How about some Bay Trail?

Cheddoleum

Member
Feb 19, 2014
97
22
8
Mine shipped with BIOS revision 1.0b installed. From dmidecode:

Code:
# dmidecode 2.12
# SMBIOS entry point at 0x000f04d0
SMBIOS 2.7 present.
49 structures occupying 1930 bytes.
Table at 0x000EBD30.

Handle 0x0000, DMI type 0, 24 bytes
BIOS Information
	Vendor: American Megatrends Inc.
	Version: 1.0b
	Release Date: 12/31/2013
There isn't so much a problem booting Linux with this platform, the issue is that it

A) will boot EFI only, will not boot any MBR BIOS type installer. Trying just gets you a continuous beep-and-reset cycle.
B) will only boot a 32-bit EFI loader, even if you're running a 64-bit OS.

But as I mentioned above, if you have an EFI boot disk and you make bootia32.efi available alongside grubx64.efi, it installs just fine.
 

ant

New Member
Jul 16, 2013
21
5
3
Thanks Cheddoleum. I wonder if "Please use bios version after 1.0b, which will fix this issue." means that the next firmware that comes out after 1.0b will allow stock linux installers to work without workarounds. Anyway - as you have shown at least there is a workaround.
 
Last edited:

Cheddoleum

Member
Feb 19, 2014
97
22
8
Yes, it's hard to be sure what they mean there. I don't know if they're going to release a BIOS that will boot from an MBR in the old-school way; seems like if they were going to do that they would have. This "32-bit EFI only" thing seems to be consistent across the BayTrail SoC platform; which is why some guidance meant for an Asus T100 convertible ultrabook came in so handy there. Supermicro may not be able to overcome it with BIOS alone.

Anyway, upgraded in-place to 13.10 -- probably should have tried installing that in the first place but I wanted to limit the variables in case something didn't work and I needed to start backtracking. I've stripped out all the desktop stuff and converted it to a server the hard way, put in LXDM and LXDE in place of the horrible Unity for my occasional minimal UI requirements since it's mostly run headless, and removed all the other Ubuntu One cloud and backup cruft and daemons, added the server components for NFS and netatalk, etc.

Everything's great, and I'm feeling confident enough that I've gone ahead and put it into service as the replacement for my old Via Epia SN10000eg; mostly an NAS and general-purpose server with 4x 3.5" SATA and 1x 2.5" for the OS; I'd like to replace that latter with an mSATA when I find a nice small cheap one. On the Epia I was using the onboard Compact Flash slot, so this will be equivalent if a million times faster.

Only one hiccup: got a hard freeze and lockup when using Firefox on the desktop -- this kind of thing has been linked to the intel xorg driver's SNA acceleration which is still pretty new and probably not fully debugged on this platform. So I made xorg fall back to UXA by adding the following file:

/usr/share/X11/xorg.conf.d/20-intel.conf:
Code:
Section "Device"
   Identifier  "Intel Graphics"
   Driver      "intel"
   Option      "AccelMethod"  "uxa"
EndSection
This will reduce video playback performance a little and prevent the new "Tearfree" SNA option from being available, but it's better than locking up.
 

PigLover

Moderator
Jan 26, 2011
2,969
1,280
113
In fairness, there's a lot to wade through at that link. Here are some simplified instructions for how to install Ubuntu on the X10SBA:

  1. Build a USB boot disk for an EFI 64-bit Ubuntu ISO. I chose 13.04 AMD64 Desktop as a known quantity, but I suspect later editions will do fine. Also per the above link, I used Rufus on a windows system to build it, with these settings: GPT for UEFI, FAT, 64kb.
  2. Obtain the file bootia32.efi from the zip file in this comment. Copy it to /EFI/BOOT/ on the USB drive.
  3. Use this USB drive to boot the X10SBA by selecting it from the BBS menu (F11 at boot) or setting up USB key as the first boot priority in Setup. It's UEFI, not legacy boot. I used the VGA adapter, I haven't tested the HDMI or DP output.
  4. Select "Try out Ubuntu" rather than install. Screen will be blank for a bit, then it'll go into the standard Ubuntu graphical startup. Eventually you'll get the Unity desktop. (You'll note that the default graphics support is crude and can't handle transitions.)
  5. Use the first icon at the top left to open a search window, type "terminal" and hit enter to get a terminal. Drag that out of the way if necessary so you can get to the install icon.
  6. Install as you normally would, BUT before rebooting, use the terminal and copy the bootia32.efi file from your thumb drive (poke around a bit to see where it's mounted) to /target/boot/efi/EFI/ubuntu/ This directory won't show up until near the end of the install process, so wait for it to prompt you to reboot.
  7. Click Reboot when it asks. Pull the USB drive as it restarts and that's it. If your new drive is anywhere in the boot priority, it should boot right into your new Ubuntu installation.
All basic credit for working this out and for making the 32bit efi binary conveniently available goes to paperWastage at xda-developers.com; I had to do very little indeed to adapt it to this.
Found some time this evening and gave it a try...works like a charm. Ubuntu 13.10 loads up quick and easy.

