SPCR

Jakoob wrote:I remember my Intel D510MO wasn't able to run Ubuntu, nor Debian and some tests even showed better power consumption on Windows.

Not sure where you got that info from but Ubuntu running on that board fine from the day it was released. I tested power consumption back then and recently and its still very low, in-line with reviews online.

As far as I'm aware the cache on modern hard drives does have ECC.

I would hope so! But unless it includes an early alert feature, ECC can't work miracles if the RAM is going bad (as opposed to handling the odd error caused by radiation or something).
Does SMART report RAM errors? Does it flag a drive with eccessive RAM errors as failing? I don't know which means it can't be a very common failure mode but it'd be interesting to know.
For system RAM, there's software that can notify you when errors are corrected by ECC (relying on features which are not necessarily included in affordable hardware).

Good RAIDs don't have the write hole issue. This includes mdadam, GEOM, Intel FakeRAID and good hardware RAIDs.

Unfortunately, this is not true - the only option to prevent write hole issue is using battery backup - which is possible only with very good hardware RAIDs.
http://www.raid-recovery-guide.com/raid ... -hole.aspx

Silent data corruption does happen, but not on hard drives.

Unfortunately, this is also not true.
http://www.necam.com/Docs/?id=54157ff5- ... 1cf2cb27d3

The chance that a drive will miss an error that ZFS would catch is in the realm of "mathematically impossible".

Again, not true.
http://www.zdnet.com/blog/storage/data- ... u-know/191
and if the blog post above is not verbose enough check this one:
http://queue.acm.org/detail.cfm?id=1866298

I'm not in any way saying ZFS is bad, just that some of it's features are overblown. If you are worried about data corruption, ZFS will not help you in 99.9% of the cases. You need something above the filesystem for that.

Yeah, obviously with some effort you can get most of the features that ZFS provides. You can use hardware RAID with redundant power source to avoid the write hole issue. You can use a good modern filesystem with checksums to identify silent data corruption. You can use LVM or dynamic disks (or how is it called in Windows) to overcome partitioning problems.

The thing is that with ZFS you have all these features 'out of the box', with some additional benefits and very easy administration.

I have personally experienced a very simple problem: on RAID1 (two disks mirrored), one chunk of data on one disk got damaged. RAID could not tell which disk holds correct data, so the read process failed. These are the kind of problems that RAID does not solve. And this would never happen on ZFS mirror, because it integrates hardware redundancy with filesystem checksums.

I am not saying that ZFS is the only possible solution, but it is really feels like a different world. Once I have tried, I never looked back.

Jakoob wrote: The Q1900B-ITX is on stock in Czech already (http://www.agem.cz/Default.aspx?content ... d_id;42400) seems to me ASRock already started the shipment. Price is around 72 Euro

It's the 'B' version, available since a few weeks if I recall correctly.

NX3 wrote:

Jakoob wrote:I remember my Intel D510MO wasn't able to run Ubuntu, nor Debian and some tests even showed better power consumption on Windows.

Not sure where you got that info from but Ubuntu running on that board fine from the day it was released. I tested power consumption back then and recently and its still very low, in-line with reviews online.

My personal experience, I was one of the first to own this board in Czech and Ubuntu after install and restart showed just black screen. There is a lot of threads around internet about it (there was some missing driver or what). I spent two days trying to fix it. Windows Server 2008 worked much better...

gorkypl wrote:

Jakoob wrote: The Q1900B-ITX is on stock in Czech already (http://www.agem.cz/Default.aspx?content ... d_id;42400) seems to me ASRock already started the shipment. Price is around 72 Euro

It's the 'B' version, available since a few weeks if I recall correctly.

Yeah sorry, I meant B version. I corrected my previous post to make it more clear.

gorkypl wrote: Unfortunately, this is also not true.
http://www.necam.com/Docs/?id=54157ff5- ... 1cf2cb27d3

That is a blatant advertisement for NEC storage. Pure advertising BS.

Again, not true.
http://www.zdnet.com/blog/storage/data- ... u-know/191
and if the blog post above is not verbose enough check this one:
http://queue.acm.org/detail.cfm?id=1866298

Sorry, but simple logic proves the CERN report (as reported in the linked article at least) is impossible in the real world. Modern storage does not have a byte error rate of 3 * 10^7. That is one error in every 30 MB. The modern world could not function if it were true. ZFS could not even deal with it, as it's own checksums would get corrupted regularly with that error rate.

