SATA Controller Cards


Recommended Posts

Okay - we have several great / cheap sources for PCIe x1 2 port controller cards.

 

Does anyone have a source for cheap 2 (or 4) port PCI cards that are compatible with unRAID.  I need 2 more ports.

 

For PCI it's hard to get much cheaper than monoprice.  I know that 4 port card works. 

 

Yep, I've used the 2 port PCI SIL3132 as well.  Works just fine.

Link to comment

Okay - we have several great / cheap sources for PCIe x1 2 port controller cards.

 

Does anyone have a source for cheap 2 (or 4) port PCI cards that are compatible with unRAID.  I need 2 more ports.

 

For PCI it's hard to get much cheaper than monoprice.  I know that 4 port card works. 

 

Thanks - I ordered one.  Do I need to update the firmware?

Link to comment

HighPoint RocketRAID 2310 PCI Express x4 (x8 and x16 slot compatible) SATA II (3.0Gb/s) Controller Card, 4 SATA Ports

 

Requirements: PCIe x4 or faster slot

 

Cables: Standard SATA (1 port per cable)

 

Where to get them: http://www.newegg.ca/Product/Product.aspx?Item=N82E16816115027&cm_re=rocketraid-_-16-115-027-_-Product

 

Special Instructions: Need to disable "INT13" and "Reallocate EBDA" by booting to DOS disk and flashing with latest BIOS. Only when Flashing do you get the option to change these settings. Afterwards, work fine. You can also simply hit the END key when the cards BIOS screen comes up, but have to do that each time.

 

Thanks!  Added to the first post.

- Raj

Just a side note that I didn't have to flash the BIOS on RR2310 and disable "INT13"/"Reallocate EBDA" to get the card working. Both settings are related to booting so if you have problems on that area, then you should flash and disable those settings.

 

You might also discover an alarming warning message in your syslog about data corruption. Back in 2009 I investigated this and got a nothing-to-worry-about message from Highpoint. More details here http://lime-technology.com/forum/index.php?topic=3896.0

 

This was ~2 years ago and the card has been running rock solid since. So though pricey, I can recommend this card if cheaper alternatives are not available.

Link to comment

Okay - we have several great / cheap sources for PCIe x1 2 port controller cards.

 

Does anyone have a source for cheap 2 (or 4) port PCI cards that are compatible with unRAID.  I need 2 more ports.

 

For PCI it's hard to get much cheaper than monoprice.  I know that 4 port card works. 

 

Thanks - I ordered one.  Do I need to update the firmware?

 

No, you shouldn't have to.  I didn't.

 

I'm hoping to receive a SuperMicro AOC-USAS2-L8i (LSI2008-based, both IT and IR firmware available from SM) soon, so I look forward to posting about my experience with that.

 

And we look forward to reading about it!

Link to comment

Okay - we have several great / cheap sources for PCIe x1 2 port controller cards.

 

Does anyone have a source for cheap 2 (or 4) port PCI cards that are compatible with unRAID.  I need 2 more ports.

 

For PCI it's hard to get much cheaper than monoprice.  I know that 4 port card works. 

 

The SIL3114 is used on lots of inexpensive 4 port PCI cards - plenty on Ebay as well for European users who can't access Newegg/Monoprice, etc.  Just search on Ebay for "SIL3114"

Link to comment

Okay - we have several great / cheap sources for PCIe x1 2 port controller cards.

 

Does anyone have a source for cheap 2 (or 4) port PCI cards that are compatible with unRAID.  I need 2 more ports.

 

For PCI it's hard to get much cheaper than monoprice.  I know that 4 port card works.  

 

The SIL3114 is used on lots of inexpensive 4 port PCI cards - plenty on Ebay as well for European users who can't access Newegg/Monoprice, etc.  Just search on Ebay for "SIL3114"

 

These guys do the SIL3114 PCI cards for a good price (UK): http://linitx.com/viewproduct.php?prodid=11624

 

I've bought a couple from them.

