Discussion of the Tabor / A1222 mainboard
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4818 from 2009/1/28
    From: Delaware, USA
    Quote:

    Kronos wrote:
    Neither Tabor nor X5000 have Altivec.


    Yes, an issue with e5500 cored cpus, but not as big a deal as the e500v2's fpu.
    "Never attribute to malice what can more readily explained by incompetence"
  • »08.10.16 - 22:41
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 10380 from 2003/5/22
    From: Germany
    > firmware updates are extra

    That was ACube, not A-Eon.

    > Efika level performance in real life.

    This would depend on the amount of floating point code in one's particular real life. Tabor's integer performance per core should be about 4 times Efika 5200B's.
  • »08.10.16 - 23:21
    Profile
  • Paladin of the Pegasos
    Paladin of the Pegasos
    redrumloa
    Posts: 1322 from 2003/4/13
    Quote:

    Andreas_Wolf wrote:

    That was ACube, not A-Eon.


    Isn't Acube a partner with A-Eonkit for the Tabor?

    [ Edited by redrumloa 09.10.2016 - 13:17 ]
  • »09.10.16 - 00:28
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 10380 from 2003/5/22
    From: Germany
    >>> firmware updates are extra

    >> That was ACube, not A-Eon.

    > Isn't Acube a partner with A-Eonkit for the Tabor?

    Yes, they are. That still doesn't mean A-Eon will adopt ACube's firmware update policy, does it?
  • »09.10.16 - 08:13
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4818 from 2009/1/28
    From: Delaware, USA
    Quote:

    Andreas_Wolf wrote:
    > The releasedate should now be importent, to keep some momentum....

    It's gotten a little bit nearer: https://www.youtube.com/watch?v=fMSvxQIfq2w


    Well, they always said they would focus on this once the X5000 was ready for delivery, and that is close, so...

    It will be interesting to see how the whole issue of the P1022's spe fpu plays out.
    In software specifically compiled for the A1222, no problem, BUT...
    "Never attribute to malice what can more readily explained by incompetence"
  • »24.11.16 - 15:17
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 10380 from 2003/5/22
    From: Germany
    > It will be interesting to see how the whole issue of the P1022's spe fpu plays out.
    > In software specifically compiled for the A1222, no problem, BUT...

    A year later and Daniel "Daytona675x" Müßener posts "a good demonstration of why you must not mix FPU and SPE code as soon as floating point values are involved" (i.e. mixing as in SPE-compiled executable calls FPU-compiled library or vice versa):

    https://www.facebook.com/daniel.mussener/posts/2086799091346527
  • »03.01.18 - 12:39
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    takemehomegrandma
    Posts: 2558 from 2003/2/24
    Quote:

    Andreas_Wolf wrote:
    > It will be interesting to see how the whole issue of the P1022's spe fpu plays out.
    > In software specifically compiled for the A1222, no problem, BUT...

    A year later and Daniel "Daytona675x" Müßener posts "a good demonstration of why you must not mix FPU and SPE code as soon as floating point values are involved" (i.e. mixing as in SPE-compiled executable calls FPU-compiled library or vice versa):

    https://www.facebook.com/daniel.mussener/posts/2086799091346527


    Daniel Müßener: "Note: VoxelNoid is barely playable (if shadowing is turned off it's okay) on the Tabor as of now, the Warp3D SI driver is simply doing to much stuff internally that involves FPU instructions and therefore constantly and frequently triggers the emulation.

    My tests revealed that this is apparently due to FPU-register-based memcpy (when drawing vertex data and when uploading textures) and RGBA-color-format-conversions from packed uint32 to unpacked float[4] done using FPU code.
    Hans has been informed, hopefully his time allows it to make the respective code more Tabor-friendly soon "


    "Barely playable", and we are not exactly talking about Star Wars Battlefront 2.

    And so it begins, the parallel development of both "normal" code and "franken-code", just to satisfy a small group of customers with a one-time-deal HW that will never see the sun again.

    It makes so much sense that only Amiga makes it possible!

