, configure the drives and that's it.
Raid can also be used with Raid drivers on the other 6 Sata ports, so 10 ports in all......just about covers everything eh ? :D


That's still software RAID, the CPU and BIOS handle the parity calculations, and migrating the array to another system (different mobo/ BIOS) is impossible.
 
My Gigabyte GA EX58 Extreme has built in hardware Raid.
There are 4 extra Sata ports to be used as 'ordinary' HDD ports, or enable Raid and boot up, configure the drives and that's it.
Raid can also be used with Raid drivers on the other 6 Sata ports, so 10 ports in all......just about covers everything eh ? :D

As vick1000 pointed out, you still have fakeRAID or hardware assisted software RAID where the CPU still does all of the parity calculations. No consumer level motherboard, including the GA-EX58 Extreme have true hardware RAID where there's a XOR engine, or CPU in simple terms, completely dedicated to parity calculations.

... which is why I dont know why mobo makers even bother putting it on the boards anymore. Half the time the 'on board' RAID is buggy or doesnt even work, and if you are a 'power user', chances are you arent going to use it anyways, opting for an actual card.

besides, with SSD's gaining speed/market, RAID's days are numbered. In some ways, maybe already gone (since if you had the money, and wanted the speed, why wouldnt you just get a SSD?). So why mobo makers still add RAID onto the boards, I dont know.

RAID 0 or RAID 1 does perfectly fine onboard RAID. An actualy hardware RAID card is not worth it for just RAID 0 or RAID 1. There are many people out there willing to put up with the slow speeds of RAID5 using onboard RAID since they don't have to shell out $300 for a 4 port true hardware RAID card. Intel's current crop of storage controllers are pretty decent when it comes to RAID 0 or RAID 1.

Right now SSDs aren't that much faster than a RAID0 setup of decently fast drives. In addition, SSDs still cost a quite a bit compared to mechanical drives. While future SSDs should outperform current Velociraptor drives in RAID0, someone out there will want a RAID 0 setup of said fast future SSDs. And finally, SSDs do not replace RAID5 or RAID 1 in any way. The need for uptime that RAID 1 and RAID5 provide will still be there even with screaming fast SSDs. Even a SSD drive can die.
 
I dunno... it seems to me that with the mechanical aspects of HDD's removed, and just having the 'pure electronic' transfer of information to and from a SSDs, that a RAID setup with SSD's wouldnt help that one bit... its like when intel was saying their procs were faster because they had a longer pipeline in the P4... which just wasnt the case. Adding extra processing stages to a SSD's I/O shouldnt speed much up. Proof? Well, the fact that defragging a SSD is pointless should tell you something... the very idea of making a SSD in raid is rather pointless. And as far as backup, considering how fast they are, you could run a software backup every time you use it in the blink of an eye... making Raid 5 and other 'backup' oriented RAIDs rather useless as well.
 
I dunno... it seems to me that with the mechanical aspects of HDD's removed, and just having the 'pure electronic' transfer of information to and from a SSDs, that a RAID setup with SSD's wouldnt help that one bit... its like when intel was saying their procs were faster because they had a longer pipeline in the P4... which just wasnt the case. Adding extra processing stages to a SSD's I/O shouldnt speed much up. Proof? Well, the fact that defragging a SSD is pointless should tell you something... the very idea of making a SSD in raid is rather pointless. And as far as backup, considering how fast they are, you could run a software backup every time you use it in the blink of an eye... making Raid 5 and other 'backup' oriented RAIDs rather useless as well.

That doesn't make any sense. The reason why defragging an SSD doesn't help is because it takes an SSD the same amount of time to read from any given sector. It can switch from reading one sector to reading another in a completely different physical location without any time penalty. Mechanical HDs on the other hand take time to move the head to a different location on the drive, so sequential reads of data that is all in one place on the drive can be considerably faster than if it has to read bits and pieces of files that are scattered all over the drive.

RAID on the other hand is a completely different story. The benefit of RAID0 is that two drives can be reading or writing different pieces of data simultaneously. Although SSDs don't have any limitations regarding the location of different pieces of data, they are still limited by their maximum read and write speeds. RAID0 can provide up to double the normal read and write speeds in theory, so there's still a distinct speed benefit from using it, even with an SSD. As for RAID1 and RAID5, the redundancy that they provide is still a benefit regardless of the type of drive used, since even an SSD can fail prematurely.
 
Ah, I stand corrected.
On 4 ports there is no need to load drivers to enable Raid, but if Raid is to be used on the other 6 ports, then drivers have to be loaded.
The manual says 'hardware Raid' so I assumed (wrongly) that's what it was....:)
 
Isnt defragging a SSD drive bad? Due to the limited number of writes per data cell?
 
Isnt defragging a SSD drive bad? Due to the limited number of writes per data cell?

SSDs have technology built in to evenly distribute the number of writes across the entire drive. Although defragging isn't great for the flash chips, the main thing is that defragging simply doesn't have an affect on SSD performance.
 
