Drobo S, eSata, Xserve G5 w/ PCI/PCI-X only

I’ve nearly exhausted my search for compatible eSata cards for an early model Xserve G5 and eSata cards. The approved Sonnet Card is technically compatible, but doesn’t fit due to the foot in front of the pci connectors; the Wiebetech card isn’t supported and doesn’t work. I’ve now found the Highpoint RocketRAID cards, the 1522A and the 1742. Has anyone used either of these cards successfully with a Drobo on an Xserve G5? And since when did eSata stop being eSata?

Due to the foot in front of the PCI connectors? That doesn’t sound right… unless by foot you mean there’s a component on the motherboard (and not the card) that is blocking the installation.

You don’t happen to have pictures, do you?

From the wisdom of others, Drobo needs not only eSATA, but eSATA with support for multiple LUNs. This entails having port-multiplier support on the eSATA controller, and not all controllers have that.

you can see a picture of the approved card here: http://sonnettech.com/product/images/Tsatae2p_0509.jpg

The “foot” I’m speaking of is the little tab in front of the PCI connectors. It physically will not fit in to an Xserve.

How can you tell if a card has multiple LUN support if it’s otherwise not listed?


That’s NOT a PCI/PCI-X card. That’s a PCI Express card.

The foot is there to prevent you from accidentally plugging it into a PCI (Conventional) or PCI-X slot.
The foot did its job well. :slight_smile:

Those RocketRAID cards are probably overkill - unless you’re planning to RAID-0 your Drobos. Otherwise it’s just extra money for stuff you’re not going to use.

Some info on Port-Multipliers

Silicon Image has a Compatibility List for their port multiplier products.

The SiI3132 (often typoed SiL3132) chip is quite common and supports port multipliers, but that’s a PCI Express (PCIe) controller. The SiI3124 also supports port multipliers, and that one is the PCI version.

Since you have PCI-X slots, I would definitely go with a PCI-X card versus a standard PCI card, as PCI-X has higher much higher bandwidth capability and will help ensure you don’t have a bottleneck at the storage controller.

Here’s a PCI-X board based on the SiI3124 that supports port multiplier

Please note that I cannot guarantee it will work, but the DRI-recommended cards (at least the ones that I could tell) seem to all be SiI3132 cards, so it should work since the 3132 shares similar driver structure to the 3124.

Hope that helps!

Whoever in their right mind named both versions of PCI E and (E)XPRESS should be shot.

Hehehe, yes…

PCI and PCI-X are the same “class” with PCI-X being an “eXtension” of PCI, literally - the base slots are the same, but PCI-X has “extra” slot. Hence the ability to plug a PCI board into a PCI-X slot (except in the case when the PCI board is 5-volt only, since PCI-X supports only 3.3-volt signaling).

PCI Express (PCIe, PCI-e), on the other hand, is completely different animal.

You can blame the PCI SIG :slight_smile:

Let me rephrase that. Whoever in their right mind named 3 versions of PCI, PCI-X and PCI e(x)press should be shot. And doubly so for the guy who said let’s change the voltage but keep the slot the same. I’m pretty sure Apple falls in that category for iPhone/iPod connector, and the MacBook Air SuperDrive, and probably a whole lot more.

We’re a little OT, but I’m gonna add a last bit of info, as it might be helpful for those with PCI in the future.

PCI spec v2.1 (or maybe v2.2) got rid of 5-volt signaling. Thus, you will find the slots look different… Most PCI cards nowadays support both voltages, so you’ll see two “gaps” in the card-edge (one more if it’s 64-bit). On the slot side, there’s only one “hump” (2 if it’s 64-bit) as the slot itself only supports one voltage.

Slide 7 of this PowerPoint presentation illustrates it well for 32-bit and 64-bit PCI cards and slots.

And I completely agree - it wasn’t very nice to not only remove one voltage, but to name the next-generation stuff in a confusing manner. :frowning: