• 1
  • 4
  • 5
  • 6(current)
  • 7
  • 8
  • 54
Bay Trail-D motherboards
#76
Then the Baytrail NUC also won't have it - http://ark.intel.com/compare/79052,78866
But the NUC seem to do well as HTPC, except SSE Yadif Deinterlacing.
| myHTPC |
Reply
#77
(2014-03-23, 08:47)-DDD- Wrote: Then the Baytrail NUC also won't have it - http://ark.intel.com/compare/79052,78866
But the NUC seem to do well as HTPC, except SSE Yadif Deinterlacing.

Seems like they don't: http://forum.xbmc.org/showthread.php?tid...pid1660779
Reply
#78
hi10p is always SW decoded as far as i know.
| myHTPC |
Reply
#79
(2014-03-23, 08:41)mrchief Wrote:
(2014-03-01, 18:34)BORIStheBLADE Wrote:
(2014-03-01, 17:59)tonwes Wrote: Gigabyte J1900 already available. Lots and lots of connectivity (4xUSB3.0, DVI, VGA2560x1600, 2xUTP, S/PDIF even 2xSerial and 2xPS2).

So what is missing? straight hmdi...If only...

http://hexus.net/tech/news/mainboard/663...therboard/

How do these CPU compare to the Celeron 1037?

J1800 doesn't support SSE which mean no hardware decoding for videos. But then again, I'm no expert and I may be wrong.

SSE is not the same as hardware decoding...

The N2820 Bay Trail does have hardware VAAPI decoding for the same types of video that most other GPUs accelerate (standard H264, MPEG2, VC-1 (*) etc.). (Like most other solutions there isn't hardware acceleration for the less widespread formats like Hi10p H264 and 4:2:2 H264 - and these will be software decoded. AIUI The N2820 is quite a low power, Atom-based but Celeron-branded CPU in comparison to more standard, and powerful, Celeron offerings like the 1007, 1037U and 2955U. (Passmark for N2820 is 1048, for 1007U is 1442, for 1037U is 1750, for 2955U is 1531 - so you can see the 1037U is 70% faster than the N2820 in CPU terms using Passmark)

However Intel drivers in Linux have lots of bugs, particularly around de-interlacing. The OpenElec devs produced some builds with Bob hardware de-interlacing (as well as some with more advanced de-interlacing like Motion and Vector Adaptive). However these were really unreliable across various processors, because of the buggy nature of the Intel Drivers AIUI.

So the developers took a different tack, that of VAAPI hardware decoding of the video codec, but software de-interlacing of the decoded video. So the H264/MPEG2/VC-1 encoded video is decoded in hardware, but the CPU then takes over to handle de-interlacing.

De-interlacing is pretty processor intensive (it's actually doing something quite complex) but the OpenElec guys have optimised the de-interlacing code to take advantage of modern CPU features that make computation faster, like SSE, allowing a Yadif 2x de-interlace to be done very efficiently.

"Normal" Celerons like the 1007U, the 1037U, the 2955U support YADIF 2x software de-interlacing with VAAPI decoding. (My 2955U Chromebook and my 1007U Revo RL80 are both fine with 1080/50i broadcast content like Live TV) However the N2820 doesn't have the processing power (maybe it has different optimisations that can be exploited? It looks as if the new Atoms - sorry "Celerons" - have had Hyperthreading removed now that they support out-of-order execution, and that they have a new SSE implementation?)

AIUI the 1007U is probably the base processor for OpenElec if you want 1080/60i de-interlacing (the 847 can just about do 1080/50i I believe - but the extra 10fps needed for 60i push it over the edge?)

I stress - this is related to de-interlacing in software, not video decoding in hardware. It's not relevant to 1080p or 720p content codec using standard H264 profiles and levels.

(*) There is also an issue with interlaced VC-1 content on Intel hardware I believe - which means that software decoding is better for this. I'm not sure what the base level of hardware required is. Hi10p stuff is always software decoded. AIUI the 1037U is fine, the N2820 struggles.
Reply
#80
Hi,

I've just build a PC with the following:
Gigabyte GA-J1800N-D2H
4gb Ram
Cheap Toshiba SSD
Windows 8.1 64 bit
Wired network (gigabit) to a QNAP NAS that stores all my media files.
Connected via HDMI to LG TV

My results:
Booting / General use = Very quick - quite amazing for such a low cost / power motherboard / CPU.
XBMC (Beta2 Release: 13.0 “Gotham”) = Again very quick - plays a 16gb MKV episode of Dr Who absolutely fine with low CPU usage.
BBC Iplayer (web site version) = Plays HD programmes fine with low CPU usage.
Amazon Instant Streaming = Won't play HD streams without judder of fast moving streams - very high CPU usage.

I guess the Amazon problem is due to Silverlight not using the GPU?

Would upgrading to a quad core J2900 motherboard be likely to cure the Amazon problem?
Reply
#81
(2014-03-21, 21:36)Tobor Wrote:
(2014-03-21, 04:18)Dougie Fresh Wrote: Man, this Gigabyte J1800 motherboard is a royal PITA so far.

Can you elaborate?

I put it in my Amazon review: http://www.amazon.com/Gigabyte-Dual-Core...B00IFE4VF8
Reply
#82
(2014-03-23, 12:08)noggin Wrote:
(2014-03-23, 08:41)mrchief Wrote:
(2014-03-01, 18:34)BORIStheBLADE Wrote: How do these CPU compare to the Celeron 1037?

J1800 doesn't support SSE which mean no hardware decoding for videos. But then again, I'm no expert and I may be wrong.