    :-)



    [ Edited by takemehomegrandma 03.01.2018 - 13:12 ]
    MorphOS is Amiga done right! :-)
    MorphOS NG will be AROS done right! :-)
  • »03.01.18 - 13:04
    Profile
  • Just looking around
    Posts: 6 from 2017/10/7
    VoxelNoid works fluently at 720p with all settings to max on my mdd g4 1.25 soooo...
  • »03.01.18 - 14:39
    Profile
  • Priest of the Order of the Butterfly
    Priest of the Order of the Butterfly
    koszer
    Posts: 877 from 2004/2/8
    From: Poland
    Quote:

    CMTX wrote:
    VoxelNoid works fluently at 720p with all settings to max on my mdd g4 1.25 soooo...


    Oh yeah? But does it get anywhere near the spectacular count of 1075037508 FPS? LOL...
    The issue described by Daytona will make porting for Tabor oh-so-more-funny for the devs.
  • »03.01.18 - 16:01
    Profile
  • Paladin of the Pegasos
    Paladin of the Pegasos
    redrumloa
    Posts: 1322 from 2003/4/13
    Quote:

    ppcamiga1 wrote:
    http://www.os4welt.de/viewtopic.php?f=4&t=1670&sid=e2c6fbaecf723b05ca59e3531417292b

    So, I was right about apes throwing shit on a-eon and powerpc technology in general.

    Some people simply don't like bitter truth.

    btw netbsd has port for e500v2 for three yrs, and users of this port also report

    that trapping fpu instructions on e500v2 is not a problem, and has no impact on performance.




    This comment has not aged well.
  • »03.01.18 - 23:22
    Profile
  • Paladin of the Pegasos
    Paladin of the Pegasos
    redrumloa
    Posts: 1322 from 2003/4/13
    Quote:

    redrumloa wrote:
    Quote:

    Cego wrote:
    dude, i think some of you dont get the point.

    the situation is not about "slow cpu" vs "faster cpu"

    its rather "slow cpu" vs. "nothing at all". so be grateful that at least somethings still coming. we're not a huge community and its not like we can choose everything we dream about. What have you done for the community? maybe thats the question you should ask yourself.


    Did I just walk into the Twilight Zone? Why exactly should a MorphOS user be grateful A-Eonkit will be selling a Raspberry Pi class motherboard for an exorbitant amount of money? A board that the MorphOS Team have rightfully stated would be a waste of time supporting?

    Also, why do you even pose a question about "done for the community"? The only thing end users need to do is vote with their dollars. A-Eonkit are running a commercial business, not a nonprofit charity.

    This board won't be for sale with a working OS4.1 port until at least 2018. Do you really expect people to pretend like a ~500Euro rpi level board in 2018 will be a good idea?


    I can see the future it seems.
  • »03.01.18 - 23:27
    Profile
  • Butterfly
    Butterfly
    Posts: 80 from 2017/9/10
    sorry but i have to make a quick reply
    tabor is not like a pi in performances just because a pi3 overkill it. the only good point of tabor is the pcie slot that the pi dont have.

    just for, this is the imx6 board from freescale with arm cpu ... dont think you will have this performances on tabor

    https://www.youtube.com/watch?v=N3QY74keH_E

    ogre demo https://www.youtube.com/watch?v=foOm9OQ9IE8


    about A-Cube and A-eon for the tabor ... are you sure than a-cube made the tabor or do something about?

    [ Edited by tlosmx 04.01.2018 - 09:34 ]
  • »04.01.18 - 10:24
    Profile
  • Moderator
    Kronos
    Posts: 1872 from 2003/2/24
    Quote:

    redrumloa wrote:
    Quote:

    redrumloa wrote:

    This board won't be for sale with a working OS4.1 port until at least 2018. Do you really expect people to pretend like a ~500Euro rpi level board in 2018 will be a good idea?


    I can see the future it seems.


    A day after the 31st December 2018 it will be the year 2019!!

    I can see the future too !!!!!



    Now tell me this weeks Lotto numbers and we might be onto something ;)
    --------------------- May the 4th be with you ------------------
    Mother Russia dance of the Zar, don't you know how lucky you are
  • »04.01.18 - 11:33
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 10380 from 2003/5/22
    From: Germany
    >> A-Eonkit will be selling a Raspberry Pi class motherboard

    > tabor is not like a pi in performances just because a pi3 overkill it.