THANK YOU!

Now - on to give XBMC a shot.
 

PigLover

Moderator
Jan 26, 2011
2,969
1,280
113
XBMC Frodo works perfect under Ubuntu, including solid stutter-free playback of Blu Rays (well, so far - I've only tested on a couple of them). Very pleased so far.
 

Cheddoleum

Member
Feb 19, 2014
97
22
8
Just need to keep in mind that that particular bootia32.efi was built for ubuntu 13.04. As we've both seen it's successful with 13.10 which is reasonable to expect: it shouldn't be a problem for a while since mostly it just contains filesystem and block device modules which are broadly stable... stable in the sense that the filesystems they refer to aren't moving targets. But there are exceptions, like btrfs. I think I'll take a shot at building my own 32bit efi for 14.04 LTS when it comes out before upgrading to it.

Edit: After a bit more reading (and a test build of the 32bit efi binary from the 13.10 grub sources) I think this is probably not a concern. Maybe one day if you're booting from some exotic new device or new kind of filesystem (and btrfs in its current form is built into the one linked above, so that one should be okay) you might need to build a new efi boot loader with up to date modules, but otherwise I wouldn't expect it to be needed any time soon.
 
Last edited:

Cheddoleum

Member
Feb 19, 2014
97
22
8
Just FYI, it (the X10SBA) Got into a state where it would not warm boot (e.g. ctrl-alt-del, shutdown -r now, etc.); it would hang before grub and I'd have to hit the reset button or power cycle it. If this happens to you, ask yourself what hardware you've connected to it recently.

I've heard of this happening with some controller cards, in my case it seemed to be because I had a particular USB2 wireless dongle plugged into the rear USB3 port. It works fine there but this problem occurs on warm boot consistently until I remove it or move it to a USB2 port.
 

NotMine999

New Member
Mar 7, 2014
7
1
3
I have this same board. I am also running 16GB of 1333 RAM (G.Skill F3-1333C9D-16GSL). The SoDIMMs for this board have to be 1.35V DDR3L types. Out of curiosity, I tried a few spare DDR3 SoDIMMs (G.Skill 4GB 1333 RAM) rated at 1.5V and they did not work at all.

I have also gotten my board stuck in a "boot fail" situation by changing the Video OpROM value from "Legacy" to "UEFI". I had to follow the procedure in the Supermicro manual to reset the BIOS to defaults. That process does work; it worked for me.

In my case the OS is Gentoo running the current "gentoo-sources" kernel (3.10.32) in x86_64 mode; Gentoo compiles decently on this board, but getting it setup for UEFI boot via GRUB2 is definitely "not for beginners". I easily built a ZFS RaidZ1 pool using HDD attached to the Marvell 9230 HDD ports onboard. In my test case, the pool was built from 4x WDC Black 750G SATA2 laptop drives. Using "proftpd" as the server package (and with very minimal "feature" tweaks, not "performance" tweaks), I am able to transfer files to the pool from a network connected laptop at rates up to ~700 megabits per second using a sustained run of various large video files in sizes from 300MB to 1.5GB. Pretty good performance and comparable to my RAID5 arrays. File deletion performance on ZFS is slower than I would want, but not enough to be "painful".

The fact these same HDD work on the onboard Marvell ports indicates any issues getting a storage controller in the PCIe slot to work will be due to BIOS settings, strange "conflicts" between the add-in card and the system, or a flaky UEFI implementation; keep reading to see this.

I have found some serious annoyances with this board:
(1) Some USB keyboards refuse to work properly with this board. A very basic BTC USB keyboard works fine everytime while an IOgear USB keyboard with a few more buttons is "hit&miss". When attached to an IOgear KVM that has USB support for the keyboard & mouse, this board acts really strange. While using the KVM to access the system BIOS, I watched the HDMI-attached screen flip off and on. Not "power off and on", but "screen bank off and on". I got around that issue by directly attaching the keyboard to the motherboard via USB 2.0.