Link to comment

About those 2-port JMB362 Cards - there have been 3 instances of these boards resulting in no POST lately. In at least two of the instances this has involved motherboards that already have a JMB 36x chip onboard. See this thread for details:

 

http://lime-technology.com/forum/index.php?topic=10165.60

 

Hopefully more people will write back about the cards in the above threads and hopefully there will be a pattern as to which motherboards work with these and which ones don't. In the mean time, I feel a caution and link to the above thread would be in order.

Link to comment

About those 2-port JMB362 Cards - there have been 3 instances of these boards resulting in no POST lately. In at least two of the instances this has involved motherboards that already have a JMB 36x chip onboard. See this thread for details:

 

http://lime-technology.com/forum/index.php?topic=10165.60

 

Hopefully more people will write back about the cards in the above threads and hopefully there will be a pattern as to which motherboards work with these and which ones don't. In the mean time, I feel a caution and link to the above thread would be in order.

 

Agreed. If it's not due to another onboard controller then it appears some newer LGA1366 boards are having issues with them. But then you are having issues on an older P35 chipset board too. So far, no reports of AMD systems having trouble with it.

 

Peter

 

Link to comment

Okay - we have several great / cheap sources for PCIe x1 2 port controller cards.

 

Does anyone have a source for cheap 2 (or 4) port PCI cards that are compatible with unRAID.  I need 2 more ports.

 

For PCI it's hard to get much cheaper than monoprice.  I know that 4 port card works.  

 

yea, that monoprice card works great for me, but a small FYI to those considering it -- it's SATA 1.0 -- a little slow.  I've got to jiggle around my drives so my smaller ones are on that card, so that a pairity check, etc gets done faster once it's past the 4 SATA 1.0 connections and moves over to just my SATA II's on the MoBo (six) -- the recommendation received was move my small drives to the SATA 1 connections, and the big drives to SATA II.

 

And it may be the case that being an old-fashioned PCI card that SATA 1.0 is just fine...   but if anyone has a lead on a similar, cheap 4-port PCI card that is SATA II ... drop a line here ;)

Link to comment

Good idea, warning added.

 

Perhaps another tweak?  SATA I, II, or III ports?   

 

As I understand it, today's HDDs don't even saturate SATA I, so SATA II or III is basically just marketing hype.  For an SSD application it certainly matters, but it shouldn't matter for a HDD application like unRAID.

Link to comment

Good idea, warning added.

 

Perhaps another tweak?  SATA I, II, or III ports?   

 

As I understand it, today's HDDs don't even saturate SATA I, so SATA II or III is basically just marketing hype.  For an SSD application it certainly matters, but it shouldn't matter for a HDD application like unRAID.

 

So far, I can see a difference between SATA I and II in performance...  I'm not quite OCD to graph it out though.    Especially if I'm doing port multiplying.... {of course}.  And my ancient SSD (ancient by modern standards) is an OCZ Agility which doesn't benefit from SATA III,  but certainly does for SATA II vs. SATA I.

 

But even internally, one-to-one, I've learned to move my "smaller" drives to the SATA I connections so that parity checks get done with the SATA I connections as soon as they can.

Link to comment

Although our current disks don't saturate the theoretical SATA I speed, it is plausible that the communication speed could be lower on SATA II for two reasons: a) practical speed on SATA I may be lower than the theoretical, and b) the faster link will have lower latency. That being said, I would expect the difference to be fairly minor - small enouth to probably not need to worry about. But I do like raw data to prove me wrong  ;)

Link to comment
  • 2 weeks later...

I have a question about these SATA expansion / controller cards with regard to PCI-e slots and bandwidth. I notice that typically products will follow the below trend:

 

PCI-e x1 slot = 2 port expansion cards available

PCI-e x4 slot = 4 port expansion cards available (and some 8 port)

PCI-e x8 slot = 8 port + expansion cards available

 

I understand that this is to do with bandwidth of each slot and not producing products that would be bottlenecked by each slot. However, the above does not make sense to me…

 

Why do we not have for example 4 port cards available for PCI-e x1 slots? PCI-e version 2.0 x1 slots, have a maximum bandwidth of 500 MB/s. Taking the assumption that each drive attached to this slot could be reading/writing at a theoretical 125MB/s each (max possible over gigabit ethernet) then we have:

 

500 / 125 = 4 possible drives

 

So we could have 4 hard disk drives attached to a PCI-e 2.0 x1 slot if they manufactured a card that had 4 ports. Yet we only see cards with 2 ports on x1 slots.