    There are various Raspberry Pi models with CPUs ranging from 700 MHz single-core ARMv6 to 1.2 GHz quad-core ARMv8, where the latter should be over twice as fast as the former on per-core basis alone. So talking about "Raspberry Pi class" or similar without specifying the model doesn't make sense anyway.
    Tabor is definitely way faster than all Raspberry Pi 1 models (700 MHz ARMv6). Regarding quad-core Raspberry Pi 2 (900 MHz, ARMv7 initially, ARMv8 in later revisions), Tabor should be faster on per-core base but is slower in SMP. Comparing against quad-core Raspberry Pi 3 (1.2 GHz ARMv8), one's as good/bad as the other on a per-core base, so Tabor obviously loses in SMP.

    > this is the imx6 board from freescale with arm cpu

    No, this is a board from FEDEVEL/VOIPAC with an ARM-based i.MX6 CPU from NXP :-)

    > dont think you will have this performances on tabor

    The i.MX6 used on the board is a quad-core, yes. Comparing the 1.2 GHz Cortex-A9 and the 1.2 GHz e500v2 on a per-core base, one's as good/bad as the other.

    > are you sure than a-cube made the tabor or do something about?

    As is well known, the Tabor has been designed by Varisys. Regarding the ACube connection:

    "Working in cooperation with ACube srl, a beta test programme is about to commence"
    http://www.a-eon.com/news/News_Release_A1222.pdf

    ...and there was this in 2013:

    http://www.a-eon.com/news/14-12-2013.html
    http://www.acube-systems.biz/news.php?id=127

    ...and this in 2015:

    http://www.a-eon.com/PDF/News_Release_Summit.pdf
    http://www.acube-systems.biz/news.php?id=135
  • »04.01.18 - 13:31
    Profile
  • Butterfly
    Butterfly
    Posts: 80 from 2017/9/10
    Tabor is all Varisys and acube dint nothing . i was thinking the same about tabor that there was acube hands on it ... but no they at the end didnt nothing on that crappy.

    about freescale or nxp ... we are speak now about the same company ;-)

    about pi3 ... blender using the 4 cores on the pi is faster in blender bench compared the x5000-20 that use two cores . tested by me... pratically if you render something you will have same time on a 35 euro computer compared a 2300 euro computer :P

    darkplace on SPE Linux on Radeonhd = 70 fps on Tabor if some one have better result because what i have is old pls post here it .Thanks

    [ Edited by tlosmx 04.01.2018 - 18:21 ]
  • »04.01.18 - 19:14
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4818 from 2009/1/28
    From: Delaware, USA
    Quote:

    tlosmx wrote:
    Tabor is all Varisys and acube dint nothing . i was thinking the same about tabor that there was acube hands on it ... but no they at the end didnt nothing on that crappy.


    Actually, I was under the impression that Acube was to develop the firmware for the platform.
    "Never attribute to malice what can more readily explained by incompetence"
  • »04.01.18 - 19:17
    Profile
  • Butterfly
    Butterfly
    Posts: 80 from 2017/9/10
    Quote:

    Actually, I was under the impression that Acube was to develop the firmware for the platform.

    I was thinking the same...but ..

    [ Edited by tlosmx 04.01.2018 - 18:25 ]
  • »04.01.18 - 19:24
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4818 from 2009/1/28
    From: Delaware, USA
    Quote:

    tlosmx wrote:
    Quote:

    Actually, I was under the impression that Acube was to develop the firmware for the platform.

    I was thinking the same...but ..


    Its not really an issue for me, as Tabor holds no appeal for me.
    But Acube would be the natural company for the T2081 laptop project to go to for firmware (after all, they've already been tapped for design work).
    "Never attribute to malice what can more readily explained by incompetence"
  • »04.01.18 - 19:30
    Profile
  • Butterfly
    Butterfly
    Posts: 80 from 2017/9/10
    i can say sure yes about openlaptop and acube :)
    i know many people involved in this project and for sure acube is involved


    [ Edited by tlosmx 04.01.2018 - 18:46 ]
  • »04.01.18 - 19:45
    Profile
  • rob
  • Acolyte of the Butterfly
    Acolyte of the Butterfly
    rob
    Posts: 110 from 2008/7/22
    Quote:

    koszer wrote:
    Quote:

    CMTX wrote:
    VoxelNoid works fluently at 720p with all settings to max on my mdd g4 1.25 soooo...