(2) I have yet to get a SATA storage controller working in the PCIe x2 slot no matter what BIOS options I select. Some options can make the system "hang hard", requiring a full power down and removal of the card. Other BIOS options cause the boot process to "hang hard" before it ever accesses the bood HDD. In both of those cases the add-on storage controller had SATA2 HDD (same as above) attached to that add-in card. When I remove those HDD and avoid using BIOS options that "hang hard", the system does boot and Gentoo comes up. Even "lspci" shows the card installed at "01:00.0"; it's a Syba PEX40064 with a Marvell 9215 chip. I find this issue could have bigger ramifications for this board since every add-in "non-RAID" storage controller that I could find on Newegg that will work in PCIe x1 or x2 slots is based on the Marvell 9215 or 9235 chipsets. Why "non RAID" chipsets? With ZFS having a "RAID capable" chipset could cause issues for ZFS and will definitely slowdown system boot times.

(3) Video output is "funky". I can rarely get video output on a directly conencted VGA monitor. I can always get video output on a directly connected HDMI monitor. I have yet to figure out how to consistently get video out the VGA port, but I have spent more time on the storage issues than diagnosing CLI output on video screens; I am not doing any fancy graphics stuff at all.

Right now using this board over using an older Supermicro X7SPA/X7SPE board is tough choice. The older X7SPA/X7SPE boards have an add-in slot that seems to work with anything I install in it, but the Atom CPU "has seen better days" especially compared to these newer Celeron SoC CPUs. An older Atom CPU is limited to 4GB of RAM, so running ZFS is only a choice if you don't mind potential throughput issues on large arrays. These newer Celeron SoC CPUs with 8GB (or even 16GB) of RAM support should have better ZFS performance with large arrays. Using other SoC-based boards would mean running Realtek NICs (ugh!) and the need to load added firmware when running most recent Realtek NICs under Linux. Boards that use Intel NICs appear to "run solid". In the case of the X10SBA, the Intel 210 NICs are easily supported by the Linux "igb" driver and Linux recognizes the queueing support in the Intel 210 NICs; "tc qdisc show" displays "mq" not the default value.

My plan is to build a 5x4TB ZFS pool (about 14TB usable) with separate cache/ZIL (via SSD) and separate (small laptop) boot drives. So a total of 7 drives. I can fit up to 6x4TB in the chosen case along with the separate cache/ZIL and boot drives, for a total of 8 drives. Perhaps my alternative to adding more SATA ports to this Supermicro board is an add-in mini PCIe 2-port SATA2 controller based on the ASmedia 1061 chipset, but that "non mSATA" mini PCIe slot is muxed via the USB controller hub to the Celeron SoC. Choices choices.

Last comment: "your mileage may vary".
 
Last edited:

cheezehead

Active Member
Sep 23, 2012
717
174
43
WI
Looks ok for a NAS or Firewall.

I just don't get it, why use it for an XBMC front-end? By the time you add ram/psu/case your north of $250...there are plenty of other XBMC builds for south of $100.
 

Cheddoleum

Member
Feb 19, 2014
97
22
8
Looks ok for a NAS or Firewall.
I was feeling brave so mine went into service as my household NAS/SMTP/HTTP/TFTP/you-name-it/etc host right away; almost two weeks ago now. Running 4x 3.5" 7200 HDDs (two each Seagate whatever-was-cheap-at-Newegg and Hitachi Ultrastar) and a 2.5" 5400 for the OS; that latter's an ancient SATA-I spare, the only real performance bottleneck so far which I'll be replacing with an mSATA drive in the handy slot when I find a smallish one on sale.

For various legacy reasons running a mix of EXT4, XFS and now a BTRFS RAID1 pair using logical volumes on two of the drives. Serving over NFS and AFS, pxeboot/nfsroot host for the diskless HTPC in the living room, doing Time Machine duty for the Macs, even running a smallish VM under kvm/libvirt. Memory is 2x Crucial CT2K2G3S1339M for a total of a mere 4G. It's dual-voltage 1.35v and 1.5v, listed as DDR3L and works great. Given that this is replacing a board on which I was only running 1G of DDR2 667 I figured 4G would be plenty, and it is: even doing all the above Linux doesn't manage to use it all and has yet to touch the swap partition.