The second article is dealing with data loss, which is not the same thing as corruption. Hard drives lose data all the time, no one is disputing that. What they don't do is magically return something different than what was originally written. They either return the requested data, or an error. As I said before, they have at least an order of magnitude more error checking than ZFS has.

Many people forget that we had RAID with built in check summing, it was called RAID 2. Is long gone since it was quickly proven to be useless with modern drives.

I have personally experienced a very simple problem: on RAID1 (two disks mirrored), one chunk of data on one disk got damaged. RAID could not tell which disk holds correct data, so the read process failed. These are the kind of problems that RAID does not solve. And this would never happen on ZFS mirror, because it integrates hardware redundancy with filesystem checksums.

How would ZFS help if both disks appear good? If both return data, how do you trust which checksum is correct? It's a major logical problem with ZFS, it claims that hard drives cannot be trusted, yet trusts them with the checksum data.

Linking to the scribblings of salesmen isn't an effective way to establish credibility.

washu wrote:Sorry, but simple logic proves the CERN report (as reported in the linked article at least) is impossible in the real world. Modern storage does not have a byte error rate of 3 * 10^7.

They weren't talking about random byte errors but mostly about more serious problems which can be troubleshooted if the admin is paying attention or prevented if she had been reading about other people's problems in the first place:

Bernd Panzer-Steindel, CERN/IT wrote:-64k regions of corrupted data, one up to 4 blocks (large correlation with the 3ware-WD disk drop-out problem) (80% of all errors)

washu wrote:How would ZFS help if both disks appear good? If both return data, how do you trust which checksum is correct? It's a major logical problem with ZFS, it claims that hard drives cannot be trusted, yet trusts them with the checksum data.

The scenario was partial data loss on one hard drive (as if some part of the disk had been overwritten with garbage), not inconsistency following power loss or something. Garbage rarely has the right checksum.
The issue is rather: how common is it for a drive to return garbage? The answer is evidently: a lot less common than it is for controllers *or indeed the system RAM in low-end systems* (notice the thread's topic?) to return garbage.

Now if the RAID implementation gorkypl "personally experienced" fails every time it encounters a read error on a drive, that's a different problem! Because the normal effect of "one chunk of data on one disk got damaged" is a read error. Surely everyone here has experienced at least one.

HFat wrote:The scenario was partial data loss on one hard drive (as if some part of the disk had been overwritten with garbage), not inconsistency following power loss or something. Garbage rarely has the right checksum.

I see what you are saying, but how would part of one disk get overwritten by garbage assuming the drive is part of a RAID set? That would require a serious failure on in the OS or the RAID system. ZFS could not deal with the OS being faulty, it has to place some trust in it.

Well yeah, it seems to me that the case in which this feature of ZFS is most useful is: solid OS, unreliable storage. In other words, expensive servers attached to loads of cheap storage.
But if all you've got is a couple of drives attached to a cheap Bay Trail board (meaning your OS is your RAID system and that it's likely to be affected by any corruption affecting your storage), running ZFS might be more risky than not.

Biostar boards added, however not different from competitors... 2x sata and ATX power connector.

ZFS is awesome. You need ECC RAM.

Wonder if you guys would mind taking your ZFS theology debate to its own thread and letting this one stay on topic regarding bay trail MBs?

On topic: any word on when/if SM will release a bios upgrade supporting proper UEFI boot options on the X10SBA(-L)? The 32-bit UEFI boot loader in bios 1.0b is a real pain. Also, anybody bot OpenElec running stable on it? Currently running with XBMCuntu and I'm still having sleep/wake issues and HDMI stability issues when the AV receiver is off. I need more stable options.

new thread for discussing the Gigabyte J1900 board BIOS:
viewtopic.php?f=28&t=67563

Gigabyte have posted the F3 bios on their website now
http://www.gigabyte.com/products/produc ... =4918#bios

gorkypl wrote:For those interested, Asrock Q1900-ITX should be available in Poland in the first days of May, with expected price ~70EUR.

Which shop
PM me please w/details.

Had some talks with ASRock and it looks like, that Q1900DC-ITX will be available pretty soon and even pricing looks reasonable.
As soon as they confirm I will post it here. I also asked for one piece for review.

Stay tuned

Jakoob wrote:Had some talks with ASRock and it looks like, that Q1900DC-ITX will be available pretty soon and even pricing looks reasonable.
As soon as they confirm I will post it here. I also asked for one piece for review.

Stay tuned

Some Japanese pricing is now showing up. It is 25% higher than J1900-ITX, ~87.5EUR.

