HP SAS Expander Owner's Thread

have many people used this with the SASLP-MV8 ?

im still a bit lost on how it works. would I plug a cable direct fromt he mv8 to the hp sas?

in theory could I use the mv8 and two of these sas in a single machine? realizing I would need to find a case(s) big enough to hold them all. heh

Shawn
 
treadstone said:
The HP expander cards obviously seem to use more than just the power connections from the PCIE connector.
Click to expand...

That would contradict the PCIe power board I have working just fine with the HP expander.

treadstone said:
I have an ASUS P7P55-WS-Supercomputer motherboard and populated the slots as follows:
<snip>

I liked this setup because it left a spare slot and enough room for cabling and air flow between the RAID controller and SAS expander cards. For some strange reason however the motherboard doesn't seem to let the SAS expander cards work properly in the PCIEX_2 and PCIEX_5 slots :(
I noticed after I moved the expander #2 into the PCIEX16_3 slot that the top most LED started to blink, where as before it would just be solid on!
Click to expand...

Those kinds of quirks is why I beg people to buy Supermicro rather than gaming motherboards when they're building big storage servers.
 
Last edited:
I wonder what J5 on the HP expander lets you do... It's an RS232 level connection:

pin 1: Ground
pin 2: TXD (data out from the card)
pin 3: RXD (data in to the card)

Might actually be used to do firmware upgrades without a controller?!?
 
It is possible. The Areca cards have a serial port for the expander as well. I'll probably build a cable when I have time to test it out. As for the server boards, I'm doing quite fine with my Asus P5BV-C/4L. I prefer Supermicro, but it has 4 gigabit ethernet (which works really well) and I got it for an excellent price ($50).
 
odditory said:
That would contradict the PCIe power board I have working just fine with the HP expander.



those kinds of quirks is why I beg people to buy Supermicro rather than gaming motherboards when they're building big storage servers.
Click to expand...

Just because the card is working on power only doesn't mean that it doesn't have any other connections to the PCIE connector. From what I can tell so far, it looks like the expander actually uses the SMCLK and SMDAT lines and the PWRGD signal.

Except I couldn't find a single useful Supermicro motherboard that had all the connections I wanted... plus there is no guarantee that stuff like what I just went through wouldn't happen with a server board too!
 
Last edited:
sure there's no guarantee but based on the experience i've had in testing it both ways, i'd say there's a lower likelihood of issues with a SM server board. i've tested the HP expander in every single PCIe slot on several SM boards including the X7DWE, X7DWN, X8SIL and couldn't find a single slot it DIDN'T work on.

conversely i've tested on 5 or 6 other 'regular' and desktop class motherboards -- even an old HP mobo from an xw8600 workstation -- and the HP expander did not work on quite a few of them, so agreed that there's something more at play here.
 
odditory said:
conversely i've tested on 5 or 6 other 'regular' and desktop class motherboards -- even an old HP mobo from an xw8600 workstation -- and the HP expander did not work on quite a few of them, so agreed that there's something more at play here.
Click to expand...

I know that a lot of the newer boards have some logic to switch the slots around depending on the configuration. Since there are only a certain number of total lanes available, if, for example, you plug an x16 card into one slot, it may disable another slot or reduce the number of link lanes on another slot.

Anyone know a PCIe expert? I was wondering if it is possible that, either through buggy logic or intentionally, certain configurations may turn off the power lines on other slots that the HP SAS Expander needs.

In treadstone's case, it seems some of the power was on, but maybe not all the lanes? What if the slot got reduced to x1 link for some reason?
 
Extra lanes do not constitute extra power (they are used for data only). The card merely takes power from the PCIe slot and does not utilize the data bus at all.
 
Blue Fox said:
Extra lanes do not constitute extra power (they are used for data only). The card merely takes power from the PCIe slot and does not utilize the data bus at all.
Click to expand...

Right but something is definitely missing here.
 
Well, the fact that it works in that industrial backplane on its own kinda says something. There are no lanes linked to anything considering the CPU board is missing.
 
I don't think that the expander cards actually use any of the data links (lanes) or the clock connection for that matter. However there are some additional connections that seem to interfere with the operation of the card when plugged into one of those slots I tried it in.

Again, I'm sure that the card would work just fine if you connected only power to it and not any of those other signals. I assume that what happens is that one of the other signals is causing the processor on the expander not to start up. I'm going to see if I can run a test with two of the signals disconnected and see what happens....
 
FYI, here is a page from that MB manual. It looks like if the black slot(s) are occupied, the corresponding blue slot(s) gets reduced from x16 to x8.

But no mention of a slot being disabled, so I guess treadstone is right that there is something on one or more of the pins that is shutting the HP SAS expander down.

screenshotow.png
 
That's the MB I used.