It's doing a perfect job and staying between 38-43C with only the slight accidental case airflow from the fans in the drive compartment. That warm-boot quirk when I'd left a USB2 wifi dongle in the USB3 port is the only issue I've had, though I'm getting a sense from the little information out there so far that you can get similar problems from certain adapters in the PCIE slot.

Tempted to use it as a transcoding DLNA server as well, for tablet viewing of things like older .AVIs which they won't play natively. Right now I delegate DLNA duties (passthrough only) to XBMC's built-in server on the living room HTPC, but this is a very decent CPU with metric shedloads of idle cycles so I'm tempted to put them to work.

I'm a little puzzled by your VGA and KB issues, NotMine999. I'm doing VGA through an old cheapo Linksys KVM that almost predates the millennium and wasn't their expensive model even then. For PS/2 USB conversion using a SIIG dongle from Fry's. I've found this arrangement a little flaky on some machines but the X10SBA seems very happy with it. Though see my comment above about using UXA instead of the newer SNA acceleration under Xorg.
 
Last edited:

NotMine999

New Member
Mar 7, 2014
7
1
3
...[snip]...

I'm a little puzzled by your VGA and KB issues, NotMine999. I'm doing VGA through an old cheapo Linksys KVM that almost predates the millennium and wasn't their expensive model even then. For PS/2 USB conversion using a SIIG dongle from Fry's. I've found this arrangement a little flaky on some machines but the X10SBA seems very happy with it. Though see my comment above about using UXA instead of the newer SNA acceleration under Xorg.
I am also puzzled by the flaky VGA output. It could be the KVM itself. As for UXA or SNA acceleration, I am only running the basic "no D" CLI console, no 2D or 3D graphics acceleration at all.

As for my add-in SATA card issues, I have a bunch of different bits of SATA2 add-in PCIe hardware arriving this week for testing. I suspect the following issues in "most to least" order: (1) onboard Marvell chipset code conflicts with add-in cards based on Marvell chipsets; (2) Supermicro UEFI implementation "has issues"; (3) "strange slot problems" (some desktop boards do not like to see anything other than video cards in the PCIe x16 slot that they expect to be used with video cards only). IMHO Supermicro has a good reputation of building boards that don't have "strange slot issues", but there is always a first time. I have verified the slot does work by plugging in an Intel PCIe NIC card (82574 chipset I think) and even the BIOS recognized that card on the page where the onboard Intel 210 NICs are configured. So I am tending towards issues (1) and (2). I am even testing the Syba mini PCIe 2-port SATA2 card for 2 reasons: (a) not a Marvell chipset [uses ASmedia 1061] and (b) it was pretty inexpensive.

Now if I can find a way to "watch" the Supermicro web site for any announcements on updated BIOS for this board. Perhaps a trouble ticket with Supermicro is in order in my case?

I would prefer to build my next revision (previous was based on SM X7SPA) of storage servers with this board over 1155 or 1150 socket boards due to the operational costs. Finding a truly low power consuming 1155 or 1150 CPU is difficult; the Intel "T" series CPUs can be found but I think you have to be lucky to find a vendor you can trust...and I don't trust most "flea bay" sellers. My own research shows the lowest power 1155 or 1150 CPU and a decent desktop mini ITX board that meets my needs (4-pin fan headers and other stuff) comes to about the same price as the Supermicro X10SBA before adding memory to either board. So that moves the evaluation to power consumption costs ("electricity costs"), possible noise issues ("silence is golden" even for a storage server that lives in a room by itself), those silly "push peg" fans that seem to come loose (my storage server boards are mounted on their side in Lian-Li PC-08 cases so the CPU fan has to be secure), and overall board reliability (I expect to run 24x7 when I am at home). I favor ASUS for my desktop boards, but I favor Supermicro for server boards. Enough of my ranting/rambling for now.

Stay tuned.....
 
Last edited:

ant

New Member
Jul 16, 2013
21
5
3
Just want to say that I really appreciate the comments from you all about your experiences running the X10SBA with linux. Thank you.

It (actually the X10SBA-L) is on my list of boards that I am considering. It is quite expensive in my area of the world especially for the purpose I would use it for, so I am hoping to also hear from others using linux on other bay trail boards as they become available.

I wonder if the USB3 and PCIe issues are related to the errata for bay trail CPUs with B3 and earlier stepping? They don't mention your issues specifically but might be related:

http://www.intel.com.au/content/dam/www/public/us/en/documents/specification-updates/pentium-n3520-j2850-celeron-n2920-n2820-n2815-n2806-j1850-j1750-spec-update.pdf

c0 stepping might not have the same issues but it is not clear to me.