    Oh yeah? But does it get anywhere near the spectacular count of 1075037508 FPS? LOL...
    The issue described by Daytona will make porting for Tabor oh-so-more-funny for the devs.


    Quote:

    Three Tower57 port Tabor fun facts:

    - the latest SPE build reaches up to 80 fps. This simply demolishs the sam460ex, factor 2.

    - when running the latest FPU (!) build on the Tabor it reaches up to 40 fps (!) (old OS version, first fpu-emu).

    - when running the latest FPU build on the sam460ex it reaches up to 40 fps.
    So the Tabor is even en par with the sam460 if running the version which is incompatible with its FPU :)

    And before you say sth. a la "then you optimized away most fpu instructions"...
    Wrong, think, man, think - and don't forget that the SPE build runs twice as fast... Yeah, the SPE is a pretty fast FPU unit indeed, but looking at the emu-speed that whole A1222 delivers (y)

    Still anybody out there making up stories about the A1222's performance compared to the SAMs or about the power of the SPE? LOL


    Quote:

    Yes, I know: VoxelNoid and VoxelBird on Tabor are nice, but I suppose most of you were curious about Wings ;)

    And here it is:
    The current Wings Remastered Demo (no, it's not finished yet) build, compiled for SPE, running on the Tabor.

    As being said before the Warp3D driver still contains FPU code (which ironically is actually inside to improve performance on systems with standard FPU) which is the cause for severe performance penalties because it constantly triggers the exception-based FPU emulation.

    But even then: as you can see everything but strafing and dogfight run smooth as butter, and even those two mission types are still playable! The Tabor is certainly a pretty powerful beast if unleashed!

    Now imagine if that driver was modified to not trigger the emulation anymore...

    As always with the Tabor videos here: the sound-driver is not done yet, therefore nothing but silence for now, unfortunately.


    So it's not all doom and gloom after all.

    [ Edited by rob 04.01.2018 - 20:19 ]
  • »04.01.18 - 21:18
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 10380 from 2003/5/22
    From: Germany
    > about freescale or nxp ... we are speak now about the same company ;-)

    My point was that the board you mentioned was not from this company, just the SoC is.

    > about pi3 ... blender using the 4 cores on the pi is faster in blender bench compared
    > the x5000-20 that use two cores . [...] if you render something you will have same time

    So the X5000/20 is about twice as fast per core as the Raspberry Pi 3. And as the X5000/20 is about twice as fast as Tabor, Tabor is about as fast per core as the Raspberry Pi 3. Just as I said :-)
  • »05.01.18 - 01:19
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 10380 from 2003/5/22
    From: Germany
    >> The issue described by Daytona will make porting for Tabor oh-so-more-funny for the devs.

    > [...]
    > So it's not all doom and gloom after all.

    The comment you replied to was about the compatibility problems developers will have to cope with, not about speed (non-)issues. What does speed help when it's not working right, except that it'll crash and burn or show nonsense faster?
  • »05.01.18 - 01:31
    Profile
  • rob
  • Acolyte of the Butterfly
    Acolyte of the Butterfly
    rob
    Posts: 110 from 2008/7/22
    Quote:

    Andreas_Wolf wrote:
    >> The issue described by Daytona will make porting for Tabor oh-so-more-funny for the devs.

    > [...]
    > So it's not all doom and gloom after all.

    The comment you replied to was about the compatibility problems developers will have to cope with, not about speed (non-)issues. What does speed help when it's not working right, except that it'll crash and burn or show nonsense faster?


    Only if they are using SPE code. If you can get the same performance from Tabor as the Sam460 with standard FPU code then you may not have a need for special binaries.
  • »05.01.18 - 08:03
    Profile