How about some Bay Trail?

ant

New Member
Jul 16, 2013
21
5
3
Yeah I think you are right. For myself, I have not bothered looking at these processor errata documents before. I'm probably just researching too much.

I have a E5-1620 and I just skimmed through that document and there is a huge amount of errata - most of which I don't understand. My E5-1620 server is running awesomely though.
 

Cheddoleum

Member
Feb 19, 2014
97
22
8
Well, it's a good find, and the fact that they'll be fixing the USB timing issues in the next stepping indicates they take that one pretty seriously.
 

NotMine999

New Member
Mar 7, 2014
7
1
3
Here is my promised follow-up to my own testing of the Supermicro X10SBA board. My testbed is Gentoo Linux x86_64, fully updated, so that means 3.12.13 kernel ("gentoo-sources", not "vanilla"). Since I am building a home storage server, video output better than "no D" CLI is not important to me; 2D and 3D acceleration have no value in my setup. I have 4x WDC 750G Black laptop drives already built as a ZFS "pool". Gentoo has the ZFS & SPL code compiled as modules, but I boot the system from a separate HDD with EXT4 partititions. I "load" the ZFS kernel module at boot but do not automount the ZFS pool; this is "testing" after all.

(1) Video can be "picky".
If HDMI is plugged in when the system boots, there is a good chance that the VGA port will not output anything. If I plan to use VGA I have to keep the HDMI port disconnected.

(2) USB can be "picky".
I saw strange USB keyboard issues in Gentoo 3.10.x but cannot reproduce them in Gentoo 3.12 using the same hardware. So assume a kernel issue.

(3) PCIe x2 slot is "picky".
I still suspect the onboard Marvell BIOS for the onboard 9230 chip (uses the white SATA plugs) does not work with other Marvell storage controllers. I have tweaked the various BIOS settings for hours and outright "locked up the system" in certain cases, "locked up the system boot process" in other cases, and gotten the system to boot in other cases.

SYBA PEX40064, PCIe x1, 4 port, SATA3, non-RAID, Marvell 88SE9215 chipset
- Can be seen in "lspci"...when NO HDD are attached to the storage controller :confused:
- "lspci" reports the controller using the "ahci" driver (fine with me...and matches SATA setup in the BIOS, but could also be coincidence)
- Controller does not have any jumpers.
- In other systems it displays a "splash" during boot but under UEFI boot on this board it does not.

SYBA PEX40062, PCIe x2, 4 port, SATA3, non-RAID, Marvell 88SE9235 chipset
- Can be seen in "lspci"...when NO HDD are attached to the storage controller :confused:
- "lspci" reports the controller using the "ahci" driver (fine with me...and matches SATA setup in the BIOS, but could also be coincidence)
- Controller does not have any jumpers.
- Have not booted this card in any other system...yet.

SYBA PEX40039, PCIe x1, 2 port, SATA3, non-RAID, ASmedia ASM1061 chipset
- Can be seen in "lspci"...with hard drives attached (per "blkid" command) :D
- "lspci" reports the controller using the "ahci" driver (fine with me...and matches SATA setup in the BIOS, but could also be coincidence)
- Controller does not have any jumpers.
- Have not booted this card in any other system...yet.
- Drives attached to this conntroller can be included in a ZFS pool without errors

SYBA MPE40056, mini PCIe card, 2 port, non-RAID, ASmedia ASM1061 chipset
- Cannot be seen in "lspci"...with or without HDD attached :(
- Controller does not have any jumpers.
- Have not booted this card in any other system...yet.
- Supermicro PCIe (not the mSATA slot) is muxed into the SoC via a USB hub, and that might have some limitations...or a missing Linux kernel driver.

Rosewill RC-211, PCIe x1, 2 port, SATA2, RAID-5 (but can be flashed to non-RAID), SiliconImage 3132 chipset
- Cannot be seen in "lspci"...with or without HDD attached
- Controller does have a BIOS "enable/disable" jumper. Tried both ways...no luck either way. :(
- Very loose SATA data cable connector ports.
- Have not booted this card in any other system...yet.

It is very hard to find an add-in card using a SiliconImage chipset (ok, so they are limited to SATA2) that is preloaded with a non-RAID BIOS. I have read reviews stating that some add-in cards that use SiliconImage BIOS no longer have Flash ROMs that SiliconImage supports, and that prevents the user from changing the factory-installed RAID BIOS to a non-RAID BIOS. SYBA has openly stated (or implied...forgot which one) that they will void your warranty if you do that to their add-in cards that use SiliconImage code while providing the excuse that they can no longer obtain Flash ROMs that are supported by SiliconImage Flash tools. :mad:

I forgot to get a SYBA PCIe x1 add-in card with a Marvell 9230 chipset, but given the perceived incompatibility issues between the onboard Marvell controller and add-in cards with Marvell controllers, it might just be a waste of money for me since I never use add-in cards that have "built-in RAID" (aka "fake RAID") functions. If I want an add-in card with a real hardware-based RAID5 function I would spend the big $$ to get one, but I am liking ZFS a lot so add-in RAID5 cards are useless in that case...and potentially "conflicting".