From the left:

Blue x16 slot - Video Card
White x1 slot - Not used
Black x8 slot - Expander #1
White x4 slot - Not used
Blue x16 slot - RAID controller
White PCI slot - Not used
Black x8 slot - Expander #2

When I moved the expander from the black slot on the left to the white x4 slot the expander worked!
 
treadstone said:
That's the MB I used.

From the left:

Blue x16 slot - Video Card
White x1 slot - Not used
Black x8 slot - Expander #1
White x4 slot - Not used
Blue x16 slot - RAID controller
White PCI slot - Not used
Black x8 slot - Expander #2

When I moved the expander from the black slot on the left to the white x4 slot the expander worked!
Click to expand...

So, to add a detail to your theory, maybe something on lanes 5-8 from the black slot caused the HP SAS Expander to stop responding.

Does the HP SAS expander even have pins for lanes 5-8? If so, can you tape them over?
 
The edge connector does have all the pins and short traces on all of the x8 lanes and the clock pair, however as far as I can tell, they aren't connected to anything. The traces end in test point/probe connections.
 
GOT IT WORKING!!! :D :D :D

In the black slots too :)

Finally figured it out what caused them not to work :D

Have to modify both expander cards but it's a super easy fix :D
 
Even though you got it working, you really made a poor motherboard choice for a large file server to be honest.
 
It's only large in the sense that it has lots of storage capability, other than that, I do not really consider this large.

It's only got 4GB of memory, no high performance network connections. Don't need high performance as I need storage capacity not throughput.
 
There is a PWRGD (Power Good) signal on the PCIe connector Side A pin 11 (A11) that is used by the expanders processor (most likely as a #RESET signal). In the two black slots this signal is always low (not sure if it's hardwired that way or simply kept low until the PnP function releases the signal). For my tests I used some Kapton tape to cover the signal pin on the edge connector. Once the PWRGD signal was floating, the processor started working and I was able to see the drives in the RAID controllers BIOS screen. Now my plan was to simply disconnect (cut the trace) just above the A11 pin, however I just found out that the expander board does not pull this signal high on it's own. Now leaving a reset signal floating is a BAD thing, so I need to add a pull up resistor somewhere on this board.
 
treadstone said:
There is a PWRGD (Power Good) signal on the PCIe connector Side A pin 11 (A11) that is used by the expanders processor (most likely as a #RESET signal). In the two black slots this signal is always low (not sure if it's hardwired that way or simply kept low until the PnP function releases the signal). For my tests I used some Kapton tape to cover the signal pin on the edge connector. Once the PWRGD signal was floating, the processor started working and I was able to see the drives in the RAID controllers BIOS screen. Now my plan was to simply disconnect (cut the trace) just above the A11 pin, however I just found out that the expander board does not pull this signal high on it's own. Now leaving a reset signal floating is a BAD thing, so I need to add a pull up resistor somewhere on this board.
Click to expand...

I'm missing something. Why would that pin be pulled high in the x4 link slot but pulled low in the x8 link slots?
 
Well, that's what I was wondering too, so I did some more testing and here is what I found out:

The PWRGD signal in the black and blue slots are by default low and on the white connector it goes high (3.3V) as soon as the board powers up. If no card is plugged into the black or blue slot, the PWRGD signal will stay low. If a card is plugged in and the BIOS has finished it's scan and figured out what kind of card it is and what resources it requires, it will switch the PWRGD signal high for that particular slot. If a reset is issued from the motherboard, the PWRGD signal is pulled low momentarily to reset whatever is on the I/O card. So I guess it functions as a power good as well as reset signal.

Now I could simply disconnect the PWRGD line on the expander from the connector and pull it up via a resistor, but to be able to reset the board, the entire server would have to be powered down versus simply reset (by pushing the reset button). Not ideal but how often do you really need to reset a server :)

Now I assume the reason that the expander cards are not being released out of their reset state is because the BIOS is unable to get any or the proper information from the expander card during the PnP setup phase. I assume that's what they use the SMBus (an I2C type of interface) for.
 
Did you try the expanders in the blue slots?

The reason I ask is that the motherboard obviously tries to detect whether there is a card in the black slots so that it can reduce the blue slots from x16 to x8. Maybe that detection is the reason for keeping that pin low, and maybe it is different in the blue slots.
 
I did try the expander cards in the blue slots without any luck.

I don't think that it has anything to do with how many lanes the mother board would allocate to each slot.
 
I think I found the solution to MY problem.

However I can NOT and do NOT recommend this to others. I am only listing this here for reference to those that have the knowledge, ability and tools to perform this kind of modification!