SSE is not the same as hardware decoding...

The N2820 Bay Trail does have hardware VAAPI decoding for the same types of video that most other GPUs accelerate (standard H264, MPEG2, VC-1 (*) etc.). (Like most other solutions there isn't hardware acceleration for the less widespread formats like Hi10p H264 and 4:2:2 H264 - and these will be software decoded. AIUI The N2820 is quite a low power, Atom-based but Celeron-branded CPU in comparison to more standard, and powerful, Celeron offerings like the 1007, 1037U and 2955U. (Passmark for N2820 is 1048, for 1007U is 1442, for 1037U is 1750, for 2955U is 1531 - so you can see the 1037U is 70% faster than the N2820 in CPU terms using Passmark)

However Intel drivers in Linux have lots of bugs, particularly around de-interlacing. The OpenElec devs produced some builds with Bob hardware de-interlacing (as well as some with more advanced de-interlacing like Motion and Vector Adaptive). However these were really unreliable across various processors, because of the buggy nature of the Intel Drivers AIUI.

So the developers took a different tack, that of VAAPI hardware decoding of the video codec, but software de-interlacing of the decoded video. So the H264/MPEG2/VC-1 encoded video is decoded in hardware, but the CPU then takes over to handle de-interlacing.

De-interlacing is pretty processor intensive (it's actually doing something quite complex) but the OpenElec guys have optimised the de-interlacing code to take advantage of modern CPU features that make computation faster, like SSE, allowing a Yadif 2x de-interlace to be done very efficiently.

"Normal" Celerons like the 1007U, the 1037U, the 2955U support YADIF 2x software de-interlacing with VAAPI decoding. (My 2955U Chromebook and my 1007U Revo RL80 are both fine with 1080/50i broadcast content like Live TV) However the N2820 doesn't have the processing power (maybe it has different optimisations that can be exploited? It looks as if the new Atoms - sorry "Celerons" - have had Hyperthreading removed now that they support out-of-order execution, and that they have a new SSE implementation?)

AIUI the 1007U is probably the base processor for OpenElec if you want 1080/60i de-interlacing (the 847 can just about do 1080/50i I believe - but the extra 10fps needed for 60i push it over the edge?)

I stress - this is related to de-interlacing in software, not video decoding in hardware. It's not relevant to 1080p or 720p content codec using standard H264 profiles and levels.

(*) There is also an issue with interlaced VC-1 content on Intel hardware I believe - which means that software decoding is better for this. I'm not sure what the base level of hardware required is. Hi10p stuff is always software decoded. AIUI the 1037U is fine, the N2820 struggles.

This should be made a wiki! Smile @noggin, thank you so much for the great info!!
Reply
#83
(2014-03-23, 19:54)Dougie Fresh Wrote:
(2014-03-21, 21:36)Tobor Wrote:
(2014-03-21, 04:18)Dougie Fresh Wrote: Man, this Gigabyte J1800 motherboard is a royal PITA so far.

Can you elaborate?

I put it in my Amazon review: http://www.amazon.com/Gigabyte-Dual-Core...B00IFE4VF8

Finally found a local shop with this mobo in stock. I have it now with RAM, SSD and Win 8.1. Will be back after the install.

Reply
#84
I totally forgot I need a 24 pin PSU. I thought my spare PSU would work but its only 20pin. Now I need to wait for a order to ship from Newegg.Sad

Reply
#85
Many boards also work with a 20 pin atx switch.
Did you checked the manual?
| myHTPC |
Reply
#86
(2014-03-24, 07:32)BORIStheBLADE Wrote: I totally forgot I need a 24 pin PSU. I thought my spare PSU would work but its only 20pin. Now I need to wait for a order to ship from Newegg.Sad

You can buy a 20 to 24pin adapter for a few dollars

Link Depot POW-ADT-2024 Power Adapter Cable
Reply
#87
Most PSUs with a 24 pin have the 21-24 pin as a separate removable plug?
Modded MK1 NUC - CLICK ----- NUC Wiki - CLICK

Bay Trail NUC FTW!

I've donated, have you?

Reply
#88
(2014-03-24, 07:47)-DDD- Wrote: Many boards also work with a 20 pin atx switch.
Did you checked the manual?
Yeah manual only mentions 24 pin, nothing about 20.

(2014-03-24, 08:39)tetsuya Wrote:
(2014-03-24, 07:32)BORIStheBLADE Wrote: I totally forgot I need a 24 pin PSU. I thought my spare PSU would work but its only 20pin. Now I need to wait for a order to ship from Newegg.Sad

You can buy a 20 to 24pin adapter for a few dollars

Link Depot POW-ADT-2024 Power Adapter Cable

Good to know for next time, thanks!

(2014-03-24, 09:33)jammyb Wrote: Most PSUs with a 24 pin have the 21-24 pin as a separate removable plug?

No this old one isn't like that.

Reply
#89
I am using a 120W DC 20-pin powerboard (wiring harness is 20-to-24pin) and 60W AC adapter with the Gigabyte J1800 board. It works fine. I can test a 20-pin picoPSU if you want.
Reply
#90
I am running the Gigabyte j1800 D2H with a Pico PSU 90W which has only 20 Pins. So far it doesn't seem to be a problem Smile
I only have one issue which could be related to the PSU. When I pull the plug to save energy over night and connect it again the next morning, the system wont start. I have to wait a minute first, before I can start the system.
Reply
  • 1
  • 4
  • 5
  • 6(current)
  • 7
  • 8
  • 54

Logout Mark Read Team Forum Stats Members Help
Bay Trail-D motherboards8