In a 'hard disk' HDD, there are physical arms that must move to find the data... this is what causes the delay in 'getting to' the information. Without the mechanical arms, the only delay is how fast the I/O is... as if the arms are always where they need to be. This is also why RAID with them is rather useless as far as read times are concerned. With Multi-layer memory and such to make the write speeds as fast as the read speeds, RAID is rather useless... that is, if your SSD is running the full bandwidth of the bus (so splitting up ops between drives is rather useless). Considering SSD's are quickly approaching the read/write times that max out the bus, RAID is going to be rather useless very fast. Using RAID on SSD's is going to be alot like using a processor on all of a microprocessor's I/O lanes to help it go faster... it just doesnt work. Lets say one SSD can do 250MB/s of data, and RAID could in theory double that to 500MB/s... well, thats great and all, but if the bus (southbridge controller) itself can only take data at 250MB/s... the RAID isnt doing anything now is it?

Does that make sense to you Zero82z?
 
Does that make sense to you Zero82z?

No, because it doesn't make sense altogether. The only thing limiting the performance of SSDs is the performance itself. If you RAID together two SSDs with 130MB/s read and 90MB/s write speeds, the theoretical maximum speed in RAID0 is still 260MB/s read and 180MB/s write.

Regarding your comments about SSDs nearing the limitations of the bus, you don't seem to be considering the fact that when you have two drives in RAID0, each one is using a separate SATA channel. Therefore, the theoretical maximum amount of bandwidth for a RAID0 array is double that which a single drive would have access to. So let's say for argument's sake that you have two SSDs with a read speed of 200MB/s each. In RAID0, the maximum theoretical bandwidth would still be 400MB/s, because although SATA2 has a maximum speed of 300MB/s, since the drives are each using one channel, the combined maximum speed is actually 600MB/s. Unless we start seeing SSDs with a read speed of over 300MB/s, the limitations of the SATA2 bus won't be an issue. And even if we do, a pair of >300MB/s SSDs in RAID0 would still be able to hit a combined read speed of 600MB/s, which is significantly more than a single drive could muster.
 
"The only thing limiting the performance of SSDs is the performance (of the SSD) itself."

Yes, exactly what I was saying... so what would RAID add to a system where the only limitation on the performance be the SSD itself? Now you are the one who doesnt make sense.

And while you are correct that each drive operates on a seperate SATA channel, you are leaving out that ALL of these channels are on one bus, and that every drive in RAID is going to use one interface with the southbridge and CPU... that is the limiting factor. Just like if you have a hardware RAID card that uses one PCI-E slot, you are limited by the I/O of the bus itself that the RAID controller uses. Considering all we are talking about is relative clock speeds, propigation delays, and the time it takes to switch the magnetic fields in each transistor of the SSD, how is adding more transistors and processes in the series of data transfer mechanisms going to speed anything up? Intel's P4 architecture with its 20 stage pipeline was an example of how that didnt work. Putting storage/RAM/etc in parallel only benefits you if the bus that they must both go through in series can actually make use of the extra bandwidth. Much like with running multiple GPU's... the drives are getting so fast that the return on running multiple in parallel is getting to be less and less.

Even EVGA has mentioned that RAID 0 with SSD's will only work if the throughput is less than the SATA-2 spec (for both drives combined) for SATA-2 boards... the RAID controller is on a bus which cant handle more than this... so if you have 200-250MB/s per drive, and you link them together to get 400-500MB/s, you will still be capped at the 200-250MB/s speed because that is all the RAID controller has to work with between it and the southbridge.
 
Your SSD will burn up quick. It's too early to walk away from the proven tech.
 
Here, see for yourself:
0,1425,i=219313,00.gif


Now, getting an extra 4.5% speed doesnt make alot of sense to me. The SSD's can operate at nearly the speed of the RAID controller... simply put, you cant speed up the speed of light by adding more stages between the drive and the bus that it operates on. That is why the memory controllers have been moved from the northbridge to the CPU itself. Until the southbridge gets boosted (or also moved onto the CPU), and RAID processors get a serious boost in speed (right now, hardware controllers arent fast enough to handle SSD's anyways), SSD's will spell the death of RAID. Actually, about a year ago I made a prediction that SSD's will spell the death of RAM as well. With on-die cache levels getting bigger and bigger, and SSD's which will quickly approach the speed of RAM itself, there will be no use for anything to operate between these two stages.

And most here arent even operating hardware (faster) RAID. With software raid, the controller for the data transfer is the CPU itself, so the bus that the controller runs on is not just running PCI-E or SATA information, but now also RAID controller data back and forth as well... choking the transfer lanes. As you can see from above, adding RAID can actually bog some of these drives down.
 
Ughh, SSDs are no where near apporaoching the speeds of volatile RAM, especially if you consider RAM is also getting faster.
 
OSAR! It has taken how long for RAM to get how fast? How many years does it take for RAM to double speed? And in contrast to that, it has taken how long for SSD's to get as fast as they are? SSD's have doubled in speed in a year, and the white papers I have seen suggest they could push the SATA3 spec to its limits within a year. SSD's might not be anywhere near as fast RIGHT NOW, but the rate that they are getting faster is many times that of RAM. Its just a matter of time. The thread where I first mentioned RAM going adios was something like 'what do you predict in the next 10 years?', so Im not talking next year or something.
 