Apparently, there are different stuffing options on the Asus P7P55-WS-SC motherboard for the power good signal next to each of the PCIe connectors. Currently there is a jumper resistor that connects the power good pin of each slot to a dedicated (programmable) power good signal. Right next to each of those resistors is an unpopulated spot for a jumper resistor that would connect the power good pin to a global power good signal. So all I need to do is move the resistors for those two black connectors over and I should be all set. Again, I do NOT recommend this to anyone unless you have experience with soldering VERY small components and you have the appropriate tools for the job. Those components are 0402 type 0Ohm resistors (aka jumper resistors). I already verified that the signal I want to move the resistors to is indeed the correct signal and will work as power good signal.
 
nice work but clearly beyond what 99+% of people reading here would care to do, which brings us back to getting the right motherboard to begin with.

on the bright side it helps explain the mystery of some motherboards people tried earlier in the thread that the expander wasn't working on. thanks for the info, look forward to seeing your build thread!
 
SUCCESS!!!! :D

Modified the motherboard and put the server back together the way I had originally planned on setting it up. Booted into the RAID BIOS and can finally see all 48 2TB drives!!!

The SGPIO still doesn't work through the expander cards but I'm pretty sure that's because the SGPIO signals don't correspond between the controller and the expander cards. I think I'm going to butcher a SFF8087 cable tomorrow and swap them from a controller-to-backplane to a controller-to-controller pinout configuration and give that a try.

Now the next question is going to be on how to setup that RAID system...
 
Do you have your HP SAS Expanders cascaded/daisy-chained, or do you have one HP connected to each port on the 1680i? If you haven't tried cascading the HPs, would you give it a try and let me know if it works?
 
odditory said:
on the bright side it helps explain the mystery of some motherboards people tried earlier in the thread that the expander wasn't working on. thanks for the info, look forward to seeing your build thread!
Click to expand...

How about adding a "compatible motherboards" and "incompatible motherboards" list to the OP? You now have at least one for the incompatible list, and I am sure you have many for the compatible list.
 
john4200 said:
Do you have your HP SAS Expanders cascaded/daisy-chained, or do you have one HP connected to each port on the 1680i? If you haven't tried cascading the HPs, would you give it a try and let me know if it works?
Click to expand...

I have one expander per controller port. No need for me to daisy chain them.

I may be able to try the daisy chain setup tomorrow. Since I am not quite done with the physical setup anyway.
 
Could someone clear this up for me?

I've been reading like crazy. but. bah lol

Okay so this sas expander is like a multiplier right? I'm trying to wrap my head around this.

It has 9 ports. each of which is for 4 hdd's? correct? I forget what the name of the cables are. but you would use a 1>4 breakout on a 4020 case or a direct cable on a 4220 case right?

If I picked up the MV8 card
Essentially 8 ports across two of those cables.

I use one of the cable ports (4 drives) and plug it into one of the ports on the SAS Expander?

So essentially leaving me with room to hook up 8*4 hdd's on a single sas?

So I could in theory hook two of these expanders to a single MV8 and get 16*4 hdd's in a single setup?

I think im catching on. but do you need to use all 4 to expand?

Could someone "SAS Expanders for Dummies" this for me? I'm trying. but I'm a bit behind on the catch for these.

Thanks

Shawn
 
unless u run the external back in. or to another case.

so the sas/mv8 hookup would use a port on each card though? thats my main point of confusion

Thanks
 
Yes. You need one SFF8087-SFF8087 cable to connect MV8 to HP SAS Expander, consuming one x4 port on each card.
 
okay thanks. thats what i gathered but didnt want to coninue with the wrong info.

And for daisy chaining the expanders. you could in theory use the external ports on each expander and have two of them hooked to a single port on the MV8. giving you ideally
7+ internal ports allowing 4 drives each. and ideally do the same for the other port on the mv8. givings you double that even.

Kinda leaves the door wide open for some pretty kck ass expansion.

Thanks

Shawn
 
That looks interesting. How to make sure it works?

izx said:
Amfeltec, a small Canadian outfit, makes PCI-E x1 "risers" for debugging. The riser has two boards, the "interface" which goes into the host and the "adapter" which holds the PCI-E card. They're connected by a flex cable.

Of interest to us is the adapter, which can be powered by a Molex floppy connector:

x1pcie-adptr-atxpw.jpg


Assuming you don't mind using the enclosure power supply's switch for on/off (which it seems the Orbitmicro board has problems with anyway), AND you can/will dremel the x1 slot's edge to make it open ended, it could be a good alternative to power the HP expander. They will sell just the adapter board for $22, but shipping via mail is a bit steep at $15.

Any folks here who'd like to try one? Could do a mini-group-buy to amortize the shipping :)
Click to expand...
 
You must log in or register to reply here.
Back
Top