HP ProLiant SE1101 PCI Express x8

[mturner-dc]

Hello all,  I have roughly 50 ProLiant SE1101 servers and would like to be able to make use of the SAS1064E cards that came in them.  Unfortunately they will not run at SATA2 speed.  This is because the cards are in an x4 slot and upon investigation I found that the servers have 2 x8 slots.  Whenever I put my RAID card in the x8 slot (PERC 5i and LSI 1064E) they are unrecognized and therefor unusable.   

I have searched high and low for information on these mchines and all I have found is a few ebay auctions.

Anyone here have any ideas?

[j4sonxp]

I am also having the same issue where my PCI x8 slots are not recognizing my ethernet NIC card. I've tried searching for the correct ROM images but no luck.

[Casper42]

You need to get help from whereever you bought them from unfortunately.

SE = Special Edition and are custom one off creations for large customers. I know for instance Yahoo had some SE model(s) created for them a few years back. Not saying your model is a Yahoo one though.

Anyway, the only people who can technically get Updates for those machines would be the customer they were created for.

Likely what happened is the large customer who had them created retired them and the company they use to scrap equipment sold them in a batch and then you bought them on eBay or similar from the person who bought the batch.

For drivers you can usually find other models with similar cards/features and make things work, but I think BIOS you may be out of luck.

[Sbrown]

PCI-E 1.0 !! Not all cards are happy about that.

Qlogic 8152 is fine with dual 10gbe @ x8 but some cards are not very happy!

Mturner-DC: Those machines are so old, the power consumption is not worth it. ebay search : L5639 DL180 G6 and see how awesome that deal is!

[Observer2018]

That is not how PCI-E or SATA works.  Your drives will connect at whatever SATA speed standards the card supports. for connecting to drives (SATA 1.0 = 1.5 Gbps, SATA 2 = 3Gbps, or SATA 3 = 6Gbps).  Getting data off or onto the card is an independent operation by PCI-E bus using the maximum combined bandwidth of all available real slot lines. But shortage of PCI-E bandwidth (lanes) can make communication with drives pause when onboard buffers fill up. 

Quite often only certain combinations of PCI-E slots can operate at the same time. Extra slots are there merely to present the same lanes in a different mechancial intereface option. Furthermore individual PCI-E slots quite often do not include all lanes the connector supports, but instead provide more mechanical support or allow high end cards to operate at lower bus speeds. So x8 slots might have only x4 real lanes. PCI-E version changes the speed of individual PCi-E lines in a way similar to SATA version increments.

 I will note that x4 PCI-E 1.0 slot actually provides 8Gbps bandwidth such that when talking to 1-2 SATA 2.0 drives that there is sufficient bandwidth for uninterrupted max performance. That is assuming non-RAID, RAID 0 or 1 operations and that the CPU can manage all striping "calculations" if software RAID 1 is used. So really only higher RAID options talking to many dirves at once are slowed.  Even then disk operations may not be slowed at the card and bus. If your advanced RAID is software driven, the load of RAID on the CPU quite often throttles disk I/O back so far that the card itself can keep up with what has become a CPU limited I/O process.

[Observer2018]

It is true that SATA board drivers might balk at version 1.0 PCI-E.  But in general backward compatibility is pretty good. Again I suggest you look more closely at how you use the SATA cards.  Do you have hardware RAID drivers or is RAID software driven. If sofware driven, then CPU splitting data into strips and calculating data redundancy error codes could well mean the bottleneck is these old CPUs. 

 If no RAID - well a x4 PCI-E 1.0 card is more than fast enough to keep up (almost 3 SATA2 drives at once continuouosly)  - unless you got a ton of simple VMs doing max I/O with little processing of data. Which is why I suggest you confirm where the bottleneck really is.

[Observer2018]

FYI - on separate issue with SE1101 servers ...memory slots.  There is a reason that why many SE1101 will only accept 4 memory sticks -- the dual CPUs have only 2 memory channels each. 16GB limit.

No docs still online.  But I think one memory slot is wired to each channel of each CPU. That is I do not think this server has an interleaved mode for the memory banks.  

 So to use all 6 slots you need a CPU that supports 3-memory channels. In general look for 56xx series CPUs. Actually I think you may need dual 56xx series.   24GB limit