Another issue listed there that concerns me is:

"VLP51 No Fix HD Audio Recording And Playback May Glitch or Stop"

So if I get a bay trail board, I might wait a while to hopefully get a later stepping.

Ant
 

ant

New Member
Jul 16, 2013
21
5
3
This looks promising for the possibility of 64bit UEFI for bay trail:

The site seems a bit mixed up mentioning H61 along with bay trail type boards - but anyways - the "bios and spec card" tab here mentions 32Bit UEFI and 64Bit UEFI versions of "bios" for some of their J1800 and J1900 motherboards:

MotherBoard,TOP 1 MotherBoard Design House - Wibtek

They even mention:

"32bit - UEFI
Update to this version of BIOS, you can only use the Windows 8/8.1 32bit OS Only.
If you want to use 64bit OS, please then reflash BIOS to 64bit-UEFI version with Flash Programming Tool."

So to me that means that there is a good chance that supermicro should be able to provide a 64bit UEFI for X10SBA.
 

ant

New Member
Jul 16, 2013
21
5
3
USB3 warm boot issue might be hinted at here?:

https://qdms.intel.com/dm/d.aspx/78314527-AF5E-42F2-985F-B49F92E6F64E/PCN112861-00.pdf

"Description of Change to the Customer:
Intel will start transitioning from B3 to C0 stepping in for Intel Celeron Processors J1800 & J1900 and Intel Pentium® J2900 Series based Windows platforms.
Summary of hardware & software changes from B to C step:
1. Improved robustness of USB3 wake after D3
... "

from https://en.wikipedia.org/wiki/Advanced_Configuration_and_Power_Interface

"D3 Off has the device powered off and unresponsive to its bus.
D3 Hot & Cold: The D3 state is further divided into D3 Hot (has aux power), and D3 Cold (no power provided). A device in D3 Hot state can assert power management requests to transition to higher power states."

So D3 might be related to warm boot?

The intel document indicates that C0 stepping J1900 won't be available till about June 2014. Probably doesn't help you. Just selfishly thinking that for myself if I want one of these boards I might have to wait at least a few months to get something with a few less issues. I guess all cpu's etc have issues of one sort or another though.
 

Cheddoleum

Member
Feb 19, 2014
97
22
8
That's a good find, ant, and your guess seems reasonable on the face of it. To anyone planning on leaving a USB3 device connected, and furthermore expecting cases where unattended reboots might be necessary, the only suggestion I might make is "try your device and see". It should at least be deterministic; i.e. if you can't reproduce the problem experimentally with any given device then it shouldn't happen in production either, since warm-boot states don't carry a lot of runtime variability.

One of the unimportant issues seems to be trivially reproducible, VLP54, Frequency Reported by CPUID Instruction May Not Match Published Frequency: "When the CPUID instruction is executed with EAX = 80000002H, 80000003H, and 80000004H, the frequency reported in the brand string may be truncated while the published frequency is rounded." This is definitely happening: in /proc/cpuinfo the nomenclature frequency (as opposed to the current or maximum frequency) is reported as 1.99GHz, when Intel publishes it everywhere as 2.0GHz.

I'm not overly concerned about most of of this personally -- an awful lot of it concerns xHCI and SD which I mostly don't care about. But I can see how the HD audio issue might concern people planning to use it as a media device, and while a BIOS fix is proposed one doesn't know when AMI will deliver and when/if Supermicro will issue a new one based on it.

However some of them concern VMX, which I am using and plan to use. If I were planning to use a VM for arbitrary workloads (i.e. install a desktop in it and run various apps at will) I'd be a little worried. As it is I'm planning to stick with embedded systems which basically have a pretty constrained workload and should run through all their ordinary states with regular testing. So I feel reasonably safe that one that's been put through its paces for a few days without problems should be stable long term.

I'm not in the habit of reading these docs so I'm not sure how serious it all is. As you say, presumably all CPUs have problems of one sort or another so there's no point in being a hypochondriac about it.
 
Last edited:

Cheddoleum

Member
Feb 19, 2014
97
22
8
(If like me you're not familiar with these processor errata documents it looks kinda daunting, but I guess this one's par for the course and even pretty lightweight. Check out this one for the Xeon E5 family for comparison. It's a slightly different case as the Bay Trail is an SoC not just a CPU so the potential range of issues covers more territory. But it's a damn short list compared to the E5, especially given that.)