 

I can understand this scenario if you take into account the cards are only manufactured for PCI-e version 1.0/1.1, as then the bandwidth is halved to 250MB/s, so the above example would check out and you could have exactly 2 HDDs attached before theoretically hitting a bottleneck. Is this all it is? That nobody has bothered making a PCI-e version 2.0 x1, 4 port card? Can I assume that the reason is due to lack of demand when people can use the already out there products for x4 slots?

 

Reason I am interested, is that my motherboard has 2 separate x1 slots (PCI-e version 2.0) and I am trying to plan for the future what the maximum amount of hard disk drives I could handle is. I have on my motherboard:

8 x SATA ports

1 x 16x pci-e slot (version 2.0)

2 x 1x pci-e slot (version 2.0)

 

I think I can have 8 + 8 + 2 + 2 = 20 which nicely maximises the current licence for Pro I believe.

 

Link to comment

Some of the wider controller cards will work in a narrower slot if the user physically removes (e.g. melts the back with a soldering iron) the plastic at the back of the slot to allow physical insertion of the card. I am pretty sure that the BR10i, is advertised to work in a x1, x4, or x8 slot.

Link to comment

It's been mentioned, that you can use an external/top mounted dock for pre-clearing for a number of reasons:

 

a) Easy to plug in--no dismantling your case/screwing a drive into a new bay.  Especially if the drive ends up being DOA.

b) If your array is full and looking to upgrade a drive to a larger/faster drive, you don't need to take your array offline to do so

c) Less risk of pre-clearing the wrong drive, as you can tell which one it is--based on it being outside of your array, so you can view the Disk Management screen to see what device slot it's loaded on..

 

May be others, but I think those are good enough reasons to pre-clear outside of your array--and outside of your case.

Link to comment

I received my Supermicro AOC-USAS2-L8i card today.

 

As a quick resume, this is an 8-port SAS/SATA card, using the LSI SAS 2008 controller chip.  It supports SATA3 6Gb/s transfers and has a PCIe 2.0 x8 host interface.  Actually, Supermicro call the interface 'UIO', but all this means is that the card is made the 'wrong way round' - the majority of the components and the backplate are on the opposite side to what they should be for a standard PCI card.  I actually see this as an advantage because, in a standard tower case, the heatsink is on the upper side of the card, where it can work effectively, rather than being on the underneath.  However, it does mean that the backplate has to be removed (and refitted with spacers, if you wish to support the card in the conventional way in a PCIe slot).  The other thing to note about this card is that it has RAID functionality build in (there is an AOC-USAS2-L8e card which is plain JBOD, but it is more expensive!).

 

Anyway, within minutes of arriving home with the card, I'd removed the backplate and plugged it into the PCIe x16 'video card slot' on my Intel mobo.  I'd already concluded that this bus slot is not a 'dedicated' video slot.

 

unRAID 5.0 beta6 booted straight up and ran with no problems.  The syslog file indicates that the card is recognised by the mpt2sas driver.  Log entries do suggest that the card is flashed with the RAID firmware (IR mode), but it still seems to be able to operate in JBOD (IT mode) - Protocol=(Initiator, Target) .