Most motherboards use software RAID. There are some exceptions such as the ASUS P5W-DH Deluxe which had one real hardware RAID controller. It used an integrated Silicon Image 4723 RAID controller. The 4723 was designed for motherboard and OEM use and was seriously stripped down in terms of features. It supported only two devices in RAID 0 or RAID 1 as well as JBOD spanning. It was also problematic to make function properly at times. This was normally due to the fact that the drivers needed unspecified Windows updates for Windows XP in order to function properly. The ICH7R and other JMicron controllers were of course software RAID as is typical of integrated motherboard solutions.

This is some of the information I found on the controller chip. It is a XOR processor, but obviously not in the same league as the ones typically found in higher end Areca or LSI controllers.
 
Here's the proof I was waiting for, as to why RAID is going to go adios:
http://www.fudzilla.com/index.php?option=com_content&task=view&id=11365&Itemid=1

"The drive is insanely fast as this internal quad raid drive can write at up to 480MB/s and read around 550MB/s, depending on the size of the files."

Because you no longer have physical platter restraints as far as size and control complexity... whatever function RAID serves can simply be used inside the drive itself to boost speeds.
 
Here's the proof I was waiting for, as to why RAID is going to go adios

Uh, except for what (true) RAID is intended to do: increase the uptime of the array. I don't know if SSD's experience the same failure as mechanical HDDs do, but applications which have a critical need for uptime use actual raid (read: not RAID 0) to do so and that is not going adios soon.
 
SSD's are transistor based, solid state... like RAM, and made from the same components as your actual RAID or SATA controllers, so by definition, they are going to have about the same reliability as the actual raid controller and other parts of your computer that also have no moving parts. The whole idea is that the striping/copying technology that RAID uses can migrate from the board to the drive. This greatly improves functionality and reliability since it HAS to work with the drive, and isnt subject to the more finicky nature of being controlled by the chipset/cpu/bios. By placing the RAID on the drive itself, you are removing any bottlenecks as well imposed by the SATA connections... you have direct access (like cache vs. system RAM). The only bottleneck is the connection from the drive itself to the board. I 'bet' you will see the SATA 4.0 spec or some other interface being talked about soon as a result, or even direct mounting of the drives right into the PCIx16 slots. SSD's in theory can catch up with RAM in speed very quickly (this combined with larger and larger on-die CPU caches is why I say RAM in its current form will just go adios as well in 8-10 years and the cpu will talk directly with the SSD's). I bet in 5 years platter drives are no longer even mainstream.

As for reliability, that can be built into the on-disk controller... perhaps some SSD's will adopt 'secirity/stability' over speed by putting the raid controller on the disk. But with SSD's, this shouldnt even be a big deal... you are eliminating the possibility of mechanical failure, which is the biggest problem with current HDD's. This leaves you with internal shorts or magnetic fields, neither of which RAID is going to do alot for you anyways should they happen.
 
SSD's are transistor based, solid state... like RAM, and made from the same components as your actual RAID or SATA controllers, so by definition, they are going to have about the same reliability as the actual raid controller and other parts of your computer that also have no moving parts. The whole idea is that the striping/copying technology that RAID uses can migrate from the board to the drive. This greatly improves functionality and reliability since it HAS to work with the drive, and isnt subject to the more finicky nature of being controlled by the chipset/cpu/bios. By placing the RAID on the drive itself, you are removing any bottlenecks as well imposed by the SATA connections... you have direct access (like cache vs. system RAM). The only bottleneck is the connection from the drive itself to the board. I 'bet' you will see the SATA 4.0 spec or some other interface being talked about soon as a result, or even direct mounting of the drives right into the PCIx16 slots. SSD's in theory can catch up with RAM in speed very quickly (this combined with larger and larger on-die CPU caches is why I say RAM in its current form will just go adios as well in 8-10 years and the cpu will talk directly with the SSD's). I bet in 5 years platter drives are no longer even mainstream.

As for reliability, that can be built into the on-disk controller... perhaps some SSD's will adopt 'secirity/stability' over speed by putting the raid controller on the disk. But with SSD's, this shouldnt even be a big deal... you are eliminating the possibility of mechanical failure, which is the biggest problem with current HDD's. This leaves you with internal shorts or magnetic fields, neither of which RAID is going to do alot for you anyways should they happen.
 
http://enthusiast.hardocp.com/article.html?art=MTYxMSwyLCxoZW50aHVzaWFzdA==

"Q. How well does RAID-0 work with SSDs? Can RAID-0 be done inside the drive?

A. Let me answer the second question first and point out that SSDs are already massively parallel. The current Intel SSDs use 10 parallel channels to access the flash media, so in a way, it's already a 10-way RAID-0 on the drive.


Doing RAID-0 with multiple SSDs works very well, but make sure your RAID card is up to the task. SSDs are individually capable of enormous amounts of throughput, potentially much more than the RAID controller was designed to handle."
 
i would never use onboard mobo raid. never.

well i did in a laptop but that was just because :D
 
Back
Top