So, for all of my efforts I have successfully added only 2x SATA3 ports via a PCIe x1 card that cost me about 20 USD before shipping costs. That brings the count of SATA ports on my system to 8: 2x SATA2, 6x SATA3. My target was 8 SATA ports, but 10 or even 12 would have been nicer.

If anyone has any contacts at Supermicro or Marvell that would like to discuss the behaviors I have documented above, something can be arranged. :cool:
 
Last edited:

exitmrhat

New Member
Aug 27, 2014
2
0
1
33
Hello!

I'm looking at the Supermicro X10SBA, but I'm wondering what power supply is needed to power 6 drives (+1 mSSD). 6 WD Reds 4TB for example alone need 10.5A from the +12V rail, so something with 14-15A would suffice?

I'm wondering if a PicoPSU 160 (15A peak on the +12V) + 192W 12V 16A power adapter will work.

Anyone running 6 3.5" drives?

Cheers.
 

Cheddoleum

Member
Feb 19, 2014
97
22
8
First, just so you know, installing an SSD in the X10SBA's mSATA slot uses one of the two Intel SATA-II channels, disabling the second SATA port on the mobo. So you can only get 5 HDDs on it if you're using the slot.

As for the power supply, this may not be the best subforum for this question since it's not strictly about the mobo so the best people to answer it may not see it. Personally I think you'd be pushing it. I did run 3 Seagate HDDs with about 1.7A peak draw at spinup from a 120W PicoPSU, on a mobo with similar power requirements to the SBA. It worked for about six months and then the brick died. The brick was only rated 80W continuous, with a higher inrush peak, so all was still within spec so I thought it'd be okay, and it wasn't, at least in this case. I just now tested that PicoPSU with another brick and still works but I switched to a more robust arrangement after that happened.
 
Last edited:

exitmrhat

New Member
Aug 27, 2014
2
0
1
33
Cheddoleum, thanks for your answer. Not being able to use 7 drives is essential info.
Is there anything else I should know about this board? I'm seriously considering it (it's either the Supermicro or the HP Microserver N54L).
 
Last edited:

Cheddoleum

Member
Feb 19, 2014
97
22
8
Nothing that isn't mentioned above, such as an erratum that can result in the USB3 port not being backward compatible with some USB2 devices, but that's easily worked around; or NotMine999's adventures in getting it to play nicely with a PCIe SATA add-in card.

One thing I might point out is the Type A USB header on top of the board. If you were planning to use the mSATA as a boot disk for relatively stable OS like an embedded system or hypervisor, you could use a low profile thumb drive installed there instead, freeing up all six SATA channels for storage. The boot might be slightly slower but on an appliance this almost never matters.

Note also when running a mix of SSDs and HDDs you want the SSDs on the SATA-III Marvell ports. One disadvantage to the SBA design is that the mSATA port is bound to the SATA-II controller, constraining the maximum disk bandwidth to something like 275MB/s, or about half of what a decent SSD can do.

Edit: make sure to check the processor features -- if you're really going to want VT-d or AES-NI then keep looking (and probably spend more money). I'm getting along fine without them, passing through one of the NICs to a VM using macvtap/virtio direct and that works fine, I get native GigE speeds with little additional overhead and that's good enough, and it does AES fast enough the hard way for my limited encryption requirements.
 
Last edited:

Roger Wilco

New Member
Sep 16, 2014
1
0
1
45
I have been considering buying the SuperMicro X10SBA for a pfSense firewall appliance. I have read with interest about the UEFI boot problems since I would like to be able to install FreeBSD/Linux on it. I noticed SuperMicro has released version 1.0c of this BIOS. Could anyone confirm whether this may solve the UEFI boot issues?
 

Patrick

Administrator
Staff member
Dec 21, 2010
11,966
4,931
113
FWIW - I have both the X10SBA and a A1SAi-2550F at my desk. The A1SAi-2550F has been running my home Hyper-V based pfsense VM for some time now. Given, I do have access to relatively a lot of hardware but that has been working well.

Since the X10SBA is $170 and the A1SRi-2558F is $230-240, I would be at least tempted to go for the C2558. The additional 2 NICs and IPMI are worth the increase in price.

pfsense uses such low CPU power on these things you can run a PXE boot server, server for virtual network functions or similar items.

The thing I really like is that you can do NIC teaming using Windows Hyper-V easily. If a VM goes wild, you can just revert. I upgraded to a random pfsense alpha version that went poorly. 2 clicks and about 90 seconds to revert and everything was back to normal.
 

Cheddoleum

Member
Feb 19, 2014
97
22
8
It's unfortunate that Supermicro doesn't post a changelist with a BIOS update (nor include one in the zip, it seems), but an email to support might tell you what the update is actually for.

Anyway, for Linux, at least, the UEFI boot issue is a solved problem; it was fire and forget after the initial install; successive upgrades have required no special attention, it just goes as normal.

I'm quite happy with this board. It's been my home server-of-all-work for over 8 months, including my DSL router/firewall under KVM using a custom 64-bit OpenWRT trunk with all the paravirtualization build options enabled, which runs very happily inside a tiny amount of memory and barely ticks over CPU wise.
 
  • Like
Reactions: Patrick