root@Tower:~# grep mpt2 /var/log/syslog
May 20 23:21:16 Tower kernel: mpt2sas version 06.100.00.00 loaded
May 20 23:21:16 Tower kernel: mpt2sas 0000:01:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
May 20 23:21:16 Tower kernel: mpt2sas 0000:01:00.0: setting latency timer to 64
May 20 23:21:16 Tower kernel: mpt2sas0: 64 BIT PCI BUS DMA ADDRESSING SUPPORTED, total mem (3942652 kB)
May 20 23:21:16 Tower kernel: mpt2sas 0000:01:00.0: irq 41 for MSI/MSI-X
May 20 23:21:16 Tower kernel: mpt2sas0: PCI-MSI-X enabled: IRQ 41
May 20 23:21:16 Tower kernel: mpt2sas0: iomem(0x00000000fe4c0000), mapped(0xf8558000), size(16384)
May 20 23:21:16 Tower kernel: mpt2sas0: ioport(0x000000000000e000), size(256)
May 20 23:21:16 Tower kernel: mpt2sas0: sending diag reset !!
May 20 23:21:16 Tower kernel: mpt2sas0: diag reset: SUCCESS
May 20 23:21:16 Tower kernel: mpt2sas0: Allocated physical memory: size(839 kB)
May 20 23:21:16 Tower kernel: mpt2sas0: Current Controller Queue Depth(339), Max Controller Queue Depth(2239)
May 20 23:21:16 Tower kernel: mpt2sas0: Scatter Gather Elements per IO(128)
May 20 23:21:16 Tower kernel: mpt2sas0: LSISAS2008: FWVersion(07.00.00.00), ChipRevision(0x02), BiosVersion(07.11.00.00)
May 20 23:21:16 Tower kernel: mpt2sas0: Protocol=(Initiator,Target), Capabilities=(Raid,TLR,EEDP,Snapshot Buffer,Diag Trace Buffer,Task Set Full,NCQ)
May 20 23:21:16 Tower kernel: mpt2sas0: sending port enable !!
May 20 23:21:16 Tower kernel: mpt2sas0: host_add: handle(0x0001), sas_addr(0x50030480041f4c00), phys(
May 20 23:21:16 Tower kernel: mpt2sas0: port enable: SUCCESS
root@Tower:~# 

 

I then shut unRAID down, and connected up a spare Samsung HD502HJ (512GB) drive.  I am in the middle of preclearing this drive and during Step 2 'Copying zeros to remainder of disk ....' it is reporting a rate in excess of 120MB/s (at 90% complete).

 

The only slight issue I've noted so far is that it is claimed not to be able to access SMART data.  I'm sure that I've read of this problem with other cards and that there is a work-around - I will research this tomorrow.

 

When the preclear has finished, I will try moving a couple of my array drives to the new card - if I understand correctly, unRAID 5b6 should be unfazed by this and should start and run as normal.

 

I've just noticed that the preclear has completed the ten steps and is now on the Post-Read, reporting 120+MB/s.

 

As far as I can tell, so far, the SM AOC-USAS2-L8i card is compatible with unRAID, with the slight proviso, that something needs to be tweaked to persuade it to give up the SMART data.

 

I will download the alternat firmwares from the SM site, and experiment to see whether there is any noticeable change from using the IT firmware.

 

If any one can think of other things I could, or should, test, then let me know.

 

To be continued ......

Link to comment

Thanks for the testing report, looks like a great card!  Where did you buy it, and how much did you pay for it?

 

By the way, the error about 'inability to access SMART data' may be the fault of the latest version of the preclear script, not the SATA card itself.  For example, I've seen the same error when preclearing certain drives on the Supermicro AOC-SASLP-MV8 card, which of course are very tried and true in unRAID by this point.  Even though I see that error when first kicking off the preclear, I still get a full SMART report at the end of the preclear cycles.  I expect that you may see the same.  I believe that Joe L. knows about this issue and is working on it.  In any case, if you see the same behavior that I do then I would blame the preclear script, not the card itself.  Of course, if you don't actually get a SMART report at the end of the preclear cycle, then that could actually indicate an issue with the card.

 

By the way, I recently wrote up a blog post on the GreenLeaf site about testing an older PCI-X card for unRAID compatibility:

 

http://www.greenleaf-technology.com/blogs/softwareandhardware/index.php

 

I encourage you to read through it and run all the same tests on this card.  If you think of any tests that I didn't run, by all means let me know.  Of course I don't expect that you would be able to test the card on as many different motherboards as I had available to me, but just running the tests on one motherboard should be enough.

Link to comment
  • JorgeB unpinned and locked this topic
Guest
This topic is now closed to further replies.