Wonder if it comes with a laptop power brick for that price. Looks like May 16th is the release date in Japan.
google translate http://www.gdm.or.jp/pressrelease/2014/0502/68925

"Of the AC adapter support "Q1900DC-ITX" will be released May 16"

Here's hoping Europe/USA is similar. Be nice if it was the c0 stepping of J1900.

I've just came across two nice blog posts about building a silent PC based on J1900 and running LInux.
http://www.rkblog.rk.edu.pl/w/p/buildin ... ron-j1900/
http://www.rkblog.rk.edu.pl/w/p/benchma ... ntu-linux/

The author claims that CPU sees whole 16GB of RAM, which is great.

Thanks gorkypl, very useful.
No mention if any bios issues at all, so maybe it's just gigabyte who screwed up their board.

Anandtech have reviewed an ASRock motherboard designed for a storage server, it uses a C2750 Atom Silvermont which is related to the Baytrail-D processors.
http://www.anandtech.com/show/7970/asro ... management

I asked a UK official reseller of Supermicro for pricing and there's a 7 to 10 day lead time, quote was

* MBD-X10SBA-L-O £115.50 + VAT + delivery
* MBD-X10SBA-O £138.00 + VAT + delivery

ASRock Q1900DC-ITX is official ... http://www.asrock.com/mb/Intel/Q1900DC-ITX/

PCIe x1 can't be used for graphic card (maybe with extra PSU), otherwise this seems to be the best Bay-Trail board so far.

Because lets face it, for boards with idle around 10-15 W, there is no classic PSU.

There is PCIe x1 Zotac nVidia GT610. There are also [mod: deleted non-functional link]Chinese funky risers for x1 to x16 where power comes from molex or SATA power cables and logic is passed from x1 to x16 riser via USB3 cable - I'll be testing that with bit old nVidia quadro FX580 and Radeon HD6450 (both with low power usage and they don't need the extra power plug) as a third part of the review I'm curious how much can pass via pcie x1 2.0 and how it will perform vs J1900 graphics (is that graphics limited by slow RAM etc.). In general they should be below HD4000 performance.

http://www.videocardbenchmark.net/gpu.p ... 450&id=267 - radeon
http://www.videocardbenchmark.net/gpu.p ... 80&id=1608 - quadro
http://www.videocardbenchmark.net/gpu.p ... +4000&id=2 - intel

but Bay Trail doesn't has 16 EU only 4EU and that slow RAM

Good day.

Does anybody know the height of the cooler on AsRock Q1900? Chieftec makes some nice cases 01B and 03B, but the cooler height without optional addon is ca 30 mm.

sihker wrote:Good day.

Does anybody know the height of the cooler on AsRock Q1900? Chieftec makes some nice cases 01B and 03B, but the cooler height without optional addon is ca 30 mm.

The highest element on Q1900B-ITX are the audio outputs - 35 mm. The radiator isn't very high, even the SODIMM end up bit higher than it.

Can someone address me to the best way to boot FreeDOS on a Gigabyte w/ F1 firmware? I prepared a FAT16 2GB usb key but it does not want to boot (black screen, needs a power cycle, I'm using DVI output). I'm not even sure that DVI output is enabled during boot, I tried to boot from a SATA hard disk with a working Win8 installation (in legacy mode) but it did the same (black screen, no disk activity).

Strange thing is, after a CMOS reset I forgot an usb key inserted and it booted, sadly it was not the correct drive

I believe the F1 bios in the Gigabyte board J1900 and j1800 are hopeless, and you'll need to upgrade. People have been reporting they won't boot without a monitor connected, perhaps they mean VGA?

Jakoob wrote: PCIe x1 can't be used for graphic card (maybe with extra PSU), otherwise this seems to be the best Bay-Trail board so far.

Because lets face it, for boards with idle around 10-15 W, there is no classic PSU.

I'm testing three x16 graphic cards connected via riser-alike board to the Asrock x1 slot. Power is provided via Molex or SATA power cable. That seems to be working, some benchmarks drop compared with x16 (results to come). The bigger problem is that for stronger GPU the 6 pin additional power plug is needed, and that usually doesn't exist in small cases with fanless power adapters (and it would have to be probably ~200W or more depending on card).

I just found that Jetway now have a Silvermont/Baytrail-D board with two variants, one with N2930, the other E3827

JNF9M-2930 Mini-ITX motherboard with Celeron N2930 Quad Core Bay Trail Processor

JNF9M-3827 Mini-ITX motherboard with Atom E3827 Dual Core Bay Trail Processor

http://www.jetwaycomputer.com/NF9M.html